ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007) 589 ndash 601

Experimental Investigation of Noise AnnoyanceCaused by High-speed Trains

Bert De Coensela Dick Botteldoorena Birgitta Berglund1 Mats E Nilsson2 Tom De Muer1Peter Lercher3

1 Acoustics Group Department of Information Technology Ghent University St Pietersnieuwstraat 41B-9000 Ghent Belgium

2 Goumlsta Ekman Laboratory Karolinska Institute and Stockholm University Sweden3 Department of Hygiene and Social Medicine Medical University of Innsbruck Austria

SummaryA field experiment was conducted to investigate the possible differences in perceived annoyance of noise causedby the traffic on a highway by conventional trains and by high-speed trains both conventional and magneticlevitation The design of the experiment was different from earlier research in many ways Most importantly itwas conducted in a realistic setting a holiday cottage and during the tests the participants were engaged in lightdaily activities Traffic noise was reproduced in an ecologically valid way through loudspeakers placed outdoorsA stepwise selection of panelists was based on a screening questionnaire that was administered at the doorstepof 1500 persons living in the test site surroundings The 100 panelists were selected to be representative of theDutch population The LAeq-annoyance relationships determined for the conventional high-speed train and forthe magnetic levitation high-speed train did not differ significantly The annoyance differences observed could beexplained in terms of train noise differences in rise time and in propagation effects due to the distance betweenthe track and the listening (recording) position

PACS no 4350Qp 4350Lj 4350Rq

1 Introduction

A difference in perceived annoyance between train andother traffic noise at the same average sound level hasbeen observed in several field studies in the past [1 2 3]In a number of countries this observation has led to lessrestrictive regulation or railway bonus for train noise rel-ative to noise from other sources such as highways ma-jor roads or aircraft (usually 5 dB(A) see eg the Ger-man French or Austrian legislation) With the introduc-tion of high-speed trains and train-like transportation sys-tems based on magnetic levitation (maglev) the questionhas arisen whether a difference in perceived annoyance oftrain and highway noise still exists In particular it is prob-able that spectral changes due to a higher fraction of aero-dynamic noise and shorter rise times due to high speedswould change the perception of high-speed train and ma-glev train noiseThe main goal of this research was to investigate the

possible differences in annoyance on the one hand be-tween magnetic levitation and conventional high speedtrains and on the other hand between highway noise andtrain noise Next to this the influence of some additional

Received 7 July 2006accepted 16 March 2007

factors on noise annoyance was studied such as the dis-tance between the source and the listener the speed of thesource and the rise time of the sound

Prior laboratory research by Fastl and Gottschling [4]showed no significant difference in noise annoyance of aTransrapid 07 maglev train at a speed of 400 kmh and aconventional high-speed train at a speed of 250 kmh ifpresented at a comparable A-weighted equivalent soundlevel Conversely Neugebauer and Ortscheid [5 6] con-cluded that maglev noise annoyance differed markedlyfrom that of a conventional train An experiment by Vos[7 8] showed that if the outdoor ASEL (A-weightedSound Exposure Level) was set equal the Transrapid 08maglev train was more annoying than a conventional in-tercity train and approximately equally annoying as roadtraffic

In addition to the fact that these previous studies wereinconclusive a few factors of potential importance werenot explicitly considered in previous work Firstly in lis-tening experiments with short fragments of noise listen-ers assess the perceived annoyance of noise Such assess-ments cover both perceived loudness and perceived char-acter of noise (eg see [9]) However for short fragmentsof sound the temporal effect may partly contribute to theannoyance differences between trains and continuous traf-fic sound Longer exposures containing several train pas-sages as well as the typical quiet periods in between were

copy S Hirzel Verlag middot EAA 589

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

necessary to include in this experiment Secondly in reallife sounds may be annoying also because they changeadversely the current soundscape or are associated with acultural change or they interfere with activities for exam-ple reading or relaxation This latter ldquoacuterdquo but importantaspect of noise annoyance is not captured in traditional lis-tening experiments but is possible to assess if the exper-iment is designed in the right way as shown in [10 11]Finally it is well known from environmental noise ques-tionnaire surveys that personal factors such as noise sensi-tivity influence annoyance reports [12 13] Some of thesefactors have also been observed in listening experiments[14 15] Therefore the results may not be valid and itmay not be possible to generalize beyond the subgroupif this subgroup had not been selected carefully to matchthe population concerning these critical factors

Recently a small annoyance survey was conducted nearthe maglev line in Shanghai [16] Such annoyance surveysare not possible in Europe because the magnetic levitationsystem has not yet been implemented but for a test facil-ity Therefore a field experiment was specially designed tosolve as many of the above mentioned issues as possibleThe experiment differed significantly from the above citedearlier research A realistic home-like setting was createdin which the panelists were asked to relax while exposed tolonger fragments of sound including quiet periods (Sec-tion 21) Traffic noise was reproduced in an ecologicallyvalid way using multiple loudspeakers outdoors to simu-late pass-by sound (Section 22) The set of panelists wasselected to be representative of the Dutch population infactors known to be important modifiers of noise annoy-ance (Sections 23) For the outline of the listening testmenus of train passages delimiting longer exposure dura-tions were used (Section 24) The method of master scal-ing by which perceived annoyance was scaled calibratedthe scales used by different participants to a common mas-ter scale (Section 25)

2 The experiment

21 Sound reproduction in a realistic setting

As a natural setting a holiday cottage in Westkapelle(Zeeland The Netherlands) was selected because of itsquiet environment and accessibility During the experi-ment subgroups of participants were seated in the liv-ing room reading a magazine engaging in light conver-sation or having something to drink Figure 1 shows thecottage and its environment Much attention was paid tocreating a realistic reproduction of the three-dimensionalindoor sound field produced by a moving train outside thehouse Observe that the goal was to obtain an ldquoecologi-cally validrdquo [17 18 19] reproduction rather than physi-cal precision ie the methods materials and setting areaimed at approximating the real-life at-home situation un-der study It is difficult to produce the effect of any houseby signal processing and playback through headphones orindoor loudspeakers and to accomplish a natural feeling

Figure 1 Entrance through the garden to the holiday cottage (atthe left) where the experiment was performed

of the sound field Therefore it was decided to reproducethe sound field as recorded outdoors outside the experi-mental cottage

A similar approach has recently been described in [11]where a laboratory test room was modified to mimick astandard living room Traffic sounds were reproduced frombehind a fake window by a 16-channel loudspeaker setupapplying the wave field synthesis technique [20] Our fieldexperiment was conducted in a real living room with thesound reproduction system installed outside the house inopen air Our setup therefore favors a more realistic andecologically valid context in exchange for a less accu-rate sound field reproduction as compared to [11] Thetwo-channel recording was however accurate enough forproducing a realistic three-dimensional representation in-doors Neither approach can however completely relateto and account for the participants earlier experiences ofnoise annoyance in their own natural home environmentIn a small field study the selected technique for realis-

tic indoor representations of train passages was checkedperceptually and acoustically for low speed trains at shortdistances In another house situated close to a densely traf-ficed railway track the indoor sound fields of real trainsand of artificially reproduced train noise were comparedTwo loudspeakers placed outdoors were used for repro-ducing the artificial passages of train noise The procedureconsisted of 2 phases Firstly during the passage of a trainthe sound was recorded outdoor by 2 BampK 4189 free fieldmicrophones separated 20m from each other along thetrack for calibration the faccedilade level was also recordedAt the same time a binaural recording was made inside thehouse Secondly the recorded sound was played back by2 loudspeakers in front of the house separated about 10mfrom each other and along the same horizontal axis as seenfrom the window The volume was adjusted to reproducethe 13-octave band spectrum at the faccedilade as accurately aspossible Simultaneously a binaural recording was againmade inside the house Ideally both binaural recordings(real train and reproduced train) should be equal For mosttrains the artificial sound could not be distinguished from

590

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

control room

openedwindow

H

6

7

3 4 5

2

1

L

faccedilade level

subwoofer

Figure 2 Schematic drawing of the experimental cottage (not toscale) The different seats of the panelists are shown (1ndash7) aswell as the seat of the experimental leader (L) and of an artificialhead (H) for binaural recordings

the real sound by audition The two spectra were in most13-octave bands within an error of ca 5 dB neverthelessit was decided to introduce an equalizer for fine-tuning anda subwoofer for reproducing more accurately the low fre-quency part of a moving high speed train

Figure 2 shows a floor plan of the living room and thecontrol room of the experimental cottage together withthe final loudspeaker setup The sounds were played backon a regular PC equipped with a high quality audio cardlocated in the control room The sound signal was thenequalized by an Allen amp Heath 12-channel mixer and 31-channel equalizer Subsequently the sound signal was am-plified by a Bose 802II amplifier and fed to 4 Bose loud-speakers which were placed stacked per 2 on 2 tripodstands at a height of ca 15m and to a HK Audio SL218Apowered subwoofer on the ground All loudspeakers wereplaced outside the house in front of the main windowThe 2 loudspeaker tripods were placed ca 10m from eachother perpendicular at 3m distance to the faccedilade Thesubwoofer was placed in front of the window in betweenboth tripods at about 50 cm from the faccedilade This loud-speaker setup was located in front of a slightly openedwindow of the experimental cottage invisible to the pan-elists entering the house

The faccedilade level was measured continuously during allexperimental sessions using a BampK Investigator 2260sound level meter with a BampK 4189 free field microphone(5 cm from the window at 75 cm height) The sound levelmeter was also used to calibrate the playback system Forthis calibration pink noise was played back and adjustedto give a faccedilade level of 91 dB with a flat 13-octave bandspectrum The equalizer accomplished a flat (plusmn 3 dB forall 13-octave bands) spectrum between 30Hz and 16 kHzThe faccedilade attenuation and the reverberation in the experi-mental room both modify the spectrum and temporal char-

acteristics of the sound Since it would not be possible tosee a train passage from the window because of plenty oftrees a visual presentation of passing trains was consid-ered not appropriate

22 Sample collection and preparation

Two-channel recordings were conducted for three typesof trains Two microphones were placed at 20m distancefrom each other along the track 15m above ground levelTGV trains at high speed were recorded in Beloeil (Bel-gium) a site near the TGV connection between Brus-sels and Lille (France) Dutch intercity (IC) trains of thenew type (duplex) were recorded in Oudenbosch nearRoosendaal (The Netherlands) at this same site the TGVtraveling at low speed from Brussels to Rotterdam wasalso recorded At the maglev test track in Lathen (Ger-many) the Transrapid 08 train was recorded at speeds ofapprox 200 kmh 300 kmh and 400 kmh For the mas-ter scaling references the sound of the E40 highway wasalso recorded near Ghent (Belgium) To be able to assessthe influence on annoyance of the distance to the track4 recording distances were included (25m 50m 100mand 200m) All recordings were made in free field withoutnoise barriers Not only the spectrum and temporal changewere reproduced exactly but also the sound level as if thehouse would have been located at the measurement site

From the many train recordings made at each site thepassages of highest quality were selected in each categoryof recording and for these 45-second single passage frag-ments were cut It was important to expose the panelists tosufficient and natural durations of noise Therefore theyhad to be exposed to ldquoexperimental soundrdquo during at least10 minutes (henceforth called a menu) To create a realis-tic exposure situation within a 10-minute menu it shouldbe composed of the same train type at the same distanceand speed Menus with 2 or 4 passages were created be-cause 4 passages in 10 minutes already represents the nat-ural time-schedule maximum and 2 passages in 10 min-utes represents a minimum passage rate with inter-passagebackground sound Less than two passages are not usefulbecause the inter-event silence is non-defined in this caseApart from the 45-second fragments recorded at the fourdistances to the track a 10-minute highway sound wasrecorded at 50m distance to the closest lane

Table I summarizes the sound exposure (ASEL) andsound levels (LAeq45s) associated with the 45-second pas-sages used in the 10-minute menus It should be mentionedthat the level of the IC train at 25m happens to be lowerthan the level at 50m This inconsistency is due to the factthat the selected high-quality sound fragments do not nec-essarily originate from identical train passages There isalways a natural spread in the speed and the number ofwagons of the different passages of the same type of trainAs an illustration Figures 3 and 4 show the A-weightedsound exposure level in 13-octave bands for some of theexperimental traffic sounds as recorded in free field

For master scaling 7 traffic-noise-like reference soundfragments of 45 seconds duration with sound pressure

591

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

Table I Sound exposure levels (ASEL) for one 45-second train passage sound level (LAeq45s) of one 45-second train passage and ofhighway traffic and sound level (LAeq10min) of the 10-minute menus of the experiment at 25m to 200m distance to track or route(all free field recordings) The train noise LAeq10min values are given for the 2-train menu to obtain the LAeq10min values for thecorresponding 4-train menu add 3 dB

Sound source Outdoor ASEL [dB(A)] Outdoor LAeq45s [dB(A)] Outdoor LAeq10min [dB(A)]25m 50m 100m 200m 25m 50m 100m 200m 25m 50m 100m 200m

Maglev 200 kmh 801 729 713 597 636 564 548 432 553 481 465 349300 kmh 863 830 803 696 698 665 638 531 615 582 555 448400 kmh 926 887 852 704 761 722 687 539 678 639 604 456

TGV 140 kmh 841 783 736 644 676 618 571 479 593 535 488 396300 kmh 928 906 869 830 763 741 704 665 680 658 621 582

IC 140 kmh 750 809 724 620 585 644 559 455 502 561 476 372

Highway free flow minus minus minus minus 716 661 626 553 minus 653 minus minus

Figure 3 Sound exposure level (ASEL) in 13-octave bands offour different types of traffic sounds all recorded during 45 sec-onds in free field at a distance of 50m to the track (or highwayroute) ( ) a passage of a maglev train traveling at 400 kmh( ) a passage of a TGV traveling at 300 kmh ( ) a passage ofan IC train traveling at 140 kmh and ( ) a highway with freeflow traffic

Figure 4 Sound exposure level (ASEL) in 13-octave bands of amaglev train traveling at 400 kmh recorded during 45 secondsin free field at various distances to the track ( ) 25m ( ) 50m( ) 100m and ( ) 200m

level spanning the whole experimental range were in-cluded in the experiment A 45-second fragment of thehighway noise recorded at 50m distance to the highwaywas used as the centre reference sound A filter which at-tenuates the sound at frequencies below 500Hz by 3 dBand above 500Hz by 6 dB was applied 3 times to produce3 reference sounds with varying level all below the levelof the centre reference sound giving the impression that

the source is further away In the same way a filter thatamplifies the sound at frequencies below 500Hz by 3 dBand above 500Hz by 6 dB was used to generate 3 refer-ence sounds with varying level higher than the level of thecentre reference sound

23 Selection of a representative panel

In contrast to previous experimental work on noise annoy-ance caused by high speed trains in which small ldquoconve-nientrdquo samples of test persons were recruited the selectionof panelists was here made to guarantee a representativesample of panelists A questionnaire was administered atthe doorstep of the homes of approximately 1500 personsall living within a distance of 15 km from the experimentalsite In an introductory letter one inhabitant of the housewas invited to participate in the study The prerequisiteswere that (s)he had to fill in and send the questionnaireback to the address on the enclosed stamped envelope Acompensation of C 100 was offered for participation

The questionnaire contained selected questions that hadbeen asked to a representative sample of the target popula-tion in a recent survey The structure of the Dutch popula-tion was inferred to be representative from a recent RIVMsurvey [21] and partly from a Eurobarometer question-naire Our questionnaire contained (standard) questions onenvironmental noise as regards perception annoyance andsleep disturbance Included were evaluations of the qualityof the neigbourhood in terms of housing and environmen-tal pollution of other types than noise as well as evalu-ations of overall satisfaction with the current living situ-ation Other questions addressed basic demographic vari-ables such as age gender education housing family sizeand work arrangements A set of questions were also in-cluded on general and mental health hearing ability en-vironmental background opinion and worry and environ-mental sensitivity

A procedure to draw panelists representative of thetarget population from the 255 replies received involvedthree stages Stage 1 removed potential panelists on thebasis of their age and hearing ability (information had al-ready been given in the introductory letter) Stage 2 furtherremoved those that were very dissimilar from the typical

592

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

Table II Comparison between the panelists and the referencepopulation on various criteria Mean and standard deviation isshown the results for the second series of criteria are on an 11-point scale and vary from 0 (not at all bad) to 10 (very good)

Criterium Participants Reference

Gender [ male female] 51 49 48 52Age [year] 451 plusmn 134 456 plusmn 177

Noise sensitivity 51 plusmn 24 46 plusmn 26

Quality of traffic noise in the 66 plusmn 24 64 plusmn 23living environment

Quality of the living 76 plusmn 14 73 plusmn 13environment

Feeling afraid or frightened 24 plusmn 20 23 plusmn 21

Dutch person on the basis of binary coding of most of theother criteria included in the questionnaire This stage im-plicitly assessed individual responses on the questions asregards their concordance with the response profile of thetypical Dutch person in the reference survey Stage 3 fi-nally selected panelists on the basis of fuzzy resemblanceto the typical Dutch person on the most critical criteria ofannoyance surveys such as age gender education noisesensitivity feeling afraid or frightened hearing train noiseat home quality of traffic noise in the living environmentquality of the living environment general health and ill-ness Finally ca 100 representative participants were se-lected Table II shows a comparison of the panelists withthe Dutch target population as regards the mean and stan-dard deviation of some of the selection criteria used andmentioned above

24 Listening test outline

Four to six panelists jointly participated in a session Theoverall structure and time schedule of the listening exper-iment was identical for each group of panelists It startedwith a 14-minute training session during which the pan-elists were asked to scale each of the 7 reference (high-way) sounds two times (in random order) Thereafter7 10-minute menus were played of which the first menualways was the highway traffic menu A short break wasthen taken and the training session was repeated afterwhich again 7 new 10-minute menus were played Afterthis experiment with menus a more conventional psychoa-coustical listening test was conducted in which the pan-elists had to scale 45-second excerpts of all transport noisestimuli used in the menu experiment The duration of anexperimental session was on average about 4 hours Toillustrate how the listening test was performed Figure 5shows the sound level in dB(A) rerecorded in front of thefaccedilade during one of the panelist groupsrsquo listening exper-imentIn all two times 6 train menus were presented to each

panelist It was decided that within one set of 6 trainmenus conventional trains (IC or high-speed) should notbe mixed with magnetic levitation trains By this separa-

tion it was possible to include a retrospective evaluationover the last hour as well From previous experience it wasknown that the order of the menu pesentations might af-fect the results Half of the panelists were therefore pre-sented the maglev train sounds first the other half the con-ventional trains first A singular session consisted of thesame number of passages inside the menus This avoidsthat panelists would concentrate on counting events Fi-nally since one distance to the track would create a nat-ural setting large distances were never mixed with shortdistances in the menus of a session

During the experimental sessions perceived noise an-noyance of all transport noises was scaled with the methodof free-number magnitude estimation [22] The panelistswere asked to write down their magnitude assessments ondifferent coloured pieces of paper Before the start of theexperiment the panelists were instructed to select an ap-propriate number and then to double this number if theyfound the next stimulus to be twice as annoying to makethe number three times larger if they found the next stim-ulus to be three times as annoying etc and to scale 0 ifthey considered it not to be annoying at all For each 45-second sound (training sessions and conventional listeningtest) a conditional question was included ldquoTo what extentwould you be annoyed by this traffic sound if you heardit while relaxingrdquo For each 10-minute menu a very simi-lar but retrospective question was asked ldquoTo what extentwere you annoyed by traffic sound during the previous pe-riodrdquo In these latter questions we explicitly did not wantto refer to train noise since we wanted the panelists to de-cide themselves whether the sound period they last heardsounded like train-contaminated or not

25 Master scaling

In all experimental sessions the 7 road-traffic-noise-likereference sounds helped the panelists to define their ownscaling context The annoyance values given to these refer-ence sounds made it possible to control for the individualpanelistsrsquo choice-of-number behaviour in scaling the tar-get train sounds It would also control the influence of per-sonal factors such as noise sensitivity To get rid of theseeffects each individual panelistrsquos annoyance scale was cal-ibrated by the aid of the reference to the common masterscale [23]

A graphical illustration of the master scale transforma-tion applied to the annoyance reference data of one of thepanelists is given in Figure 6 The average annoyance re-ported for each of the 7 reference sound levels of roadtraffic noise is plotted in lin-log coordinates against theirsound levels LAeq45s measured at the faccedilade Individualpsychophysical functions are fitted to the reference data(open circles) They are of the form

Ar = a + b logSr (1)

where Ar is the reported annoyance during the train-ing session and logSr is the corresponding ldquoroad trafficnoiserdquo reference (r) sound level in dB(A) The constants a

593

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

Figure 5 Sound pressure level rerecorded in front of the faccedilade during one panelist groupsrsquo participation in the whole listening testtwo training sessions two menu sessions and one conventional psychoacoustical experiment with references

0

10

20

30

40

50

60

70

80

40 50 60 70 80 90

R = 55 master-scaled annoyance

A = 70 assessment

annoyance

road traffic reference sound level [dB(A)]

Figure 6 Calculation of master-scaled annoyance using one pan-elistrsquos empirical psychophysical function of the reference sounds(data points with solid line) and the master function for the samesounds (dashed line obtained as average function for all pan-elists)

and b will be different for each panelist and will depend ontheir choice-of-number behaviour in the particular scalingcontext The empirically derived master functions for thegroup of 100 panelists (dashed line in Figure 6) were thenused to transform the free number magnitude estimationsof the train or road traffic menus for each individual Aeto the corresponding annoyance values R in master scaleunits

R = minus629 + 145Ae minus a

b (2)

The slope of the master function was set to 145 whichis the average slope of all the individual psychophysicalfunctions whereas the intercept was set to produce a valueof ldquozerordquo for the most quiet train menu The reason for thelatter choice was that a majority of the panelists (84)reported their annoyance to be zero for this menu and amajority of the panelists reported annoyance to be greaterthan zero for all other menus

The choice of a logarithmic psychophysical function(equation 1) was a compromise In previous magnitudeestimation experiments of loudness [23 9] a power func-tion of the form logA = c + d logS was found to fit the

Table III Test-retest reliability of panelistsrsquo perceived noise an-noyance of the 7 reference road traffic sounds Each cell containsan arithmetic mean of Pearsonrsquos coefficient (r) and its standarderror Ts1 Training session 1 Ts2 Training session 2 Ct Con-ventional test

Ts1 Ts2 CtSet 1 Set 2 Set 1 Set 2 Set 1

Ts22 082plusmn0015

Ts21 086 087plusmn0014 plusmn0016

Ts22 086 088 087plusmn0017 plusmn0020 plusmn0019

Ct1 083 083 082 085plusmn0015 plusmn0021 plusmn0019 plusmn0020

Ct2 084 085 081 084 082plusmn0015 plusmn0019 plusmn0016 plusmn0019 plusmn0015

empirical data best However in this experiment noise an-noyance rather than loudness was scaled and thus obvi-ously also a value of zero (= not at all annoyed) had tobe handled although the noise was heard and its loudnesswas above zero The power function (after removal of ze-ros) did not fit the data better than the chosen logarithmicfunction

26 Data quality analysis

The master scaling made it possible to investigate the qual-ity of the experimental data in two ways as panelistsrsquo test-retest reliability and as their scaling ability The 7 refer-ence sounds were presented 6 times to each panelist twicein the two training sessions and twice in the last conven-tional listening test The set of 6 reference scale valueswere used to determine each panelistrsquos test-retest reliabil-ity of annoyance Table III shows the Pearsonrsquos coefficientof correlation for these 6 annoyance scales averaged overall panelists The test-retest reliability was very good be-tween 081 and 088 and the standard error was low be-tween 0014 and 0019

The deviation from the proposed master function (equa-tion 1) was used to assess the data quality and annoyance

594

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

Table IV Distribution of constants of the panelistsrsquo individualpsychophysical functions (Eq 1) The number of data sets refersto the average of 4 or 2 raw annoyance values which was takenfor each of the 7 reference sounds to calculate the psychophysicalfunctions

Data Psychophysical functionSets r2 a b

Training session 1amp2 4 0947 minus6727 1449plusmn0077 plusmn6128 plusmn1230

Conventional test 2 0881 minus4757 1105plusmn0118 plusmn4817 plusmn0948

scaling ability for each panelist and to trace errors and in-accuracies Table IV shows the distribution of constants ofthe panelistsrsquo individual psychophysical functions (equa-tion 1) The average annoyance variance explained bysound level (LAeq45s) of the reference road traffic soundswas found to range from 88 to 95 All panelists wereable to produce acceptable individual logarithmic func-tions of annoyance as a function of sound level to the refer-ence They have thus produced acceptable annoyance datain order to transform these to a common master scale ofannoyance no panelists were excluded from further dataanalysis

3 Results

The main listening experiment with menus differed fromprevious laboratory experiments in a number of aspectsOne important novelty is that participants were asked tojudge annoyance over a longer period of time mdash Fastl andGottschlingrsquos experiment [4] forms an exception Duringthe 10-minute periods the panelists were engaged in lowattention relaxing activities such as reading a magazinemaking a conversation or having something to drink In or-der to find out how this new approach affected the resultsa subsequent experiment was included which was morecomparable to earlier experiments on train noise (eg [8])

31 Main field experiment with 10-minute menus

The panelistsrsquo master scale values of annoyance were av-eraged for each menu in the field experiment A step-wise multiple linear regression analysis was performedwith average master scaled annoyance as dependent vari-able and (a) time averaged A-weighted faccedilade exposureLAeq10min (b) distance to the source (logarithmic) and(c) source type as independent variables Because of itslegislative importance in the Netherlands the faccedilade ex-posure was preferred to the actual panelistsrsquo noise expo-sure Faccedilade exposure was calculated from the sound lev-els measured on the recording sites since the faccedilade levelsmeasured during the experiment also contain noise fromwind and rain It has to be noted that the actual soundexposure levels experienced by panelists participating ina single experimental session may differ because of dif-ferent seating positions However personal characteristicssuch as noise sensitivity will have a much larger influence

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq10min

(a)

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq10min

(b)

Figure 7 Average master scaled annoyance of the menus ver-sus LAeq10min (a) for 2 events per 10-minute menu and (b) for4 events per 10-minute menu for different types of train sounds( ) IC train ( ) TGV and ( ) maglev train In comparisonthe annoyance for the highway traffic ( ) is also shown Stan-dard error on means is indicated as well as the master function(dashed line)

on perceived annoyance as compared to the influence of aslightly different exposure

Table V summarizes the results In the first modelsound level was the only independent variable this modelexplained 80 of the variance in annoyance In the secondmodel distance to track was added to sound level as an in-dependent variable this model increased the variance ex-plained to 85 (F-change = 1449 df1 = 1 df2 = 46 p lt0001) Thus distance to source explained a significant ad-ditional part of the annoyance variance not accounted forby sound level In the third model source type was in-cluded as a third independent variable along with soundlevel and distance Source type was defined on a nominalscale MAGLEV TGV IC and HIGHWAY It was intro-duced in the analysis as three dummy variables coded 0and 1 (the highway noise source type corresponds to thecase that these variables are all zero) The inclusion ofsource type did not increase significantly (F-change lt 10)the proportion of variance explained This suggests thatstatistically there is no additional contribution of sourcetype on perceived annoyance over and above the effects ofsound level and distance It can therefore be concluded thatmagnetic-levitation based transportation systems are notsignificantly more annoying than conventional rail basedsystems (same faccedilade LAeq and same distance are pre-requisites) Moreover railway noise was not found to be

595

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

Table V Stepwise multiple regression analysis of acoustic variables on perceived annoyance of train and highway traffic sounds for themain field experiment with 10-minute menus The Pearsonrsquos correlation coefficients of the variables entered in the regression analysisare shown at the bottom lowastp lt 005 lowastlowastp lt 001 lowastlowastlowastp lt 0001

Model Model Model fit increase F-change Independent Variables Coefficient t-valuefit (r2) (r2-change)

1 080 080 18748lowastlowastlowast LAeq10min [dB(A)] 118 1369lowastlowastlowast

2 085 005 1449lowastlowastlowast LAeq10min [dB(A)] 092 917lowastlowastlowast

log10(distance [m]) minus1074 minus381lowastlowastlowast

3 085 000 013 LAeq10min [dB(A)] 096 822lowastlowastlowast

log10(distance [m]) minus1017 minus333lowastlowastMAGLEV [01] 145 027TGV [01] 085 016IC [01] 227 040

Label Variable ANN LEQ DIST MAG TGV

ANN Annoyance [master scale]LEQ LAeq10min [dB(A)] 0894DIST log10(distance [m]) minus0754 minus0659MAG MAGLEV [01] minus0023 minus0038 0009TGV TGV [01] 0132 0224 0006 minus0682IC IC [01] minus0179 minus0290 0004 minus0433 minus0308

systematically less annoying than highway traffic noiseThis means that no support for a railway bonus was foundin this experiment at least it was not as obvious that itcould be observed using linear statistics Figure 7 givesan overview of the annoyance functions for the 10-minutemenus as a function of LAeq10min The dashed line indi-cates the master function of annoyance for the road-traffic-like sounds used as referencesThe shorter rise time of the noise of arriving high speed

trains may create more annoyance than a conventionaltrain can do Figure 8 Panel a shows the rise speeds indB(A)s in proportion to circle sizes These values werecalculated for all sound events included in this experi-ment by fitting a straight line through the initial increasein sound level The accelerating growth of annoyance withincreasing LAeq may be explained by the rise time In Fig-ure 8 Panel b the size of the circles is instead propor-tional to the distance to the track For LAeq in the intervalbetween 50 and 65 dB(A) annoyance is clearly lower fortrain passages at larger distances than for train passagesat closer distances or road traffic noise (dashed line) Thiscould indicate that a possible noise annoyance bonus fortrain noise would only hold at larger distances from thetrack and only in the latter LAeq interval

32 Conventional listening test

In the conventional listening experiment the sounds werepresented as short 45-second fragments containing thesound of one train passage and highway excerpts Figure 9shows the results of these master scaled annoyance val-ues as a function of time averaged A-weighted faccedilade ex-posure LAeq45s A railway penalty can be observed bothin regard to the artificial reference sounds as well as to

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq10min

(a)

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq10min

(b)

Figure 8 Average master scaled annoyance of the menus versusLAeq10min showing (a) the noise event rise speed and (b) the dis-tance to the track as the size of the circles The master functionis also indicated (dashed line)

the highway sounds Figure 10 shows the annoyance as afunction of rise speed ( Panel a) and distance to the track(Panel b)

596

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

Table VI Stepwise multiple regression analysis of acoustic variables on perceived annoyance of train sounds (no highway trafficsounds) for the conventional listening test (45-second passages) The Pearsonrsquos correlation coefficients of the variables entered in theregression analysis are shown at the bottom lowastp lt 005 lowastlowastp lt 001 lowastlowastlowastp lt 0001

Model Model Model fit increase F-change Independent Variables Coefficient t-valuefit (r2) (r2-change)

1 095 095 42017lowastlowastlowast LAeq45s [dB(A)] 167 2050lowastlowastlowast

2 098 003 1250lowastlowastlowast LAeq45s [dB(A)] 123 1104lowastlowastlowast

Speed [kmh] 002 203log10(distance [m]) minus178 minus077Rise speed [dB(A)s] 063 365lowastlowast

3 099 001 098 LAeq45s [dB(A)] 108 685lowastlowastlowast

Speed [kmh] 003 209log10(distance [m]) minus476 minus146Rise speed [dB(A)s] 058 330lowastlowast

MAGLEV [01] minus070 minus037TGV [01] 166 104

Label Variable ANN LEQ SPD DIST RISE MAG

ANN Annoyance [master scale]LEQ LAeq45s [dB(A)] 0975SPD Speed [kmh] 0646 0541DIST log10(distance [m]) minus0613 minus0667 minus0001RISE Rise speed [dB(A)s] 0885 0804 0708 minus0437MAG MAGLEV [01] 0070 minus0017 0552 0000 0188TGV TGV [01] 0207 0246 minus0190 0000 0100 minus0707

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq45sec

Figure 9 Average master scaled annoyance versus LAeq45s forthe conventional listening test for different types of train sounds( ) IC train ( ) TGV and ( ) maglev train In comparison theannoyance for the highway traffic ( ) is also shown Standard er-ror on means is indicated as well as the master function (dashedline)

A stepwise multiple linear regression analysis was alsoperformed separately for the train noises (Table VI) Thefirst model in which sound level LAeq45s was included asthe only independent variable explained 95 of the vari-ance in annoyance In the second model train speed dis-tance to the track and rise speed were added to sound levelas independent variables This increased the variance ex-plained to 98 (F-change = 1250 df1 = 3 df2 = 19p lt 0001) Apart from sound level also rise speed con-

tributed significantly to the variance explained The thirdmodel in which train type was added as an independentdummy variable did not significantly increase the propor-tion explained variance (F-change lt 10) These resultssuggest that in this conventional listening test there is nodifference in perceived annoyance between different typesof trains over and above the effect of sound level and risespeed

One has to note that the number of responses to eachstimulus was smaller in the main experiment (10-minutemenus) than in the conventional listenig test (45-secondpassages) This explains why the standard errors are lowerand the explained variance is higher in the latter experi-ment

4 Discussion

The annoyance results of the present field experiment areclose to residentsrsquo everyday reality although comparisonwith published studies is somewhat limited Previous lab-oratory experiments on noise annoyance of conventionalIC and high-speed trains specifically magnetic levitationtrains [8 5] report significant differences for these typesof sound In particular the results have shown that forthe same LAeq high-speed trains were more annoyingthan other trains Compared to road traffic noise the citedstudies claimed a lower annoyance level for conventionaltrains In the present field experiment we did not find sup-port for any annoyance difference between various types

597

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

0

10

20

30

40

50

60

70

80

40 50 60 70 80 90

annoyance

[maste

rscale

units]

L [dB(A)]Aeq45sec

(a)

0

10

20

30

40

50

60

70

80

40 50 60 70 80 90

annoyance

[maste

rscale

units]

L [dB(A)]Aeq45sec

(b)

Figure 10 Average master scaled annoyance versus LAeq45s

showing (a) the noise event rise speed and (b) the distance tothe track as the size of the circles The master function is alsoindicated (dashed line)

of trains and road traffic Some possible explanations willbe given in the following subsections

41 Realistic listening situation with 10-minutemenus

The experiment was performed in a realistic setting inwhich outdoor transportation noise was reproduced andnatural outdoor-to-indoor sound propagation characteris-tics were utilized (slightly open window) This setting pro-vided a realistic sound environment indoors Subgroups ofpanelists were kept indoors during the four-hour experi-ments and upon request annoyance to transport noise wasreported with reference to 10-minute periods

Because trains run on expected schedules to which peo-ple habituate the experimental situation in classical exper-iments is rather unrealistic The experimental one-passagesituation [8] requires full attention and will have a largevariation of train sounds compared with a particular rail-way track The outcome will to a large extent depend onthe experimental context that is the variation introducedin the experiment by selecting stimuli and using randompresentation orders Random orders of recordings can beselected and arranged so that annoyance judgments on cat-egory scales plotted against sound level differentiate wellor not well on type of transport In the present field experi-ment sub-context in sessions was kept invariant similar tothe situation on a real railway track The judgmental con-

text will then be much more restricted as is the case whenliving along one railway track

Next to this the annoyance reports of the one 45-secondtrain passage were higher than those of two passages ofthe same train within the 10-minute menus This is all inorder because the two types of annoyance were masterscaled in order to become comparable over experimentalsessions When judging 45-second train passages imme-diately after exposure it is quiet clear that the task is toassess the annoyance of that particular train passage (orother sounds that were presented) However when askedto assess the annoyance retrospectively of the transportnoise during the last 10-minute menu (eg two train pas-sages) the panelist will have to choose a strategy on howto go about this For example the annoyance may only bereferred to the two noise-stimulus periods or to the whole10-minute period (menu) It has been shown that the noiseannoyance of two overlapping (equal) noises would be ex-pected always to be less than the arithmetic sum of the twoannoyances (for a review see [24]) It is more uncertainhow total annoyance of two train passages separated intime will actually be acquired A laboratory experimentwhich included long sound fragments [4] has not foundthe above-mentioned annoyance difference between differ-ent train types which is in line with our results

42 Advanced scaling methodology

Long-term retrospective annoyance asked for in question-naire surveys has typically been assessed on categoryscales (eg [3]) A response category is then implicitlypostulated to be identical for every participant by verballabeling of the two end points or of every response boxalso the intervals between categories are assumed to be thesame However this assumption does not hold true [25]eg in questionnaire surveys the response criteria (scalevalue or category borders) for annoyance are much higherfor respondents in low noise areas as compared to those inhighly exposed areas The most well known scaling bias inlaboratory experiments is the context effect in which par-ticipants distribute their responses over the ldquofullrdquo range ofcategories independent of the size of the exposure range(for a review see [22]) In the process of using categoryscales floor and ceiling effects on annoyance may also ap-pearTo avoid uncontrolled context effects an invariant sound

level range of references was used as the annoyance con-text in the present field experiment Continuous road trafficnoise was chosen as a reference instead of multiple eventsounds because it is simpler to reproduce in future stud-ies To avoid the scaling bias of category scales the methodof magnitude estimation was chosen in which participantswere free to use the range of numbers they felt comfortablewith Master scaling was applied to these individual an-noyance estimates involving a transformation function toa common master scale defined by the references whichsound levels defined the scaling context In theory thismaster scale transformation will calibrate the loudness-dependence of noise annoyance whereas the relative con-

598

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

tribution to noise annoyance from qualitative content (egthe type of sound the time pattern and cues for speed anddistance) will hopefully be unchanged

Earlier research has shown that master scaling with ref-erences works well for loudness or annoyance of a one-occasion target exposure that is when repeated exposureis unfeasible (eg experiments with long duration expo-sures) or impossible (eg questionnaire surveys in fieldstudies) an example can be found in [23] The results ob-tained from the present field experiment are probably morereliable than the results that would have been obtained bycategory scaling The test-retest reliability of the panelistsrsquomagnitude estimates of annoyance of the reference soundswas found to be very good (above 08) compared to thereliability of 072 obtained in [8] for a group of 12 muchyounger subjects Considering that our panelists all werenaiumlve participants they also each produced high qualitypsychophysical functions for the reference as discussedin Section 26

43 Other possible explanations

There are several reasons why other investigators havefound a railway bonus (for a review see [3]) which wasnot found in this field experiment One of the reasons forfinding a railway bonus for short (one minute) noises inlistening experiments may be that the relation betweenloudness and LAeq is inherently different for train androad traffic noise Indeed some researchers have arguedthat noise annoyance evaluation in listening tests of shortsounds actually is close to a perceptual loudness evalua-tion (however see [26] on differences between loudness-based and quality-based perceived annoyance) If Zwickerloudness is a good first estimate of perceptual loudnessthe difference between train noise (of different types) andhighway noise would be seen in a Zwicker loudness ver-sus LAeq plot (Figure 11) Because the IC train noise usedin the present experiment was the noise of modern ratherquiet trains a few older and noisier IC train models wereadded in this acoustic analysis At levels above 65 dB(A)TGV and maglev trains seem to be a little louder thanhighway traffic or older IC trains However this effect onZwicker loudness is not significant and does therefore notsupport a railway bonus of 5 dB(A) stipulated in severalcountriesrsquo legislation Rather it seems to be a good actionto start to replace old IC trains by new ones The railwaybonus was originally based on studies with rather old low-speed trains and with much less dense traffic intensity thannowadaysThe intermittent character of railway noise could also be

an explanation for the railway bonus However this doesnot hold for aircraft noise which is also intermittent thiscan be explained by a difference in exposure In the case ofaircraft noise the exposure is on top of buildings and on allfaccedilades In the case of road traffic noise the probability ishigh that there are local roads also but there is a possibil-ity for a ldquoquiet siderdquo people are less annoyed if quiet sidesare available [27] In the case of railway noise there is alow probability for the presence of more than one track

Zw

ickerlo

udness

[sone]

L [dB(A)]Aeq45sec

0

10

20

30

40

50

40 50 60 70 80

Figure 11 Zwicker loudness versus LAeq45s for different types oftransportation sounds ( ) IC train ( ) TGV ( ) maglev train( ) highway traffic and ( ) some additional noisier IC trains(older type)

so the exposure will also be directed at only one faccediladeIn comparing road traffic and trains the faccedilade insulationwill be more effective in the case of train noise because ofthe smaller low-frequency proportion associated to trainnoise In comparing aircraft and trains which are both in-termittent the indoor exposure is certainly more intensivefor aircraft Considering these arguments it seems obviousthat aircraft is more annoying than road traffic which ismore annoying than train However faccedilade reduction wastaken into account in the present field experiment and stillthere was no clear railway bonus found Compared withthe field condition with closed windows and the faccediladefilter used in [8] a partially open window was used in thepresent field experiment which could explain this

In surveys questioning people at their home a lowerreported annoyance for train noise compared to highwaytraffic noise was observed in a particular range of noiselevels Most of the possible explanations proposed in liter-ature conflict with the fact that this railway bonus wouldbe observed in experiments based on single passages Wemention just a few The typical character of train noise andthe concentration of the sound energy in short time inter-vals may be advantageous with regard to activity distur-bance If the level is sufficiently low the probability ofnoticing the train noise is small compared to the prob-ability of noticing the sound of a continuous source Inaddition to physical differences in the sound the ldquogreenimagerdquo of trains as a means of transportation may add tothe acceptability of the source and thus increase the tol-erance to its noise that is as long as train passages arenot too frequent However a more recent hedonic pric-ing study found that householders in Birmingham placea greater value on reductions in railway noise than in roadtraffic noise [28] Cross-cultural studies (in field and lab-oratory context) have shown that a railway bonus is notuniversal [29 30] which would favor the argument aboveIt has further been shown that the bonus varies depend-ing on the (multiple) exposure situation [31] Based on theabove only part of the effect is supposed to be visible in

599

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

field experiments such as the one reported of in this paperPart of the effect is precisely what is observed

5 Conclusion

This study has shown that in an ldquoat homerdquo like contextnoise annoyance caused by different types of trains at thesame average outdoor faccedilade exposure level is not signifi-cantly different In particular magnetic levitation systemsare not more annoying than conventional high speed trainswhich is in agreement with earlier research Noise annoy-ance caused by conventional trains was not found to besignificantly lower than annoyance caused by TGVrsquos ormaglev trains at the same average faccedilade exposure Fieldsurveys have shown that for the same average sound levelrailway noise causes less annoyance or highly annoyedpersons than highway traffic noise Although our field ex-periment included several factors that may contribute tothis effect we could not observe it

More insight may be gained by taking into account thepsychoacoustic characteristics of the noise exposure andthe relevant personal factors of the panelists This paperhas focused on discussing the experimental methodologyin great detail and on presenting the results as a functionof the average outdoor faccedilade exposure level since thisis the main noise legislation indicator used in the Nether-lands Results of a detailed psychoacoustic analysis willbe reported in a future paper

Acknowledgments

This research was financed by the Project Group Zuider-zeelijn of the Ministry of Public Transport Public Worksand Water Management in the Netherlands The membersof the project steering committeemdashGilles Janssen (dBVi-sion) Annemarie Ruysbroek (RIVM) Martin van denBerg (VROM) and Pieter Jansse (Project Group Zuider-zeelijn) mdash are acknowledged for their valuable input Wealso appreciate the experimental assistance provided byMs Ingrid Decoster Finally we would like to thank thereviewers for their valuable comments and suggestions

References

[1] J M Fields J G Walker Comparing the relationships be-tween noise level and annoyance in different surveys a rail-way vs aircraft and road traffic comparison J Sound Vib81 (1982) 51ndash80

[2] U Moehler Community response to railway noise A re-view of social surveys J Sound Vib 120 (1988) 321ndash332

[3] H M E Miedema H Vos Exposure-response relation-ships for transportation noise J Acoust Soc Am 104(1998) 3432ndash3445

[4] H Fastl G Gottschling Subjective evaluation of noiseimmissions from transrapid Proceedings of the 25thInternational Congress on Noise Control Engineering(Intermiddotnoise) Liverpool UK Juli 1996 F A Hill R Law-rence (eds)

[5] D Neugebauer J Ortscheid Geraumluschbewertung des Tran-srapid Ist der Transrapid wie die Bahn zu beurteilenBerlin Umweltbundesamt Texte 25 1997

[6] D Neugebauer J Ortscheid Geraumluschbewertung des Tran-srapid Fortschritte der Akustik mdash DAGA 97 Kiel Ger-many 1997

[7] J Vos Annoyance caused by magnetic levitation trainTransrapid 08 mdash a laboratory study TNO report TM-03-C001 The Netherlands Organisation for Applied ScientificResearch (TNO) Soesterberg The Netherlands Jan 2003

[8] J Vos Annoyance caused by the sounds of a magnetic lev-itation train J Acoust Soc Am 115 (2004) 1597ndash1608

[9] B Berglund A Preis Is perceived annoyance more sub-ject-dependent than perceived loudness Acta Acusticaunited with Acustica 83 (1997) 313ndash319

[10] M Ishibashi K Ueno H Tachibana M Watanabe Sub-jective experiment on annoying sounds in living environ-ments Proceedings of the 18th International Congress onAcoustics (ICA) Kyoto Japan 2004

[11] J Maillard J Martin P Champelovier J Lambert Percep-tive evaluation of road traffic noise inside buildings using acombined image and wave field synthesis system Proceed-ings of CFADAGArsquo04 Strasbourg France Maumlrz 2004

[12] R F S Job Community response to noise A review offactors influencing the relationship between noise exposureand reaction J Acoust Soc Am 83 (1988) 991ndash1001

[13] H M E Miedema H Vos Demographic and attitudinalfactors that modify annoyance from transportation noise JAcoust Soc Am 105 (1999) 3336ndash3344

[14] W Ellermeier M Eigenstetter K Zimmer Psychoacousticcorrelates of individual noise sensitivity J Acoust SocAm 109 (2001) 1464ndash1473

[15] D Vaumlstfjaumlll M Kleiner T Gaumlrling Affective reactions tointerior aircraft sounds Acta Acustica united with Acustica89 (2003) 693ndash701

[16] T Feng C Xiao-hong L Tan-feng A social survey onthe effects of high-speed maglev noise on residents inShangai Proceedings of The 2005 Congress and Expo-sition on Noise Control Engineering (Intermiddotnoise) Rio deJaneiro Brazil Aug 2005

[17] J J Gibson The ecological approach to visual perceptionLawrence Erlbaum Associates Inc Hillsdale New JerseyUSA 1979

[18] J G Neuhoff Ecological psychoacoustics Introductionand history ndash In Ecological Psychoacoustics J G Neuhoff(ed) Elsevier Academic Press London UK 2004

[19] C Guastavino B F G Katz J-D Polack D J LevitinD Dubois Ecological validity of soundscape reproductionActa Acustica united with Acustica 91 (2005) 333ndash341

[20] A J Berkhout D de Vries P Vogel Acoustic control bywave field synthesis J Acoust Soc Am 93 (1993) 2764ndash2778

[21] E A M Franssen J E F van Dongen J M H Ruys-broek H Vos R K Stellato Hinder door milieufactorenen de beoordeling van de leefomgeving in Nederland In-ventarisatie verstoringen 2003 RIVM report 815120 001National Institute for Public Health and the Environment(RIVM) Bilthoven The Netherlands 2004

[22] L E Marks D Algom Psychophysical scaling ndash In Mea-surement Judgment and Decision Making M H Birn-baum (ed) Academic Press New York 1998

[23] B Berglund Quality assurance in environmental psycho-physics ndash In Ratio Scaling of Psychological Magnitudes -In Honor of the Memory of S S Stevens S J BolanowskiG A Gescheider (eds) Erlbaum Hillsdale N J 1991

600

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

necessary to include in this experiment Secondly in reallife sounds may be annoying also because they changeadversely the current soundscape or are associated with acultural change or they interfere with activities for exam-ple reading or relaxation This latter ldquoacuterdquo but importantaspect of noise annoyance is not captured in traditional lis-tening experiments but is possible to assess if the exper-iment is designed in the right way as shown in [10 11]Finally it is well known from environmental noise ques-tionnaire surveys that personal factors such as noise sensi-tivity influence annoyance reports [12 13] Some of thesefactors have also been observed in listening experiments[14 15] Therefore the results may not be valid and itmay not be possible to generalize beyond the subgroupif this subgroup had not been selected carefully to matchthe population concerning these critical factors

Recently a small annoyance survey was conducted nearthe maglev line in Shanghai [16] Such annoyance surveysare not possible in Europe because the magnetic levitationsystem has not yet been implemented but for a test facil-ity Therefore a field experiment was specially designed tosolve as many of the above mentioned issues as possibleThe experiment differed significantly from the above citedearlier research A realistic home-like setting was createdin which the panelists were asked to relax while exposed tolonger fragments of sound including quiet periods (Sec-tion 21) Traffic noise was reproduced in an ecologicallyvalid way using multiple loudspeakers outdoors to simu-late pass-by sound (Section 22) The set of panelists wasselected to be representative of the Dutch population infactors known to be important modifiers of noise annoy-ance (Sections 23) For the outline of the listening testmenus of train passages delimiting longer exposure dura-tions were used (Section 24) The method of master scal-ing by which perceived annoyance was scaled calibratedthe scales used by different participants to a common mas-ter scale (Section 25)

2 The experiment

21 Sound reproduction in a realistic setting

As a natural setting a holiday cottage in Westkapelle(Zeeland The Netherlands) was selected because of itsquiet environment and accessibility During the experi-ment subgroups of participants were seated in the liv-ing room reading a magazine engaging in light conver-sation or having something to drink Figure 1 shows thecottage and its environment Much attention was paid tocreating a realistic reproduction of the three-dimensionalindoor sound field produced by a moving train outside thehouse Observe that the goal was to obtain an ldquoecologi-cally validrdquo [17 18 19] reproduction rather than physi-cal precision ie the methods materials and setting areaimed at approximating the real-life at-home situation un-der study It is difficult to produce the effect of any houseby signal processing and playback through headphones orindoor loudspeakers and to accomplish a natural feeling

Figure 1 Entrance through the garden to the holiday cottage (atthe left) where the experiment was performed

of the sound field Therefore it was decided to reproducethe sound field as recorded outdoors outside the experi-mental cottage

A similar approach has recently been described in [11]where a laboratory test room was modified to mimick astandard living room Traffic sounds were reproduced frombehind a fake window by a 16-channel loudspeaker setupapplying the wave field synthesis technique [20] Our fieldexperiment was conducted in a real living room with thesound reproduction system installed outside the house inopen air Our setup therefore favors a more realistic andecologically valid context in exchange for a less accu-rate sound field reproduction as compared to [11] Thetwo-channel recording was however accurate enough forproducing a realistic three-dimensional representation in-doors Neither approach can however completely relateto and account for the participants earlier experiences ofnoise annoyance in their own natural home environmentIn a small field study the selected technique for realis-

tic indoor representations of train passages was checkedperceptually and acoustically for low speed trains at shortdistances In another house situated close to a densely traf-ficed railway track the indoor sound fields of real trainsand of artificially reproduced train noise were comparedTwo loudspeakers placed outdoors were used for repro-ducing the artificial passages of train noise The procedureconsisted of 2 phases Firstly during the passage of a trainthe sound was recorded outdoor by 2 BampK 4189 free fieldmicrophones separated 20m from each other along thetrack for calibration the faccedilade level was also recordedAt the same time a binaural recording was made inside thehouse Secondly the recorded sound was played back by2 loudspeakers in front of the house separated about 10mfrom each other and along the same horizontal axis as seenfrom the window The volume was adjusted to reproducethe 13-octave band spectrum at the faccedilade as accurately aspossible Simultaneously a binaural recording was againmade inside the house Ideally both binaural recordings(real train and reproduced train) should be equal For mosttrains the artificial sound could not be distinguished from

590

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

control room

openedwindow

H

6

7

3 4 5

2

1

L

faccedilade level

subwoofer

Figure 2 Schematic drawing of the experimental cottage (not toscale) The different seats of the panelists are shown (1ndash7) aswell as the seat of the experimental leader (L) and of an artificialhead (H) for binaural recordings

the real sound by audition The two spectra were in most13-octave bands within an error of ca 5 dB neverthelessit was decided to introduce an equalizer for fine-tuning anda subwoofer for reproducing more accurately the low fre-quency part of a moving high speed train

Figure 2 shows a floor plan of the living room and thecontrol room of the experimental cottage together withthe final loudspeaker setup The sounds were played backon a regular PC equipped with a high quality audio cardlocated in the control room The sound signal was thenequalized by an Allen amp Heath 12-channel mixer and 31-channel equalizer Subsequently the sound signal was am-plified by a Bose 802II amplifier and fed to 4 Bose loud-speakers which were placed stacked per 2 on 2 tripodstands at a height of ca 15m and to a HK Audio SL218Apowered subwoofer on the ground All loudspeakers wereplaced outside the house in front of the main windowThe 2 loudspeaker tripods were placed ca 10m from eachother perpendicular at 3m distance to the faccedilade Thesubwoofer was placed in front of the window in betweenboth tripods at about 50 cm from the faccedilade This loud-speaker setup was located in front of a slightly openedwindow of the experimental cottage invisible to the pan-elists entering the house

The faccedilade level was measured continuously during allexperimental sessions using a BampK Investigator 2260sound level meter with a BampK 4189 free field microphone(5 cm from the window at 75 cm height) The sound levelmeter was also used to calibrate the playback system Forthis calibration pink noise was played back and adjustedto give a faccedilade level of 91 dB with a flat 13-octave bandspectrum The equalizer accomplished a flat (plusmn 3 dB forall 13-octave bands) spectrum between 30Hz and 16 kHzThe faccedilade attenuation and the reverberation in the experi-mental room both modify the spectrum and temporal char-

acteristics of the sound Since it would not be possible tosee a train passage from the window because of plenty oftrees a visual presentation of passing trains was consid-ered not appropriate

22 Sample collection and preparation

Two-channel recordings were conducted for three typesof trains Two microphones were placed at 20m distancefrom each other along the track 15m above ground levelTGV trains at high speed were recorded in Beloeil (Bel-gium) a site near the TGV connection between Brus-sels and Lille (France) Dutch intercity (IC) trains of thenew type (duplex) were recorded in Oudenbosch nearRoosendaal (The Netherlands) at this same site the TGVtraveling at low speed from Brussels to Rotterdam wasalso recorded At the maglev test track in Lathen (Ger-many) the Transrapid 08 train was recorded at speeds ofapprox 200 kmh 300 kmh and 400 kmh For the mas-ter scaling references the sound of the E40 highway wasalso recorded near Ghent (Belgium) To be able to assessthe influence on annoyance of the distance to the track4 recording distances were included (25m 50m 100mand 200m) All recordings were made in free field withoutnoise barriers Not only the spectrum and temporal changewere reproduced exactly but also the sound level as if thehouse would have been located at the measurement site

From the many train recordings made at each site thepassages of highest quality were selected in each categoryof recording and for these 45-second single passage frag-ments were cut It was important to expose the panelists tosufficient and natural durations of noise Therefore theyhad to be exposed to ldquoexperimental soundrdquo during at least10 minutes (henceforth called a menu) To create a realis-tic exposure situation within a 10-minute menu it shouldbe composed of the same train type at the same distanceand speed Menus with 2 or 4 passages were created be-cause 4 passages in 10 minutes already represents the nat-ural time-schedule maximum and 2 passages in 10 min-utes represents a minimum passage rate with inter-passagebackground sound Less than two passages are not usefulbecause the inter-event silence is non-defined in this caseApart from the 45-second fragments recorded at the fourdistances to the track a 10-minute highway sound wasrecorded at 50m distance to the closest lane

Table I summarizes the sound exposure (ASEL) andsound levels (LAeq45s) associated with the 45-second pas-sages used in the 10-minute menus It should be mentionedthat the level of the IC train at 25m happens to be lowerthan the level at 50m This inconsistency is due to the factthat the selected high-quality sound fragments do not nec-essarily originate from identical train passages There isalways a natural spread in the speed and the number ofwagons of the different passages of the same type of trainAs an illustration Figures 3 and 4 show the A-weightedsound exposure level in 13-octave bands for some of theexperimental traffic sounds as recorded in free field

For master scaling 7 traffic-noise-like reference soundfragments of 45 seconds duration with sound pressure

591

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

Table I Sound exposure levels (ASEL) for one 45-second train passage sound level (LAeq45s) of one 45-second train passage and ofhighway traffic and sound level (LAeq10min) of the 10-minute menus of the experiment at 25m to 200m distance to track or route(all free field recordings) The train noise LAeq10min values are given for the 2-train menu to obtain the LAeq10min values for thecorresponding 4-train menu add 3 dB

Sound source Outdoor ASEL [dB(A)] Outdoor LAeq45s [dB(A)] Outdoor LAeq10min [dB(A)]25m 50m 100m 200m 25m 50m 100m 200m 25m 50m 100m 200m

Maglev 200 kmh 801 729 713 597 636 564 548 432 553 481 465 349300 kmh 863 830 803 696 698 665 638 531 615 582 555 448400 kmh 926 887 852 704 761 722 687 539 678 639 604 456

TGV 140 kmh 841 783 736 644 676 618 571 479 593 535 488 396300 kmh 928 906 869 830 763 741 704 665 680 658 621 582

IC 140 kmh 750 809 724 620 585 644 559 455 502 561 476 372

Highway free flow minus minus minus minus 716 661 626 553 minus 653 minus minus

Figure 3 Sound exposure level (ASEL) in 13-octave bands offour different types of traffic sounds all recorded during 45 sec-onds in free field at a distance of 50m to the track (or highwayroute) ( ) a passage of a maglev train traveling at 400 kmh( ) a passage of a TGV traveling at 300 kmh ( ) a passage ofan IC train traveling at 140 kmh and ( ) a highway with freeflow traffic

Figure 4 Sound exposure level (ASEL) in 13-octave bands of amaglev train traveling at 400 kmh recorded during 45 secondsin free field at various distances to the track ( ) 25m ( ) 50m( ) 100m and ( ) 200m

level spanning the whole experimental range were in-cluded in the experiment A 45-second fragment of thehighway noise recorded at 50m distance to the highwaywas used as the centre reference sound A filter which at-tenuates the sound at frequencies below 500Hz by 3 dBand above 500Hz by 6 dB was applied 3 times to produce3 reference sounds with varying level all below the levelof the centre reference sound giving the impression that

the source is further away In the same way a filter thatamplifies the sound at frequencies below 500Hz by 3 dBand above 500Hz by 6 dB was used to generate 3 refer-ence sounds with varying level higher than the level of thecentre reference sound

23 Selection of a representative panel

In contrast to previous experimental work on noise annoy-ance caused by high speed trains in which small ldquoconve-nientrdquo samples of test persons were recruited the selectionof panelists was here made to guarantee a representativesample of panelists A questionnaire was administered atthe doorstep of the homes of approximately 1500 personsall living within a distance of 15 km from the experimentalsite In an introductory letter one inhabitant of the housewas invited to participate in the study The prerequisiteswere that (s)he had to fill in and send the questionnaireback to the address on the enclosed stamped envelope Acompensation of C 100 was offered for participation

The questionnaire contained selected questions that hadbeen asked to a representative sample of the target popula-tion in a recent survey The structure of the Dutch popula-tion was inferred to be representative from a recent RIVMsurvey [21] and partly from a Eurobarometer question-naire Our questionnaire contained (standard) questions onenvironmental noise as regards perception annoyance andsleep disturbance Included were evaluations of the qualityof the neigbourhood in terms of housing and environmen-tal pollution of other types than noise as well as evalu-ations of overall satisfaction with the current living situ-ation Other questions addressed basic demographic vari-ables such as age gender education housing family sizeand work arrangements A set of questions were also in-cluded on general and mental health hearing ability en-vironmental background opinion and worry and environ-mental sensitivity

A procedure to draw panelists representative of thetarget population from the 255 replies received involvedthree stages Stage 1 removed potential panelists on thebasis of their age and hearing ability (information had al-ready been given in the introductory letter) Stage 2 furtherremoved those that were very dissimilar from the typical

592

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

Table II Comparison between the panelists and the referencepopulation on various criteria Mean and standard deviation isshown the results for the second series of criteria are on an 11-point scale and vary from 0 (not at all bad) to 10 (very good)

Criterium Participants Reference

Gender [ male female] 51 49 48 52Age [year] 451 plusmn 134 456 plusmn 177

Noise sensitivity 51 plusmn 24 46 plusmn 26

Quality of traffic noise in the 66 plusmn 24 64 plusmn 23living environment

Quality of the living 76 plusmn 14 73 plusmn 13environment

Feeling afraid or frightened 24 plusmn 20 23 plusmn 21

Dutch person on the basis of binary coding of most of theother criteria included in the questionnaire This stage im-plicitly assessed individual responses on the questions asregards their concordance with the response profile of thetypical Dutch person in the reference survey Stage 3 fi-nally selected panelists on the basis of fuzzy resemblanceto the typical Dutch person on the most critical criteria ofannoyance surveys such as age gender education noisesensitivity feeling afraid or frightened hearing train noiseat home quality of traffic noise in the living environmentquality of the living environment general health and ill-ness Finally ca 100 representative participants were se-lected Table II shows a comparison of the panelists withthe Dutch target population as regards the mean and stan-dard deviation of some of the selection criteria used andmentioned above

24 Listening test outline

Four to six panelists jointly participated in a session Theoverall structure and time schedule of the listening exper-iment was identical for each group of panelists It startedwith a 14-minute training session during which the pan-elists were asked to scale each of the 7 reference (high-way) sounds two times (in random order) Thereafter7 10-minute menus were played of which the first menualways was the highway traffic menu A short break wasthen taken and the training session was repeated afterwhich again 7 new 10-minute menus were played Afterthis experiment with menus a more conventional psychoa-coustical listening test was conducted in which the pan-elists had to scale 45-second excerpts of all transport noisestimuli used in the menu experiment The duration of anexperimental session was on average about 4 hours Toillustrate how the listening test was performed Figure 5shows the sound level in dB(A) rerecorded in front of thefaccedilade during one of the panelist groupsrsquo listening exper-imentIn all two times 6 train menus were presented to each

panelist It was decided that within one set of 6 trainmenus conventional trains (IC or high-speed) should notbe mixed with magnetic levitation trains By this separa-

tion it was possible to include a retrospective evaluationover the last hour as well From previous experience it wasknown that the order of the menu pesentations might af-fect the results Half of the panelists were therefore pre-sented the maglev train sounds first the other half the con-ventional trains first A singular session consisted of thesame number of passages inside the menus This avoidsthat panelists would concentrate on counting events Fi-nally since one distance to the track would create a nat-ural setting large distances were never mixed with shortdistances in the menus of a session

During the experimental sessions perceived noise an-noyance of all transport noises was scaled with the methodof free-number magnitude estimation [22] The panelistswere asked to write down their magnitude assessments ondifferent coloured pieces of paper Before the start of theexperiment the panelists were instructed to select an ap-propriate number and then to double this number if theyfound the next stimulus to be twice as annoying to makethe number three times larger if they found the next stim-ulus to be three times as annoying etc and to scale 0 ifthey considered it not to be annoying at all For each 45-second sound (training sessions and conventional listeningtest) a conditional question was included ldquoTo what extentwould you be annoyed by this traffic sound if you heardit while relaxingrdquo For each 10-minute menu a very simi-lar but retrospective question was asked ldquoTo what extentwere you annoyed by traffic sound during the previous pe-riodrdquo In these latter questions we explicitly did not wantto refer to train noise since we wanted the panelists to de-cide themselves whether the sound period they last heardsounded like train-contaminated or not

25 Master scaling

In all experimental sessions the 7 road-traffic-noise-likereference sounds helped the panelists to define their ownscaling context The annoyance values given to these refer-ence sounds made it possible to control for the individualpanelistsrsquo choice-of-number behaviour in scaling the tar-get train sounds It would also control the influence of per-sonal factors such as noise sensitivity To get rid of theseeffects each individual panelistrsquos annoyance scale was cal-ibrated by the aid of the reference to the common masterscale [23]

A graphical illustration of the master scale transforma-tion applied to the annoyance reference data of one of thepanelists is given in Figure 6 The average annoyance re-ported for each of the 7 reference sound levels of roadtraffic noise is plotted in lin-log coordinates against theirsound levels LAeq45s measured at the faccedilade Individualpsychophysical functions are fitted to the reference data(open circles) They are of the form

Ar = a + b logSr (1)

where Ar is the reported annoyance during the train-ing session and logSr is the corresponding ldquoroad trafficnoiserdquo reference (r) sound level in dB(A) The constants a

593

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

Figure 5 Sound pressure level rerecorded in front of the faccedilade during one panelist groupsrsquo participation in the whole listening testtwo training sessions two menu sessions and one conventional psychoacoustical experiment with references

0

10

20

30

40

50

60

70

80

40 50 60 70 80 90

R = 55 master-scaled annoyance

A = 70 assessment

annoyance

road traffic reference sound level [dB(A)]

Figure 6 Calculation of master-scaled annoyance using one pan-elistrsquos empirical psychophysical function of the reference sounds(data points with solid line) and the master function for the samesounds (dashed line obtained as average function for all pan-elists)

and b will be different for each panelist and will depend ontheir choice-of-number behaviour in the particular scalingcontext The empirically derived master functions for thegroup of 100 panelists (dashed line in Figure 6) were thenused to transform the free number magnitude estimationsof the train or road traffic menus for each individual Aeto the corresponding annoyance values R in master scaleunits

R = minus629 + 145Ae minus a

b (2)

The slope of the master function was set to 145 whichis the average slope of all the individual psychophysicalfunctions whereas the intercept was set to produce a valueof ldquozerordquo for the most quiet train menu The reason for thelatter choice was that a majority of the panelists (84)reported their annoyance to be zero for this menu and amajority of the panelists reported annoyance to be greaterthan zero for all other menus

The choice of a logarithmic psychophysical function(equation 1) was a compromise In previous magnitudeestimation experiments of loudness [23 9] a power func-tion of the form logA = c + d logS was found to fit the

Table III Test-retest reliability of panelistsrsquo perceived noise an-noyance of the 7 reference road traffic sounds Each cell containsan arithmetic mean of Pearsonrsquos coefficient (r) and its standarderror Ts1 Training session 1 Ts2 Training session 2 Ct Con-ventional test

Ts1 Ts2 CtSet 1 Set 2 Set 1 Set 2 Set 1

Ts22 082plusmn0015

Ts21 086 087plusmn0014 plusmn0016

Ts22 086 088 087plusmn0017 plusmn0020 plusmn0019

Ct1 083 083 082 085plusmn0015 plusmn0021 plusmn0019 plusmn0020

Ct2 084 085 081 084 082plusmn0015 plusmn0019 plusmn0016 plusmn0019 plusmn0015

empirical data best However in this experiment noise an-noyance rather than loudness was scaled and thus obvi-ously also a value of zero (= not at all annoyed) had tobe handled although the noise was heard and its loudnesswas above zero The power function (after removal of ze-ros) did not fit the data better than the chosen logarithmicfunction

26 Data quality analysis

The master scaling made it possible to investigate the qual-ity of the experimental data in two ways as panelistsrsquo test-retest reliability and as their scaling ability The 7 refer-ence sounds were presented 6 times to each panelist twicein the two training sessions and twice in the last conven-tional listening test The set of 6 reference scale valueswere used to determine each panelistrsquos test-retest reliabil-ity of annoyance Table III shows the Pearsonrsquos coefficientof correlation for these 6 annoyance scales averaged overall panelists The test-retest reliability was very good be-tween 081 and 088 and the standard error was low be-tween 0014 and 0019

The deviation from the proposed master function (equa-tion 1) was used to assess the data quality and annoyance

594

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

Table IV Distribution of constants of the panelistsrsquo individualpsychophysical functions (Eq 1) The number of data sets refersto the average of 4 or 2 raw annoyance values which was takenfor each of the 7 reference sounds to calculate the psychophysicalfunctions

Data Psychophysical functionSets r2 a b

Training session 1amp2 4 0947 minus6727 1449plusmn0077 plusmn6128 plusmn1230

Conventional test 2 0881 minus4757 1105plusmn0118 plusmn4817 plusmn0948

scaling ability for each panelist and to trace errors and in-accuracies Table IV shows the distribution of constants ofthe panelistsrsquo individual psychophysical functions (equa-tion 1) The average annoyance variance explained bysound level (LAeq45s) of the reference road traffic soundswas found to range from 88 to 95 All panelists wereable to produce acceptable individual logarithmic func-tions of annoyance as a function of sound level to the refer-ence They have thus produced acceptable annoyance datain order to transform these to a common master scale ofannoyance no panelists were excluded from further dataanalysis

3 Results

The main listening experiment with menus differed fromprevious laboratory experiments in a number of aspectsOne important novelty is that participants were asked tojudge annoyance over a longer period of time mdash Fastl andGottschlingrsquos experiment [4] forms an exception Duringthe 10-minute periods the panelists were engaged in lowattention relaxing activities such as reading a magazinemaking a conversation or having something to drink In or-der to find out how this new approach affected the resultsa subsequent experiment was included which was morecomparable to earlier experiments on train noise (eg [8])

31 Main field experiment with 10-minute menus

The panelistsrsquo master scale values of annoyance were av-eraged for each menu in the field experiment A step-wise multiple linear regression analysis was performedwith average master scaled annoyance as dependent vari-able and (a) time averaged A-weighted faccedilade exposureLAeq10min (b) distance to the source (logarithmic) and(c) source type as independent variables Because of itslegislative importance in the Netherlands the faccedilade ex-posure was preferred to the actual panelistsrsquo noise expo-sure Faccedilade exposure was calculated from the sound lev-els measured on the recording sites since the faccedilade levelsmeasured during the experiment also contain noise fromwind and rain It has to be noted that the actual soundexposure levels experienced by panelists participating ina single experimental session may differ because of dif-ferent seating positions However personal characteristicssuch as noise sensitivity will have a much larger influence

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq10min

(a)

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq10min

(b)

Figure 7 Average master scaled annoyance of the menus ver-sus LAeq10min (a) for 2 events per 10-minute menu and (b) for4 events per 10-minute menu for different types of train sounds( ) IC train ( ) TGV and ( ) maglev train In comparisonthe annoyance for the highway traffic ( ) is also shown Stan-dard error on means is indicated as well as the master function(dashed line)

on perceived annoyance as compared to the influence of aslightly different exposure

Table V summarizes the results In the first modelsound level was the only independent variable this modelexplained 80 of the variance in annoyance In the secondmodel distance to track was added to sound level as an in-dependent variable this model increased the variance ex-plained to 85 (F-change = 1449 df1 = 1 df2 = 46 p lt0001) Thus distance to source explained a significant ad-ditional part of the annoyance variance not accounted forby sound level In the third model source type was in-cluded as a third independent variable along with soundlevel and distance Source type was defined on a nominalscale MAGLEV TGV IC and HIGHWAY It was intro-duced in the analysis as three dummy variables coded 0and 1 (the highway noise source type corresponds to thecase that these variables are all zero) The inclusion ofsource type did not increase significantly (F-change lt 10)the proportion of variance explained This suggests thatstatistically there is no additional contribution of sourcetype on perceived annoyance over and above the effects ofsound level and distance It can therefore be concluded thatmagnetic-levitation based transportation systems are notsignificantly more annoying than conventional rail basedsystems (same faccedilade LAeq and same distance are pre-requisites) Moreover railway noise was not found to be

595

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

Table V Stepwise multiple regression analysis of acoustic variables on perceived annoyance of train and highway traffic sounds for themain field experiment with 10-minute menus The Pearsonrsquos correlation coefficients of the variables entered in the regression analysisare shown at the bottom lowastp lt 005 lowastlowastp lt 001 lowastlowastlowastp lt 0001

Model Model Model fit increase F-change Independent Variables Coefficient t-valuefit (r2) (r2-change)

1 080 080 18748lowastlowastlowast LAeq10min [dB(A)] 118 1369lowastlowastlowast

2 085 005 1449lowastlowastlowast LAeq10min [dB(A)] 092 917lowastlowastlowast

log10(distance [m]) minus1074 minus381lowastlowastlowast

3 085 000 013 LAeq10min [dB(A)] 096 822lowastlowastlowast

log10(distance [m]) minus1017 minus333lowastlowastMAGLEV [01] 145 027TGV [01] 085 016IC [01] 227 040

Label Variable ANN LEQ DIST MAG TGV

ANN Annoyance [master scale]LEQ LAeq10min [dB(A)] 0894DIST log10(distance [m]) minus0754 minus0659MAG MAGLEV [01] minus0023 minus0038 0009TGV TGV [01] 0132 0224 0006 minus0682IC IC [01] minus0179 minus0290 0004 minus0433 minus0308

systematically less annoying than highway traffic noiseThis means that no support for a railway bonus was foundin this experiment at least it was not as obvious that itcould be observed using linear statistics Figure 7 givesan overview of the annoyance functions for the 10-minutemenus as a function of LAeq10min The dashed line indi-cates the master function of annoyance for the road-traffic-like sounds used as referencesThe shorter rise time of the noise of arriving high speed

trains may create more annoyance than a conventionaltrain can do Figure 8 Panel a shows the rise speeds indB(A)s in proportion to circle sizes These values werecalculated for all sound events included in this experi-ment by fitting a straight line through the initial increasein sound level The accelerating growth of annoyance withincreasing LAeq may be explained by the rise time In Fig-ure 8 Panel b the size of the circles is instead propor-tional to the distance to the track For LAeq in the intervalbetween 50 and 65 dB(A) annoyance is clearly lower fortrain passages at larger distances than for train passagesat closer distances or road traffic noise (dashed line) Thiscould indicate that a possible noise annoyance bonus fortrain noise would only hold at larger distances from thetrack and only in the latter LAeq interval

32 Conventional listening test

In the conventional listening experiment the sounds werepresented as short 45-second fragments containing thesound of one train passage and highway excerpts Figure 9shows the results of these master scaled annoyance val-ues as a function of time averaged A-weighted faccedilade ex-posure LAeq45s A railway penalty can be observed bothin regard to the artificial reference sounds as well as to

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq10min

(a)

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq10min

(b)

Figure 8 Average master scaled annoyance of the menus versusLAeq10min showing (a) the noise event rise speed and (b) the dis-tance to the track as the size of the circles The master functionis also indicated (dashed line)

the highway sounds Figure 10 shows the annoyance as afunction of rise speed ( Panel a) and distance to the track(Panel b)

596

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

Table VI Stepwise multiple regression analysis of acoustic variables on perceived annoyance of train sounds (no highway trafficsounds) for the conventional listening test (45-second passages) The Pearsonrsquos correlation coefficients of the variables entered in theregression analysis are shown at the bottom lowastp lt 005 lowastlowastp lt 001 lowastlowastlowastp lt 0001

Model Model Model fit increase F-change Independent Variables Coefficient t-valuefit (r2) (r2-change)

1 095 095 42017lowastlowastlowast LAeq45s [dB(A)] 167 2050lowastlowastlowast

2 098 003 1250lowastlowastlowast LAeq45s [dB(A)] 123 1104lowastlowastlowast

Speed [kmh] 002 203log10(distance [m]) minus178 minus077Rise speed [dB(A)s] 063 365lowastlowast

3 099 001 098 LAeq45s [dB(A)] 108 685lowastlowastlowast

Speed [kmh] 003 209log10(distance [m]) minus476 minus146Rise speed [dB(A)s] 058 330lowastlowast

MAGLEV [01] minus070 minus037TGV [01] 166 104

Label Variable ANN LEQ SPD DIST RISE MAG

ANN Annoyance [master scale]LEQ LAeq45s [dB(A)] 0975SPD Speed [kmh] 0646 0541DIST log10(distance [m]) minus0613 minus0667 minus0001RISE Rise speed [dB(A)s] 0885 0804 0708 minus0437MAG MAGLEV [01] 0070 minus0017 0552 0000 0188TGV TGV [01] 0207 0246 minus0190 0000 0100 minus0707

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq45sec

Figure 9 Average master scaled annoyance versus LAeq45s forthe conventional listening test for different types of train sounds( ) IC train ( ) TGV and ( ) maglev train In comparison theannoyance for the highway traffic ( ) is also shown Standard er-ror on means is indicated as well as the master function (dashedline)

A stepwise multiple linear regression analysis was alsoperformed separately for the train noises (Table VI) Thefirst model in which sound level LAeq45s was included asthe only independent variable explained 95 of the vari-ance in annoyance In the second model train speed dis-tance to the track and rise speed were added to sound levelas independent variables This increased the variance ex-plained to 98 (F-change = 1250 df1 = 3 df2 = 19p lt 0001) Apart from sound level also rise speed con-

tributed significantly to the variance explained The thirdmodel in which train type was added as an independentdummy variable did not significantly increase the propor-tion explained variance (F-change lt 10) These resultssuggest that in this conventional listening test there is nodifference in perceived annoyance between different typesof trains over and above the effect of sound level and risespeed

One has to note that the number of responses to eachstimulus was smaller in the main experiment (10-minutemenus) than in the conventional listenig test (45-secondpassages) This explains why the standard errors are lowerand the explained variance is higher in the latter experi-ment

4 Discussion

The annoyance results of the present field experiment areclose to residentsrsquo everyday reality although comparisonwith published studies is somewhat limited Previous lab-oratory experiments on noise annoyance of conventionalIC and high-speed trains specifically magnetic levitationtrains [8 5] report significant differences for these typesof sound In particular the results have shown that forthe same LAeq high-speed trains were more annoyingthan other trains Compared to road traffic noise the citedstudies claimed a lower annoyance level for conventionaltrains In the present field experiment we did not find sup-port for any annoyance difference between various types

597

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

0

10

20

30

40

50

60

70

80

40 50 60 70 80 90

annoyance

[maste

rscale

units]

L [dB(A)]Aeq45sec

(a)

0

10

20

30

40

50

60

70

80

40 50 60 70 80 90

annoyance

[maste

rscale

units]

L [dB(A)]Aeq45sec

(b)

Figure 10 Average master scaled annoyance versus LAeq45s

showing (a) the noise event rise speed and (b) the distance tothe track as the size of the circles The master function is alsoindicated (dashed line)

of trains and road traffic Some possible explanations willbe given in the following subsections

41 Realistic listening situation with 10-minutemenus

The experiment was performed in a realistic setting inwhich outdoor transportation noise was reproduced andnatural outdoor-to-indoor sound propagation characteris-tics were utilized (slightly open window) This setting pro-vided a realistic sound environment indoors Subgroups ofpanelists were kept indoors during the four-hour experi-ments and upon request annoyance to transport noise wasreported with reference to 10-minute periods

Because trains run on expected schedules to which peo-ple habituate the experimental situation in classical exper-iments is rather unrealistic The experimental one-passagesituation [8] requires full attention and will have a largevariation of train sounds compared with a particular rail-way track The outcome will to a large extent depend onthe experimental context that is the variation introducedin the experiment by selecting stimuli and using randompresentation orders Random orders of recordings can beselected and arranged so that annoyance judgments on cat-egory scales plotted against sound level differentiate wellor not well on type of transport In the present field experi-ment sub-context in sessions was kept invariant similar tothe situation on a real railway track The judgmental con-

text will then be much more restricted as is the case whenliving along one railway track

Next to this the annoyance reports of the one 45-secondtrain passage were higher than those of two passages ofthe same train within the 10-minute menus This is all inorder because the two types of annoyance were masterscaled in order to become comparable over experimentalsessions When judging 45-second train passages imme-diately after exposure it is quiet clear that the task is toassess the annoyance of that particular train passage (orother sounds that were presented) However when askedto assess the annoyance retrospectively of the transportnoise during the last 10-minute menu (eg two train pas-sages) the panelist will have to choose a strategy on howto go about this For example the annoyance may only bereferred to the two noise-stimulus periods or to the whole10-minute period (menu) It has been shown that the noiseannoyance of two overlapping (equal) noises would be ex-pected always to be less than the arithmetic sum of the twoannoyances (for a review see [24]) It is more uncertainhow total annoyance of two train passages separated intime will actually be acquired A laboratory experimentwhich included long sound fragments [4] has not foundthe above-mentioned annoyance difference between differ-ent train types which is in line with our results

42 Advanced scaling methodology

Long-term retrospective annoyance asked for in question-naire surveys has typically been assessed on categoryscales (eg [3]) A response category is then implicitlypostulated to be identical for every participant by verballabeling of the two end points or of every response boxalso the intervals between categories are assumed to be thesame However this assumption does not hold true [25]eg in questionnaire surveys the response criteria (scalevalue or category borders) for annoyance are much higherfor respondents in low noise areas as compared to those inhighly exposed areas The most well known scaling bias inlaboratory experiments is the context effect in which par-ticipants distribute their responses over the ldquofullrdquo range ofcategories independent of the size of the exposure range(for a review see [22]) In the process of using categoryscales floor and ceiling effects on annoyance may also ap-pearTo avoid uncontrolled context effects an invariant sound

level range of references was used as the annoyance con-text in the present field experiment Continuous road trafficnoise was chosen as a reference instead of multiple eventsounds because it is simpler to reproduce in future stud-ies To avoid the scaling bias of category scales the methodof magnitude estimation was chosen in which participantswere free to use the range of numbers they felt comfortablewith Master scaling was applied to these individual an-noyance estimates involving a transformation function toa common master scale defined by the references whichsound levels defined the scaling context In theory thismaster scale transformation will calibrate the loudness-dependence of noise annoyance whereas the relative con-

598

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

tribution to noise annoyance from qualitative content (egthe type of sound the time pattern and cues for speed anddistance) will hopefully be unchanged

Earlier research has shown that master scaling with ref-erences works well for loudness or annoyance of a one-occasion target exposure that is when repeated exposureis unfeasible (eg experiments with long duration expo-sures) or impossible (eg questionnaire surveys in fieldstudies) an example can be found in [23] The results ob-tained from the present field experiment are probably morereliable than the results that would have been obtained bycategory scaling The test-retest reliability of the panelistsrsquomagnitude estimates of annoyance of the reference soundswas found to be very good (above 08) compared to thereliability of 072 obtained in [8] for a group of 12 muchyounger subjects Considering that our panelists all werenaiumlve participants they also each produced high qualitypsychophysical functions for the reference as discussedin Section 26

43 Other possible explanations

There are several reasons why other investigators havefound a railway bonus (for a review see [3]) which wasnot found in this field experiment One of the reasons forfinding a railway bonus for short (one minute) noises inlistening experiments may be that the relation betweenloudness and LAeq is inherently different for train androad traffic noise Indeed some researchers have arguedthat noise annoyance evaluation in listening tests of shortsounds actually is close to a perceptual loudness evalua-tion (however see [26] on differences between loudness-based and quality-based perceived annoyance) If Zwickerloudness is a good first estimate of perceptual loudnessthe difference between train noise (of different types) andhighway noise would be seen in a Zwicker loudness ver-sus LAeq plot (Figure 11) Because the IC train noise usedin the present experiment was the noise of modern ratherquiet trains a few older and noisier IC train models wereadded in this acoustic analysis At levels above 65 dB(A)TGV and maglev trains seem to be a little louder thanhighway traffic or older IC trains However this effect onZwicker loudness is not significant and does therefore notsupport a railway bonus of 5 dB(A) stipulated in severalcountriesrsquo legislation Rather it seems to be a good actionto start to replace old IC trains by new ones The railwaybonus was originally based on studies with rather old low-speed trains and with much less dense traffic intensity thannowadaysThe intermittent character of railway noise could also be

an explanation for the railway bonus However this doesnot hold for aircraft noise which is also intermittent thiscan be explained by a difference in exposure In the case ofaircraft noise the exposure is on top of buildings and on allfaccedilades In the case of road traffic noise the probability ishigh that there are local roads also but there is a possibil-ity for a ldquoquiet siderdquo people are less annoyed if quiet sidesare available [27] In the case of railway noise there is alow probability for the presence of more than one track

Zw

ickerlo

udness

[sone]

L [dB(A)]Aeq45sec

0

10

20

30

40

50

40 50 60 70 80

Figure 11 Zwicker loudness versus LAeq45s for different types oftransportation sounds ( ) IC train ( ) TGV ( ) maglev train( ) highway traffic and ( ) some additional noisier IC trains(older type)

so the exposure will also be directed at only one faccediladeIn comparing road traffic and trains the faccedilade insulationwill be more effective in the case of train noise because ofthe smaller low-frequency proportion associated to trainnoise In comparing aircraft and trains which are both in-termittent the indoor exposure is certainly more intensivefor aircraft Considering these arguments it seems obviousthat aircraft is more annoying than road traffic which ismore annoying than train However faccedilade reduction wastaken into account in the present field experiment and stillthere was no clear railway bonus found Compared withthe field condition with closed windows and the faccediladefilter used in [8] a partially open window was used in thepresent field experiment which could explain this

In surveys questioning people at their home a lowerreported annoyance for train noise compared to highwaytraffic noise was observed in a particular range of noiselevels Most of the possible explanations proposed in liter-ature conflict with the fact that this railway bonus wouldbe observed in experiments based on single passages Wemention just a few The typical character of train noise andthe concentration of the sound energy in short time inter-vals may be advantageous with regard to activity distur-bance If the level is sufficiently low the probability ofnoticing the train noise is small compared to the prob-ability of noticing the sound of a continuous source Inaddition to physical differences in the sound the ldquogreenimagerdquo of trains as a means of transportation may add tothe acceptability of the source and thus increase the tol-erance to its noise that is as long as train passages arenot too frequent However a more recent hedonic pric-ing study found that householders in Birmingham placea greater value on reductions in railway noise than in roadtraffic noise [28] Cross-cultural studies (in field and lab-oratory context) have shown that a railway bonus is notuniversal [29 30] which would favor the argument aboveIt has further been shown that the bonus varies depend-ing on the (multiple) exposure situation [31] Based on theabove only part of the effect is supposed to be visible in

599

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

field experiments such as the one reported of in this paperPart of the effect is precisely what is observed

5 Conclusion

This study has shown that in an ldquoat homerdquo like contextnoise annoyance caused by different types of trains at thesame average outdoor faccedilade exposure level is not signifi-cantly different In particular magnetic levitation systemsare not more annoying than conventional high speed trainswhich is in agreement with earlier research Noise annoy-ance caused by conventional trains was not found to besignificantly lower than annoyance caused by TGVrsquos ormaglev trains at the same average faccedilade exposure Fieldsurveys have shown that for the same average sound levelrailway noise causes less annoyance or highly annoyedpersons than highway traffic noise Although our field ex-periment included several factors that may contribute tothis effect we could not observe it

More insight may be gained by taking into account thepsychoacoustic characteristics of the noise exposure andthe relevant personal factors of the panelists This paperhas focused on discussing the experimental methodologyin great detail and on presenting the results as a functionof the average outdoor faccedilade exposure level since thisis the main noise legislation indicator used in the Nether-lands Results of a detailed psychoacoustic analysis willbe reported in a future paper

Acknowledgments

This research was financed by the Project Group Zuider-zeelijn of the Ministry of Public Transport Public Worksand Water Management in the Netherlands The membersof the project steering committeemdashGilles Janssen (dBVi-sion) Annemarie Ruysbroek (RIVM) Martin van denBerg (VROM) and Pieter Jansse (Project Group Zuider-zeelijn) mdash are acknowledged for their valuable input Wealso appreciate the experimental assistance provided byMs Ingrid Decoster Finally we would like to thank thereviewers for their valuable comments and suggestions

References

[1] J M Fields J G Walker Comparing the relationships be-tween noise level and annoyance in different surveys a rail-way vs aircraft and road traffic comparison J Sound Vib81 (1982) 51ndash80

[2] U Moehler Community response to railway noise A re-view of social surveys J Sound Vib 120 (1988) 321ndash332

[3] H M E Miedema H Vos Exposure-response relation-ships for transportation noise J Acoust Soc Am 104(1998) 3432ndash3445

[4] H Fastl G Gottschling Subjective evaluation of noiseimmissions from transrapid Proceedings of the 25thInternational Congress on Noise Control Engineering(Intermiddotnoise) Liverpool UK Juli 1996 F A Hill R Law-rence (eds)

[5] D Neugebauer J Ortscheid Geraumluschbewertung des Tran-srapid Ist der Transrapid wie die Bahn zu beurteilenBerlin Umweltbundesamt Texte 25 1997

[6] D Neugebauer J Ortscheid Geraumluschbewertung des Tran-srapid Fortschritte der Akustik mdash DAGA 97 Kiel Ger-many 1997

[7] J Vos Annoyance caused by magnetic levitation trainTransrapid 08 mdash a laboratory study TNO report TM-03-C001 The Netherlands Organisation for Applied ScientificResearch (TNO) Soesterberg The Netherlands Jan 2003

[8] J Vos Annoyance caused by the sounds of a magnetic lev-itation train J Acoust Soc Am 115 (2004) 1597ndash1608

[9] B Berglund A Preis Is perceived annoyance more sub-ject-dependent than perceived loudness Acta Acusticaunited with Acustica 83 (1997) 313ndash319

[10] M Ishibashi K Ueno H Tachibana M Watanabe Sub-jective experiment on annoying sounds in living environ-ments Proceedings of the 18th International Congress onAcoustics (ICA) Kyoto Japan 2004

[11] J Maillard J Martin P Champelovier J Lambert Percep-tive evaluation of road traffic noise inside buildings using acombined image and wave field synthesis system Proceed-ings of CFADAGArsquo04 Strasbourg France Maumlrz 2004

[12] R F S Job Community response to noise A review offactors influencing the relationship between noise exposureand reaction J Acoust Soc Am 83 (1988) 991ndash1001

[13] H M E Miedema H Vos Demographic and attitudinalfactors that modify annoyance from transportation noise JAcoust Soc Am 105 (1999) 3336ndash3344

[14] W Ellermeier M Eigenstetter K Zimmer Psychoacousticcorrelates of individual noise sensitivity J Acoust SocAm 109 (2001) 1464ndash1473

[15] D Vaumlstfjaumlll M Kleiner T Gaumlrling Affective reactions tointerior aircraft sounds Acta Acustica united with Acustica89 (2003) 693ndash701

[16] T Feng C Xiao-hong L Tan-feng A social survey onthe effects of high-speed maglev noise on residents inShangai Proceedings of The 2005 Congress and Expo-sition on Noise Control Engineering (Intermiddotnoise) Rio deJaneiro Brazil Aug 2005

[17] J J Gibson The ecological approach to visual perceptionLawrence Erlbaum Associates Inc Hillsdale New JerseyUSA 1979

[18] J G Neuhoff Ecological psychoacoustics Introductionand history ndash In Ecological Psychoacoustics J G Neuhoff(ed) Elsevier Academic Press London UK 2004

[19] C Guastavino B F G Katz J-D Polack D J LevitinD Dubois Ecological validity of soundscape reproductionActa Acustica united with Acustica 91 (2005) 333ndash341

[20] A J Berkhout D de Vries P Vogel Acoustic control bywave field synthesis J Acoust Soc Am 93 (1993) 2764ndash2778

[21] E A M Franssen J E F van Dongen J M H Ruys-broek H Vos R K Stellato Hinder door milieufactorenen de beoordeling van de leefomgeving in Nederland In-ventarisatie verstoringen 2003 RIVM report 815120 001National Institute for Public Health and the Environment(RIVM) Bilthoven The Netherlands 2004

[22] L E Marks D Algom Psychophysical scaling ndash In Mea-surement Judgment and Decision Making M H Birn-baum (ed) Academic Press New York 1998

[23] B Berglund Quality assurance in environmental psycho-physics ndash In Ratio Scaling of Psychological Magnitudes -In Honor of the Memory of S S Stevens S J BolanowskiG A Gescheider (eds) Erlbaum Hillsdale N J 1991

600

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

control room

openedwindow

H

6

7

3 4 5

2

1

L

faccedilade level

subwoofer

Figure 2 Schematic drawing of the experimental cottage (not toscale) The different seats of the panelists are shown (1ndash7) aswell as the seat of the experimental leader (L) and of an artificialhead (H) for binaural recordings

the real sound by audition The two spectra were in most13-octave bands within an error of ca 5 dB neverthelessit was decided to introduce an equalizer for fine-tuning anda subwoofer for reproducing more accurately the low fre-quency part of a moving high speed train

Figure 2 shows a floor plan of the living room and thecontrol room of the experimental cottage together withthe final loudspeaker setup The sounds were played backon a regular PC equipped with a high quality audio cardlocated in the control room The sound signal was thenequalized by an Allen amp Heath 12-channel mixer and 31-channel equalizer Subsequently the sound signal was am-plified by a Bose 802II amplifier and fed to 4 Bose loud-speakers which were placed stacked per 2 on 2 tripodstands at a height of ca 15m and to a HK Audio SL218Apowered subwoofer on the ground All loudspeakers wereplaced outside the house in front of the main windowThe 2 loudspeaker tripods were placed ca 10m from eachother perpendicular at 3m distance to the faccedilade Thesubwoofer was placed in front of the window in betweenboth tripods at about 50 cm from the faccedilade This loud-speaker setup was located in front of a slightly openedwindow of the experimental cottage invisible to the pan-elists entering the house

The faccedilade level was measured continuously during allexperimental sessions using a BampK Investigator 2260sound level meter with a BampK 4189 free field microphone(5 cm from the window at 75 cm height) The sound levelmeter was also used to calibrate the playback system Forthis calibration pink noise was played back and adjustedto give a faccedilade level of 91 dB with a flat 13-octave bandspectrum The equalizer accomplished a flat (plusmn 3 dB forall 13-octave bands) spectrum between 30Hz and 16 kHzThe faccedilade attenuation and the reverberation in the experi-mental room both modify the spectrum and temporal char-

acteristics of the sound Since it would not be possible tosee a train passage from the window because of plenty oftrees a visual presentation of passing trains was consid-ered not appropriate

22 Sample collection and preparation

Two-channel recordings were conducted for three typesof trains Two microphones were placed at 20m distancefrom each other along the track 15m above ground levelTGV trains at high speed were recorded in Beloeil (Bel-gium) a site near the TGV connection between Brus-sels and Lille (France) Dutch intercity (IC) trains of thenew type (duplex) were recorded in Oudenbosch nearRoosendaal (The Netherlands) at this same site the TGVtraveling at low speed from Brussels to Rotterdam wasalso recorded At the maglev test track in Lathen (Ger-many) the Transrapid 08 train was recorded at speeds ofapprox 200 kmh 300 kmh and 400 kmh For the mas-ter scaling references the sound of the E40 highway wasalso recorded near Ghent (Belgium) To be able to assessthe influence on annoyance of the distance to the track4 recording distances were included (25m 50m 100mand 200m) All recordings were made in free field withoutnoise barriers Not only the spectrum and temporal changewere reproduced exactly but also the sound level as if thehouse would have been located at the measurement site

From the many train recordings made at each site thepassages of highest quality were selected in each categoryof recording and for these 45-second single passage frag-ments were cut It was important to expose the panelists tosufficient and natural durations of noise Therefore theyhad to be exposed to ldquoexperimental soundrdquo during at least10 minutes (henceforth called a menu) To create a realis-tic exposure situation within a 10-minute menu it shouldbe composed of the same train type at the same distanceand speed Menus with 2 or 4 passages were created be-cause 4 passages in 10 minutes already represents the nat-ural time-schedule maximum and 2 passages in 10 min-utes represents a minimum passage rate with inter-passagebackground sound Less than two passages are not usefulbecause the inter-event silence is non-defined in this caseApart from the 45-second fragments recorded at the fourdistances to the track a 10-minute highway sound wasrecorded at 50m distance to the closest lane

Table I summarizes the sound exposure (ASEL) andsound levels (LAeq45s) associated with the 45-second pas-sages used in the 10-minute menus It should be mentionedthat the level of the IC train at 25m happens to be lowerthan the level at 50m This inconsistency is due to the factthat the selected high-quality sound fragments do not nec-essarily originate from identical train passages There isalways a natural spread in the speed and the number ofwagons of the different passages of the same type of trainAs an illustration Figures 3 and 4 show the A-weightedsound exposure level in 13-octave bands for some of theexperimental traffic sounds as recorded in free field

For master scaling 7 traffic-noise-like reference soundfragments of 45 seconds duration with sound pressure

591

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

Table I Sound exposure levels (ASEL) for one 45-second train passage sound level (LAeq45s) of one 45-second train passage and ofhighway traffic and sound level (LAeq10min) of the 10-minute menus of the experiment at 25m to 200m distance to track or route(all free field recordings) The train noise LAeq10min values are given for the 2-train menu to obtain the LAeq10min values for thecorresponding 4-train menu add 3 dB

Sound source Outdoor ASEL [dB(A)] Outdoor LAeq45s [dB(A)] Outdoor LAeq10min [dB(A)]25m 50m 100m 200m 25m 50m 100m 200m 25m 50m 100m 200m

Maglev 200 kmh 801 729 713 597 636 564 548 432 553 481 465 349300 kmh 863 830 803 696 698 665 638 531 615 582 555 448400 kmh 926 887 852 704 761 722 687 539 678 639 604 456

TGV 140 kmh 841 783 736 644 676 618 571 479 593 535 488 396300 kmh 928 906 869 830 763 741 704 665 680 658 621 582

IC 140 kmh 750 809 724 620 585 644 559 455 502 561 476 372

Highway free flow minus minus minus minus 716 661 626 553 minus 653 minus minus

Figure 3 Sound exposure level (ASEL) in 13-octave bands offour different types of traffic sounds all recorded during 45 sec-onds in free field at a distance of 50m to the track (or highwayroute) ( ) a passage of a maglev train traveling at 400 kmh( ) a passage of a TGV traveling at 300 kmh ( ) a passage ofan IC train traveling at 140 kmh and ( ) a highway with freeflow traffic

Figure 4 Sound exposure level (ASEL) in 13-octave bands of amaglev train traveling at 400 kmh recorded during 45 secondsin free field at various distances to the track ( ) 25m ( ) 50m( ) 100m and ( ) 200m

level spanning the whole experimental range were in-cluded in the experiment A 45-second fragment of thehighway noise recorded at 50m distance to the highwaywas used as the centre reference sound A filter which at-tenuates the sound at frequencies below 500Hz by 3 dBand above 500Hz by 6 dB was applied 3 times to produce3 reference sounds with varying level all below the levelof the centre reference sound giving the impression that

the source is further away In the same way a filter thatamplifies the sound at frequencies below 500Hz by 3 dBand above 500Hz by 6 dB was used to generate 3 refer-ence sounds with varying level higher than the level of thecentre reference sound

23 Selection of a representative panel

In contrast to previous experimental work on noise annoy-ance caused by high speed trains in which small ldquoconve-nientrdquo samples of test persons were recruited the selectionof panelists was here made to guarantee a representativesample of panelists A questionnaire was administered atthe doorstep of the homes of approximately 1500 personsall living within a distance of 15 km from the experimentalsite In an introductory letter one inhabitant of the housewas invited to participate in the study The prerequisiteswere that (s)he had to fill in and send the questionnaireback to the address on the enclosed stamped envelope Acompensation of C 100 was offered for participation

The questionnaire contained selected questions that hadbeen asked to a representative sample of the target popula-tion in a recent survey The structure of the Dutch popula-tion was inferred to be representative from a recent RIVMsurvey [21] and partly from a Eurobarometer question-naire Our questionnaire contained (standard) questions onenvironmental noise as regards perception annoyance andsleep disturbance Included were evaluations of the qualityof the neigbourhood in terms of housing and environmen-tal pollution of other types than noise as well as evalu-ations of overall satisfaction with the current living situ-ation Other questions addressed basic demographic vari-ables such as age gender education housing family sizeand work arrangements A set of questions were also in-cluded on general and mental health hearing ability en-vironmental background opinion and worry and environ-mental sensitivity

A procedure to draw panelists representative of thetarget population from the 255 replies received involvedthree stages Stage 1 removed potential panelists on thebasis of their age and hearing ability (information had al-ready been given in the introductory letter) Stage 2 furtherremoved those that were very dissimilar from the typical

592

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

Table II Comparison between the panelists and the referencepopulation on various criteria Mean and standard deviation isshown the results for the second series of criteria are on an 11-point scale and vary from 0 (not at all bad) to 10 (very good)

Criterium Participants Reference

Gender [ male female] 51 49 48 52Age [year] 451 plusmn 134 456 plusmn 177

Noise sensitivity 51 plusmn 24 46 plusmn 26

Quality of traffic noise in the 66 plusmn 24 64 plusmn 23living environment

Quality of the living 76 plusmn 14 73 plusmn 13environment

Feeling afraid or frightened 24 plusmn 20 23 plusmn 21

Dutch person on the basis of binary coding of most of theother criteria included in the questionnaire This stage im-plicitly assessed individual responses on the questions asregards their concordance with the response profile of thetypical Dutch person in the reference survey Stage 3 fi-nally selected panelists on the basis of fuzzy resemblanceto the typical Dutch person on the most critical criteria ofannoyance surveys such as age gender education noisesensitivity feeling afraid or frightened hearing train noiseat home quality of traffic noise in the living environmentquality of the living environment general health and ill-ness Finally ca 100 representative participants were se-lected Table II shows a comparison of the panelists withthe Dutch target population as regards the mean and stan-dard deviation of some of the selection criteria used andmentioned above

24 Listening test outline

Four to six panelists jointly participated in a session Theoverall structure and time schedule of the listening exper-iment was identical for each group of panelists It startedwith a 14-minute training session during which the pan-elists were asked to scale each of the 7 reference (high-way) sounds two times (in random order) Thereafter7 10-minute menus were played of which the first menualways was the highway traffic menu A short break wasthen taken and the training session was repeated afterwhich again 7 new 10-minute menus were played Afterthis experiment with menus a more conventional psychoa-coustical listening test was conducted in which the pan-elists had to scale 45-second excerpts of all transport noisestimuli used in the menu experiment The duration of anexperimental session was on average about 4 hours Toillustrate how the listening test was performed Figure 5shows the sound level in dB(A) rerecorded in front of thefaccedilade during one of the panelist groupsrsquo listening exper-imentIn all two times 6 train menus were presented to each

panelist It was decided that within one set of 6 trainmenus conventional trains (IC or high-speed) should notbe mixed with magnetic levitation trains By this separa-

tion it was possible to include a retrospective evaluationover the last hour as well From previous experience it wasknown that the order of the menu pesentations might af-fect the results Half of the panelists were therefore pre-sented the maglev train sounds first the other half the con-ventional trains first A singular session consisted of thesame number of passages inside the menus This avoidsthat panelists would concentrate on counting events Fi-nally since one distance to the track would create a nat-ural setting large distances were never mixed with shortdistances in the menus of a session

During the experimental sessions perceived noise an-noyance of all transport noises was scaled with the methodof free-number magnitude estimation [22] The panelistswere asked to write down their magnitude assessments ondifferent coloured pieces of paper Before the start of theexperiment the panelists were instructed to select an ap-propriate number and then to double this number if theyfound the next stimulus to be twice as annoying to makethe number three times larger if they found the next stim-ulus to be three times as annoying etc and to scale 0 ifthey considered it not to be annoying at all For each 45-second sound (training sessions and conventional listeningtest) a conditional question was included ldquoTo what extentwould you be annoyed by this traffic sound if you heardit while relaxingrdquo For each 10-minute menu a very simi-lar but retrospective question was asked ldquoTo what extentwere you annoyed by traffic sound during the previous pe-riodrdquo In these latter questions we explicitly did not wantto refer to train noise since we wanted the panelists to de-cide themselves whether the sound period they last heardsounded like train-contaminated or not

25 Master scaling

In all experimental sessions the 7 road-traffic-noise-likereference sounds helped the panelists to define their ownscaling context The annoyance values given to these refer-ence sounds made it possible to control for the individualpanelistsrsquo choice-of-number behaviour in scaling the tar-get train sounds It would also control the influence of per-sonal factors such as noise sensitivity To get rid of theseeffects each individual panelistrsquos annoyance scale was cal-ibrated by the aid of the reference to the common masterscale [23]

A graphical illustration of the master scale transforma-tion applied to the annoyance reference data of one of thepanelists is given in Figure 6 The average annoyance re-ported for each of the 7 reference sound levels of roadtraffic noise is plotted in lin-log coordinates against theirsound levels LAeq45s measured at the faccedilade Individualpsychophysical functions are fitted to the reference data(open circles) They are of the form

Ar = a + b logSr (1)

where Ar is the reported annoyance during the train-ing session and logSr is the corresponding ldquoroad trafficnoiserdquo reference (r) sound level in dB(A) The constants a

593

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

Figure 5 Sound pressure level rerecorded in front of the faccedilade during one panelist groupsrsquo participation in the whole listening testtwo training sessions two menu sessions and one conventional psychoacoustical experiment with references

0

10

20

30

40

50

60

70

80

40 50 60 70 80 90

R = 55 master-scaled annoyance

A = 70 assessment

annoyance

road traffic reference sound level [dB(A)]

Figure 6 Calculation of master-scaled annoyance using one pan-elistrsquos empirical psychophysical function of the reference sounds(data points with solid line) and the master function for the samesounds (dashed line obtained as average function for all pan-elists)

and b will be different for each panelist and will depend ontheir choice-of-number behaviour in the particular scalingcontext The empirically derived master functions for thegroup of 100 panelists (dashed line in Figure 6) were thenused to transform the free number magnitude estimationsof the train or road traffic menus for each individual Aeto the corresponding annoyance values R in master scaleunits

R = minus629 + 145Ae minus a

b (2)

The slope of the master function was set to 145 whichis the average slope of all the individual psychophysicalfunctions whereas the intercept was set to produce a valueof ldquozerordquo for the most quiet train menu The reason for thelatter choice was that a majority of the panelists (84)reported their annoyance to be zero for this menu and amajority of the panelists reported annoyance to be greaterthan zero for all other menus

The choice of a logarithmic psychophysical function(equation 1) was a compromise In previous magnitudeestimation experiments of loudness [23 9] a power func-tion of the form logA = c + d logS was found to fit the

Table III Test-retest reliability of panelistsrsquo perceived noise an-noyance of the 7 reference road traffic sounds Each cell containsan arithmetic mean of Pearsonrsquos coefficient (r) and its standarderror Ts1 Training session 1 Ts2 Training session 2 Ct Con-ventional test

Ts1 Ts2 CtSet 1 Set 2 Set 1 Set 2 Set 1

Ts22 082plusmn0015

Ts21 086 087plusmn0014 plusmn0016

Ts22 086 088 087plusmn0017 plusmn0020 plusmn0019

Ct1 083 083 082 085plusmn0015 plusmn0021 plusmn0019 plusmn0020

Ct2 084 085 081 084 082plusmn0015 plusmn0019 plusmn0016 plusmn0019 plusmn0015

empirical data best However in this experiment noise an-noyance rather than loudness was scaled and thus obvi-ously also a value of zero (= not at all annoyed) had tobe handled although the noise was heard and its loudnesswas above zero The power function (after removal of ze-ros) did not fit the data better than the chosen logarithmicfunction

26 Data quality analysis

The master scaling made it possible to investigate the qual-ity of the experimental data in two ways as panelistsrsquo test-retest reliability and as their scaling ability The 7 refer-ence sounds were presented 6 times to each panelist twicein the two training sessions and twice in the last conven-tional listening test The set of 6 reference scale valueswere used to determine each panelistrsquos test-retest reliabil-ity of annoyance Table III shows the Pearsonrsquos coefficientof correlation for these 6 annoyance scales averaged overall panelists The test-retest reliability was very good be-tween 081 and 088 and the standard error was low be-tween 0014 and 0019

The deviation from the proposed master function (equa-tion 1) was used to assess the data quality and annoyance

594

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

Table IV Distribution of constants of the panelistsrsquo individualpsychophysical functions (Eq 1) The number of data sets refersto the average of 4 or 2 raw annoyance values which was takenfor each of the 7 reference sounds to calculate the psychophysicalfunctions

Data Psychophysical functionSets r2 a b

Training session 1amp2 4 0947 minus6727 1449plusmn0077 plusmn6128 plusmn1230

Conventional test 2 0881 minus4757 1105plusmn0118 plusmn4817 plusmn0948

scaling ability for each panelist and to trace errors and in-accuracies Table IV shows the distribution of constants ofthe panelistsrsquo individual psychophysical functions (equa-tion 1) The average annoyance variance explained bysound level (LAeq45s) of the reference road traffic soundswas found to range from 88 to 95 All panelists wereable to produce acceptable individual logarithmic func-tions of annoyance as a function of sound level to the refer-ence They have thus produced acceptable annoyance datain order to transform these to a common master scale ofannoyance no panelists were excluded from further dataanalysis

3 Results

The main listening experiment with menus differed fromprevious laboratory experiments in a number of aspectsOne important novelty is that participants were asked tojudge annoyance over a longer period of time mdash Fastl andGottschlingrsquos experiment [4] forms an exception Duringthe 10-minute periods the panelists were engaged in lowattention relaxing activities such as reading a magazinemaking a conversation or having something to drink In or-der to find out how this new approach affected the resultsa subsequent experiment was included which was morecomparable to earlier experiments on train noise (eg [8])

31 Main field experiment with 10-minute menus

The panelistsrsquo master scale values of annoyance were av-eraged for each menu in the field experiment A step-wise multiple linear regression analysis was performedwith average master scaled annoyance as dependent vari-able and (a) time averaged A-weighted faccedilade exposureLAeq10min (b) distance to the source (logarithmic) and(c) source type as independent variables Because of itslegislative importance in the Netherlands the faccedilade ex-posure was preferred to the actual panelistsrsquo noise expo-sure Faccedilade exposure was calculated from the sound lev-els measured on the recording sites since the faccedilade levelsmeasured during the experiment also contain noise fromwind and rain It has to be noted that the actual soundexposure levels experienced by panelists participating ina single experimental session may differ because of dif-ferent seating positions However personal characteristicssuch as noise sensitivity will have a much larger influence

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq10min

(a)

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq10min

(b)

Figure 7 Average master scaled annoyance of the menus ver-sus LAeq10min (a) for 2 events per 10-minute menu and (b) for4 events per 10-minute menu for different types of train sounds( ) IC train ( ) TGV and ( ) maglev train In comparisonthe annoyance for the highway traffic ( ) is also shown Stan-dard error on means is indicated as well as the master function(dashed line)

on perceived annoyance as compared to the influence of aslightly different exposure

Table V summarizes the results In the first modelsound level was the only independent variable this modelexplained 80 of the variance in annoyance In the secondmodel distance to track was added to sound level as an in-dependent variable this model increased the variance ex-plained to 85 (F-change = 1449 df1 = 1 df2 = 46 p lt0001) Thus distance to source explained a significant ad-ditional part of the annoyance variance not accounted forby sound level In the third model source type was in-cluded as a third independent variable along with soundlevel and distance Source type was defined on a nominalscale MAGLEV TGV IC and HIGHWAY It was intro-duced in the analysis as three dummy variables coded 0and 1 (the highway noise source type corresponds to thecase that these variables are all zero) The inclusion ofsource type did not increase significantly (F-change lt 10)the proportion of variance explained This suggests thatstatistically there is no additional contribution of sourcetype on perceived annoyance over and above the effects ofsound level and distance It can therefore be concluded thatmagnetic-levitation based transportation systems are notsignificantly more annoying than conventional rail basedsystems (same faccedilade LAeq and same distance are pre-requisites) Moreover railway noise was not found to be

595

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

Table V Stepwise multiple regression analysis of acoustic variables on perceived annoyance of train and highway traffic sounds for themain field experiment with 10-minute menus The Pearsonrsquos correlation coefficients of the variables entered in the regression analysisare shown at the bottom lowastp lt 005 lowastlowastp lt 001 lowastlowastlowastp lt 0001

Model Model Model fit increase F-change Independent Variables Coefficient t-valuefit (r2) (r2-change)

1 080 080 18748lowastlowastlowast LAeq10min [dB(A)] 118 1369lowastlowastlowast

2 085 005 1449lowastlowastlowast LAeq10min [dB(A)] 092 917lowastlowastlowast

log10(distance [m]) minus1074 minus381lowastlowastlowast

3 085 000 013 LAeq10min [dB(A)] 096 822lowastlowastlowast

log10(distance [m]) minus1017 minus333lowastlowastMAGLEV [01] 145 027TGV [01] 085 016IC [01] 227 040

Label Variable ANN LEQ DIST MAG TGV

ANN Annoyance [master scale]LEQ LAeq10min [dB(A)] 0894DIST log10(distance [m]) minus0754 minus0659MAG MAGLEV [01] minus0023 minus0038 0009TGV TGV [01] 0132 0224 0006 minus0682IC IC [01] minus0179 minus0290 0004 minus0433 minus0308

systematically less annoying than highway traffic noiseThis means that no support for a railway bonus was foundin this experiment at least it was not as obvious that itcould be observed using linear statistics Figure 7 givesan overview of the annoyance functions for the 10-minutemenus as a function of LAeq10min The dashed line indi-cates the master function of annoyance for the road-traffic-like sounds used as referencesThe shorter rise time of the noise of arriving high speed

trains may create more annoyance than a conventionaltrain can do Figure 8 Panel a shows the rise speeds indB(A)s in proportion to circle sizes These values werecalculated for all sound events included in this experi-ment by fitting a straight line through the initial increasein sound level The accelerating growth of annoyance withincreasing LAeq may be explained by the rise time In Fig-ure 8 Panel b the size of the circles is instead propor-tional to the distance to the track For LAeq in the intervalbetween 50 and 65 dB(A) annoyance is clearly lower fortrain passages at larger distances than for train passagesat closer distances or road traffic noise (dashed line) Thiscould indicate that a possible noise annoyance bonus fortrain noise would only hold at larger distances from thetrack and only in the latter LAeq interval

32 Conventional listening test

In the conventional listening experiment the sounds werepresented as short 45-second fragments containing thesound of one train passage and highway excerpts Figure 9shows the results of these master scaled annoyance val-ues as a function of time averaged A-weighted faccedilade ex-posure LAeq45s A railway penalty can be observed bothin regard to the artificial reference sounds as well as to

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq10min

(a)

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq10min

(b)

Figure 8 Average master scaled annoyance of the menus versusLAeq10min showing (a) the noise event rise speed and (b) the dis-tance to the track as the size of the circles The master functionis also indicated (dashed line)

the highway sounds Figure 10 shows the annoyance as afunction of rise speed ( Panel a) and distance to the track(Panel b)

596

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

Table VI Stepwise multiple regression analysis of acoustic variables on perceived annoyance of train sounds (no highway trafficsounds) for the conventional listening test (45-second passages) The Pearsonrsquos correlation coefficients of the variables entered in theregression analysis are shown at the bottom lowastp lt 005 lowastlowastp lt 001 lowastlowastlowastp lt 0001

Model Model Model fit increase F-change Independent Variables Coefficient t-valuefit (r2) (r2-change)

1 095 095 42017lowastlowastlowast LAeq45s [dB(A)] 167 2050lowastlowastlowast

2 098 003 1250lowastlowastlowast LAeq45s [dB(A)] 123 1104lowastlowastlowast

Speed [kmh] 002 203log10(distance [m]) minus178 minus077Rise speed [dB(A)s] 063 365lowastlowast

3 099 001 098 LAeq45s [dB(A)] 108 685lowastlowastlowast

Speed [kmh] 003 209log10(distance [m]) minus476 minus146Rise speed [dB(A)s] 058 330lowastlowast

MAGLEV [01] minus070 minus037TGV [01] 166 104

Label Variable ANN LEQ SPD DIST RISE MAG

ANN Annoyance [master scale]LEQ LAeq45s [dB(A)] 0975SPD Speed [kmh] 0646 0541DIST log10(distance [m]) minus0613 minus0667 minus0001RISE Rise speed [dB(A)s] 0885 0804 0708 minus0437MAG MAGLEV [01] 0070 minus0017 0552 0000 0188TGV TGV [01] 0207 0246 minus0190 0000 0100 minus0707

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq45sec

Figure 9 Average master scaled annoyance versus LAeq45s forthe conventional listening test for different types of train sounds( ) IC train ( ) TGV and ( ) maglev train In comparison theannoyance for the highway traffic ( ) is also shown Standard er-ror on means is indicated as well as the master function (dashedline)

A stepwise multiple linear regression analysis was alsoperformed separately for the train noises (Table VI) Thefirst model in which sound level LAeq45s was included asthe only independent variable explained 95 of the vari-ance in annoyance In the second model train speed dis-tance to the track and rise speed were added to sound levelas independent variables This increased the variance ex-plained to 98 (F-change = 1250 df1 = 3 df2 = 19p lt 0001) Apart from sound level also rise speed con-

tributed significantly to the variance explained The thirdmodel in which train type was added as an independentdummy variable did not significantly increase the propor-tion explained variance (F-change lt 10) These resultssuggest that in this conventional listening test there is nodifference in perceived annoyance between different typesof trains over and above the effect of sound level and risespeed

One has to note that the number of responses to eachstimulus was smaller in the main experiment (10-minutemenus) than in the conventional listenig test (45-secondpassages) This explains why the standard errors are lowerand the explained variance is higher in the latter experi-ment

4 Discussion

The annoyance results of the present field experiment areclose to residentsrsquo everyday reality although comparisonwith published studies is somewhat limited Previous lab-oratory experiments on noise annoyance of conventionalIC and high-speed trains specifically magnetic levitationtrains [8 5] report significant differences for these typesof sound In particular the results have shown that forthe same LAeq high-speed trains were more annoyingthan other trains Compared to road traffic noise the citedstudies claimed a lower annoyance level for conventionaltrains In the present field experiment we did not find sup-port for any annoyance difference between various types

597

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

0

10

20

30

40

50

60

70

80

40 50 60 70 80 90

annoyance

[maste

rscale

units]

L [dB(A)]Aeq45sec

(a)

0

10

20

30

40

50

60

70

80

40 50 60 70 80 90

annoyance

[maste

rscale

units]

L [dB(A)]Aeq45sec

(b)

Figure 10 Average master scaled annoyance versus LAeq45s

showing (a) the noise event rise speed and (b) the distance tothe track as the size of the circles The master function is alsoindicated (dashed line)

of trains and road traffic Some possible explanations willbe given in the following subsections

41 Realistic listening situation with 10-minutemenus

The experiment was performed in a realistic setting inwhich outdoor transportation noise was reproduced andnatural outdoor-to-indoor sound propagation characteris-tics were utilized (slightly open window) This setting pro-vided a realistic sound environment indoors Subgroups ofpanelists were kept indoors during the four-hour experi-ments and upon request annoyance to transport noise wasreported with reference to 10-minute periods

Because trains run on expected schedules to which peo-ple habituate the experimental situation in classical exper-iments is rather unrealistic The experimental one-passagesituation [8] requires full attention and will have a largevariation of train sounds compared with a particular rail-way track The outcome will to a large extent depend onthe experimental context that is the variation introducedin the experiment by selecting stimuli and using randompresentation orders Random orders of recordings can beselected and arranged so that annoyance judgments on cat-egory scales plotted against sound level differentiate wellor not well on type of transport In the present field experi-ment sub-context in sessions was kept invariant similar tothe situation on a real railway track The judgmental con-

text will then be much more restricted as is the case whenliving along one railway track

Next to this the annoyance reports of the one 45-secondtrain passage were higher than those of two passages ofthe same train within the 10-minute menus This is all inorder because the two types of annoyance were masterscaled in order to become comparable over experimentalsessions When judging 45-second train passages imme-diately after exposure it is quiet clear that the task is toassess the annoyance of that particular train passage (orother sounds that were presented) However when askedto assess the annoyance retrospectively of the transportnoise during the last 10-minute menu (eg two train pas-sages) the panelist will have to choose a strategy on howto go about this For example the annoyance may only bereferred to the two noise-stimulus periods or to the whole10-minute period (menu) It has been shown that the noiseannoyance of two overlapping (equal) noises would be ex-pected always to be less than the arithmetic sum of the twoannoyances (for a review see [24]) It is more uncertainhow total annoyance of two train passages separated intime will actually be acquired A laboratory experimentwhich included long sound fragments [4] has not foundthe above-mentioned annoyance difference between differ-ent train types which is in line with our results

42 Advanced scaling methodology

Long-term retrospective annoyance asked for in question-naire surveys has typically been assessed on categoryscales (eg [3]) A response category is then implicitlypostulated to be identical for every participant by verballabeling of the two end points or of every response boxalso the intervals between categories are assumed to be thesame However this assumption does not hold true [25]eg in questionnaire surveys the response criteria (scalevalue or category borders) for annoyance are much higherfor respondents in low noise areas as compared to those inhighly exposed areas The most well known scaling bias inlaboratory experiments is the context effect in which par-ticipants distribute their responses over the ldquofullrdquo range ofcategories independent of the size of the exposure range(for a review see [22]) In the process of using categoryscales floor and ceiling effects on annoyance may also ap-pearTo avoid uncontrolled context effects an invariant sound

level range of references was used as the annoyance con-text in the present field experiment Continuous road trafficnoise was chosen as a reference instead of multiple eventsounds because it is simpler to reproduce in future stud-ies To avoid the scaling bias of category scales the methodof magnitude estimation was chosen in which participantswere free to use the range of numbers they felt comfortablewith Master scaling was applied to these individual an-noyance estimates involving a transformation function toa common master scale defined by the references whichsound levels defined the scaling context In theory thismaster scale transformation will calibrate the loudness-dependence of noise annoyance whereas the relative con-

598

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

tribution to noise annoyance from qualitative content (egthe type of sound the time pattern and cues for speed anddistance) will hopefully be unchanged

Earlier research has shown that master scaling with ref-erences works well for loudness or annoyance of a one-occasion target exposure that is when repeated exposureis unfeasible (eg experiments with long duration expo-sures) or impossible (eg questionnaire surveys in fieldstudies) an example can be found in [23] The results ob-tained from the present field experiment are probably morereliable than the results that would have been obtained bycategory scaling The test-retest reliability of the panelistsrsquomagnitude estimates of annoyance of the reference soundswas found to be very good (above 08) compared to thereliability of 072 obtained in [8] for a group of 12 muchyounger subjects Considering that our panelists all werenaiumlve participants they also each produced high qualitypsychophysical functions for the reference as discussedin Section 26

43 Other possible explanations

There are several reasons why other investigators havefound a railway bonus (for a review see [3]) which wasnot found in this field experiment One of the reasons forfinding a railway bonus for short (one minute) noises inlistening experiments may be that the relation betweenloudness and LAeq is inherently different for train androad traffic noise Indeed some researchers have arguedthat noise annoyance evaluation in listening tests of shortsounds actually is close to a perceptual loudness evalua-tion (however see [26] on differences between loudness-based and quality-based perceived annoyance) If Zwickerloudness is a good first estimate of perceptual loudnessthe difference between train noise (of different types) andhighway noise would be seen in a Zwicker loudness ver-sus LAeq plot (Figure 11) Because the IC train noise usedin the present experiment was the noise of modern ratherquiet trains a few older and noisier IC train models wereadded in this acoustic analysis At levels above 65 dB(A)TGV and maglev trains seem to be a little louder thanhighway traffic or older IC trains However this effect onZwicker loudness is not significant and does therefore notsupport a railway bonus of 5 dB(A) stipulated in severalcountriesrsquo legislation Rather it seems to be a good actionto start to replace old IC trains by new ones The railwaybonus was originally based on studies with rather old low-speed trains and with much less dense traffic intensity thannowadaysThe intermittent character of railway noise could also be

an explanation for the railway bonus However this doesnot hold for aircraft noise which is also intermittent thiscan be explained by a difference in exposure In the case ofaircraft noise the exposure is on top of buildings and on allfaccedilades In the case of road traffic noise the probability ishigh that there are local roads also but there is a possibil-ity for a ldquoquiet siderdquo people are less annoyed if quiet sidesare available [27] In the case of railway noise there is alow probability for the presence of more than one track

Zw

ickerlo

udness

[sone]

L [dB(A)]Aeq45sec

0

10

20

30

40

50

40 50 60 70 80

Figure 11 Zwicker loudness versus LAeq45s for different types oftransportation sounds ( ) IC train ( ) TGV ( ) maglev train( ) highway traffic and ( ) some additional noisier IC trains(older type)

so the exposure will also be directed at only one faccediladeIn comparing road traffic and trains the faccedilade insulationwill be more effective in the case of train noise because ofthe smaller low-frequency proportion associated to trainnoise In comparing aircraft and trains which are both in-termittent the indoor exposure is certainly more intensivefor aircraft Considering these arguments it seems obviousthat aircraft is more annoying than road traffic which ismore annoying than train However faccedilade reduction wastaken into account in the present field experiment and stillthere was no clear railway bonus found Compared withthe field condition with closed windows and the faccediladefilter used in [8] a partially open window was used in thepresent field experiment which could explain this

In surveys questioning people at their home a lowerreported annoyance for train noise compared to highwaytraffic noise was observed in a particular range of noiselevels Most of the possible explanations proposed in liter-ature conflict with the fact that this railway bonus wouldbe observed in experiments based on single passages Wemention just a few The typical character of train noise andthe concentration of the sound energy in short time inter-vals may be advantageous with regard to activity distur-bance If the level is sufficiently low the probability ofnoticing the train noise is small compared to the prob-ability of noticing the sound of a continuous source Inaddition to physical differences in the sound the ldquogreenimagerdquo of trains as a means of transportation may add tothe acceptability of the source and thus increase the tol-erance to its noise that is as long as train passages arenot too frequent However a more recent hedonic pric-ing study found that householders in Birmingham placea greater value on reductions in railway noise than in roadtraffic noise [28] Cross-cultural studies (in field and lab-oratory context) have shown that a railway bonus is notuniversal [29 30] which would favor the argument aboveIt has further been shown that the bonus varies depend-ing on the (multiple) exposure situation [31] Based on theabove only part of the effect is supposed to be visible in

599

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

field experiments such as the one reported of in this paperPart of the effect is precisely what is observed

5 Conclusion

This study has shown that in an ldquoat homerdquo like contextnoise annoyance caused by different types of trains at thesame average outdoor faccedilade exposure level is not signifi-cantly different In particular magnetic levitation systemsare not more annoying than conventional high speed trainswhich is in agreement with earlier research Noise annoy-ance caused by conventional trains was not found to besignificantly lower than annoyance caused by TGVrsquos ormaglev trains at the same average faccedilade exposure Fieldsurveys have shown that for the same average sound levelrailway noise causes less annoyance or highly annoyedpersons than highway traffic noise Although our field ex-periment included several factors that may contribute tothis effect we could not observe it

More insight may be gained by taking into account thepsychoacoustic characteristics of the noise exposure andthe relevant personal factors of the panelists This paperhas focused on discussing the experimental methodologyin great detail and on presenting the results as a functionof the average outdoor faccedilade exposure level since thisis the main noise legislation indicator used in the Nether-lands Results of a detailed psychoacoustic analysis willbe reported in a future paper

Acknowledgments

This research was financed by the Project Group Zuider-zeelijn of the Ministry of Public Transport Public Worksand Water Management in the Netherlands The membersof the project steering committeemdashGilles Janssen (dBVi-sion) Annemarie Ruysbroek (RIVM) Martin van denBerg (VROM) and Pieter Jansse (Project Group Zuider-zeelijn) mdash are acknowledged for their valuable input Wealso appreciate the experimental assistance provided byMs Ingrid Decoster Finally we would like to thank thereviewers for their valuable comments and suggestions

References

[1] J M Fields J G Walker Comparing the relationships be-tween noise level and annoyance in different surveys a rail-way vs aircraft and road traffic comparison J Sound Vib81 (1982) 51ndash80

[2] U Moehler Community response to railway noise A re-view of social surveys J Sound Vib 120 (1988) 321ndash332

[3] H M E Miedema H Vos Exposure-response relation-ships for transportation noise J Acoust Soc Am 104(1998) 3432ndash3445

[4] H Fastl G Gottschling Subjective evaluation of noiseimmissions from transrapid Proceedings of the 25thInternational Congress on Noise Control Engineering(Intermiddotnoise) Liverpool UK Juli 1996 F A Hill R Law-rence (eds)

[5] D Neugebauer J Ortscheid Geraumluschbewertung des Tran-srapid Ist der Transrapid wie die Bahn zu beurteilenBerlin Umweltbundesamt Texte 25 1997

[6] D Neugebauer J Ortscheid Geraumluschbewertung des Tran-srapid Fortschritte der Akustik mdash DAGA 97 Kiel Ger-many 1997

[7] J Vos Annoyance caused by magnetic levitation trainTransrapid 08 mdash a laboratory study TNO report TM-03-C001 The Netherlands Organisation for Applied ScientificResearch (TNO) Soesterberg The Netherlands Jan 2003

[8] J Vos Annoyance caused by the sounds of a magnetic lev-itation train J Acoust Soc Am 115 (2004) 1597ndash1608

[9] B Berglund A Preis Is perceived annoyance more sub-ject-dependent than perceived loudness Acta Acusticaunited with Acustica 83 (1997) 313ndash319

[10] M Ishibashi K Ueno H Tachibana M Watanabe Sub-jective experiment on annoying sounds in living environ-ments Proceedings of the 18th International Congress onAcoustics (ICA) Kyoto Japan 2004

[11] J Maillard J Martin P Champelovier J Lambert Percep-tive evaluation of road traffic noise inside buildings using acombined image and wave field synthesis system Proceed-ings of CFADAGArsquo04 Strasbourg France Maumlrz 2004

[12] R F S Job Community response to noise A review offactors influencing the relationship between noise exposureand reaction J Acoust Soc Am 83 (1988) 991ndash1001

[13] H M E Miedema H Vos Demographic and attitudinalfactors that modify annoyance from transportation noise JAcoust Soc Am 105 (1999) 3336ndash3344

[14] W Ellermeier M Eigenstetter K Zimmer Psychoacousticcorrelates of individual noise sensitivity J Acoust SocAm 109 (2001) 1464ndash1473

[15] D Vaumlstfjaumlll M Kleiner T Gaumlrling Affective reactions tointerior aircraft sounds Acta Acustica united with Acustica89 (2003) 693ndash701

[16] T Feng C Xiao-hong L Tan-feng A social survey onthe effects of high-speed maglev noise on residents inShangai Proceedings of The 2005 Congress and Expo-sition on Noise Control Engineering (Intermiddotnoise) Rio deJaneiro Brazil Aug 2005

[17] J J Gibson The ecological approach to visual perceptionLawrence Erlbaum Associates Inc Hillsdale New JerseyUSA 1979

[18] J G Neuhoff Ecological psychoacoustics Introductionand history ndash In Ecological Psychoacoustics J G Neuhoff(ed) Elsevier Academic Press London UK 2004

[19] C Guastavino B F G Katz J-D Polack D J LevitinD Dubois Ecological validity of soundscape reproductionActa Acustica united with Acustica 91 (2005) 333ndash341

[20] A J Berkhout D de Vries P Vogel Acoustic control bywave field synthesis J Acoust Soc Am 93 (1993) 2764ndash2778

[21] E A M Franssen J E F van Dongen J M H Ruys-broek H Vos R K Stellato Hinder door milieufactorenen de beoordeling van de leefomgeving in Nederland In-ventarisatie verstoringen 2003 RIVM report 815120 001National Institute for Public Health and the Environment(RIVM) Bilthoven The Netherlands 2004

[22] L E Marks D Algom Psychophysical scaling ndash In Mea-surement Judgment and Decision Making M H Birn-baum (ed) Academic Press New York 1998

[23] B Berglund Quality assurance in environmental psycho-physics ndash In Ratio Scaling of Psychological Magnitudes -In Honor of the Memory of S S Stevens S J BolanowskiG A Gescheider (eds) Erlbaum Hillsdale N J 1991

600

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

Table I Sound exposure levels (ASEL) for one 45-second train passage sound level (LAeq45s) of one 45-second train passage and ofhighway traffic and sound level (LAeq10min) of the 10-minute menus of the experiment at 25m to 200m distance to track or route(all free field recordings) The train noise LAeq10min values are given for the 2-train menu to obtain the LAeq10min values for thecorresponding 4-train menu add 3 dB

Sound source Outdoor ASEL [dB(A)] Outdoor LAeq45s [dB(A)] Outdoor LAeq10min [dB(A)]25m 50m 100m 200m 25m 50m 100m 200m 25m 50m 100m 200m

Maglev 200 kmh 801 729 713 597 636 564 548 432 553 481 465 349300 kmh 863 830 803 696 698 665 638 531 615 582 555 448400 kmh 926 887 852 704 761 722 687 539 678 639 604 456

TGV 140 kmh 841 783 736 644 676 618 571 479 593 535 488 396300 kmh 928 906 869 830 763 741 704 665 680 658 621 582

IC 140 kmh 750 809 724 620 585 644 559 455 502 561 476 372

Highway free flow minus minus minus minus 716 661 626 553 minus 653 minus minus

Figure 3 Sound exposure level (ASEL) in 13-octave bands offour different types of traffic sounds all recorded during 45 sec-onds in free field at a distance of 50m to the track (or highwayroute) ( ) a passage of a maglev train traveling at 400 kmh( ) a passage of a TGV traveling at 300 kmh ( ) a passage ofan IC train traveling at 140 kmh and ( ) a highway with freeflow traffic

Figure 4 Sound exposure level (ASEL) in 13-octave bands of amaglev train traveling at 400 kmh recorded during 45 secondsin free field at various distances to the track ( ) 25m ( ) 50m( ) 100m and ( ) 200m

level spanning the whole experimental range were in-cluded in the experiment A 45-second fragment of thehighway noise recorded at 50m distance to the highwaywas used as the centre reference sound A filter which at-tenuates the sound at frequencies below 500Hz by 3 dBand above 500Hz by 6 dB was applied 3 times to produce3 reference sounds with varying level all below the levelof the centre reference sound giving the impression that

the source is further away In the same way a filter thatamplifies the sound at frequencies below 500Hz by 3 dBand above 500Hz by 6 dB was used to generate 3 refer-ence sounds with varying level higher than the level of thecentre reference sound

23 Selection of a representative panel

In contrast to previous experimental work on noise annoy-ance caused by high speed trains in which small ldquoconve-nientrdquo samples of test persons were recruited the selectionof panelists was here made to guarantee a representativesample of panelists A questionnaire was administered atthe doorstep of the homes of approximately 1500 personsall living within a distance of 15 km from the experimentalsite In an introductory letter one inhabitant of the housewas invited to participate in the study The prerequisiteswere that (s)he had to fill in and send the questionnaireback to the address on the enclosed stamped envelope Acompensation of C 100 was offered for participation

The questionnaire contained selected questions that hadbeen asked to a representative sample of the target popula-tion in a recent survey The structure of the Dutch popula-tion was inferred to be representative from a recent RIVMsurvey [21] and partly from a Eurobarometer question-naire Our questionnaire contained (standard) questions onenvironmental noise as regards perception annoyance andsleep disturbance Included were evaluations of the qualityof the neigbourhood in terms of housing and environmen-tal pollution of other types than noise as well as evalu-ations of overall satisfaction with the current living situ-ation Other questions addressed basic demographic vari-ables such as age gender education housing family sizeand work arrangements A set of questions were also in-cluded on general and mental health hearing ability en-vironmental background opinion and worry and environ-mental sensitivity

A procedure to draw panelists representative of thetarget population from the 255 replies received involvedthree stages Stage 1 removed potential panelists on thebasis of their age and hearing ability (information had al-ready been given in the introductory letter) Stage 2 furtherremoved those that were very dissimilar from the typical

592

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

Table II Comparison between the panelists and the referencepopulation on various criteria Mean and standard deviation isshown the results for the second series of criteria are on an 11-point scale and vary from 0 (not at all bad) to 10 (very good)

Criterium Participants Reference

Gender [ male female] 51 49 48 52Age [year] 451 plusmn 134 456 plusmn 177

Noise sensitivity 51 plusmn 24 46 plusmn 26

Quality of traffic noise in the 66 plusmn 24 64 plusmn 23living environment

Quality of the living 76 plusmn 14 73 plusmn 13environment

Feeling afraid or frightened 24 plusmn 20 23 plusmn 21

Dutch person on the basis of binary coding of most of theother criteria included in the questionnaire This stage im-plicitly assessed individual responses on the questions asregards their concordance with the response profile of thetypical Dutch person in the reference survey Stage 3 fi-nally selected panelists on the basis of fuzzy resemblanceto the typical Dutch person on the most critical criteria ofannoyance surveys such as age gender education noisesensitivity feeling afraid or frightened hearing train noiseat home quality of traffic noise in the living environmentquality of the living environment general health and ill-ness Finally ca 100 representative participants were se-lected Table II shows a comparison of the panelists withthe Dutch target population as regards the mean and stan-dard deviation of some of the selection criteria used andmentioned above

24 Listening test outline

Four to six panelists jointly participated in a session Theoverall structure and time schedule of the listening exper-iment was identical for each group of panelists It startedwith a 14-minute training session during which the pan-elists were asked to scale each of the 7 reference (high-way) sounds two times (in random order) Thereafter7 10-minute menus were played of which the first menualways was the highway traffic menu A short break wasthen taken and the training session was repeated afterwhich again 7 new 10-minute menus were played Afterthis experiment with menus a more conventional psychoa-coustical listening test was conducted in which the pan-elists had to scale 45-second excerpts of all transport noisestimuli used in the menu experiment The duration of anexperimental session was on average about 4 hours Toillustrate how the listening test was performed Figure 5shows the sound level in dB(A) rerecorded in front of thefaccedilade during one of the panelist groupsrsquo listening exper-imentIn all two times 6 train menus were presented to each

panelist It was decided that within one set of 6 trainmenus conventional trains (IC or high-speed) should notbe mixed with magnetic levitation trains By this separa-

tion it was possible to include a retrospective evaluationover the last hour as well From previous experience it wasknown that the order of the menu pesentations might af-fect the results Half of the panelists were therefore pre-sented the maglev train sounds first the other half the con-ventional trains first A singular session consisted of thesame number of passages inside the menus This avoidsthat panelists would concentrate on counting events Fi-nally since one distance to the track would create a nat-ural setting large distances were never mixed with shortdistances in the menus of a session

During the experimental sessions perceived noise an-noyance of all transport noises was scaled with the methodof free-number magnitude estimation [22] The panelistswere asked to write down their magnitude assessments ondifferent coloured pieces of paper Before the start of theexperiment the panelists were instructed to select an ap-propriate number and then to double this number if theyfound the next stimulus to be twice as annoying to makethe number three times larger if they found the next stim-ulus to be three times as annoying etc and to scale 0 ifthey considered it not to be annoying at all For each 45-second sound (training sessions and conventional listeningtest) a conditional question was included ldquoTo what extentwould you be annoyed by this traffic sound if you heardit while relaxingrdquo For each 10-minute menu a very simi-lar but retrospective question was asked ldquoTo what extentwere you annoyed by traffic sound during the previous pe-riodrdquo In these latter questions we explicitly did not wantto refer to train noise since we wanted the panelists to de-cide themselves whether the sound period they last heardsounded like train-contaminated or not

25 Master scaling

In all experimental sessions the 7 road-traffic-noise-likereference sounds helped the panelists to define their ownscaling context The annoyance values given to these refer-ence sounds made it possible to control for the individualpanelistsrsquo choice-of-number behaviour in scaling the tar-get train sounds It would also control the influence of per-sonal factors such as noise sensitivity To get rid of theseeffects each individual panelistrsquos annoyance scale was cal-ibrated by the aid of the reference to the common masterscale [23]

A graphical illustration of the master scale transforma-tion applied to the annoyance reference data of one of thepanelists is given in Figure 6 The average annoyance re-ported for each of the 7 reference sound levels of roadtraffic noise is plotted in lin-log coordinates against theirsound levels LAeq45s measured at the faccedilade Individualpsychophysical functions are fitted to the reference data(open circles) They are of the form

Ar = a + b logSr (1)

where Ar is the reported annoyance during the train-ing session and logSr is the corresponding ldquoroad trafficnoiserdquo reference (r) sound level in dB(A) The constants a

593

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

Figure 5 Sound pressure level rerecorded in front of the faccedilade during one panelist groupsrsquo participation in the whole listening testtwo training sessions two menu sessions and one conventional psychoacoustical experiment with references

0

10

20

30

40

50

60

70

80

40 50 60 70 80 90

R = 55 master-scaled annoyance

A = 70 assessment

annoyance

road traffic reference sound level [dB(A)]

Figure 6 Calculation of master-scaled annoyance using one pan-elistrsquos empirical psychophysical function of the reference sounds(data points with solid line) and the master function for the samesounds (dashed line obtained as average function for all pan-elists)

and b will be different for each panelist and will depend ontheir choice-of-number behaviour in the particular scalingcontext The empirically derived master functions for thegroup of 100 panelists (dashed line in Figure 6) were thenused to transform the free number magnitude estimationsof the train or road traffic menus for each individual Aeto the corresponding annoyance values R in master scaleunits

R = minus629 + 145Ae minus a

b (2)

The slope of the master function was set to 145 whichis the average slope of all the individual psychophysicalfunctions whereas the intercept was set to produce a valueof ldquozerordquo for the most quiet train menu The reason for thelatter choice was that a majority of the panelists (84)reported their annoyance to be zero for this menu and amajority of the panelists reported annoyance to be greaterthan zero for all other menus

The choice of a logarithmic psychophysical function(equation 1) was a compromise In previous magnitudeestimation experiments of loudness [23 9] a power func-tion of the form logA = c + d logS was found to fit the

Table III Test-retest reliability of panelistsrsquo perceived noise an-noyance of the 7 reference road traffic sounds Each cell containsan arithmetic mean of Pearsonrsquos coefficient (r) and its standarderror Ts1 Training session 1 Ts2 Training session 2 Ct Con-ventional test

Ts1 Ts2 CtSet 1 Set 2 Set 1 Set 2 Set 1

Ts22 082plusmn0015

Ts21 086 087plusmn0014 plusmn0016

Ts22 086 088 087plusmn0017 plusmn0020 plusmn0019

Ct1 083 083 082 085plusmn0015 plusmn0021 plusmn0019 plusmn0020

Ct2 084 085 081 084 082plusmn0015 plusmn0019 plusmn0016 plusmn0019 plusmn0015

empirical data best However in this experiment noise an-noyance rather than loudness was scaled and thus obvi-ously also a value of zero (= not at all annoyed) had tobe handled although the noise was heard and its loudnesswas above zero The power function (after removal of ze-ros) did not fit the data better than the chosen logarithmicfunction

26 Data quality analysis

The master scaling made it possible to investigate the qual-ity of the experimental data in two ways as panelistsrsquo test-retest reliability and as their scaling ability The 7 refer-ence sounds were presented 6 times to each panelist twicein the two training sessions and twice in the last conven-tional listening test The set of 6 reference scale valueswere used to determine each panelistrsquos test-retest reliabil-ity of annoyance Table III shows the Pearsonrsquos coefficientof correlation for these 6 annoyance scales averaged overall panelists The test-retest reliability was very good be-tween 081 and 088 and the standard error was low be-tween 0014 and 0019

The deviation from the proposed master function (equa-tion 1) was used to assess the data quality and annoyance

594

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

Table IV Distribution of constants of the panelistsrsquo individualpsychophysical functions (Eq 1) The number of data sets refersto the average of 4 or 2 raw annoyance values which was takenfor each of the 7 reference sounds to calculate the psychophysicalfunctions

Data Psychophysical functionSets r2 a b

Training session 1amp2 4 0947 minus6727 1449plusmn0077 plusmn6128 plusmn1230

Conventional test 2 0881 minus4757 1105plusmn0118 plusmn4817 plusmn0948

scaling ability for each panelist and to trace errors and in-accuracies Table IV shows the distribution of constants ofthe panelistsrsquo individual psychophysical functions (equa-tion 1) The average annoyance variance explained bysound level (LAeq45s) of the reference road traffic soundswas found to range from 88 to 95 All panelists wereable to produce acceptable individual logarithmic func-tions of annoyance as a function of sound level to the refer-ence They have thus produced acceptable annoyance datain order to transform these to a common master scale ofannoyance no panelists were excluded from further dataanalysis

3 Results

The main listening experiment with menus differed fromprevious laboratory experiments in a number of aspectsOne important novelty is that participants were asked tojudge annoyance over a longer period of time mdash Fastl andGottschlingrsquos experiment [4] forms an exception Duringthe 10-minute periods the panelists were engaged in lowattention relaxing activities such as reading a magazinemaking a conversation or having something to drink In or-der to find out how this new approach affected the resultsa subsequent experiment was included which was morecomparable to earlier experiments on train noise (eg [8])

31 Main field experiment with 10-minute menus

The panelistsrsquo master scale values of annoyance were av-eraged for each menu in the field experiment A step-wise multiple linear regression analysis was performedwith average master scaled annoyance as dependent vari-able and (a) time averaged A-weighted faccedilade exposureLAeq10min (b) distance to the source (logarithmic) and(c) source type as independent variables Because of itslegislative importance in the Netherlands the faccedilade ex-posure was preferred to the actual panelistsrsquo noise expo-sure Faccedilade exposure was calculated from the sound lev-els measured on the recording sites since the faccedilade levelsmeasured during the experiment also contain noise fromwind and rain It has to be noted that the actual soundexposure levels experienced by panelists participating ina single experimental session may differ because of dif-ferent seating positions However personal characteristicssuch as noise sensitivity will have a much larger influence

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq10min

(a)

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq10min

(b)

Figure 7 Average master scaled annoyance of the menus ver-sus LAeq10min (a) for 2 events per 10-minute menu and (b) for4 events per 10-minute menu for different types of train sounds( ) IC train ( ) TGV and ( ) maglev train In comparisonthe annoyance for the highway traffic ( ) is also shown Stan-dard error on means is indicated as well as the master function(dashed line)

on perceived annoyance as compared to the influence of aslightly different exposure

Table V summarizes the results In the first modelsound level was the only independent variable this modelexplained 80 of the variance in annoyance In the secondmodel distance to track was added to sound level as an in-dependent variable this model increased the variance ex-plained to 85 (F-change = 1449 df1 = 1 df2 = 46 p lt0001) Thus distance to source explained a significant ad-ditional part of the annoyance variance not accounted forby sound level In the third model source type was in-cluded as a third independent variable along with soundlevel and distance Source type was defined on a nominalscale MAGLEV TGV IC and HIGHWAY It was intro-duced in the analysis as three dummy variables coded 0and 1 (the highway noise source type corresponds to thecase that these variables are all zero) The inclusion ofsource type did not increase significantly (F-change lt 10)the proportion of variance explained This suggests thatstatistically there is no additional contribution of sourcetype on perceived annoyance over and above the effects ofsound level and distance It can therefore be concluded thatmagnetic-levitation based transportation systems are notsignificantly more annoying than conventional rail basedsystems (same faccedilade LAeq and same distance are pre-requisites) Moreover railway noise was not found to be

595

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

Table V Stepwise multiple regression analysis of acoustic variables on perceived annoyance of train and highway traffic sounds for themain field experiment with 10-minute menus The Pearsonrsquos correlation coefficients of the variables entered in the regression analysisare shown at the bottom lowastp lt 005 lowastlowastp lt 001 lowastlowastlowastp lt 0001

Model Model Model fit increase F-change Independent Variables Coefficient t-valuefit (r2) (r2-change)

1 080 080 18748lowastlowastlowast LAeq10min [dB(A)] 118 1369lowastlowastlowast

2 085 005 1449lowastlowastlowast LAeq10min [dB(A)] 092 917lowastlowastlowast

log10(distance [m]) minus1074 minus381lowastlowastlowast

3 085 000 013 LAeq10min [dB(A)] 096 822lowastlowastlowast

log10(distance [m]) minus1017 minus333lowastlowastMAGLEV [01] 145 027TGV [01] 085 016IC [01] 227 040

Label Variable ANN LEQ DIST MAG TGV

ANN Annoyance [master scale]LEQ LAeq10min [dB(A)] 0894DIST log10(distance [m]) minus0754 minus0659MAG MAGLEV [01] minus0023 minus0038 0009TGV TGV [01] 0132 0224 0006 minus0682IC IC [01] minus0179 minus0290 0004 minus0433 minus0308

systematically less annoying than highway traffic noiseThis means that no support for a railway bonus was foundin this experiment at least it was not as obvious that itcould be observed using linear statistics Figure 7 givesan overview of the annoyance functions for the 10-minutemenus as a function of LAeq10min The dashed line indi-cates the master function of annoyance for the road-traffic-like sounds used as referencesThe shorter rise time of the noise of arriving high speed

trains may create more annoyance than a conventionaltrain can do Figure 8 Panel a shows the rise speeds indB(A)s in proportion to circle sizes These values werecalculated for all sound events included in this experi-ment by fitting a straight line through the initial increasein sound level The accelerating growth of annoyance withincreasing LAeq may be explained by the rise time In Fig-ure 8 Panel b the size of the circles is instead propor-tional to the distance to the track For LAeq in the intervalbetween 50 and 65 dB(A) annoyance is clearly lower fortrain passages at larger distances than for train passagesat closer distances or road traffic noise (dashed line) Thiscould indicate that a possible noise annoyance bonus fortrain noise would only hold at larger distances from thetrack and only in the latter LAeq interval

32 Conventional listening test

In the conventional listening experiment the sounds werepresented as short 45-second fragments containing thesound of one train passage and highway excerpts Figure 9shows the results of these master scaled annoyance val-ues as a function of time averaged A-weighted faccedilade ex-posure LAeq45s A railway penalty can be observed bothin regard to the artificial reference sounds as well as to

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq10min

(a)

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq10min

(b)

Figure 8 Average master scaled annoyance of the menus versusLAeq10min showing (a) the noise event rise speed and (b) the dis-tance to the track as the size of the circles The master functionis also indicated (dashed line)

the highway sounds Figure 10 shows the annoyance as afunction of rise speed ( Panel a) and distance to the track(Panel b)

596

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

Table VI Stepwise multiple regression analysis of acoustic variables on perceived annoyance of train sounds (no highway trafficsounds) for the conventional listening test (45-second passages) The Pearsonrsquos correlation coefficients of the variables entered in theregression analysis are shown at the bottom lowastp lt 005 lowastlowastp lt 001 lowastlowastlowastp lt 0001

Model Model Model fit increase F-change Independent Variables Coefficient t-valuefit (r2) (r2-change)

1 095 095 42017lowastlowastlowast LAeq45s [dB(A)] 167 2050lowastlowastlowast

2 098 003 1250lowastlowastlowast LAeq45s [dB(A)] 123 1104lowastlowastlowast

Speed [kmh] 002 203log10(distance [m]) minus178 minus077Rise speed [dB(A)s] 063 365lowastlowast

3 099 001 098 LAeq45s [dB(A)] 108 685lowastlowastlowast

Speed [kmh] 003 209log10(distance [m]) minus476 minus146Rise speed [dB(A)s] 058 330lowastlowast

MAGLEV [01] minus070 minus037TGV [01] 166 104

Label Variable ANN LEQ SPD DIST RISE MAG

ANN Annoyance [master scale]LEQ LAeq45s [dB(A)] 0975SPD Speed [kmh] 0646 0541DIST log10(distance [m]) minus0613 minus0667 minus0001RISE Rise speed [dB(A)s] 0885 0804 0708 minus0437MAG MAGLEV [01] 0070 minus0017 0552 0000 0188TGV TGV [01] 0207 0246 minus0190 0000 0100 minus0707

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq45sec

Figure 9 Average master scaled annoyance versus LAeq45s forthe conventional listening test for different types of train sounds( ) IC train ( ) TGV and ( ) maglev train In comparison theannoyance for the highway traffic ( ) is also shown Standard er-ror on means is indicated as well as the master function (dashedline)

A stepwise multiple linear regression analysis was alsoperformed separately for the train noises (Table VI) Thefirst model in which sound level LAeq45s was included asthe only independent variable explained 95 of the vari-ance in annoyance In the second model train speed dis-tance to the track and rise speed were added to sound levelas independent variables This increased the variance ex-plained to 98 (F-change = 1250 df1 = 3 df2 = 19p lt 0001) Apart from sound level also rise speed con-

tributed significantly to the variance explained The thirdmodel in which train type was added as an independentdummy variable did not significantly increase the propor-tion explained variance (F-change lt 10) These resultssuggest that in this conventional listening test there is nodifference in perceived annoyance between different typesof trains over and above the effect of sound level and risespeed

One has to note that the number of responses to eachstimulus was smaller in the main experiment (10-minutemenus) than in the conventional listenig test (45-secondpassages) This explains why the standard errors are lowerand the explained variance is higher in the latter experi-ment

4 Discussion

The annoyance results of the present field experiment areclose to residentsrsquo everyday reality although comparisonwith published studies is somewhat limited Previous lab-oratory experiments on noise annoyance of conventionalIC and high-speed trains specifically magnetic levitationtrains [8 5] report significant differences for these typesof sound In particular the results have shown that forthe same LAeq high-speed trains were more annoyingthan other trains Compared to road traffic noise the citedstudies claimed a lower annoyance level for conventionaltrains In the present field experiment we did not find sup-port for any annoyance difference between various types

597

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

0

10

20

30

40

50

60

70

80

40 50 60 70 80 90

annoyance

[maste

rscale

units]

L [dB(A)]Aeq45sec

(a)

0

10

20

30

40

50

60

70

80

40 50 60 70 80 90

annoyance

[maste

rscale

units]

L [dB(A)]Aeq45sec

(b)

Figure 10 Average master scaled annoyance versus LAeq45s

showing (a) the noise event rise speed and (b) the distance tothe track as the size of the circles The master function is alsoindicated (dashed line)

of trains and road traffic Some possible explanations willbe given in the following subsections

41 Realistic listening situation with 10-minutemenus

The experiment was performed in a realistic setting inwhich outdoor transportation noise was reproduced andnatural outdoor-to-indoor sound propagation characteris-tics were utilized (slightly open window) This setting pro-vided a realistic sound environment indoors Subgroups ofpanelists were kept indoors during the four-hour experi-ments and upon request annoyance to transport noise wasreported with reference to 10-minute periods

Because trains run on expected schedules to which peo-ple habituate the experimental situation in classical exper-iments is rather unrealistic The experimental one-passagesituation [8] requires full attention and will have a largevariation of train sounds compared with a particular rail-way track The outcome will to a large extent depend onthe experimental context that is the variation introducedin the experiment by selecting stimuli and using randompresentation orders Random orders of recordings can beselected and arranged so that annoyance judgments on cat-egory scales plotted against sound level differentiate wellor not well on type of transport In the present field experi-ment sub-context in sessions was kept invariant similar tothe situation on a real railway track The judgmental con-

text will then be much more restricted as is the case whenliving along one railway track

Next to this the annoyance reports of the one 45-secondtrain passage were higher than those of two passages ofthe same train within the 10-minute menus This is all inorder because the two types of annoyance were masterscaled in order to become comparable over experimentalsessions When judging 45-second train passages imme-diately after exposure it is quiet clear that the task is toassess the annoyance of that particular train passage (orother sounds that were presented) However when askedto assess the annoyance retrospectively of the transportnoise during the last 10-minute menu (eg two train pas-sages) the panelist will have to choose a strategy on howto go about this For example the annoyance may only bereferred to the two noise-stimulus periods or to the whole10-minute period (menu) It has been shown that the noiseannoyance of two overlapping (equal) noises would be ex-pected always to be less than the arithmetic sum of the twoannoyances (for a review see [24]) It is more uncertainhow total annoyance of two train passages separated intime will actually be acquired A laboratory experimentwhich included long sound fragments [4] has not foundthe above-mentioned annoyance difference between differ-ent train types which is in line with our results

42 Advanced scaling methodology

Long-term retrospective annoyance asked for in question-naire surveys has typically been assessed on categoryscales (eg [3]) A response category is then implicitlypostulated to be identical for every participant by verballabeling of the two end points or of every response boxalso the intervals between categories are assumed to be thesame However this assumption does not hold true [25]eg in questionnaire surveys the response criteria (scalevalue or category borders) for annoyance are much higherfor respondents in low noise areas as compared to those inhighly exposed areas The most well known scaling bias inlaboratory experiments is the context effect in which par-ticipants distribute their responses over the ldquofullrdquo range ofcategories independent of the size of the exposure range(for a review see [22]) In the process of using categoryscales floor and ceiling effects on annoyance may also ap-pearTo avoid uncontrolled context effects an invariant sound

level range of references was used as the annoyance con-text in the present field experiment Continuous road trafficnoise was chosen as a reference instead of multiple eventsounds because it is simpler to reproduce in future stud-ies To avoid the scaling bias of category scales the methodof magnitude estimation was chosen in which participantswere free to use the range of numbers they felt comfortablewith Master scaling was applied to these individual an-noyance estimates involving a transformation function toa common master scale defined by the references whichsound levels defined the scaling context In theory thismaster scale transformation will calibrate the loudness-dependence of noise annoyance whereas the relative con-

598

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

tribution to noise annoyance from qualitative content (egthe type of sound the time pattern and cues for speed anddistance) will hopefully be unchanged

Earlier research has shown that master scaling with ref-erences works well for loudness or annoyance of a one-occasion target exposure that is when repeated exposureis unfeasible (eg experiments with long duration expo-sures) or impossible (eg questionnaire surveys in fieldstudies) an example can be found in [23] The results ob-tained from the present field experiment are probably morereliable than the results that would have been obtained bycategory scaling The test-retest reliability of the panelistsrsquomagnitude estimates of annoyance of the reference soundswas found to be very good (above 08) compared to thereliability of 072 obtained in [8] for a group of 12 muchyounger subjects Considering that our panelists all werenaiumlve participants they also each produced high qualitypsychophysical functions for the reference as discussedin Section 26

43 Other possible explanations

There are several reasons why other investigators havefound a railway bonus (for a review see [3]) which wasnot found in this field experiment One of the reasons forfinding a railway bonus for short (one minute) noises inlistening experiments may be that the relation betweenloudness and LAeq is inherently different for train androad traffic noise Indeed some researchers have arguedthat noise annoyance evaluation in listening tests of shortsounds actually is close to a perceptual loudness evalua-tion (however see [26] on differences between loudness-based and quality-based perceived annoyance) If Zwickerloudness is a good first estimate of perceptual loudnessthe difference between train noise (of different types) andhighway noise would be seen in a Zwicker loudness ver-sus LAeq plot (Figure 11) Because the IC train noise usedin the present experiment was the noise of modern ratherquiet trains a few older and noisier IC train models wereadded in this acoustic analysis At levels above 65 dB(A)TGV and maglev trains seem to be a little louder thanhighway traffic or older IC trains However this effect onZwicker loudness is not significant and does therefore notsupport a railway bonus of 5 dB(A) stipulated in severalcountriesrsquo legislation Rather it seems to be a good actionto start to replace old IC trains by new ones The railwaybonus was originally based on studies with rather old low-speed trains and with much less dense traffic intensity thannowadaysThe intermittent character of railway noise could also be

an explanation for the railway bonus However this doesnot hold for aircraft noise which is also intermittent thiscan be explained by a difference in exposure In the case ofaircraft noise the exposure is on top of buildings and on allfaccedilades In the case of road traffic noise the probability ishigh that there are local roads also but there is a possibil-ity for a ldquoquiet siderdquo people are less annoyed if quiet sidesare available [27] In the case of railway noise there is alow probability for the presence of more than one track

Zw

ickerlo

udness

[sone]

L [dB(A)]Aeq45sec

0

10

20

30

40

50

40 50 60 70 80

Figure 11 Zwicker loudness versus LAeq45s for different types oftransportation sounds ( ) IC train ( ) TGV ( ) maglev train( ) highway traffic and ( ) some additional noisier IC trains(older type)

so the exposure will also be directed at only one faccediladeIn comparing road traffic and trains the faccedilade insulationwill be more effective in the case of train noise because ofthe smaller low-frequency proportion associated to trainnoise In comparing aircraft and trains which are both in-termittent the indoor exposure is certainly more intensivefor aircraft Considering these arguments it seems obviousthat aircraft is more annoying than road traffic which ismore annoying than train However faccedilade reduction wastaken into account in the present field experiment and stillthere was no clear railway bonus found Compared withthe field condition with closed windows and the faccediladefilter used in [8] a partially open window was used in thepresent field experiment which could explain this

In surveys questioning people at their home a lowerreported annoyance for train noise compared to highwaytraffic noise was observed in a particular range of noiselevels Most of the possible explanations proposed in liter-ature conflict with the fact that this railway bonus wouldbe observed in experiments based on single passages Wemention just a few The typical character of train noise andthe concentration of the sound energy in short time inter-vals may be advantageous with regard to activity distur-bance If the level is sufficiently low the probability ofnoticing the train noise is small compared to the prob-ability of noticing the sound of a continuous source Inaddition to physical differences in the sound the ldquogreenimagerdquo of trains as a means of transportation may add tothe acceptability of the source and thus increase the tol-erance to its noise that is as long as train passages arenot too frequent However a more recent hedonic pric-ing study found that householders in Birmingham placea greater value on reductions in railway noise than in roadtraffic noise [28] Cross-cultural studies (in field and lab-oratory context) have shown that a railway bonus is notuniversal [29 30] which would favor the argument aboveIt has further been shown that the bonus varies depend-ing on the (multiple) exposure situation [31] Based on theabove only part of the effect is supposed to be visible in

599

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

field experiments such as the one reported of in this paperPart of the effect is precisely what is observed

5 Conclusion

This study has shown that in an ldquoat homerdquo like contextnoise annoyance caused by different types of trains at thesame average outdoor faccedilade exposure level is not signifi-cantly different In particular magnetic levitation systemsare not more annoying than conventional high speed trainswhich is in agreement with earlier research Noise annoy-ance caused by conventional trains was not found to besignificantly lower than annoyance caused by TGVrsquos ormaglev trains at the same average faccedilade exposure Fieldsurveys have shown that for the same average sound levelrailway noise causes less annoyance or highly annoyedpersons than highway traffic noise Although our field ex-periment included several factors that may contribute tothis effect we could not observe it

More insight may be gained by taking into account thepsychoacoustic characteristics of the noise exposure andthe relevant personal factors of the panelists This paperhas focused on discussing the experimental methodologyin great detail and on presenting the results as a functionof the average outdoor faccedilade exposure level since thisis the main noise legislation indicator used in the Nether-lands Results of a detailed psychoacoustic analysis willbe reported in a future paper

Acknowledgments

This research was financed by the Project Group Zuider-zeelijn of the Ministry of Public Transport Public Worksand Water Management in the Netherlands The membersof the project steering committeemdashGilles Janssen (dBVi-sion) Annemarie Ruysbroek (RIVM) Martin van denBerg (VROM) and Pieter Jansse (Project Group Zuider-zeelijn) mdash are acknowledged for their valuable input Wealso appreciate the experimental assistance provided byMs Ingrid Decoster Finally we would like to thank thereviewers for their valuable comments and suggestions

References

[1] J M Fields J G Walker Comparing the relationships be-tween noise level and annoyance in different surveys a rail-way vs aircraft and road traffic comparison J Sound Vib81 (1982) 51ndash80

[2] U Moehler Community response to railway noise A re-view of social surveys J Sound Vib 120 (1988) 321ndash332

[3] H M E Miedema H Vos Exposure-response relation-ships for transportation noise J Acoust Soc Am 104(1998) 3432ndash3445

[4] H Fastl G Gottschling Subjective evaluation of noiseimmissions from transrapid Proceedings of the 25thInternational Congress on Noise Control Engineering(Intermiddotnoise) Liverpool UK Juli 1996 F A Hill R Law-rence (eds)

[5] D Neugebauer J Ortscheid Geraumluschbewertung des Tran-srapid Ist der Transrapid wie die Bahn zu beurteilenBerlin Umweltbundesamt Texte 25 1997

[6] D Neugebauer J Ortscheid Geraumluschbewertung des Tran-srapid Fortschritte der Akustik mdash DAGA 97 Kiel Ger-many 1997

[7] J Vos Annoyance caused by magnetic levitation trainTransrapid 08 mdash a laboratory study TNO report TM-03-C001 The Netherlands Organisation for Applied ScientificResearch (TNO) Soesterberg The Netherlands Jan 2003

[8] J Vos Annoyance caused by the sounds of a magnetic lev-itation train J Acoust Soc Am 115 (2004) 1597ndash1608

[9] B Berglund A Preis Is perceived annoyance more sub-ject-dependent than perceived loudness Acta Acusticaunited with Acustica 83 (1997) 313ndash319

[10] M Ishibashi K Ueno H Tachibana M Watanabe Sub-jective experiment on annoying sounds in living environ-ments Proceedings of the 18th International Congress onAcoustics (ICA) Kyoto Japan 2004

[11] J Maillard J Martin P Champelovier J Lambert Percep-tive evaluation of road traffic noise inside buildings using acombined image and wave field synthesis system Proceed-ings of CFADAGArsquo04 Strasbourg France Maumlrz 2004

[12] R F S Job Community response to noise A review offactors influencing the relationship between noise exposureand reaction J Acoust Soc Am 83 (1988) 991ndash1001

[13] H M E Miedema H Vos Demographic and attitudinalfactors that modify annoyance from transportation noise JAcoust Soc Am 105 (1999) 3336ndash3344

[14] W Ellermeier M Eigenstetter K Zimmer Psychoacousticcorrelates of individual noise sensitivity J Acoust SocAm 109 (2001) 1464ndash1473

[15] D Vaumlstfjaumlll M Kleiner T Gaumlrling Affective reactions tointerior aircraft sounds Acta Acustica united with Acustica89 (2003) 693ndash701

[16] T Feng C Xiao-hong L Tan-feng A social survey onthe effects of high-speed maglev noise on residents inShangai Proceedings of The 2005 Congress and Expo-sition on Noise Control Engineering (Intermiddotnoise) Rio deJaneiro Brazil Aug 2005

[17] J J Gibson The ecological approach to visual perceptionLawrence Erlbaum Associates Inc Hillsdale New JerseyUSA 1979

[18] J G Neuhoff Ecological psychoacoustics Introductionand history ndash In Ecological Psychoacoustics J G Neuhoff(ed) Elsevier Academic Press London UK 2004

[19] C Guastavino B F G Katz J-D Polack D J LevitinD Dubois Ecological validity of soundscape reproductionActa Acustica united with Acustica 91 (2005) 333ndash341

[20] A J Berkhout D de Vries P Vogel Acoustic control bywave field synthesis J Acoust Soc Am 93 (1993) 2764ndash2778

[21] E A M Franssen J E F van Dongen J M H Ruys-broek H Vos R K Stellato Hinder door milieufactorenen de beoordeling van de leefomgeving in Nederland In-ventarisatie verstoringen 2003 RIVM report 815120 001National Institute for Public Health and the Environment(RIVM) Bilthoven The Netherlands 2004

[22] L E Marks D Algom Psychophysical scaling ndash In Mea-surement Judgment and Decision Making M H Birn-baum (ed) Academic Press New York 1998

[23] B Berglund Quality assurance in environmental psycho-physics ndash In Ratio Scaling of Psychological Magnitudes -In Honor of the Memory of S S Stevens S J BolanowskiG A Gescheider (eds) Erlbaum Hillsdale N J 1991

600

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

Table II Comparison between the panelists and the referencepopulation on various criteria Mean and standard deviation isshown the results for the second series of criteria are on an 11-point scale and vary from 0 (not at all bad) to 10 (very good)

Criterium Participants Reference

Gender [ male female] 51 49 48 52Age [year] 451 plusmn 134 456 plusmn 177

Noise sensitivity 51 plusmn 24 46 plusmn 26

Quality of traffic noise in the 66 plusmn 24 64 plusmn 23living environment

Quality of the living 76 plusmn 14 73 plusmn 13environment

Feeling afraid or frightened 24 plusmn 20 23 plusmn 21

Dutch person on the basis of binary coding of most of theother criteria included in the questionnaire This stage im-plicitly assessed individual responses on the questions asregards their concordance with the response profile of thetypical Dutch person in the reference survey Stage 3 fi-nally selected panelists on the basis of fuzzy resemblanceto the typical Dutch person on the most critical criteria ofannoyance surveys such as age gender education noisesensitivity feeling afraid or frightened hearing train noiseat home quality of traffic noise in the living environmentquality of the living environment general health and ill-ness Finally ca 100 representative participants were se-lected Table II shows a comparison of the panelists withthe Dutch target population as regards the mean and stan-dard deviation of some of the selection criteria used andmentioned above

24 Listening test outline

Four to six panelists jointly participated in a session Theoverall structure and time schedule of the listening exper-iment was identical for each group of panelists It startedwith a 14-minute training session during which the pan-elists were asked to scale each of the 7 reference (high-way) sounds two times (in random order) Thereafter7 10-minute menus were played of which the first menualways was the highway traffic menu A short break wasthen taken and the training session was repeated afterwhich again 7 new 10-minute menus were played Afterthis experiment with menus a more conventional psychoa-coustical listening test was conducted in which the pan-elists had to scale 45-second excerpts of all transport noisestimuli used in the menu experiment The duration of anexperimental session was on average about 4 hours Toillustrate how the listening test was performed Figure 5shows the sound level in dB(A) rerecorded in front of thefaccedilade during one of the panelist groupsrsquo listening exper-imentIn all two times 6 train menus were presented to each

panelist It was decided that within one set of 6 trainmenus conventional trains (IC or high-speed) should notbe mixed with magnetic levitation trains By this separa-

tion it was possible to include a retrospective evaluationover the last hour as well From previous experience it wasknown that the order of the menu pesentations might af-fect the results Half of the panelists were therefore pre-sented the maglev train sounds first the other half the con-ventional trains first A singular session consisted of thesame number of passages inside the menus This avoidsthat panelists would concentrate on counting events Fi-nally since one distance to the track would create a nat-ural setting large distances were never mixed with shortdistances in the menus of a session

During the experimental sessions perceived noise an-noyance of all transport noises was scaled with the methodof free-number magnitude estimation [22] The panelistswere asked to write down their magnitude assessments ondifferent coloured pieces of paper Before the start of theexperiment the panelists were instructed to select an ap-propriate number and then to double this number if theyfound the next stimulus to be twice as annoying to makethe number three times larger if they found the next stim-ulus to be three times as annoying etc and to scale 0 ifthey considered it not to be annoying at all For each 45-second sound (training sessions and conventional listeningtest) a conditional question was included ldquoTo what extentwould you be annoyed by this traffic sound if you heardit while relaxingrdquo For each 10-minute menu a very simi-lar but retrospective question was asked ldquoTo what extentwere you annoyed by traffic sound during the previous pe-riodrdquo In these latter questions we explicitly did not wantto refer to train noise since we wanted the panelists to de-cide themselves whether the sound period they last heardsounded like train-contaminated or not

25 Master scaling

In all experimental sessions the 7 road-traffic-noise-likereference sounds helped the panelists to define their ownscaling context The annoyance values given to these refer-ence sounds made it possible to control for the individualpanelistsrsquo choice-of-number behaviour in scaling the tar-get train sounds It would also control the influence of per-sonal factors such as noise sensitivity To get rid of theseeffects each individual panelistrsquos annoyance scale was cal-ibrated by the aid of the reference to the common masterscale [23]

A graphical illustration of the master scale transforma-tion applied to the annoyance reference data of one of thepanelists is given in Figure 6 The average annoyance re-ported for each of the 7 reference sound levels of roadtraffic noise is plotted in lin-log coordinates against theirsound levels LAeq45s measured at the faccedilade Individualpsychophysical functions are fitted to the reference data(open circles) They are of the form

Ar = a + b logSr (1)

where Ar is the reported annoyance during the train-ing session and logSr is the corresponding ldquoroad trafficnoiserdquo reference (r) sound level in dB(A) The constants a

593

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

Figure 5 Sound pressure level rerecorded in front of the faccedilade during one panelist groupsrsquo participation in the whole listening testtwo training sessions two menu sessions and one conventional psychoacoustical experiment with references

0

10

20

30

40

50

60

70

80

40 50 60 70 80 90

R = 55 master-scaled annoyance

A = 70 assessment

annoyance

road traffic reference sound level [dB(A)]

Figure 6 Calculation of master-scaled annoyance using one pan-elistrsquos empirical psychophysical function of the reference sounds(data points with solid line) and the master function for the samesounds (dashed line obtained as average function for all pan-elists)

and b will be different for each panelist and will depend ontheir choice-of-number behaviour in the particular scalingcontext The empirically derived master functions for thegroup of 100 panelists (dashed line in Figure 6) were thenused to transform the free number magnitude estimationsof the train or road traffic menus for each individual Aeto the corresponding annoyance values R in master scaleunits

R = minus629 + 145Ae minus a

b (2)

The slope of the master function was set to 145 whichis the average slope of all the individual psychophysicalfunctions whereas the intercept was set to produce a valueof ldquozerordquo for the most quiet train menu The reason for thelatter choice was that a majority of the panelists (84)reported their annoyance to be zero for this menu and amajority of the panelists reported annoyance to be greaterthan zero for all other menus

The choice of a logarithmic psychophysical function(equation 1) was a compromise In previous magnitudeestimation experiments of loudness [23 9] a power func-tion of the form logA = c + d logS was found to fit the

Table III Test-retest reliability of panelistsrsquo perceived noise an-noyance of the 7 reference road traffic sounds Each cell containsan arithmetic mean of Pearsonrsquos coefficient (r) and its standarderror Ts1 Training session 1 Ts2 Training session 2 Ct Con-ventional test

Ts1 Ts2 CtSet 1 Set 2 Set 1 Set 2 Set 1

Ts22 082plusmn0015

Ts21 086 087plusmn0014 plusmn0016

Ts22 086 088 087plusmn0017 plusmn0020 plusmn0019

Ct1 083 083 082 085plusmn0015 plusmn0021 plusmn0019 plusmn0020

Ct2 084 085 081 084 082plusmn0015 plusmn0019 plusmn0016 plusmn0019 plusmn0015

empirical data best However in this experiment noise an-noyance rather than loudness was scaled and thus obvi-ously also a value of zero (= not at all annoyed) had tobe handled although the noise was heard and its loudnesswas above zero The power function (after removal of ze-ros) did not fit the data better than the chosen logarithmicfunction

26 Data quality analysis

The master scaling made it possible to investigate the qual-ity of the experimental data in two ways as panelistsrsquo test-retest reliability and as their scaling ability The 7 refer-ence sounds were presented 6 times to each panelist twicein the two training sessions and twice in the last conven-tional listening test The set of 6 reference scale valueswere used to determine each panelistrsquos test-retest reliabil-ity of annoyance Table III shows the Pearsonrsquos coefficientof correlation for these 6 annoyance scales averaged overall panelists The test-retest reliability was very good be-tween 081 and 088 and the standard error was low be-tween 0014 and 0019

The deviation from the proposed master function (equa-tion 1) was used to assess the data quality and annoyance

594

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

Table IV Distribution of constants of the panelistsrsquo individualpsychophysical functions (Eq 1) The number of data sets refersto the average of 4 or 2 raw annoyance values which was takenfor each of the 7 reference sounds to calculate the psychophysicalfunctions

Data Psychophysical functionSets r2 a b

Training session 1amp2 4 0947 minus6727 1449plusmn0077 plusmn6128 plusmn1230

Conventional test 2 0881 minus4757 1105plusmn0118 plusmn4817 plusmn0948

scaling ability for each panelist and to trace errors and in-accuracies Table IV shows the distribution of constants ofthe panelistsrsquo individual psychophysical functions (equa-tion 1) The average annoyance variance explained bysound level (LAeq45s) of the reference road traffic soundswas found to range from 88 to 95 All panelists wereable to produce acceptable individual logarithmic func-tions of annoyance as a function of sound level to the refer-ence They have thus produced acceptable annoyance datain order to transform these to a common master scale ofannoyance no panelists were excluded from further dataanalysis

3 Results

The main listening experiment with menus differed fromprevious laboratory experiments in a number of aspectsOne important novelty is that participants were asked tojudge annoyance over a longer period of time mdash Fastl andGottschlingrsquos experiment [4] forms an exception Duringthe 10-minute periods the panelists were engaged in lowattention relaxing activities such as reading a magazinemaking a conversation or having something to drink In or-der to find out how this new approach affected the resultsa subsequent experiment was included which was morecomparable to earlier experiments on train noise (eg [8])

31 Main field experiment with 10-minute menus

The panelistsrsquo master scale values of annoyance were av-eraged for each menu in the field experiment A step-wise multiple linear regression analysis was performedwith average master scaled annoyance as dependent vari-able and (a) time averaged A-weighted faccedilade exposureLAeq10min (b) distance to the source (logarithmic) and(c) source type as independent variables Because of itslegislative importance in the Netherlands the faccedilade ex-posure was preferred to the actual panelistsrsquo noise expo-sure Faccedilade exposure was calculated from the sound lev-els measured on the recording sites since the faccedilade levelsmeasured during the experiment also contain noise fromwind and rain It has to be noted that the actual soundexposure levels experienced by panelists participating ina single experimental session may differ because of dif-ferent seating positions However personal characteristicssuch as noise sensitivity will have a much larger influence

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq10min

(a)

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq10min

(b)

Figure 7 Average master scaled annoyance of the menus ver-sus LAeq10min (a) for 2 events per 10-minute menu and (b) for4 events per 10-minute menu for different types of train sounds( ) IC train ( ) TGV and ( ) maglev train In comparisonthe annoyance for the highway traffic ( ) is also shown Stan-dard error on means is indicated as well as the master function(dashed line)

on perceived annoyance as compared to the influence of aslightly different exposure

Table V summarizes the results In the first modelsound level was the only independent variable this modelexplained 80 of the variance in annoyance In the secondmodel distance to track was added to sound level as an in-dependent variable this model increased the variance ex-plained to 85 (F-change = 1449 df1 = 1 df2 = 46 p lt0001) Thus distance to source explained a significant ad-ditional part of the annoyance variance not accounted forby sound level In the third model source type was in-cluded as a third independent variable along with soundlevel and distance Source type was defined on a nominalscale MAGLEV TGV IC and HIGHWAY It was intro-duced in the analysis as three dummy variables coded 0and 1 (the highway noise source type corresponds to thecase that these variables are all zero) The inclusion ofsource type did not increase significantly (F-change lt 10)the proportion of variance explained This suggests thatstatistically there is no additional contribution of sourcetype on perceived annoyance over and above the effects ofsound level and distance It can therefore be concluded thatmagnetic-levitation based transportation systems are notsignificantly more annoying than conventional rail basedsystems (same faccedilade LAeq and same distance are pre-requisites) Moreover railway noise was not found to be

595

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

Table V Stepwise multiple regression analysis of acoustic variables on perceived annoyance of train and highway traffic sounds for themain field experiment with 10-minute menus The Pearsonrsquos correlation coefficients of the variables entered in the regression analysisare shown at the bottom lowastp lt 005 lowastlowastp lt 001 lowastlowastlowastp lt 0001

Model Model Model fit increase F-change Independent Variables Coefficient t-valuefit (r2) (r2-change)

1 080 080 18748lowastlowastlowast LAeq10min [dB(A)] 118 1369lowastlowastlowast

2 085 005 1449lowastlowastlowast LAeq10min [dB(A)] 092 917lowastlowastlowast

log10(distance [m]) minus1074 minus381lowastlowastlowast

3 085 000 013 LAeq10min [dB(A)] 096 822lowastlowastlowast

log10(distance [m]) minus1017 minus333lowastlowastMAGLEV [01] 145 027TGV [01] 085 016IC [01] 227 040

Label Variable ANN LEQ DIST MAG TGV

ANN Annoyance [master scale]LEQ LAeq10min [dB(A)] 0894DIST log10(distance [m]) minus0754 minus0659MAG MAGLEV [01] minus0023 minus0038 0009TGV TGV [01] 0132 0224 0006 minus0682IC IC [01] minus0179 minus0290 0004 minus0433 minus0308

systematically less annoying than highway traffic noiseThis means that no support for a railway bonus was foundin this experiment at least it was not as obvious that itcould be observed using linear statistics Figure 7 givesan overview of the annoyance functions for the 10-minutemenus as a function of LAeq10min The dashed line indi-cates the master function of annoyance for the road-traffic-like sounds used as referencesThe shorter rise time of the noise of arriving high speed

trains may create more annoyance than a conventionaltrain can do Figure 8 Panel a shows the rise speeds indB(A)s in proportion to circle sizes These values werecalculated for all sound events included in this experi-ment by fitting a straight line through the initial increasein sound level The accelerating growth of annoyance withincreasing LAeq may be explained by the rise time In Fig-ure 8 Panel b the size of the circles is instead propor-tional to the distance to the track For LAeq in the intervalbetween 50 and 65 dB(A) annoyance is clearly lower fortrain passages at larger distances than for train passagesat closer distances or road traffic noise (dashed line) Thiscould indicate that a possible noise annoyance bonus fortrain noise would only hold at larger distances from thetrack and only in the latter LAeq interval

32 Conventional listening test

In the conventional listening experiment the sounds werepresented as short 45-second fragments containing thesound of one train passage and highway excerpts Figure 9shows the results of these master scaled annoyance val-ues as a function of time averaged A-weighted faccedilade ex-posure LAeq45s A railway penalty can be observed bothin regard to the artificial reference sounds as well as to

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq10min

(a)

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq10min

(b)

Figure 8 Average master scaled annoyance of the menus versusLAeq10min showing (a) the noise event rise speed and (b) the dis-tance to the track as the size of the circles The master functionis also indicated (dashed line)

the highway sounds Figure 10 shows the annoyance as afunction of rise speed ( Panel a) and distance to the track(Panel b)

596

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

Table VI Stepwise multiple regression analysis of acoustic variables on perceived annoyance of train sounds (no highway trafficsounds) for the conventional listening test (45-second passages) The Pearsonrsquos correlation coefficients of the variables entered in theregression analysis are shown at the bottom lowastp lt 005 lowastlowastp lt 001 lowastlowastlowastp lt 0001

Model Model Model fit increase F-change Independent Variables Coefficient t-valuefit (r2) (r2-change)

1 095 095 42017lowastlowastlowast LAeq45s [dB(A)] 167 2050lowastlowastlowast

2 098 003 1250lowastlowastlowast LAeq45s [dB(A)] 123 1104lowastlowastlowast

Speed [kmh] 002 203log10(distance [m]) minus178 minus077Rise speed [dB(A)s] 063 365lowastlowast

3 099 001 098 LAeq45s [dB(A)] 108 685lowastlowastlowast

Speed [kmh] 003 209log10(distance [m]) minus476 minus146Rise speed [dB(A)s] 058 330lowastlowast

MAGLEV [01] minus070 minus037TGV [01] 166 104

Label Variable ANN LEQ SPD DIST RISE MAG

ANN Annoyance [master scale]LEQ LAeq45s [dB(A)] 0975SPD Speed [kmh] 0646 0541DIST log10(distance [m]) minus0613 minus0667 minus0001RISE Rise speed [dB(A)s] 0885 0804 0708 minus0437MAG MAGLEV [01] 0070 minus0017 0552 0000 0188TGV TGV [01] 0207 0246 minus0190 0000 0100 minus0707

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq45sec

Figure 9 Average master scaled annoyance versus LAeq45s forthe conventional listening test for different types of train sounds( ) IC train ( ) TGV and ( ) maglev train In comparison theannoyance for the highway traffic ( ) is also shown Standard er-ror on means is indicated as well as the master function (dashedline)

A stepwise multiple linear regression analysis was alsoperformed separately for the train noises (Table VI) Thefirst model in which sound level LAeq45s was included asthe only independent variable explained 95 of the vari-ance in annoyance In the second model train speed dis-tance to the track and rise speed were added to sound levelas independent variables This increased the variance ex-plained to 98 (F-change = 1250 df1 = 3 df2 = 19p lt 0001) Apart from sound level also rise speed con-

tributed significantly to the variance explained The thirdmodel in which train type was added as an independentdummy variable did not significantly increase the propor-tion explained variance (F-change lt 10) These resultssuggest that in this conventional listening test there is nodifference in perceived annoyance between different typesof trains over and above the effect of sound level and risespeed

One has to note that the number of responses to eachstimulus was smaller in the main experiment (10-minutemenus) than in the conventional listenig test (45-secondpassages) This explains why the standard errors are lowerand the explained variance is higher in the latter experi-ment

4 Discussion

The annoyance results of the present field experiment areclose to residentsrsquo everyday reality although comparisonwith published studies is somewhat limited Previous lab-oratory experiments on noise annoyance of conventionalIC and high-speed trains specifically magnetic levitationtrains [8 5] report significant differences for these typesof sound In particular the results have shown that forthe same LAeq high-speed trains were more annoyingthan other trains Compared to road traffic noise the citedstudies claimed a lower annoyance level for conventionaltrains In the present field experiment we did not find sup-port for any annoyance difference between various types

597

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

0

10

20

30

40

50

60

70

80

40 50 60 70 80 90

annoyance

[maste

rscale

units]

L [dB(A)]Aeq45sec

(a)

0

10

20

30

40

50

60

70

80

40 50 60 70 80 90

annoyance

[maste

rscale

units]

L [dB(A)]Aeq45sec

(b)

Figure 10 Average master scaled annoyance versus LAeq45s

showing (a) the noise event rise speed and (b) the distance tothe track as the size of the circles The master function is alsoindicated (dashed line)

of trains and road traffic Some possible explanations willbe given in the following subsections

41 Realistic listening situation with 10-minutemenus

The experiment was performed in a realistic setting inwhich outdoor transportation noise was reproduced andnatural outdoor-to-indoor sound propagation characteris-tics were utilized (slightly open window) This setting pro-vided a realistic sound environment indoors Subgroups ofpanelists were kept indoors during the four-hour experi-ments and upon request annoyance to transport noise wasreported with reference to 10-minute periods

Because trains run on expected schedules to which peo-ple habituate the experimental situation in classical exper-iments is rather unrealistic The experimental one-passagesituation [8] requires full attention and will have a largevariation of train sounds compared with a particular rail-way track The outcome will to a large extent depend onthe experimental context that is the variation introducedin the experiment by selecting stimuli and using randompresentation orders Random orders of recordings can beselected and arranged so that annoyance judgments on cat-egory scales plotted against sound level differentiate wellor not well on type of transport In the present field experi-ment sub-context in sessions was kept invariant similar tothe situation on a real railway track The judgmental con-

text will then be much more restricted as is the case whenliving along one railway track

Next to this the annoyance reports of the one 45-secondtrain passage were higher than those of two passages ofthe same train within the 10-minute menus This is all inorder because the two types of annoyance were masterscaled in order to become comparable over experimentalsessions When judging 45-second train passages imme-diately after exposure it is quiet clear that the task is toassess the annoyance of that particular train passage (orother sounds that were presented) However when askedto assess the annoyance retrospectively of the transportnoise during the last 10-minute menu (eg two train pas-sages) the panelist will have to choose a strategy on howto go about this For example the annoyance may only bereferred to the two noise-stimulus periods or to the whole10-minute period (menu) It has been shown that the noiseannoyance of two overlapping (equal) noises would be ex-pected always to be less than the arithmetic sum of the twoannoyances (for a review see [24]) It is more uncertainhow total annoyance of two train passages separated intime will actually be acquired A laboratory experimentwhich included long sound fragments [4] has not foundthe above-mentioned annoyance difference between differ-ent train types which is in line with our results

42 Advanced scaling methodology

Long-term retrospective annoyance asked for in question-naire surveys has typically been assessed on categoryscales (eg [3]) A response category is then implicitlypostulated to be identical for every participant by verballabeling of the two end points or of every response boxalso the intervals between categories are assumed to be thesame However this assumption does not hold true [25]eg in questionnaire surveys the response criteria (scalevalue or category borders) for annoyance are much higherfor respondents in low noise areas as compared to those inhighly exposed areas The most well known scaling bias inlaboratory experiments is the context effect in which par-ticipants distribute their responses over the ldquofullrdquo range ofcategories independent of the size of the exposure range(for a review see [22]) In the process of using categoryscales floor and ceiling effects on annoyance may also ap-pearTo avoid uncontrolled context effects an invariant sound

level range of references was used as the annoyance con-text in the present field experiment Continuous road trafficnoise was chosen as a reference instead of multiple eventsounds because it is simpler to reproduce in future stud-ies To avoid the scaling bias of category scales the methodof magnitude estimation was chosen in which participantswere free to use the range of numbers they felt comfortablewith Master scaling was applied to these individual an-noyance estimates involving a transformation function toa common master scale defined by the references whichsound levels defined the scaling context In theory thismaster scale transformation will calibrate the loudness-dependence of noise annoyance whereas the relative con-

598

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

tribution to noise annoyance from qualitative content (egthe type of sound the time pattern and cues for speed anddistance) will hopefully be unchanged

Earlier research has shown that master scaling with ref-erences works well for loudness or annoyance of a one-occasion target exposure that is when repeated exposureis unfeasible (eg experiments with long duration expo-sures) or impossible (eg questionnaire surveys in fieldstudies) an example can be found in [23] The results ob-tained from the present field experiment are probably morereliable than the results that would have been obtained bycategory scaling The test-retest reliability of the panelistsrsquomagnitude estimates of annoyance of the reference soundswas found to be very good (above 08) compared to thereliability of 072 obtained in [8] for a group of 12 muchyounger subjects Considering that our panelists all werenaiumlve participants they also each produced high qualitypsychophysical functions for the reference as discussedin Section 26

43 Other possible explanations

There are several reasons why other investigators havefound a railway bonus (for a review see [3]) which wasnot found in this field experiment One of the reasons forfinding a railway bonus for short (one minute) noises inlistening experiments may be that the relation betweenloudness and LAeq is inherently different for train androad traffic noise Indeed some researchers have arguedthat noise annoyance evaluation in listening tests of shortsounds actually is close to a perceptual loudness evalua-tion (however see [26] on differences between loudness-based and quality-based perceived annoyance) If Zwickerloudness is a good first estimate of perceptual loudnessthe difference between train noise (of different types) andhighway noise would be seen in a Zwicker loudness ver-sus LAeq plot (Figure 11) Because the IC train noise usedin the present experiment was the noise of modern ratherquiet trains a few older and noisier IC train models wereadded in this acoustic analysis At levels above 65 dB(A)TGV and maglev trains seem to be a little louder thanhighway traffic or older IC trains However this effect onZwicker loudness is not significant and does therefore notsupport a railway bonus of 5 dB(A) stipulated in severalcountriesrsquo legislation Rather it seems to be a good actionto start to replace old IC trains by new ones The railwaybonus was originally based on studies with rather old low-speed trains and with much less dense traffic intensity thannowadaysThe intermittent character of railway noise could also be

an explanation for the railway bonus However this doesnot hold for aircraft noise which is also intermittent thiscan be explained by a difference in exposure In the case ofaircraft noise the exposure is on top of buildings and on allfaccedilades In the case of road traffic noise the probability ishigh that there are local roads also but there is a possibil-ity for a ldquoquiet siderdquo people are less annoyed if quiet sidesare available [27] In the case of railway noise there is alow probability for the presence of more than one track

Zw

ickerlo

udness

[sone]

L [dB(A)]Aeq45sec

0

10

20

30

40

50

40 50 60 70 80

Figure 11 Zwicker loudness versus LAeq45s for different types oftransportation sounds ( ) IC train ( ) TGV ( ) maglev train( ) highway traffic and ( ) some additional noisier IC trains(older type)

so the exposure will also be directed at only one faccediladeIn comparing road traffic and trains the faccedilade insulationwill be more effective in the case of train noise because ofthe smaller low-frequency proportion associated to trainnoise In comparing aircraft and trains which are both in-termittent the indoor exposure is certainly more intensivefor aircraft Considering these arguments it seems obviousthat aircraft is more annoying than road traffic which ismore annoying than train However faccedilade reduction wastaken into account in the present field experiment and stillthere was no clear railway bonus found Compared withthe field condition with closed windows and the faccediladefilter used in [8] a partially open window was used in thepresent field experiment which could explain this

In surveys questioning people at their home a lowerreported annoyance for train noise compared to highwaytraffic noise was observed in a particular range of noiselevels Most of the possible explanations proposed in liter-ature conflict with the fact that this railway bonus wouldbe observed in experiments based on single passages Wemention just a few The typical character of train noise andthe concentration of the sound energy in short time inter-vals may be advantageous with regard to activity distur-bance If the level is sufficiently low the probability ofnoticing the train noise is small compared to the prob-ability of noticing the sound of a continuous source Inaddition to physical differences in the sound the ldquogreenimagerdquo of trains as a means of transportation may add tothe acceptability of the source and thus increase the tol-erance to its noise that is as long as train passages arenot too frequent However a more recent hedonic pric-ing study found that householders in Birmingham placea greater value on reductions in railway noise than in roadtraffic noise [28] Cross-cultural studies (in field and lab-oratory context) have shown that a railway bonus is notuniversal [29 30] which would favor the argument aboveIt has further been shown that the bonus varies depend-ing on the (multiple) exposure situation [31] Based on theabove only part of the effect is supposed to be visible in

599

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

field experiments such as the one reported of in this paperPart of the effect is precisely what is observed

5 Conclusion

This study has shown that in an ldquoat homerdquo like contextnoise annoyance caused by different types of trains at thesame average outdoor faccedilade exposure level is not signifi-cantly different In particular magnetic levitation systemsare not more annoying than conventional high speed trainswhich is in agreement with earlier research Noise annoy-ance caused by conventional trains was not found to besignificantly lower than annoyance caused by TGVrsquos ormaglev trains at the same average faccedilade exposure Fieldsurveys have shown that for the same average sound levelrailway noise causes less annoyance or highly annoyedpersons than highway traffic noise Although our field ex-periment included several factors that may contribute tothis effect we could not observe it

More insight may be gained by taking into account thepsychoacoustic characteristics of the noise exposure andthe relevant personal factors of the panelists This paperhas focused on discussing the experimental methodologyin great detail and on presenting the results as a functionof the average outdoor faccedilade exposure level since thisis the main noise legislation indicator used in the Nether-lands Results of a detailed psychoacoustic analysis willbe reported in a future paper

Acknowledgments

This research was financed by the Project Group Zuider-zeelijn of the Ministry of Public Transport Public Worksand Water Management in the Netherlands The membersof the project steering committeemdashGilles Janssen (dBVi-sion) Annemarie Ruysbroek (RIVM) Martin van denBerg (VROM) and Pieter Jansse (Project Group Zuider-zeelijn) mdash are acknowledged for their valuable input Wealso appreciate the experimental assistance provided byMs Ingrid Decoster Finally we would like to thank thereviewers for their valuable comments and suggestions

References

[1] J M Fields J G Walker Comparing the relationships be-tween noise level and annoyance in different surveys a rail-way vs aircraft and road traffic comparison J Sound Vib81 (1982) 51ndash80

[2] U Moehler Community response to railway noise A re-view of social surveys J Sound Vib 120 (1988) 321ndash332

[3] H M E Miedema H Vos Exposure-response relation-ships for transportation noise J Acoust Soc Am 104(1998) 3432ndash3445

[4] H Fastl G Gottschling Subjective evaluation of noiseimmissions from transrapid Proceedings of the 25thInternational Congress on Noise Control Engineering(Intermiddotnoise) Liverpool UK Juli 1996 F A Hill R Law-rence (eds)

[5] D Neugebauer J Ortscheid Geraumluschbewertung des Tran-srapid Ist der Transrapid wie die Bahn zu beurteilenBerlin Umweltbundesamt Texte 25 1997

[6] D Neugebauer J Ortscheid Geraumluschbewertung des Tran-srapid Fortschritte der Akustik mdash DAGA 97 Kiel Ger-many 1997

[7] J Vos Annoyance caused by magnetic levitation trainTransrapid 08 mdash a laboratory study TNO report TM-03-C001 The Netherlands Organisation for Applied ScientificResearch (TNO) Soesterberg The Netherlands Jan 2003

[8] J Vos Annoyance caused by the sounds of a magnetic lev-itation train J Acoust Soc Am 115 (2004) 1597ndash1608

[9] B Berglund A Preis Is perceived annoyance more sub-ject-dependent than perceived loudness Acta Acusticaunited with Acustica 83 (1997) 313ndash319

[10] M Ishibashi K Ueno H Tachibana M Watanabe Sub-jective experiment on annoying sounds in living environ-ments Proceedings of the 18th International Congress onAcoustics (ICA) Kyoto Japan 2004

[11] J Maillard J Martin P Champelovier J Lambert Percep-tive evaluation of road traffic noise inside buildings using acombined image and wave field synthesis system Proceed-ings of CFADAGArsquo04 Strasbourg France Maumlrz 2004

[12] R F S Job Community response to noise A review offactors influencing the relationship between noise exposureand reaction J Acoust Soc Am 83 (1988) 991ndash1001

[13] H M E Miedema H Vos Demographic and attitudinalfactors that modify annoyance from transportation noise JAcoust Soc Am 105 (1999) 3336ndash3344

[14] W Ellermeier M Eigenstetter K Zimmer Psychoacousticcorrelates of individual noise sensitivity J Acoust SocAm 109 (2001) 1464ndash1473

[15] D Vaumlstfjaumlll M Kleiner T Gaumlrling Affective reactions tointerior aircraft sounds Acta Acustica united with Acustica89 (2003) 693ndash701

[16] T Feng C Xiao-hong L Tan-feng A social survey onthe effects of high-speed maglev noise on residents inShangai Proceedings of The 2005 Congress and Expo-sition on Noise Control Engineering (Intermiddotnoise) Rio deJaneiro Brazil Aug 2005

[17] J J Gibson The ecological approach to visual perceptionLawrence Erlbaum Associates Inc Hillsdale New JerseyUSA 1979

[18] J G Neuhoff Ecological psychoacoustics Introductionand history ndash In Ecological Psychoacoustics J G Neuhoff(ed) Elsevier Academic Press London UK 2004

[19] C Guastavino B F G Katz J-D Polack D J LevitinD Dubois Ecological validity of soundscape reproductionActa Acustica united with Acustica 91 (2005) 333ndash341

[20] A J Berkhout D de Vries P Vogel Acoustic control bywave field synthesis J Acoust Soc Am 93 (1993) 2764ndash2778

[21] E A M Franssen J E F van Dongen J M H Ruys-broek H Vos R K Stellato Hinder door milieufactorenen de beoordeling van de leefomgeving in Nederland In-ventarisatie verstoringen 2003 RIVM report 815120 001National Institute for Public Health and the Environment(RIVM) Bilthoven The Netherlands 2004

[22] L E Marks D Algom Psychophysical scaling ndash In Mea-surement Judgment and Decision Making M H Birn-baum (ed) Academic Press New York 1998

[23] B Berglund Quality assurance in environmental psycho-physics ndash In Ratio Scaling of Psychological Magnitudes -In Honor of the Memory of S S Stevens S J BolanowskiG A Gescheider (eds) Erlbaum Hillsdale N J 1991

600

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

Figure 5 Sound pressure level rerecorded in front of the faccedilade during one panelist groupsrsquo participation in the whole listening testtwo training sessions two menu sessions and one conventional psychoacoustical experiment with references

0

10

20

30

40

50

60

70

80

40 50 60 70 80 90

R = 55 master-scaled annoyance

A = 70 assessment

annoyance

road traffic reference sound level [dB(A)]

Figure 6 Calculation of master-scaled annoyance using one pan-elistrsquos empirical psychophysical function of the reference sounds(data points with solid line) and the master function for the samesounds (dashed line obtained as average function for all pan-elists)

and b will be different for each panelist and will depend ontheir choice-of-number behaviour in the particular scalingcontext The empirically derived master functions for thegroup of 100 panelists (dashed line in Figure 6) were thenused to transform the free number magnitude estimationsof the train or road traffic menus for each individual Aeto the corresponding annoyance values R in master scaleunits

R = minus629 + 145Ae minus a

b (2)

The slope of the master function was set to 145 whichis the average slope of all the individual psychophysicalfunctions whereas the intercept was set to produce a valueof ldquozerordquo for the most quiet train menu The reason for thelatter choice was that a majority of the panelists (84)reported their annoyance to be zero for this menu and amajority of the panelists reported annoyance to be greaterthan zero for all other menus

The choice of a logarithmic psychophysical function(equation 1) was a compromise In previous magnitudeestimation experiments of loudness [23 9] a power func-tion of the form logA = c + d logS was found to fit the

Table III Test-retest reliability of panelistsrsquo perceived noise an-noyance of the 7 reference road traffic sounds Each cell containsan arithmetic mean of Pearsonrsquos coefficient (r) and its standarderror Ts1 Training session 1 Ts2 Training session 2 Ct Con-ventional test

Ts1 Ts2 CtSet 1 Set 2 Set 1 Set 2 Set 1

Ts22 082plusmn0015

Ts21 086 087plusmn0014 plusmn0016

Ts22 086 088 087plusmn0017 plusmn0020 plusmn0019

Ct1 083 083 082 085plusmn0015 plusmn0021 plusmn0019 plusmn0020

Ct2 084 085 081 084 082plusmn0015 plusmn0019 plusmn0016 plusmn0019 plusmn0015

empirical data best However in this experiment noise an-noyance rather than loudness was scaled and thus obvi-ously also a value of zero (= not at all annoyed) had tobe handled although the noise was heard and its loudnesswas above zero The power function (after removal of ze-ros) did not fit the data better than the chosen logarithmicfunction

26 Data quality analysis

The master scaling made it possible to investigate the qual-ity of the experimental data in two ways as panelistsrsquo test-retest reliability and as their scaling ability The 7 refer-ence sounds were presented 6 times to each panelist twicein the two training sessions and twice in the last conven-tional listening test The set of 6 reference scale valueswere used to determine each panelistrsquos test-retest reliabil-ity of annoyance Table III shows the Pearsonrsquos coefficientof correlation for these 6 annoyance scales averaged overall panelists The test-retest reliability was very good be-tween 081 and 088 and the standard error was low be-tween 0014 and 0019

The deviation from the proposed master function (equa-tion 1) was used to assess the data quality and annoyance

594

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

Table IV Distribution of constants of the panelistsrsquo individualpsychophysical functions (Eq 1) The number of data sets refersto the average of 4 or 2 raw annoyance values which was takenfor each of the 7 reference sounds to calculate the psychophysicalfunctions

Data Psychophysical functionSets r2 a b

Training session 1amp2 4 0947 minus6727 1449plusmn0077 plusmn6128 plusmn1230

Conventional test 2 0881 minus4757 1105plusmn0118 plusmn4817 plusmn0948

scaling ability for each panelist and to trace errors and in-accuracies Table IV shows the distribution of constants ofthe panelistsrsquo individual psychophysical functions (equa-tion 1) The average annoyance variance explained bysound level (LAeq45s) of the reference road traffic soundswas found to range from 88 to 95 All panelists wereable to produce acceptable individual logarithmic func-tions of annoyance as a function of sound level to the refer-ence They have thus produced acceptable annoyance datain order to transform these to a common master scale ofannoyance no panelists were excluded from further dataanalysis

3 Results

The main listening experiment with menus differed fromprevious laboratory experiments in a number of aspectsOne important novelty is that participants were asked tojudge annoyance over a longer period of time mdash Fastl andGottschlingrsquos experiment [4] forms an exception Duringthe 10-minute periods the panelists were engaged in lowattention relaxing activities such as reading a magazinemaking a conversation or having something to drink In or-der to find out how this new approach affected the resultsa subsequent experiment was included which was morecomparable to earlier experiments on train noise (eg [8])

31 Main field experiment with 10-minute menus

The panelistsrsquo master scale values of annoyance were av-eraged for each menu in the field experiment A step-wise multiple linear regression analysis was performedwith average master scaled annoyance as dependent vari-able and (a) time averaged A-weighted faccedilade exposureLAeq10min (b) distance to the source (logarithmic) and(c) source type as independent variables Because of itslegislative importance in the Netherlands the faccedilade ex-posure was preferred to the actual panelistsrsquo noise expo-sure Faccedilade exposure was calculated from the sound lev-els measured on the recording sites since the faccedilade levelsmeasured during the experiment also contain noise fromwind and rain It has to be noted that the actual soundexposure levels experienced by panelists participating ina single experimental session may differ because of dif-ferent seating positions However personal characteristicssuch as noise sensitivity will have a much larger influence

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq10min

(a)

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq10min

(b)

Figure 7 Average master scaled annoyance of the menus ver-sus LAeq10min (a) for 2 events per 10-minute menu and (b) for4 events per 10-minute menu for different types of train sounds( ) IC train ( ) TGV and ( ) maglev train In comparisonthe annoyance for the highway traffic ( ) is also shown Stan-dard error on means is indicated as well as the master function(dashed line)

on perceived annoyance as compared to the influence of aslightly different exposure

Table V summarizes the results In the first modelsound level was the only independent variable this modelexplained 80 of the variance in annoyance In the secondmodel distance to track was added to sound level as an in-dependent variable this model increased the variance ex-plained to 85 (F-change = 1449 df1 = 1 df2 = 46 p lt0001) Thus distance to source explained a significant ad-ditional part of the annoyance variance not accounted forby sound level In the third model source type was in-cluded as a third independent variable along with soundlevel and distance Source type was defined on a nominalscale MAGLEV TGV IC and HIGHWAY It was intro-duced in the analysis as three dummy variables coded 0and 1 (the highway noise source type corresponds to thecase that these variables are all zero) The inclusion ofsource type did not increase significantly (F-change lt 10)the proportion of variance explained This suggests thatstatistically there is no additional contribution of sourcetype on perceived annoyance over and above the effects ofsound level and distance It can therefore be concluded thatmagnetic-levitation based transportation systems are notsignificantly more annoying than conventional rail basedsystems (same faccedilade LAeq and same distance are pre-requisites) Moreover railway noise was not found to be

595

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

Table V Stepwise multiple regression analysis of acoustic variables on perceived annoyance of train and highway traffic sounds for themain field experiment with 10-minute menus The Pearsonrsquos correlation coefficients of the variables entered in the regression analysisare shown at the bottom lowastp lt 005 lowastlowastp lt 001 lowastlowastlowastp lt 0001

Model Model Model fit increase F-change Independent Variables Coefficient t-valuefit (r2) (r2-change)

1 080 080 18748lowastlowastlowast LAeq10min [dB(A)] 118 1369lowastlowastlowast

2 085 005 1449lowastlowastlowast LAeq10min [dB(A)] 092 917lowastlowastlowast

log10(distance [m]) minus1074 minus381lowastlowastlowast

3 085 000 013 LAeq10min [dB(A)] 096 822lowastlowastlowast

log10(distance [m]) minus1017 minus333lowastlowastMAGLEV [01] 145 027TGV [01] 085 016IC [01] 227 040

Label Variable ANN LEQ DIST MAG TGV

ANN Annoyance [master scale]LEQ LAeq10min [dB(A)] 0894DIST log10(distance [m]) minus0754 minus0659MAG MAGLEV [01] minus0023 minus0038 0009TGV TGV [01] 0132 0224 0006 minus0682IC IC [01] minus0179 minus0290 0004 minus0433 minus0308

systematically less annoying than highway traffic noiseThis means that no support for a railway bonus was foundin this experiment at least it was not as obvious that itcould be observed using linear statistics Figure 7 givesan overview of the annoyance functions for the 10-minutemenus as a function of LAeq10min The dashed line indi-cates the master function of annoyance for the road-traffic-like sounds used as referencesThe shorter rise time of the noise of arriving high speed

trains may create more annoyance than a conventionaltrain can do Figure 8 Panel a shows the rise speeds indB(A)s in proportion to circle sizes These values werecalculated for all sound events included in this experi-ment by fitting a straight line through the initial increasein sound level The accelerating growth of annoyance withincreasing LAeq may be explained by the rise time In Fig-ure 8 Panel b the size of the circles is instead propor-tional to the distance to the track For LAeq in the intervalbetween 50 and 65 dB(A) annoyance is clearly lower fortrain passages at larger distances than for train passagesat closer distances or road traffic noise (dashed line) Thiscould indicate that a possible noise annoyance bonus fortrain noise would only hold at larger distances from thetrack and only in the latter LAeq interval

32 Conventional listening test

In the conventional listening experiment the sounds werepresented as short 45-second fragments containing thesound of one train passage and highway excerpts Figure 9shows the results of these master scaled annoyance val-ues as a function of time averaged A-weighted faccedilade ex-posure LAeq45s A railway penalty can be observed bothin regard to the artificial reference sounds as well as to

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq10min

(a)

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq10min

(b)

Figure 8 Average master scaled annoyance of the menus versusLAeq10min showing (a) the noise event rise speed and (b) the dis-tance to the track as the size of the circles The master functionis also indicated (dashed line)

the highway sounds Figure 10 shows the annoyance as afunction of rise speed ( Panel a) and distance to the track(Panel b)

596

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

Table VI Stepwise multiple regression analysis of acoustic variables on perceived annoyance of train sounds (no highway trafficsounds) for the conventional listening test (45-second passages) The Pearsonrsquos correlation coefficients of the variables entered in theregression analysis are shown at the bottom lowastp lt 005 lowastlowastp lt 001 lowastlowastlowastp lt 0001

Model Model Model fit increase F-change Independent Variables Coefficient t-valuefit (r2) (r2-change)

1 095 095 42017lowastlowastlowast LAeq45s [dB(A)] 167 2050lowastlowastlowast

2 098 003 1250lowastlowastlowast LAeq45s [dB(A)] 123 1104lowastlowastlowast

Speed [kmh] 002 203log10(distance [m]) minus178 minus077Rise speed [dB(A)s] 063 365lowastlowast

3 099 001 098 LAeq45s [dB(A)] 108 685lowastlowastlowast

Speed [kmh] 003 209log10(distance [m]) minus476 minus146Rise speed [dB(A)s] 058 330lowastlowast

MAGLEV [01] minus070 minus037TGV [01] 166 104

Label Variable ANN LEQ SPD DIST RISE MAG

ANN Annoyance [master scale]LEQ LAeq45s [dB(A)] 0975SPD Speed [kmh] 0646 0541DIST log10(distance [m]) minus0613 minus0667 minus0001RISE Rise speed [dB(A)s] 0885 0804 0708 minus0437MAG MAGLEV [01] 0070 minus0017 0552 0000 0188TGV TGV [01] 0207 0246 minus0190 0000 0100 minus0707

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq45sec

Figure 9 Average master scaled annoyance versus LAeq45s forthe conventional listening test for different types of train sounds( ) IC train ( ) TGV and ( ) maglev train In comparison theannoyance for the highway traffic ( ) is also shown Standard er-ror on means is indicated as well as the master function (dashedline)

A stepwise multiple linear regression analysis was alsoperformed separately for the train noises (Table VI) Thefirst model in which sound level LAeq45s was included asthe only independent variable explained 95 of the vari-ance in annoyance In the second model train speed dis-tance to the track and rise speed were added to sound levelas independent variables This increased the variance ex-plained to 98 (F-change = 1250 df1 = 3 df2 = 19p lt 0001) Apart from sound level also rise speed con-

tributed significantly to the variance explained The thirdmodel in which train type was added as an independentdummy variable did not significantly increase the propor-tion explained variance (F-change lt 10) These resultssuggest that in this conventional listening test there is nodifference in perceived annoyance between different typesof trains over and above the effect of sound level and risespeed

One has to note that the number of responses to eachstimulus was smaller in the main experiment (10-minutemenus) than in the conventional listenig test (45-secondpassages) This explains why the standard errors are lowerand the explained variance is higher in the latter experi-ment

4 Discussion

The annoyance results of the present field experiment areclose to residentsrsquo everyday reality although comparisonwith published studies is somewhat limited Previous lab-oratory experiments on noise annoyance of conventionalIC and high-speed trains specifically magnetic levitationtrains [8 5] report significant differences for these typesof sound In particular the results have shown that forthe same LAeq high-speed trains were more annoyingthan other trains Compared to road traffic noise the citedstudies claimed a lower annoyance level for conventionaltrains In the present field experiment we did not find sup-port for any annoyance difference between various types

597

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

0

10

20

30

40

50

60

70

80

40 50 60 70 80 90

annoyance

[maste

rscale

units]

L [dB(A)]Aeq45sec

(a)

0

10

20

30

40

50

60

70

80

40 50 60 70 80 90

annoyance

[maste

rscale

units]

L [dB(A)]Aeq45sec

(b)

Figure 10 Average master scaled annoyance versus LAeq45s

showing (a) the noise event rise speed and (b) the distance tothe track as the size of the circles The master function is alsoindicated (dashed line)

of trains and road traffic Some possible explanations willbe given in the following subsections

41 Realistic listening situation with 10-minutemenus

The experiment was performed in a realistic setting inwhich outdoor transportation noise was reproduced andnatural outdoor-to-indoor sound propagation characteris-tics were utilized (slightly open window) This setting pro-vided a realistic sound environment indoors Subgroups ofpanelists were kept indoors during the four-hour experi-ments and upon request annoyance to transport noise wasreported with reference to 10-minute periods

Because trains run on expected schedules to which peo-ple habituate the experimental situation in classical exper-iments is rather unrealistic The experimental one-passagesituation [8] requires full attention and will have a largevariation of train sounds compared with a particular rail-way track The outcome will to a large extent depend onthe experimental context that is the variation introducedin the experiment by selecting stimuli and using randompresentation orders Random orders of recordings can beselected and arranged so that annoyance judgments on cat-egory scales plotted against sound level differentiate wellor not well on type of transport In the present field experi-ment sub-context in sessions was kept invariant similar tothe situation on a real railway track The judgmental con-

text will then be much more restricted as is the case whenliving along one railway track

Next to this the annoyance reports of the one 45-secondtrain passage were higher than those of two passages ofthe same train within the 10-minute menus This is all inorder because the two types of annoyance were masterscaled in order to become comparable over experimentalsessions When judging 45-second train passages imme-diately after exposure it is quiet clear that the task is toassess the annoyance of that particular train passage (orother sounds that were presented) However when askedto assess the annoyance retrospectively of the transportnoise during the last 10-minute menu (eg two train pas-sages) the panelist will have to choose a strategy on howto go about this For example the annoyance may only bereferred to the two noise-stimulus periods or to the whole10-minute period (menu) It has been shown that the noiseannoyance of two overlapping (equal) noises would be ex-pected always to be less than the arithmetic sum of the twoannoyances (for a review see [24]) It is more uncertainhow total annoyance of two train passages separated intime will actually be acquired A laboratory experimentwhich included long sound fragments [4] has not foundthe above-mentioned annoyance difference between differ-ent train types which is in line with our results

42 Advanced scaling methodology

Long-term retrospective annoyance asked for in question-naire surveys has typically been assessed on categoryscales (eg [3]) A response category is then implicitlypostulated to be identical for every participant by verballabeling of the two end points or of every response boxalso the intervals between categories are assumed to be thesame However this assumption does not hold true [25]eg in questionnaire surveys the response criteria (scalevalue or category borders) for annoyance are much higherfor respondents in low noise areas as compared to those inhighly exposed areas The most well known scaling bias inlaboratory experiments is the context effect in which par-ticipants distribute their responses over the ldquofullrdquo range ofcategories independent of the size of the exposure range(for a review see [22]) In the process of using categoryscales floor and ceiling effects on annoyance may also ap-pearTo avoid uncontrolled context effects an invariant sound

level range of references was used as the annoyance con-text in the present field experiment Continuous road trafficnoise was chosen as a reference instead of multiple eventsounds because it is simpler to reproduce in future stud-ies To avoid the scaling bias of category scales the methodof magnitude estimation was chosen in which participantswere free to use the range of numbers they felt comfortablewith Master scaling was applied to these individual an-noyance estimates involving a transformation function toa common master scale defined by the references whichsound levels defined the scaling context In theory thismaster scale transformation will calibrate the loudness-dependence of noise annoyance whereas the relative con-

598

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

tribution to noise annoyance from qualitative content (egthe type of sound the time pattern and cues for speed anddistance) will hopefully be unchanged

Earlier research has shown that master scaling with ref-erences works well for loudness or annoyance of a one-occasion target exposure that is when repeated exposureis unfeasible (eg experiments with long duration expo-sures) or impossible (eg questionnaire surveys in fieldstudies) an example can be found in [23] The results ob-tained from the present field experiment are probably morereliable than the results that would have been obtained bycategory scaling The test-retest reliability of the panelistsrsquomagnitude estimates of annoyance of the reference soundswas found to be very good (above 08) compared to thereliability of 072 obtained in [8] for a group of 12 muchyounger subjects Considering that our panelists all werenaiumlve participants they also each produced high qualitypsychophysical functions for the reference as discussedin Section 26

43 Other possible explanations

There are several reasons why other investigators havefound a railway bonus (for a review see [3]) which wasnot found in this field experiment One of the reasons forfinding a railway bonus for short (one minute) noises inlistening experiments may be that the relation betweenloudness and LAeq is inherently different for train androad traffic noise Indeed some researchers have arguedthat noise annoyance evaluation in listening tests of shortsounds actually is close to a perceptual loudness evalua-tion (however see [26] on differences between loudness-based and quality-based perceived annoyance) If Zwickerloudness is a good first estimate of perceptual loudnessthe difference between train noise (of different types) andhighway noise would be seen in a Zwicker loudness ver-sus LAeq plot (Figure 11) Because the IC train noise usedin the present experiment was the noise of modern ratherquiet trains a few older and noisier IC train models wereadded in this acoustic analysis At levels above 65 dB(A)TGV and maglev trains seem to be a little louder thanhighway traffic or older IC trains However this effect onZwicker loudness is not significant and does therefore notsupport a railway bonus of 5 dB(A) stipulated in severalcountriesrsquo legislation Rather it seems to be a good actionto start to replace old IC trains by new ones The railwaybonus was originally based on studies with rather old low-speed trains and with much less dense traffic intensity thannowadaysThe intermittent character of railway noise could also be

an explanation for the railway bonus However this doesnot hold for aircraft noise which is also intermittent thiscan be explained by a difference in exposure In the case ofaircraft noise the exposure is on top of buildings and on allfaccedilades In the case of road traffic noise the probability ishigh that there are local roads also but there is a possibil-ity for a ldquoquiet siderdquo people are less annoyed if quiet sidesare available [27] In the case of railway noise there is alow probability for the presence of more than one track

Zw

ickerlo

udness

[sone]

L [dB(A)]Aeq45sec

0

10

20

30

40

50

40 50 60 70 80

Figure 11 Zwicker loudness versus LAeq45s for different types oftransportation sounds ( ) IC train ( ) TGV ( ) maglev train( ) highway traffic and ( ) some additional noisier IC trains(older type)

so the exposure will also be directed at only one faccediladeIn comparing road traffic and trains the faccedilade insulationwill be more effective in the case of train noise because ofthe smaller low-frequency proportion associated to trainnoise In comparing aircraft and trains which are both in-termittent the indoor exposure is certainly more intensivefor aircraft Considering these arguments it seems obviousthat aircraft is more annoying than road traffic which ismore annoying than train However faccedilade reduction wastaken into account in the present field experiment and stillthere was no clear railway bonus found Compared withthe field condition with closed windows and the faccediladefilter used in [8] a partially open window was used in thepresent field experiment which could explain this

In surveys questioning people at their home a lowerreported annoyance for train noise compared to highwaytraffic noise was observed in a particular range of noiselevels Most of the possible explanations proposed in liter-ature conflict with the fact that this railway bonus wouldbe observed in experiments based on single passages Wemention just a few The typical character of train noise andthe concentration of the sound energy in short time inter-vals may be advantageous with regard to activity distur-bance If the level is sufficiently low the probability ofnoticing the train noise is small compared to the prob-ability of noticing the sound of a continuous source Inaddition to physical differences in the sound the ldquogreenimagerdquo of trains as a means of transportation may add tothe acceptability of the source and thus increase the tol-erance to its noise that is as long as train passages arenot too frequent However a more recent hedonic pric-ing study found that householders in Birmingham placea greater value on reductions in railway noise than in roadtraffic noise [28] Cross-cultural studies (in field and lab-oratory context) have shown that a railway bonus is notuniversal [29 30] which would favor the argument aboveIt has further been shown that the bonus varies depend-ing on the (multiple) exposure situation [31] Based on theabove only part of the effect is supposed to be visible in

599

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

field experiments such as the one reported of in this paperPart of the effect is precisely what is observed

5 Conclusion

This study has shown that in an ldquoat homerdquo like contextnoise annoyance caused by different types of trains at thesame average outdoor faccedilade exposure level is not signifi-cantly different In particular magnetic levitation systemsare not more annoying than conventional high speed trainswhich is in agreement with earlier research Noise annoy-ance caused by conventional trains was not found to besignificantly lower than annoyance caused by TGVrsquos ormaglev trains at the same average faccedilade exposure Fieldsurveys have shown that for the same average sound levelrailway noise causes less annoyance or highly annoyedpersons than highway traffic noise Although our field ex-periment included several factors that may contribute tothis effect we could not observe it

More insight may be gained by taking into account thepsychoacoustic characteristics of the noise exposure andthe relevant personal factors of the panelists This paperhas focused on discussing the experimental methodologyin great detail and on presenting the results as a functionof the average outdoor faccedilade exposure level since thisis the main noise legislation indicator used in the Nether-lands Results of a detailed psychoacoustic analysis willbe reported in a future paper

Acknowledgments

This research was financed by the Project Group Zuider-zeelijn of the Ministry of Public Transport Public Worksand Water Management in the Netherlands The membersof the project steering committeemdashGilles Janssen (dBVi-sion) Annemarie Ruysbroek (RIVM) Martin van denBerg (VROM) and Pieter Jansse (Project Group Zuider-zeelijn) mdash are acknowledged for their valuable input Wealso appreciate the experimental assistance provided byMs Ingrid Decoster Finally we would like to thank thereviewers for their valuable comments and suggestions

References

[1] J M Fields J G Walker Comparing the relationships be-tween noise level and annoyance in different surveys a rail-way vs aircraft and road traffic comparison J Sound Vib81 (1982) 51ndash80

[2] U Moehler Community response to railway noise A re-view of social surveys J Sound Vib 120 (1988) 321ndash332

[3] H M E Miedema H Vos Exposure-response relation-ships for transportation noise J Acoust Soc Am 104(1998) 3432ndash3445

[4] H Fastl G Gottschling Subjective evaluation of noiseimmissions from transrapid Proceedings of the 25thInternational Congress on Noise Control Engineering(Intermiddotnoise) Liverpool UK Juli 1996 F A Hill R Law-rence (eds)

[5] D Neugebauer J Ortscheid Geraumluschbewertung des Tran-srapid Ist der Transrapid wie die Bahn zu beurteilenBerlin Umweltbundesamt Texte 25 1997

[6] D Neugebauer J Ortscheid Geraumluschbewertung des Tran-srapid Fortschritte der Akustik mdash DAGA 97 Kiel Ger-many 1997

[7] J Vos Annoyance caused by magnetic levitation trainTransrapid 08 mdash a laboratory study TNO report TM-03-C001 The Netherlands Organisation for Applied ScientificResearch (TNO) Soesterberg The Netherlands Jan 2003

[8] J Vos Annoyance caused by the sounds of a magnetic lev-itation train J Acoust Soc Am 115 (2004) 1597ndash1608

[9] B Berglund A Preis Is perceived annoyance more sub-ject-dependent than perceived loudness Acta Acusticaunited with Acustica 83 (1997) 313ndash319

[10] M Ishibashi K Ueno H Tachibana M Watanabe Sub-jective experiment on annoying sounds in living environ-ments Proceedings of the 18th International Congress onAcoustics (ICA) Kyoto Japan 2004

[11] J Maillard J Martin P Champelovier J Lambert Percep-tive evaluation of road traffic noise inside buildings using acombined image and wave field synthesis system Proceed-ings of CFADAGArsquo04 Strasbourg France Maumlrz 2004

[12] R F S Job Community response to noise A review offactors influencing the relationship between noise exposureand reaction J Acoust Soc Am 83 (1988) 991ndash1001

[13] H M E Miedema H Vos Demographic and attitudinalfactors that modify annoyance from transportation noise JAcoust Soc Am 105 (1999) 3336ndash3344

[14] W Ellermeier M Eigenstetter K Zimmer Psychoacousticcorrelates of individual noise sensitivity J Acoust SocAm 109 (2001) 1464ndash1473

[15] D Vaumlstfjaumlll M Kleiner T Gaumlrling Affective reactions tointerior aircraft sounds Acta Acustica united with Acustica89 (2003) 693ndash701

[16] T Feng C Xiao-hong L Tan-feng A social survey onthe effects of high-speed maglev noise on residents inShangai Proceedings of The 2005 Congress and Expo-sition on Noise Control Engineering (Intermiddotnoise) Rio deJaneiro Brazil Aug 2005

[17] J J Gibson The ecological approach to visual perceptionLawrence Erlbaum Associates Inc Hillsdale New JerseyUSA 1979

[18] J G Neuhoff Ecological psychoacoustics Introductionand history ndash In Ecological Psychoacoustics J G Neuhoff(ed) Elsevier Academic Press London UK 2004

[19] C Guastavino B F G Katz J-D Polack D J LevitinD Dubois Ecological validity of soundscape reproductionActa Acustica united with Acustica 91 (2005) 333ndash341

[20] A J Berkhout D de Vries P Vogel Acoustic control bywave field synthesis J Acoust Soc Am 93 (1993) 2764ndash2778

[21] E A M Franssen J E F van Dongen J M H Ruys-broek H Vos R K Stellato Hinder door milieufactorenen de beoordeling van de leefomgeving in Nederland In-ventarisatie verstoringen 2003 RIVM report 815120 001National Institute for Public Health and the Environment(RIVM) Bilthoven The Netherlands 2004

[22] L E Marks D Algom Psychophysical scaling ndash In Mea-surement Judgment and Decision Making M H Birn-baum (ed) Academic Press New York 1998

[23] B Berglund Quality assurance in environmental psycho-physics ndash In Ratio Scaling of Psychological Magnitudes -In Honor of the Memory of S S Stevens S J BolanowskiG A Gescheider (eds) Erlbaum Hillsdale N J 1991

600

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

Table IV Distribution of constants of the panelistsrsquo individualpsychophysical functions (Eq 1) The number of data sets refersto the average of 4 or 2 raw annoyance values which was takenfor each of the 7 reference sounds to calculate the psychophysicalfunctions

Data Psychophysical functionSets r2 a b

Training session 1amp2 4 0947 minus6727 1449plusmn0077 plusmn6128 plusmn1230

Conventional test 2 0881 minus4757 1105plusmn0118 plusmn4817 plusmn0948

scaling ability for each panelist and to trace errors and in-accuracies Table IV shows the distribution of constants ofthe panelistsrsquo individual psychophysical functions (equa-tion 1) The average annoyance variance explained bysound level (LAeq45s) of the reference road traffic soundswas found to range from 88 to 95 All panelists wereable to produce acceptable individual logarithmic func-tions of annoyance as a function of sound level to the refer-ence They have thus produced acceptable annoyance datain order to transform these to a common master scale ofannoyance no panelists were excluded from further dataanalysis

3 Results

The main listening experiment with menus differed fromprevious laboratory experiments in a number of aspectsOne important novelty is that participants were asked tojudge annoyance over a longer period of time mdash Fastl andGottschlingrsquos experiment [4] forms an exception Duringthe 10-minute periods the panelists were engaged in lowattention relaxing activities such as reading a magazinemaking a conversation or having something to drink In or-der to find out how this new approach affected the resultsa subsequent experiment was included which was morecomparable to earlier experiments on train noise (eg [8])

31 Main field experiment with 10-minute menus

The panelistsrsquo master scale values of annoyance were av-eraged for each menu in the field experiment A step-wise multiple linear regression analysis was performedwith average master scaled annoyance as dependent vari-able and (a) time averaged A-weighted faccedilade exposureLAeq10min (b) distance to the source (logarithmic) and(c) source type as independent variables Because of itslegislative importance in the Netherlands the faccedilade ex-posure was preferred to the actual panelistsrsquo noise expo-sure Faccedilade exposure was calculated from the sound lev-els measured on the recording sites since the faccedilade levelsmeasured during the experiment also contain noise fromwind and rain It has to be noted that the actual soundexposure levels experienced by panelists participating ina single experimental session may differ because of dif-ferent seating positions However personal characteristicssuch as noise sensitivity will have a much larger influence

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq10min

(a)

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq10min

(b)

Figure 7 Average master scaled annoyance of the menus ver-sus LAeq10min (a) for 2 events per 10-minute menu and (b) for4 events per 10-minute menu for different types of train sounds( ) IC train ( ) TGV and ( ) maglev train In comparisonthe annoyance for the highway traffic ( ) is also shown Stan-dard error on means is indicated as well as the master function(dashed line)

on perceived annoyance as compared to the influence of aslightly different exposure

Table V summarizes the results In the first modelsound level was the only independent variable this modelexplained 80 of the variance in annoyance In the secondmodel distance to track was added to sound level as an in-dependent variable this model increased the variance ex-plained to 85 (F-change = 1449 df1 = 1 df2 = 46 p lt0001) Thus distance to source explained a significant ad-ditional part of the annoyance variance not accounted forby sound level In the third model source type was in-cluded as a third independent variable along with soundlevel and distance Source type was defined on a nominalscale MAGLEV TGV IC and HIGHWAY It was intro-duced in the analysis as three dummy variables coded 0and 1 (the highway noise source type corresponds to thecase that these variables are all zero) The inclusion ofsource type did not increase significantly (F-change lt 10)the proportion of variance explained This suggests thatstatistically there is no additional contribution of sourcetype on perceived annoyance over and above the effects ofsound level and distance It can therefore be concluded thatmagnetic-levitation based transportation systems are notsignificantly more annoying than conventional rail basedsystems (same faccedilade LAeq and same distance are pre-requisites) Moreover railway noise was not found to be

595

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

Table V Stepwise multiple regression analysis of acoustic variables on perceived annoyance of train and highway traffic sounds for themain field experiment with 10-minute menus The Pearsonrsquos correlation coefficients of the variables entered in the regression analysisare shown at the bottom lowastp lt 005 lowastlowastp lt 001 lowastlowastlowastp lt 0001

Model Model Model fit increase F-change Independent Variables Coefficient t-valuefit (r2) (r2-change)

1 080 080 18748lowastlowastlowast LAeq10min [dB(A)] 118 1369lowastlowastlowast

2 085 005 1449lowastlowastlowast LAeq10min [dB(A)] 092 917lowastlowastlowast

log10(distance [m]) minus1074 minus381lowastlowastlowast

3 085 000 013 LAeq10min [dB(A)] 096 822lowastlowastlowast

log10(distance [m]) minus1017 minus333lowastlowastMAGLEV [01] 145 027TGV [01] 085 016IC [01] 227 040

Label Variable ANN LEQ DIST MAG TGV

ANN Annoyance [master scale]LEQ LAeq10min [dB(A)] 0894DIST log10(distance [m]) minus0754 minus0659MAG MAGLEV [01] minus0023 minus0038 0009TGV TGV [01] 0132 0224 0006 minus0682IC IC [01] minus0179 minus0290 0004 minus0433 minus0308

systematically less annoying than highway traffic noiseThis means that no support for a railway bonus was foundin this experiment at least it was not as obvious that itcould be observed using linear statistics Figure 7 givesan overview of the annoyance functions for the 10-minutemenus as a function of LAeq10min The dashed line indi-cates the master function of annoyance for the road-traffic-like sounds used as referencesThe shorter rise time of the noise of arriving high speed

trains may create more annoyance than a conventionaltrain can do Figure 8 Panel a shows the rise speeds indB(A)s in proportion to circle sizes These values werecalculated for all sound events included in this experi-ment by fitting a straight line through the initial increasein sound level The accelerating growth of annoyance withincreasing LAeq may be explained by the rise time In Fig-ure 8 Panel b the size of the circles is instead propor-tional to the distance to the track For LAeq in the intervalbetween 50 and 65 dB(A) annoyance is clearly lower fortrain passages at larger distances than for train passagesat closer distances or road traffic noise (dashed line) Thiscould indicate that a possible noise annoyance bonus fortrain noise would only hold at larger distances from thetrack and only in the latter LAeq interval

32 Conventional listening test

In the conventional listening experiment the sounds werepresented as short 45-second fragments containing thesound of one train passage and highway excerpts Figure 9shows the results of these master scaled annoyance val-ues as a function of time averaged A-weighted faccedilade ex-posure LAeq45s A railway penalty can be observed bothin regard to the artificial reference sounds as well as to

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq10min

(a)

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq10min

(b)

Figure 8 Average master scaled annoyance of the menus versusLAeq10min showing (a) the noise event rise speed and (b) the dis-tance to the track as the size of the circles The master functionis also indicated (dashed line)

the highway sounds Figure 10 shows the annoyance as afunction of rise speed ( Panel a) and distance to the track(Panel b)

596

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

Table VI Stepwise multiple regression analysis of acoustic variables on perceived annoyance of train sounds (no highway trafficsounds) for the conventional listening test (45-second passages) The Pearsonrsquos correlation coefficients of the variables entered in theregression analysis are shown at the bottom lowastp lt 005 lowastlowastp lt 001 lowastlowastlowastp lt 0001

Model Model Model fit increase F-change Independent Variables Coefficient t-valuefit (r2) (r2-change)

1 095 095 42017lowastlowastlowast LAeq45s [dB(A)] 167 2050lowastlowastlowast

2 098 003 1250lowastlowastlowast LAeq45s [dB(A)] 123 1104lowastlowastlowast

Speed [kmh] 002 203log10(distance [m]) minus178 minus077Rise speed [dB(A)s] 063 365lowastlowast

3 099 001 098 LAeq45s [dB(A)] 108 685lowastlowastlowast

Speed [kmh] 003 209log10(distance [m]) minus476 minus146Rise speed [dB(A)s] 058 330lowastlowast

MAGLEV [01] minus070 minus037TGV [01] 166 104

Label Variable ANN LEQ SPD DIST RISE MAG

ANN Annoyance [master scale]LEQ LAeq45s [dB(A)] 0975SPD Speed [kmh] 0646 0541DIST log10(distance [m]) minus0613 minus0667 minus0001RISE Rise speed [dB(A)s] 0885 0804 0708 minus0437MAG MAGLEV [01] 0070 minus0017 0552 0000 0188TGV TGV [01] 0207 0246 minus0190 0000 0100 minus0707

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq45sec

Figure 9 Average master scaled annoyance versus LAeq45s forthe conventional listening test for different types of train sounds( ) IC train ( ) TGV and ( ) maglev train In comparison theannoyance for the highway traffic ( ) is also shown Standard er-ror on means is indicated as well as the master function (dashedline)

A stepwise multiple linear regression analysis was alsoperformed separately for the train noises (Table VI) Thefirst model in which sound level LAeq45s was included asthe only independent variable explained 95 of the vari-ance in annoyance In the second model train speed dis-tance to the track and rise speed were added to sound levelas independent variables This increased the variance ex-plained to 98 (F-change = 1250 df1 = 3 df2 = 19p lt 0001) Apart from sound level also rise speed con-

tributed significantly to the variance explained The thirdmodel in which train type was added as an independentdummy variable did not significantly increase the propor-tion explained variance (F-change lt 10) These resultssuggest that in this conventional listening test there is nodifference in perceived annoyance between different typesof trains over and above the effect of sound level and risespeed

One has to note that the number of responses to eachstimulus was smaller in the main experiment (10-minutemenus) than in the conventional listenig test (45-secondpassages) This explains why the standard errors are lowerand the explained variance is higher in the latter experi-ment

4 Discussion

The annoyance results of the present field experiment areclose to residentsrsquo everyday reality although comparisonwith published studies is somewhat limited Previous lab-oratory experiments on noise annoyance of conventionalIC and high-speed trains specifically magnetic levitationtrains [8 5] report significant differences for these typesof sound In particular the results have shown that forthe same LAeq high-speed trains were more annoyingthan other trains Compared to road traffic noise the citedstudies claimed a lower annoyance level for conventionaltrains In the present field experiment we did not find sup-port for any annoyance difference between various types

597

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

0

10

20

30

40

50

60

70

80

40 50 60 70 80 90

annoyance

[maste

rscale

units]

L [dB(A)]Aeq45sec

(a)

0

10

20

30

40

50

60

70

80

40 50 60 70 80 90

annoyance

[maste

rscale

units]

L [dB(A)]Aeq45sec

(b)

Figure 10 Average master scaled annoyance versus LAeq45s

showing (a) the noise event rise speed and (b) the distance tothe track as the size of the circles The master function is alsoindicated (dashed line)

of trains and road traffic Some possible explanations willbe given in the following subsections

41 Realistic listening situation with 10-minutemenus

The experiment was performed in a realistic setting inwhich outdoor transportation noise was reproduced andnatural outdoor-to-indoor sound propagation characteris-tics were utilized (slightly open window) This setting pro-vided a realistic sound environment indoors Subgroups ofpanelists were kept indoors during the four-hour experi-ments and upon request annoyance to transport noise wasreported with reference to 10-minute periods

Because trains run on expected schedules to which peo-ple habituate the experimental situation in classical exper-iments is rather unrealistic The experimental one-passagesituation [8] requires full attention and will have a largevariation of train sounds compared with a particular rail-way track The outcome will to a large extent depend onthe experimental context that is the variation introducedin the experiment by selecting stimuli and using randompresentation orders Random orders of recordings can beselected and arranged so that annoyance judgments on cat-egory scales plotted against sound level differentiate wellor not well on type of transport In the present field experi-ment sub-context in sessions was kept invariant similar tothe situation on a real railway track The judgmental con-

text will then be much more restricted as is the case whenliving along one railway track

Next to this the annoyance reports of the one 45-secondtrain passage were higher than those of two passages ofthe same train within the 10-minute menus This is all inorder because the two types of annoyance were masterscaled in order to become comparable over experimentalsessions When judging 45-second train passages imme-diately after exposure it is quiet clear that the task is toassess the annoyance of that particular train passage (orother sounds that were presented) However when askedto assess the annoyance retrospectively of the transportnoise during the last 10-minute menu (eg two train pas-sages) the panelist will have to choose a strategy on howto go about this For example the annoyance may only bereferred to the two noise-stimulus periods or to the whole10-minute period (menu) It has been shown that the noiseannoyance of two overlapping (equal) noises would be ex-pected always to be less than the arithmetic sum of the twoannoyances (for a review see [24]) It is more uncertainhow total annoyance of two train passages separated intime will actually be acquired A laboratory experimentwhich included long sound fragments [4] has not foundthe above-mentioned annoyance difference between differ-ent train types which is in line with our results

42 Advanced scaling methodology

Long-term retrospective annoyance asked for in question-naire surveys has typically been assessed on categoryscales (eg [3]) A response category is then implicitlypostulated to be identical for every participant by verballabeling of the two end points or of every response boxalso the intervals between categories are assumed to be thesame However this assumption does not hold true [25]eg in questionnaire surveys the response criteria (scalevalue or category borders) for annoyance are much higherfor respondents in low noise areas as compared to those inhighly exposed areas The most well known scaling bias inlaboratory experiments is the context effect in which par-ticipants distribute their responses over the ldquofullrdquo range ofcategories independent of the size of the exposure range(for a review see [22]) In the process of using categoryscales floor and ceiling effects on annoyance may also ap-pearTo avoid uncontrolled context effects an invariant sound

level range of references was used as the annoyance con-text in the present field experiment Continuous road trafficnoise was chosen as a reference instead of multiple eventsounds because it is simpler to reproduce in future stud-ies To avoid the scaling bias of category scales the methodof magnitude estimation was chosen in which participantswere free to use the range of numbers they felt comfortablewith Master scaling was applied to these individual an-noyance estimates involving a transformation function toa common master scale defined by the references whichsound levels defined the scaling context In theory thismaster scale transformation will calibrate the loudness-dependence of noise annoyance whereas the relative con-

598

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

tribution to noise annoyance from qualitative content (egthe type of sound the time pattern and cues for speed anddistance) will hopefully be unchanged

Earlier research has shown that master scaling with ref-erences works well for loudness or annoyance of a one-occasion target exposure that is when repeated exposureis unfeasible (eg experiments with long duration expo-sures) or impossible (eg questionnaire surveys in fieldstudies) an example can be found in [23] The results ob-tained from the present field experiment are probably morereliable than the results that would have been obtained bycategory scaling The test-retest reliability of the panelistsrsquomagnitude estimates of annoyance of the reference soundswas found to be very good (above 08) compared to thereliability of 072 obtained in [8] for a group of 12 muchyounger subjects Considering that our panelists all werenaiumlve participants they also each produced high qualitypsychophysical functions for the reference as discussedin Section 26

43 Other possible explanations

There are several reasons why other investigators havefound a railway bonus (for a review see [3]) which wasnot found in this field experiment One of the reasons forfinding a railway bonus for short (one minute) noises inlistening experiments may be that the relation betweenloudness and LAeq is inherently different for train androad traffic noise Indeed some researchers have arguedthat noise annoyance evaluation in listening tests of shortsounds actually is close to a perceptual loudness evalua-tion (however see [26] on differences between loudness-based and quality-based perceived annoyance) If Zwickerloudness is a good first estimate of perceptual loudnessthe difference between train noise (of different types) andhighway noise would be seen in a Zwicker loudness ver-sus LAeq plot (Figure 11) Because the IC train noise usedin the present experiment was the noise of modern ratherquiet trains a few older and noisier IC train models wereadded in this acoustic analysis At levels above 65 dB(A)TGV and maglev trains seem to be a little louder thanhighway traffic or older IC trains However this effect onZwicker loudness is not significant and does therefore notsupport a railway bonus of 5 dB(A) stipulated in severalcountriesrsquo legislation Rather it seems to be a good actionto start to replace old IC trains by new ones The railwaybonus was originally based on studies with rather old low-speed trains and with much less dense traffic intensity thannowadaysThe intermittent character of railway noise could also be

an explanation for the railway bonus However this doesnot hold for aircraft noise which is also intermittent thiscan be explained by a difference in exposure In the case ofaircraft noise the exposure is on top of buildings and on allfaccedilades In the case of road traffic noise the probability ishigh that there are local roads also but there is a possibil-ity for a ldquoquiet siderdquo people are less annoyed if quiet sidesare available [27] In the case of railway noise there is alow probability for the presence of more than one track

Zw

ickerlo

udness

[sone]

L [dB(A)]Aeq45sec

0

10

20

30

40

50

40 50 60 70 80

Figure 11 Zwicker loudness versus LAeq45s for different types oftransportation sounds ( ) IC train ( ) TGV ( ) maglev train( ) highway traffic and ( ) some additional noisier IC trains(older type)

so the exposure will also be directed at only one faccediladeIn comparing road traffic and trains the faccedilade insulationwill be more effective in the case of train noise because ofthe smaller low-frequency proportion associated to trainnoise In comparing aircraft and trains which are both in-termittent the indoor exposure is certainly more intensivefor aircraft Considering these arguments it seems obviousthat aircraft is more annoying than road traffic which ismore annoying than train However faccedilade reduction wastaken into account in the present field experiment and stillthere was no clear railway bonus found Compared withthe field condition with closed windows and the faccediladefilter used in [8] a partially open window was used in thepresent field experiment which could explain this

In surveys questioning people at their home a lowerreported annoyance for train noise compared to highwaytraffic noise was observed in a particular range of noiselevels Most of the possible explanations proposed in liter-ature conflict with the fact that this railway bonus wouldbe observed in experiments based on single passages Wemention just a few The typical character of train noise andthe concentration of the sound energy in short time inter-vals may be advantageous with regard to activity distur-bance If the level is sufficiently low the probability ofnoticing the train noise is small compared to the prob-ability of noticing the sound of a continuous source Inaddition to physical differences in the sound the ldquogreenimagerdquo of trains as a means of transportation may add tothe acceptability of the source and thus increase the tol-erance to its noise that is as long as train passages arenot too frequent However a more recent hedonic pric-ing study found that householders in Birmingham placea greater value on reductions in railway noise than in roadtraffic noise [28] Cross-cultural studies (in field and lab-oratory context) have shown that a railway bonus is notuniversal [29 30] which would favor the argument aboveIt has further been shown that the bonus varies depend-ing on the (multiple) exposure situation [31] Based on theabove only part of the effect is supposed to be visible in

599

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

field experiments such as the one reported of in this paperPart of the effect is precisely what is observed

5 Conclusion

This study has shown that in an ldquoat homerdquo like contextnoise annoyance caused by different types of trains at thesame average outdoor faccedilade exposure level is not signifi-cantly different In particular magnetic levitation systemsare not more annoying than conventional high speed trainswhich is in agreement with earlier research Noise annoy-ance caused by conventional trains was not found to besignificantly lower than annoyance caused by TGVrsquos ormaglev trains at the same average faccedilade exposure Fieldsurveys have shown that for the same average sound levelrailway noise causes less annoyance or highly annoyedpersons than highway traffic noise Although our field ex-periment included several factors that may contribute tothis effect we could not observe it

More insight may be gained by taking into account thepsychoacoustic characteristics of the noise exposure andthe relevant personal factors of the panelists This paperhas focused on discussing the experimental methodologyin great detail and on presenting the results as a functionof the average outdoor faccedilade exposure level since thisis the main noise legislation indicator used in the Nether-lands Results of a detailed psychoacoustic analysis willbe reported in a future paper

Acknowledgments

This research was financed by the Project Group Zuider-zeelijn of the Ministry of Public Transport Public Worksand Water Management in the Netherlands The membersof the project steering committeemdashGilles Janssen (dBVi-sion) Annemarie Ruysbroek (RIVM) Martin van denBerg (VROM) and Pieter Jansse (Project Group Zuider-zeelijn) mdash are acknowledged for their valuable input Wealso appreciate the experimental assistance provided byMs Ingrid Decoster Finally we would like to thank thereviewers for their valuable comments and suggestions

References

[1] J M Fields J G Walker Comparing the relationships be-tween noise level and annoyance in different surveys a rail-way vs aircraft and road traffic comparison J Sound Vib81 (1982) 51ndash80

[2] U Moehler Community response to railway noise A re-view of social surveys J Sound Vib 120 (1988) 321ndash332

[3] H M E Miedema H Vos Exposure-response relation-ships for transportation noise J Acoust Soc Am 104(1998) 3432ndash3445

[4] H Fastl G Gottschling Subjective evaluation of noiseimmissions from transrapid Proceedings of the 25thInternational Congress on Noise Control Engineering(Intermiddotnoise) Liverpool UK Juli 1996 F A Hill R Law-rence (eds)

[5] D Neugebauer J Ortscheid Geraumluschbewertung des Tran-srapid Ist der Transrapid wie die Bahn zu beurteilenBerlin Umweltbundesamt Texte 25 1997

[6] D Neugebauer J Ortscheid Geraumluschbewertung des Tran-srapid Fortschritte der Akustik mdash DAGA 97 Kiel Ger-many 1997

[7] J Vos Annoyance caused by magnetic levitation trainTransrapid 08 mdash a laboratory study TNO report TM-03-C001 The Netherlands Organisation for Applied ScientificResearch (TNO) Soesterberg The Netherlands Jan 2003

[8] J Vos Annoyance caused by the sounds of a magnetic lev-itation train J Acoust Soc Am 115 (2004) 1597ndash1608

[9] B Berglund A Preis Is perceived annoyance more sub-ject-dependent than perceived loudness Acta Acusticaunited with Acustica 83 (1997) 313ndash319

[10] M Ishibashi K Ueno H Tachibana M Watanabe Sub-jective experiment on annoying sounds in living environ-ments Proceedings of the 18th International Congress onAcoustics (ICA) Kyoto Japan 2004

[11] J Maillard J Martin P Champelovier J Lambert Percep-tive evaluation of road traffic noise inside buildings using acombined image and wave field synthesis system Proceed-ings of CFADAGArsquo04 Strasbourg France Maumlrz 2004

[12] R F S Job Community response to noise A review offactors influencing the relationship between noise exposureand reaction J Acoust Soc Am 83 (1988) 991ndash1001

[13] H M E Miedema H Vos Demographic and attitudinalfactors that modify annoyance from transportation noise JAcoust Soc Am 105 (1999) 3336ndash3344

[14] W Ellermeier M Eigenstetter K Zimmer Psychoacousticcorrelates of individual noise sensitivity J Acoust SocAm 109 (2001) 1464ndash1473

[15] D Vaumlstfjaumlll M Kleiner T Gaumlrling Affective reactions tointerior aircraft sounds Acta Acustica united with Acustica89 (2003) 693ndash701

[16] T Feng C Xiao-hong L Tan-feng A social survey onthe effects of high-speed maglev noise on residents inShangai Proceedings of The 2005 Congress and Expo-sition on Noise Control Engineering (Intermiddotnoise) Rio deJaneiro Brazil Aug 2005

[17] J J Gibson The ecological approach to visual perceptionLawrence Erlbaum Associates Inc Hillsdale New JerseyUSA 1979

[18] J G Neuhoff Ecological psychoacoustics Introductionand history ndash In Ecological Psychoacoustics J G Neuhoff(ed) Elsevier Academic Press London UK 2004

[19] C Guastavino B F G Katz J-D Polack D J LevitinD Dubois Ecological validity of soundscape reproductionActa Acustica united with Acustica 91 (2005) 333ndash341

[20] A J Berkhout D de Vries P Vogel Acoustic control bywave field synthesis J Acoust Soc Am 93 (1993) 2764ndash2778

[21] E A M Franssen J E F van Dongen J M H Ruys-broek H Vos R K Stellato Hinder door milieufactorenen de beoordeling van de leefomgeving in Nederland In-ventarisatie verstoringen 2003 RIVM report 815120 001National Institute for Public Health and the Environment(RIVM) Bilthoven The Netherlands 2004

[22] L E Marks D Algom Psychophysical scaling ndash In Mea-surement Judgment and Decision Making M H Birn-baum (ed) Academic Press New York 1998

[23] B Berglund Quality assurance in environmental psycho-physics ndash In Ratio Scaling of Psychological Magnitudes -In Honor of the Memory of S S Stevens S J BolanowskiG A Gescheider (eds) Erlbaum Hillsdale N J 1991

600

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

Table V Stepwise multiple regression analysis of acoustic variables on perceived annoyance of train and highway traffic sounds for themain field experiment with 10-minute menus The Pearsonrsquos correlation coefficients of the variables entered in the regression analysisare shown at the bottom lowastp lt 005 lowastlowastp lt 001 lowastlowastlowastp lt 0001

Model Model Model fit increase F-change Independent Variables Coefficient t-valuefit (r2) (r2-change)

1 080 080 18748lowastlowastlowast LAeq10min [dB(A)] 118 1369lowastlowastlowast

2 085 005 1449lowastlowastlowast LAeq10min [dB(A)] 092 917lowastlowastlowast

log10(distance [m]) minus1074 minus381lowastlowastlowast

3 085 000 013 LAeq10min [dB(A)] 096 822lowastlowastlowast

log10(distance [m]) minus1017 minus333lowastlowastMAGLEV [01] 145 027TGV [01] 085 016IC [01] 227 040

Label Variable ANN LEQ DIST MAG TGV

ANN Annoyance [master scale]LEQ LAeq10min [dB(A)] 0894DIST log10(distance [m]) minus0754 minus0659MAG MAGLEV [01] minus0023 minus0038 0009TGV TGV [01] 0132 0224 0006 minus0682IC IC [01] minus0179 minus0290 0004 minus0433 minus0308

systematically less annoying than highway traffic noiseThis means that no support for a railway bonus was foundin this experiment at least it was not as obvious that itcould be observed using linear statistics Figure 7 givesan overview of the annoyance functions for the 10-minutemenus as a function of LAeq10min The dashed line indi-cates the master function of annoyance for the road-traffic-like sounds used as referencesThe shorter rise time of the noise of arriving high speed

trains may create more annoyance than a conventionaltrain can do Figure 8 Panel a shows the rise speeds indB(A)s in proportion to circle sizes These values werecalculated for all sound events included in this experi-ment by fitting a straight line through the initial increasein sound level The accelerating growth of annoyance withincreasing LAeq may be explained by the rise time In Fig-ure 8 Panel b the size of the circles is instead propor-tional to the distance to the track For LAeq in the intervalbetween 50 and 65 dB(A) annoyance is clearly lower fortrain passages at larger distances than for train passagesat closer distances or road traffic noise (dashed line) Thiscould indicate that a possible noise annoyance bonus fortrain noise would only hold at larger distances from thetrack and only in the latter LAeq interval

32 Conventional listening test

In the conventional listening experiment the sounds werepresented as short 45-second fragments containing thesound of one train passage and highway excerpts Figure 9shows the results of these master scaled annoyance val-ues as a function of time averaged A-weighted faccedilade ex-posure LAeq45s A railway penalty can be observed bothin regard to the artificial reference sounds as well as to

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq10min

(a)

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq10min

(b)

Figure 8 Average master scaled annoyance of the menus versusLAeq10min showing (a) the noise event rise speed and (b) the dis-tance to the track as the size of the circles The master functionis also indicated (dashed line)

the highway sounds Figure 10 shows the annoyance as afunction of rise speed ( Panel a) and distance to the track(Panel b)

596

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

Table VI Stepwise multiple regression analysis of acoustic variables on perceived annoyance of train sounds (no highway trafficsounds) for the conventional listening test (45-second passages) The Pearsonrsquos correlation coefficients of the variables entered in theregression analysis are shown at the bottom lowastp lt 005 lowastlowastp lt 001 lowastlowastlowastp lt 0001

Model Model Model fit increase F-change Independent Variables Coefficient t-valuefit (r2) (r2-change)

1 095 095 42017lowastlowastlowast LAeq45s [dB(A)] 167 2050lowastlowastlowast

2 098 003 1250lowastlowastlowast LAeq45s [dB(A)] 123 1104lowastlowastlowast

Speed [kmh] 002 203log10(distance [m]) minus178 minus077Rise speed [dB(A)s] 063 365lowastlowast

3 099 001 098 LAeq45s [dB(A)] 108 685lowastlowastlowast

Speed [kmh] 003 209log10(distance [m]) minus476 minus146Rise speed [dB(A)s] 058 330lowastlowast

MAGLEV [01] minus070 minus037TGV [01] 166 104

Label Variable ANN LEQ SPD DIST RISE MAG

ANN Annoyance [master scale]LEQ LAeq45s [dB(A)] 0975SPD Speed [kmh] 0646 0541DIST log10(distance [m]) minus0613 minus0667 minus0001RISE Rise speed [dB(A)s] 0885 0804 0708 minus0437MAG MAGLEV [01] 0070 minus0017 0552 0000 0188TGV TGV [01] 0207 0246 minus0190 0000 0100 minus0707

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq45sec

Figure 9 Average master scaled annoyance versus LAeq45s forthe conventional listening test for different types of train sounds( ) IC train ( ) TGV and ( ) maglev train In comparison theannoyance for the highway traffic ( ) is also shown Standard er-ror on means is indicated as well as the master function (dashedline)

A stepwise multiple linear regression analysis was alsoperformed separately for the train noises (Table VI) Thefirst model in which sound level LAeq45s was included asthe only independent variable explained 95 of the vari-ance in annoyance In the second model train speed dis-tance to the track and rise speed were added to sound levelas independent variables This increased the variance ex-plained to 98 (F-change = 1250 df1 = 3 df2 = 19p lt 0001) Apart from sound level also rise speed con-

tributed significantly to the variance explained The thirdmodel in which train type was added as an independentdummy variable did not significantly increase the propor-tion explained variance (F-change lt 10) These resultssuggest that in this conventional listening test there is nodifference in perceived annoyance between different typesof trains over and above the effect of sound level and risespeed

One has to note that the number of responses to eachstimulus was smaller in the main experiment (10-minutemenus) than in the conventional listenig test (45-secondpassages) This explains why the standard errors are lowerand the explained variance is higher in the latter experi-ment

4 Discussion

The annoyance results of the present field experiment areclose to residentsrsquo everyday reality although comparisonwith published studies is somewhat limited Previous lab-oratory experiments on noise annoyance of conventionalIC and high-speed trains specifically magnetic levitationtrains [8 5] report significant differences for these typesof sound In particular the results have shown that forthe same LAeq high-speed trains were more annoyingthan other trains Compared to road traffic noise the citedstudies claimed a lower annoyance level for conventionaltrains In the present field experiment we did not find sup-port for any annoyance difference between various types

597

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

0

10

20

30

40

50

60

70

80

40 50 60 70 80 90

annoyance

[maste

rscale

units]

L [dB(A)]Aeq45sec

(a)

0

10

20

30

40

50

60

70

80

40 50 60 70 80 90

annoyance

[maste

rscale

units]

L [dB(A)]Aeq45sec

(b)

Figure 10 Average master scaled annoyance versus LAeq45s

showing (a) the noise event rise speed and (b) the distance tothe track as the size of the circles The master function is alsoindicated (dashed line)

of trains and road traffic Some possible explanations willbe given in the following subsections

41 Realistic listening situation with 10-minutemenus

The experiment was performed in a realistic setting inwhich outdoor transportation noise was reproduced andnatural outdoor-to-indoor sound propagation characteris-tics were utilized (slightly open window) This setting pro-vided a realistic sound environment indoors Subgroups ofpanelists were kept indoors during the four-hour experi-ments and upon request annoyance to transport noise wasreported with reference to 10-minute periods

Because trains run on expected schedules to which peo-ple habituate the experimental situation in classical exper-iments is rather unrealistic The experimental one-passagesituation [8] requires full attention and will have a largevariation of train sounds compared with a particular rail-way track The outcome will to a large extent depend onthe experimental context that is the variation introducedin the experiment by selecting stimuli and using randompresentation orders Random orders of recordings can beselected and arranged so that annoyance judgments on cat-egory scales plotted against sound level differentiate wellor not well on type of transport In the present field experi-ment sub-context in sessions was kept invariant similar tothe situation on a real railway track The judgmental con-

text will then be much more restricted as is the case whenliving along one railway track

Next to this the annoyance reports of the one 45-secondtrain passage were higher than those of two passages ofthe same train within the 10-minute menus This is all inorder because the two types of annoyance were masterscaled in order to become comparable over experimentalsessions When judging 45-second train passages imme-diately after exposure it is quiet clear that the task is toassess the annoyance of that particular train passage (orother sounds that were presented) However when askedto assess the annoyance retrospectively of the transportnoise during the last 10-minute menu (eg two train pas-sages) the panelist will have to choose a strategy on howto go about this For example the annoyance may only bereferred to the two noise-stimulus periods or to the whole10-minute period (menu) It has been shown that the noiseannoyance of two overlapping (equal) noises would be ex-pected always to be less than the arithmetic sum of the twoannoyances (for a review see [24]) It is more uncertainhow total annoyance of two train passages separated intime will actually be acquired A laboratory experimentwhich included long sound fragments [4] has not foundthe above-mentioned annoyance difference between differ-ent train types which is in line with our results

42 Advanced scaling methodology

Long-term retrospective annoyance asked for in question-naire surveys has typically been assessed on categoryscales (eg [3]) A response category is then implicitlypostulated to be identical for every participant by verballabeling of the two end points or of every response boxalso the intervals between categories are assumed to be thesame However this assumption does not hold true [25]eg in questionnaire surveys the response criteria (scalevalue or category borders) for annoyance are much higherfor respondents in low noise areas as compared to those inhighly exposed areas The most well known scaling bias inlaboratory experiments is the context effect in which par-ticipants distribute their responses over the ldquofullrdquo range ofcategories independent of the size of the exposure range(for a review see [22]) In the process of using categoryscales floor and ceiling effects on annoyance may also ap-pearTo avoid uncontrolled context effects an invariant sound

level range of references was used as the annoyance con-text in the present field experiment Continuous road trafficnoise was chosen as a reference instead of multiple eventsounds because it is simpler to reproduce in future stud-ies To avoid the scaling bias of category scales the methodof magnitude estimation was chosen in which participantswere free to use the range of numbers they felt comfortablewith Master scaling was applied to these individual an-noyance estimates involving a transformation function toa common master scale defined by the references whichsound levels defined the scaling context In theory thismaster scale transformation will calibrate the loudness-dependence of noise annoyance whereas the relative con-

598

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

tribution to noise annoyance from qualitative content (egthe type of sound the time pattern and cues for speed anddistance) will hopefully be unchanged

Earlier research has shown that master scaling with ref-erences works well for loudness or annoyance of a one-occasion target exposure that is when repeated exposureis unfeasible (eg experiments with long duration expo-sures) or impossible (eg questionnaire surveys in fieldstudies) an example can be found in [23] The results ob-tained from the present field experiment are probably morereliable than the results that would have been obtained bycategory scaling The test-retest reliability of the panelistsrsquomagnitude estimates of annoyance of the reference soundswas found to be very good (above 08) compared to thereliability of 072 obtained in [8] for a group of 12 muchyounger subjects Considering that our panelists all werenaiumlve participants they also each produced high qualitypsychophysical functions for the reference as discussedin Section 26

43 Other possible explanations

There are several reasons why other investigators havefound a railway bonus (for a review see [3]) which wasnot found in this field experiment One of the reasons forfinding a railway bonus for short (one minute) noises inlistening experiments may be that the relation betweenloudness and LAeq is inherently different for train androad traffic noise Indeed some researchers have arguedthat noise annoyance evaluation in listening tests of shortsounds actually is close to a perceptual loudness evalua-tion (however see [26] on differences between loudness-based and quality-based perceived annoyance) If Zwickerloudness is a good first estimate of perceptual loudnessthe difference between train noise (of different types) andhighway noise would be seen in a Zwicker loudness ver-sus LAeq plot (Figure 11) Because the IC train noise usedin the present experiment was the noise of modern ratherquiet trains a few older and noisier IC train models wereadded in this acoustic analysis At levels above 65 dB(A)TGV and maglev trains seem to be a little louder thanhighway traffic or older IC trains However this effect onZwicker loudness is not significant and does therefore notsupport a railway bonus of 5 dB(A) stipulated in severalcountriesrsquo legislation Rather it seems to be a good actionto start to replace old IC trains by new ones The railwaybonus was originally based on studies with rather old low-speed trains and with much less dense traffic intensity thannowadaysThe intermittent character of railway noise could also be

an explanation for the railway bonus However this doesnot hold for aircraft noise which is also intermittent thiscan be explained by a difference in exposure In the case ofaircraft noise the exposure is on top of buildings and on allfaccedilades In the case of road traffic noise the probability ishigh that there are local roads also but there is a possibil-ity for a ldquoquiet siderdquo people are less annoyed if quiet sidesare available [27] In the case of railway noise there is alow probability for the presence of more than one track

Zw

ickerlo

udness

[sone]

L [dB(A)]Aeq45sec

0

10

20

30

40

50

40 50 60 70 80

Figure 11 Zwicker loudness versus LAeq45s for different types oftransportation sounds ( ) IC train ( ) TGV ( ) maglev train( ) highway traffic and ( ) some additional noisier IC trains(older type)

so the exposure will also be directed at only one faccediladeIn comparing road traffic and trains the faccedilade insulationwill be more effective in the case of train noise because ofthe smaller low-frequency proportion associated to trainnoise In comparing aircraft and trains which are both in-termittent the indoor exposure is certainly more intensivefor aircraft Considering these arguments it seems obviousthat aircraft is more annoying than road traffic which ismore annoying than train However faccedilade reduction wastaken into account in the present field experiment and stillthere was no clear railway bonus found Compared withthe field condition with closed windows and the faccediladefilter used in [8] a partially open window was used in thepresent field experiment which could explain this

In surveys questioning people at their home a lowerreported annoyance for train noise compared to highwaytraffic noise was observed in a particular range of noiselevels Most of the possible explanations proposed in liter-ature conflict with the fact that this railway bonus wouldbe observed in experiments based on single passages Wemention just a few The typical character of train noise andthe concentration of the sound energy in short time inter-vals may be advantageous with regard to activity distur-bance If the level is sufficiently low the probability ofnoticing the train noise is small compared to the prob-ability of noticing the sound of a continuous source Inaddition to physical differences in the sound the ldquogreenimagerdquo of trains as a means of transportation may add tothe acceptability of the source and thus increase the tol-erance to its noise that is as long as train passages arenot too frequent However a more recent hedonic pric-ing study found that householders in Birmingham placea greater value on reductions in railway noise than in roadtraffic noise [28] Cross-cultural studies (in field and lab-oratory context) have shown that a railway bonus is notuniversal [29 30] which would favor the argument aboveIt has further been shown that the bonus varies depend-ing on the (multiple) exposure situation [31] Based on theabove only part of the effect is supposed to be visible in

599

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

field experiments such as the one reported of in this paperPart of the effect is precisely what is observed

5 Conclusion

This study has shown that in an ldquoat homerdquo like contextnoise annoyance caused by different types of trains at thesame average outdoor faccedilade exposure level is not signifi-cantly different In particular magnetic levitation systemsare not more annoying than conventional high speed trainswhich is in agreement with earlier research Noise annoy-ance caused by conventional trains was not found to besignificantly lower than annoyance caused by TGVrsquos ormaglev trains at the same average faccedilade exposure Fieldsurveys have shown that for the same average sound levelrailway noise causes less annoyance or highly annoyedpersons than highway traffic noise Although our field ex-periment included several factors that may contribute tothis effect we could not observe it

More insight may be gained by taking into account thepsychoacoustic characteristics of the noise exposure andthe relevant personal factors of the panelists This paperhas focused on discussing the experimental methodologyin great detail and on presenting the results as a functionof the average outdoor faccedilade exposure level since thisis the main noise legislation indicator used in the Nether-lands Results of a detailed psychoacoustic analysis willbe reported in a future paper

Acknowledgments

This research was financed by the Project Group Zuider-zeelijn of the Ministry of Public Transport Public Worksand Water Management in the Netherlands The membersof the project steering committeemdashGilles Janssen (dBVi-sion) Annemarie Ruysbroek (RIVM) Martin van denBerg (VROM) and Pieter Jansse (Project Group Zuider-zeelijn) mdash are acknowledged for their valuable input Wealso appreciate the experimental assistance provided byMs Ingrid Decoster Finally we would like to thank thereviewers for their valuable comments and suggestions

References

[1] J M Fields J G Walker Comparing the relationships be-tween noise level and annoyance in different surveys a rail-way vs aircraft and road traffic comparison J Sound Vib81 (1982) 51ndash80

[2] U Moehler Community response to railway noise A re-view of social surveys J Sound Vib 120 (1988) 321ndash332

[3] H M E Miedema H Vos Exposure-response relation-ships for transportation noise J Acoust Soc Am 104(1998) 3432ndash3445

[4] H Fastl G Gottschling Subjective evaluation of noiseimmissions from transrapid Proceedings of the 25thInternational Congress on Noise Control Engineering(Intermiddotnoise) Liverpool UK Juli 1996 F A Hill R Law-rence (eds)

[5] D Neugebauer J Ortscheid Geraumluschbewertung des Tran-srapid Ist der Transrapid wie die Bahn zu beurteilenBerlin Umweltbundesamt Texte 25 1997

[6] D Neugebauer J Ortscheid Geraumluschbewertung des Tran-srapid Fortschritte der Akustik mdash DAGA 97 Kiel Ger-many 1997

[7] J Vos Annoyance caused by magnetic levitation trainTransrapid 08 mdash a laboratory study TNO report TM-03-C001 The Netherlands Organisation for Applied ScientificResearch (TNO) Soesterberg The Netherlands Jan 2003

[8] J Vos Annoyance caused by the sounds of a magnetic lev-itation train J Acoust Soc Am 115 (2004) 1597ndash1608

[9] B Berglund A Preis Is perceived annoyance more sub-ject-dependent than perceived loudness Acta Acusticaunited with Acustica 83 (1997) 313ndash319

[10] M Ishibashi K Ueno H Tachibana M Watanabe Sub-jective experiment on annoying sounds in living environ-ments Proceedings of the 18th International Congress onAcoustics (ICA) Kyoto Japan 2004

[11] J Maillard J Martin P Champelovier J Lambert Percep-tive evaluation of road traffic noise inside buildings using acombined image and wave field synthesis system Proceed-ings of CFADAGArsquo04 Strasbourg France Maumlrz 2004

[12] R F S Job Community response to noise A review offactors influencing the relationship between noise exposureand reaction J Acoust Soc Am 83 (1988) 991ndash1001

[13] H M E Miedema H Vos Demographic and attitudinalfactors that modify annoyance from transportation noise JAcoust Soc Am 105 (1999) 3336ndash3344

[14] W Ellermeier M Eigenstetter K Zimmer Psychoacousticcorrelates of individual noise sensitivity J Acoust SocAm 109 (2001) 1464ndash1473

[15] D Vaumlstfjaumlll M Kleiner T Gaumlrling Affective reactions tointerior aircraft sounds Acta Acustica united with Acustica89 (2003) 693ndash701

[16] T Feng C Xiao-hong L Tan-feng A social survey onthe effects of high-speed maglev noise on residents inShangai Proceedings of The 2005 Congress and Expo-sition on Noise Control Engineering (Intermiddotnoise) Rio deJaneiro Brazil Aug 2005

[17] J J Gibson The ecological approach to visual perceptionLawrence Erlbaum Associates Inc Hillsdale New JerseyUSA 1979

[18] J G Neuhoff Ecological psychoacoustics Introductionand history ndash In Ecological Psychoacoustics J G Neuhoff(ed) Elsevier Academic Press London UK 2004

[19] C Guastavino B F G Katz J-D Polack D J LevitinD Dubois Ecological validity of soundscape reproductionActa Acustica united with Acustica 91 (2005) 333ndash341

[20] A J Berkhout D de Vries P Vogel Acoustic control bywave field synthesis J Acoust Soc Am 93 (1993) 2764ndash2778

[21] E A M Franssen J E F van Dongen J M H Ruys-broek H Vos R K Stellato Hinder door milieufactorenen de beoordeling van de leefomgeving in Nederland In-ventarisatie verstoringen 2003 RIVM report 815120 001National Institute for Public Health and the Environment(RIVM) Bilthoven The Netherlands 2004

[22] L E Marks D Algom Psychophysical scaling ndash In Mea-surement Judgment and Decision Making M H Birn-baum (ed) Academic Press New York 1998

[23] B Berglund Quality assurance in environmental psycho-physics ndash In Ratio Scaling of Psychological Magnitudes -In Honor of the Memory of S S Stevens S J BolanowskiG A Gescheider (eds) Erlbaum Hillsdale N J 1991

600

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

Table VI Stepwise multiple regression analysis of acoustic variables on perceived annoyance of train sounds (no highway trafficsounds) for the conventional listening test (45-second passages) The Pearsonrsquos correlation coefficients of the variables entered in theregression analysis are shown at the bottom lowastp lt 005 lowastlowastp lt 001 lowastlowastlowastp lt 0001

Model Model Model fit increase F-change Independent Variables Coefficient t-valuefit (r2) (r2-change)

1 095 095 42017lowastlowastlowast LAeq45s [dB(A)] 167 2050lowastlowastlowast

2 098 003 1250lowastlowastlowast LAeq45s [dB(A)] 123 1104lowastlowastlowast

Speed [kmh] 002 203log10(distance [m]) minus178 minus077Rise speed [dB(A)s] 063 365lowastlowast

3 099 001 098 LAeq45s [dB(A)] 108 685lowastlowastlowast

Speed [kmh] 003 209log10(distance [m]) minus476 minus146Rise speed [dB(A)s] 058 330lowastlowast

MAGLEV [01] minus070 minus037TGV [01] 166 104

Label Variable ANN LEQ SPD DIST RISE MAG

ANN Annoyance [master scale]LEQ LAeq45s [dB(A)] 0975SPD Speed [kmh] 0646 0541DIST log10(distance [m]) minus0613 minus0667 minus0001RISE Rise speed [dB(A)s] 0885 0804 0708 minus0437MAG MAGLEV [01] 0070 minus0017 0552 0000 0188TGV TGV [01] 0207 0246 minus0190 0000 0100 minus0707

-10

0

10

20

30

40

50

60

70

30 40 50 60 70 80

annoyance

[maste

rscale

units]

L [dB(A)]Aeq45sec

Figure 9 Average master scaled annoyance versus LAeq45s forthe conventional listening test for different types of train sounds( ) IC train ( ) TGV and ( ) maglev train In comparison theannoyance for the highway traffic ( ) is also shown Standard er-ror on means is indicated as well as the master function (dashedline)

A stepwise multiple linear regression analysis was alsoperformed separately for the train noises (Table VI) Thefirst model in which sound level LAeq45s was included asthe only independent variable explained 95 of the vari-ance in annoyance In the second model train speed dis-tance to the track and rise speed were added to sound levelas independent variables This increased the variance ex-plained to 98 (F-change = 1250 df1 = 3 df2 = 19p lt 0001) Apart from sound level also rise speed con-

tributed significantly to the variance explained The thirdmodel in which train type was added as an independentdummy variable did not significantly increase the propor-tion explained variance (F-change lt 10) These resultssuggest that in this conventional listening test there is nodifference in perceived annoyance between different typesof trains over and above the effect of sound level and risespeed

One has to note that the number of responses to eachstimulus was smaller in the main experiment (10-minutemenus) than in the conventional listenig test (45-secondpassages) This explains why the standard errors are lowerand the explained variance is higher in the latter experi-ment

4 Discussion

The annoyance results of the present field experiment areclose to residentsrsquo everyday reality although comparisonwith published studies is somewhat limited Previous lab-oratory experiments on noise annoyance of conventionalIC and high-speed trains specifically magnetic levitationtrains [8 5] report significant differences for these typesof sound In particular the results have shown that forthe same LAeq high-speed trains were more annoyingthan other trains Compared to road traffic noise the citedstudies claimed a lower annoyance level for conventionaltrains In the present field experiment we did not find sup-port for any annoyance difference between various types

597

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

0

10

20

30

40

50

60

70

80

40 50 60 70 80 90

annoyance

[maste

rscale

units]

L [dB(A)]Aeq45sec

(a)

0

10

20

30

40

50

60

70

80

40 50 60 70 80 90

annoyance

[maste

rscale

units]

L [dB(A)]Aeq45sec

(b)

Figure 10 Average master scaled annoyance versus LAeq45s

showing (a) the noise event rise speed and (b) the distance tothe track as the size of the circles The master function is alsoindicated (dashed line)

of trains and road traffic Some possible explanations willbe given in the following subsections

41 Realistic listening situation with 10-minutemenus

The experiment was performed in a realistic setting inwhich outdoor transportation noise was reproduced andnatural outdoor-to-indoor sound propagation characteris-tics were utilized (slightly open window) This setting pro-vided a realistic sound environment indoors Subgroups ofpanelists were kept indoors during the four-hour experi-ments and upon request annoyance to transport noise wasreported with reference to 10-minute periods

Because trains run on expected schedules to which peo-ple habituate the experimental situation in classical exper-iments is rather unrealistic The experimental one-passagesituation [8] requires full attention and will have a largevariation of train sounds compared with a particular rail-way track The outcome will to a large extent depend onthe experimental context that is the variation introducedin the experiment by selecting stimuli and using randompresentation orders Random orders of recordings can beselected and arranged so that annoyance judgments on cat-egory scales plotted against sound level differentiate wellor not well on type of transport In the present field experi-ment sub-context in sessions was kept invariant similar tothe situation on a real railway track The judgmental con-

text will then be much more restricted as is the case whenliving along one railway track

Next to this the annoyance reports of the one 45-secondtrain passage were higher than those of two passages ofthe same train within the 10-minute menus This is all inorder because the two types of annoyance were masterscaled in order to become comparable over experimentalsessions When judging 45-second train passages imme-diately after exposure it is quiet clear that the task is toassess the annoyance of that particular train passage (orother sounds that were presented) However when askedto assess the annoyance retrospectively of the transportnoise during the last 10-minute menu (eg two train pas-sages) the panelist will have to choose a strategy on howto go about this For example the annoyance may only bereferred to the two noise-stimulus periods or to the whole10-minute period (menu) It has been shown that the noiseannoyance of two overlapping (equal) noises would be ex-pected always to be less than the arithmetic sum of the twoannoyances (for a review see [24]) It is more uncertainhow total annoyance of two train passages separated intime will actually be acquired A laboratory experimentwhich included long sound fragments [4] has not foundthe above-mentioned annoyance difference between differ-ent train types which is in line with our results

42 Advanced scaling methodology

Long-term retrospective annoyance asked for in question-naire surveys has typically been assessed on categoryscales (eg [3]) A response category is then implicitlypostulated to be identical for every participant by verballabeling of the two end points or of every response boxalso the intervals between categories are assumed to be thesame However this assumption does not hold true [25]eg in questionnaire surveys the response criteria (scalevalue or category borders) for annoyance are much higherfor respondents in low noise areas as compared to those inhighly exposed areas The most well known scaling bias inlaboratory experiments is the context effect in which par-ticipants distribute their responses over the ldquofullrdquo range ofcategories independent of the size of the exposure range(for a review see [22]) In the process of using categoryscales floor and ceiling effects on annoyance may also ap-pearTo avoid uncontrolled context effects an invariant sound

level range of references was used as the annoyance con-text in the present field experiment Continuous road trafficnoise was chosen as a reference instead of multiple eventsounds because it is simpler to reproduce in future stud-ies To avoid the scaling bias of category scales the methodof magnitude estimation was chosen in which participantswere free to use the range of numbers they felt comfortablewith Master scaling was applied to these individual an-noyance estimates involving a transformation function toa common master scale defined by the references whichsound levels defined the scaling context In theory thismaster scale transformation will calibrate the loudness-dependence of noise annoyance whereas the relative con-

598

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

tribution to noise annoyance from qualitative content (egthe type of sound the time pattern and cues for speed anddistance) will hopefully be unchanged

Earlier research has shown that master scaling with ref-erences works well for loudness or annoyance of a one-occasion target exposure that is when repeated exposureis unfeasible (eg experiments with long duration expo-sures) or impossible (eg questionnaire surveys in fieldstudies) an example can be found in [23] The results ob-tained from the present field experiment are probably morereliable than the results that would have been obtained bycategory scaling The test-retest reliability of the panelistsrsquomagnitude estimates of annoyance of the reference soundswas found to be very good (above 08) compared to thereliability of 072 obtained in [8] for a group of 12 muchyounger subjects Considering that our panelists all werenaiumlve participants they also each produced high qualitypsychophysical functions for the reference as discussedin Section 26

43 Other possible explanations

There are several reasons why other investigators havefound a railway bonus (for a review see [3]) which wasnot found in this field experiment One of the reasons forfinding a railway bonus for short (one minute) noises inlistening experiments may be that the relation betweenloudness and LAeq is inherently different for train androad traffic noise Indeed some researchers have arguedthat noise annoyance evaluation in listening tests of shortsounds actually is close to a perceptual loudness evalua-tion (however see [26] on differences between loudness-based and quality-based perceived annoyance) If Zwickerloudness is a good first estimate of perceptual loudnessthe difference between train noise (of different types) andhighway noise would be seen in a Zwicker loudness ver-sus LAeq plot (Figure 11) Because the IC train noise usedin the present experiment was the noise of modern ratherquiet trains a few older and noisier IC train models wereadded in this acoustic analysis At levels above 65 dB(A)TGV and maglev trains seem to be a little louder thanhighway traffic or older IC trains However this effect onZwicker loudness is not significant and does therefore notsupport a railway bonus of 5 dB(A) stipulated in severalcountriesrsquo legislation Rather it seems to be a good actionto start to replace old IC trains by new ones The railwaybonus was originally based on studies with rather old low-speed trains and with much less dense traffic intensity thannowadaysThe intermittent character of railway noise could also be

an explanation for the railway bonus However this doesnot hold for aircraft noise which is also intermittent thiscan be explained by a difference in exposure In the case ofaircraft noise the exposure is on top of buildings and on allfaccedilades In the case of road traffic noise the probability ishigh that there are local roads also but there is a possibil-ity for a ldquoquiet siderdquo people are less annoyed if quiet sidesare available [27] In the case of railway noise there is alow probability for the presence of more than one track

Zw

ickerlo

udness

[sone]

L [dB(A)]Aeq45sec

0

10

20

30

40

50

40 50 60 70 80

Figure 11 Zwicker loudness versus LAeq45s for different types oftransportation sounds ( ) IC train ( ) TGV ( ) maglev train( ) highway traffic and ( ) some additional noisier IC trains(older type)

so the exposure will also be directed at only one faccediladeIn comparing road traffic and trains the faccedilade insulationwill be more effective in the case of train noise because ofthe smaller low-frequency proportion associated to trainnoise In comparing aircraft and trains which are both in-termittent the indoor exposure is certainly more intensivefor aircraft Considering these arguments it seems obviousthat aircraft is more annoying than road traffic which ismore annoying than train However faccedilade reduction wastaken into account in the present field experiment and stillthere was no clear railway bonus found Compared withthe field condition with closed windows and the faccediladefilter used in [8] a partially open window was used in thepresent field experiment which could explain this

In surveys questioning people at their home a lowerreported annoyance for train noise compared to highwaytraffic noise was observed in a particular range of noiselevels Most of the possible explanations proposed in liter-ature conflict with the fact that this railway bonus wouldbe observed in experiments based on single passages Wemention just a few The typical character of train noise andthe concentration of the sound energy in short time inter-vals may be advantageous with regard to activity distur-bance If the level is sufficiently low the probability ofnoticing the train noise is small compared to the prob-ability of noticing the sound of a continuous source Inaddition to physical differences in the sound the ldquogreenimagerdquo of trains as a means of transportation may add tothe acceptability of the source and thus increase the tol-erance to its noise that is as long as train passages arenot too frequent However a more recent hedonic pric-ing study found that householders in Birmingham placea greater value on reductions in railway noise than in roadtraffic noise [28] Cross-cultural studies (in field and lab-oratory context) have shown that a railway bonus is notuniversal [29 30] which would favor the argument aboveIt has further been shown that the bonus varies depend-ing on the (multiple) exposure situation [31] Based on theabove only part of the effect is supposed to be visible in

599

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

field experiments such as the one reported of in this paperPart of the effect is precisely what is observed

5 Conclusion

This study has shown that in an ldquoat homerdquo like contextnoise annoyance caused by different types of trains at thesame average outdoor faccedilade exposure level is not signifi-cantly different In particular magnetic levitation systemsare not more annoying than conventional high speed trainswhich is in agreement with earlier research Noise annoy-ance caused by conventional trains was not found to besignificantly lower than annoyance caused by TGVrsquos ormaglev trains at the same average faccedilade exposure Fieldsurveys have shown that for the same average sound levelrailway noise causes less annoyance or highly annoyedpersons than highway traffic noise Although our field ex-periment included several factors that may contribute tothis effect we could not observe it

More insight may be gained by taking into account thepsychoacoustic characteristics of the noise exposure andthe relevant personal factors of the panelists This paperhas focused on discussing the experimental methodologyin great detail and on presenting the results as a functionof the average outdoor faccedilade exposure level since thisis the main noise legislation indicator used in the Nether-lands Results of a detailed psychoacoustic analysis willbe reported in a future paper

Acknowledgments

This research was financed by the Project Group Zuider-zeelijn of the Ministry of Public Transport Public Worksand Water Management in the Netherlands The membersof the project steering committeemdashGilles Janssen (dBVi-sion) Annemarie Ruysbroek (RIVM) Martin van denBerg (VROM) and Pieter Jansse (Project Group Zuider-zeelijn) mdash are acknowledged for their valuable input Wealso appreciate the experimental assistance provided byMs Ingrid Decoster Finally we would like to thank thereviewers for their valuable comments and suggestions

References

[1] J M Fields J G Walker Comparing the relationships be-tween noise level and annoyance in different surveys a rail-way vs aircraft and road traffic comparison J Sound Vib81 (1982) 51ndash80

[2] U Moehler Community response to railway noise A re-view of social surveys J Sound Vib 120 (1988) 321ndash332

[3] H M E Miedema H Vos Exposure-response relation-ships for transportation noise J Acoust Soc Am 104(1998) 3432ndash3445

[4] H Fastl G Gottschling Subjective evaluation of noiseimmissions from transrapid Proceedings of the 25thInternational Congress on Noise Control Engineering(Intermiddotnoise) Liverpool UK Juli 1996 F A Hill R Law-rence (eds)

[5] D Neugebauer J Ortscheid Geraumluschbewertung des Tran-srapid Ist der Transrapid wie die Bahn zu beurteilenBerlin Umweltbundesamt Texte 25 1997

[6] D Neugebauer J Ortscheid Geraumluschbewertung des Tran-srapid Fortschritte der Akustik mdash DAGA 97 Kiel Ger-many 1997

[7] J Vos Annoyance caused by magnetic levitation trainTransrapid 08 mdash a laboratory study TNO report TM-03-C001 The Netherlands Organisation for Applied ScientificResearch (TNO) Soesterberg The Netherlands Jan 2003

[8] J Vos Annoyance caused by the sounds of a magnetic lev-itation train J Acoust Soc Am 115 (2004) 1597ndash1608

[9] B Berglund A Preis Is perceived annoyance more sub-ject-dependent than perceived loudness Acta Acusticaunited with Acustica 83 (1997) 313ndash319

[10] M Ishibashi K Ueno H Tachibana M Watanabe Sub-jective experiment on annoying sounds in living environ-ments Proceedings of the 18th International Congress onAcoustics (ICA) Kyoto Japan 2004

[11] J Maillard J Martin P Champelovier J Lambert Percep-tive evaluation of road traffic noise inside buildings using acombined image and wave field synthesis system Proceed-ings of CFADAGArsquo04 Strasbourg France Maumlrz 2004

[12] R F S Job Community response to noise A review offactors influencing the relationship between noise exposureand reaction J Acoust Soc Am 83 (1988) 991ndash1001

[13] H M E Miedema H Vos Demographic and attitudinalfactors that modify annoyance from transportation noise JAcoust Soc Am 105 (1999) 3336ndash3344

[14] W Ellermeier M Eigenstetter K Zimmer Psychoacousticcorrelates of individual noise sensitivity J Acoust SocAm 109 (2001) 1464ndash1473

[15] D Vaumlstfjaumlll M Kleiner T Gaumlrling Affective reactions tointerior aircraft sounds Acta Acustica united with Acustica89 (2003) 693ndash701

[16] T Feng C Xiao-hong L Tan-feng A social survey onthe effects of high-speed maglev noise on residents inShangai Proceedings of The 2005 Congress and Expo-sition on Noise Control Engineering (Intermiddotnoise) Rio deJaneiro Brazil Aug 2005

[17] J J Gibson The ecological approach to visual perceptionLawrence Erlbaum Associates Inc Hillsdale New JerseyUSA 1979

[18] J G Neuhoff Ecological psychoacoustics Introductionand history ndash In Ecological Psychoacoustics J G Neuhoff(ed) Elsevier Academic Press London UK 2004

[19] C Guastavino B F G Katz J-D Polack D J LevitinD Dubois Ecological validity of soundscape reproductionActa Acustica united with Acustica 91 (2005) 333ndash341

[20] A J Berkhout D de Vries P Vogel Acoustic control bywave field synthesis J Acoust Soc Am 93 (1993) 2764ndash2778

[21] E A M Franssen J E F van Dongen J M H Ruys-broek H Vos R K Stellato Hinder door milieufactorenen de beoordeling van de leefomgeving in Nederland In-ventarisatie verstoringen 2003 RIVM report 815120 001National Institute for Public Health and the Environment(RIVM) Bilthoven The Netherlands 2004

[22] L E Marks D Algom Psychophysical scaling ndash In Mea-surement Judgment and Decision Making M H Birn-baum (ed) Academic Press New York 1998

[23] B Berglund Quality assurance in environmental psycho-physics ndash In Ratio Scaling of Psychological Magnitudes -In Honor of the Memory of S S Stevens S J BolanowskiG A Gescheider (eds) Erlbaum Hillsdale N J 1991

600

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

0

10

20

30

40

50

60

70

80

40 50 60 70 80 90

annoyance

[maste

rscale

units]

L [dB(A)]Aeq45sec

(a)

0

10

20

30

40

50

60

70

80

40 50 60 70 80 90

annoyance

[maste

rscale

units]

L [dB(A)]Aeq45sec

(b)

Figure 10 Average master scaled annoyance versus LAeq45s

showing (a) the noise event rise speed and (b) the distance tothe track as the size of the circles The master function is alsoindicated (dashed line)

of trains and road traffic Some possible explanations willbe given in the following subsections

41 Realistic listening situation with 10-minutemenus

The experiment was performed in a realistic setting inwhich outdoor transportation noise was reproduced andnatural outdoor-to-indoor sound propagation characteris-tics were utilized (slightly open window) This setting pro-vided a realistic sound environment indoors Subgroups ofpanelists were kept indoors during the four-hour experi-ments and upon request annoyance to transport noise wasreported with reference to 10-minute periods

Because trains run on expected schedules to which peo-ple habituate the experimental situation in classical exper-iments is rather unrealistic The experimental one-passagesituation [8] requires full attention and will have a largevariation of train sounds compared with a particular rail-way track The outcome will to a large extent depend onthe experimental context that is the variation introducedin the experiment by selecting stimuli and using randompresentation orders Random orders of recordings can beselected and arranged so that annoyance judgments on cat-egory scales plotted against sound level differentiate wellor not well on type of transport In the present field experi-ment sub-context in sessions was kept invariant similar tothe situation on a real railway track The judgmental con-

text will then be much more restricted as is the case whenliving along one railway track

Next to this the annoyance reports of the one 45-secondtrain passage were higher than those of two passages ofthe same train within the 10-minute menus This is all inorder because the two types of annoyance were masterscaled in order to become comparable over experimentalsessions When judging 45-second train passages imme-diately after exposure it is quiet clear that the task is toassess the annoyance of that particular train passage (orother sounds that were presented) However when askedto assess the annoyance retrospectively of the transportnoise during the last 10-minute menu (eg two train pas-sages) the panelist will have to choose a strategy on howto go about this For example the annoyance may only bereferred to the two noise-stimulus periods or to the whole10-minute period (menu) It has been shown that the noiseannoyance of two overlapping (equal) noises would be ex-pected always to be less than the arithmetic sum of the twoannoyances (for a review see [24]) It is more uncertainhow total annoyance of two train passages separated intime will actually be acquired A laboratory experimentwhich included long sound fragments [4] has not foundthe above-mentioned annoyance difference between differ-ent train types which is in line with our results

42 Advanced scaling methodology

Long-term retrospective annoyance asked for in question-naire surveys has typically been assessed on categoryscales (eg [3]) A response category is then implicitlypostulated to be identical for every participant by verballabeling of the two end points or of every response boxalso the intervals between categories are assumed to be thesame However this assumption does not hold true [25]eg in questionnaire surveys the response criteria (scalevalue or category borders) for annoyance are much higherfor respondents in low noise areas as compared to those inhighly exposed areas The most well known scaling bias inlaboratory experiments is the context effect in which par-ticipants distribute their responses over the ldquofullrdquo range ofcategories independent of the size of the exposure range(for a review see [22]) In the process of using categoryscales floor and ceiling effects on annoyance may also ap-pearTo avoid uncontrolled context effects an invariant sound

level range of references was used as the annoyance con-text in the present field experiment Continuous road trafficnoise was chosen as a reference instead of multiple eventsounds because it is simpler to reproduce in future stud-ies To avoid the scaling bias of category scales the methodof magnitude estimation was chosen in which participantswere free to use the range of numbers they felt comfortablewith Master scaling was applied to these individual an-noyance estimates involving a transformation function toa common master scale defined by the references whichsound levels defined the scaling context In theory thismaster scale transformation will calibrate the loudness-dependence of noise annoyance whereas the relative con-

598

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

tribution to noise annoyance from qualitative content (egthe type of sound the time pattern and cues for speed anddistance) will hopefully be unchanged

Earlier research has shown that master scaling with ref-erences works well for loudness or annoyance of a one-occasion target exposure that is when repeated exposureis unfeasible (eg experiments with long duration expo-sures) or impossible (eg questionnaire surveys in fieldstudies) an example can be found in [23] The results ob-tained from the present field experiment are probably morereliable than the results that would have been obtained bycategory scaling The test-retest reliability of the panelistsrsquomagnitude estimates of annoyance of the reference soundswas found to be very good (above 08) compared to thereliability of 072 obtained in [8] for a group of 12 muchyounger subjects Considering that our panelists all werenaiumlve participants they also each produced high qualitypsychophysical functions for the reference as discussedin Section 26

43 Other possible explanations

There are several reasons why other investigators havefound a railway bonus (for a review see [3]) which wasnot found in this field experiment One of the reasons forfinding a railway bonus for short (one minute) noises inlistening experiments may be that the relation betweenloudness and LAeq is inherently different for train androad traffic noise Indeed some researchers have arguedthat noise annoyance evaluation in listening tests of shortsounds actually is close to a perceptual loudness evalua-tion (however see [26] on differences between loudness-based and quality-based perceived annoyance) If Zwickerloudness is a good first estimate of perceptual loudnessthe difference between train noise (of different types) andhighway noise would be seen in a Zwicker loudness ver-sus LAeq plot (Figure 11) Because the IC train noise usedin the present experiment was the noise of modern ratherquiet trains a few older and noisier IC train models wereadded in this acoustic analysis At levels above 65 dB(A)TGV and maglev trains seem to be a little louder thanhighway traffic or older IC trains However this effect onZwicker loudness is not significant and does therefore notsupport a railway bonus of 5 dB(A) stipulated in severalcountriesrsquo legislation Rather it seems to be a good actionto start to replace old IC trains by new ones The railwaybonus was originally based on studies with rather old low-speed trains and with much less dense traffic intensity thannowadaysThe intermittent character of railway noise could also be

an explanation for the railway bonus However this doesnot hold for aircraft noise which is also intermittent thiscan be explained by a difference in exposure In the case ofaircraft noise the exposure is on top of buildings and on allfaccedilades In the case of road traffic noise the probability ishigh that there are local roads also but there is a possibil-ity for a ldquoquiet siderdquo people are less annoyed if quiet sidesare available [27] In the case of railway noise there is alow probability for the presence of more than one track

Zw

ickerlo

udness

[sone]

L [dB(A)]Aeq45sec

0

10

20

30

40

50

40 50 60 70 80

Figure 11 Zwicker loudness versus LAeq45s for different types oftransportation sounds ( ) IC train ( ) TGV ( ) maglev train( ) highway traffic and ( ) some additional noisier IC trains(older type)

so the exposure will also be directed at only one faccediladeIn comparing road traffic and trains the faccedilade insulationwill be more effective in the case of train noise because ofthe smaller low-frequency proportion associated to trainnoise In comparing aircraft and trains which are both in-termittent the indoor exposure is certainly more intensivefor aircraft Considering these arguments it seems obviousthat aircraft is more annoying than road traffic which ismore annoying than train However faccedilade reduction wastaken into account in the present field experiment and stillthere was no clear railway bonus found Compared withthe field condition with closed windows and the faccediladefilter used in [8] a partially open window was used in thepresent field experiment which could explain this

In surveys questioning people at their home a lowerreported annoyance for train noise compared to highwaytraffic noise was observed in a particular range of noiselevels Most of the possible explanations proposed in liter-ature conflict with the fact that this railway bonus wouldbe observed in experiments based on single passages Wemention just a few The typical character of train noise andthe concentration of the sound energy in short time inter-vals may be advantageous with regard to activity distur-bance If the level is sufficiently low the probability ofnoticing the train noise is small compared to the prob-ability of noticing the sound of a continuous source Inaddition to physical differences in the sound the ldquogreenimagerdquo of trains as a means of transportation may add tothe acceptability of the source and thus increase the tol-erance to its noise that is as long as train passages arenot too frequent However a more recent hedonic pric-ing study found that householders in Birmingham placea greater value on reductions in railway noise than in roadtraffic noise [28] Cross-cultural studies (in field and lab-oratory context) have shown that a railway bonus is notuniversal [29 30] which would favor the argument aboveIt has further been shown that the bonus varies depend-ing on the (multiple) exposure situation [31] Based on theabove only part of the effect is supposed to be visible in

599

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

field experiments such as the one reported of in this paperPart of the effect is precisely what is observed

5 Conclusion

This study has shown that in an ldquoat homerdquo like contextnoise annoyance caused by different types of trains at thesame average outdoor faccedilade exposure level is not signifi-cantly different In particular magnetic levitation systemsare not more annoying than conventional high speed trainswhich is in agreement with earlier research Noise annoy-ance caused by conventional trains was not found to besignificantly lower than annoyance caused by TGVrsquos ormaglev trains at the same average faccedilade exposure Fieldsurveys have shown that for the same average sound levelrailway noise causes less annoyance or highly annoyedpersons than highway traffic noise Although our field ex-periment included several factors that may contribute tothis effect we could not observe it

More insight may be gained by taking into account thepsychoacoustic characteristics of the noise exposure andthe relevant personal factors of the panelists This paperhas focused on discussing the experimental methodologyin great detail and on presenting the results as a functionof the average outdoor faccedilade exposure level since thisis the main noise legislation indicator used in the Nether-lands Results of a detailed psychoacoustic analysis willbe reported in a future paper

Acknowledgments

This research was financed by the Project Group Zuider-zeelijn of the Ministry of Public Transport Public Worksand Water Management in the Netherlands The membersof the project steering committeemdashGilles Janssen (dBVi-sion) Annemarie Ruysbroek (RIVM) Martin van denBerg (VROM) and Pieter Jansse (Project Group Zuider-zeelijn) mdash are acknowledged for their valuable input Wealso appreciate the experimental assistance provided byMs Ingrid Decoster Finally we would like to thank thereviewers for their valuable comments and suggestions

References

[1] J M Fields J G Walker Comparing the relationships be-tween noise level and annoyance in different surveys a rail-way vs aircraft and road traffic comparison J Sound Vib81 (1982) 51ndash80

[2] U Moehler Community response to railway noise A re-view of social surveys J Sound Vib 120 (1988) 321ndash332

[3] H M E Miedema H Vos Exposure-response relation-ships for transportation noise J Acoust Soc Am 104(1998) 3432ndash3445

[4] H Fastl G Gottschling Subjective evaluation of noiseimmissions from transrapid Proceedings of the 25thInternational Congress on Noise Control Engineering(Intermiddotnoise) Liverpool UK Juli 1996 F A Hill R Law-rence (eds)

[5] D Neugebauer J Ortscheid Geraumluschbewertung des Tran-srapid Ist der Transrapid wie die Bahn zu beurteilenBerlin Umweltbundesamt Texte 25 1997

[6] D Neugebauer J Ortscheid Geraumluschbewertung des Tran-srapid Fortschritte der Akustik mdash DAGA 97 Kiel Ger-many 1997

[7] J Vos Annoyance caused by magnetic levitation trainTransrapid 08 mdash a laboratory study TNO report TM-03-C001 The Netherlands Organisation for Applied ScientificResearch (TNO) Soesterberg The Netherlands Jan 2003

[8] J Vos Annoyance caused by the sounds of a magnetic lev-itation train J Acoust Soc Am 115 (2004) 1597ndash1608

[9] B Berglund A Preis Is perceived annoyance more sub-ject-dependent than perceived loudness Acta Acusticaunited with Acustica 83 (1997) 313ndash319

[10] M Ishibashi K Ueno H Tachibana M Watanabe Sub-jective experiment on annoying sounds in living environ-ments Proceedings of the 18th International Congress onAcoustics (ICA) Kyoto Japan 2004

[11] J Maillard J Martin P Champelovier J Lambert Percep-tive evaluation of road traffic noise inside buildings using acombined image and wave field synthesis system Proceed-ings of CFADAGArsquo04 Strasbourg France Maumlrz 2004

[12] R F S Job Community response to noise A review offactors influencing the relationship between noise exposureand reaction J Acoust Soc Am 83 (1988) 991ndash1001

[13] H M E Miedema H Vos Demographic and attitudinalfactors that modify annoyance from transportation noise JAcoust Soc Am 105 (1999) 3336ndash3344

[14] W Ellermeier M Eigenstetter K Zimmer Psychoacousticcorrelates of individual noise sensitivity J Acoust SocAm 109 (2001) 1464ndash1473

[15] D Vaumlstfjaumlll M Kleiner T Gaumlrling Affective reactions tointerior aircraft sounds Acta Acustica united with Acustica89 (2003) 693ndash701

[16] T Feng C Xiao-hong L Tan-feng A social survey onthe effects of high-speed maglev noise on residents inShangai Proceedings of The 2005 Congress and Expo-sition on Noise Control Engineering (Intermiddotnoise) Rio deJaneiro Brazil Aug 2005

[17] J J Gibson The ecological approach to visual perceptionLawrence Erlbaum Associates Inc Hillsdale New JerseyUSA 1979

[18] J G Neuhoff Ecological psychoacoustics Introductionand history ndash In Ecological Psychoacoustics J G Neuhoff(ed) Elsevier Academic Press London UK 2004

[19] C Guastavino B F G Katz J-D Polack D J LevitinD Dubois Ecological validity of soundscape reproductionActa Acustica united with Acustica 91 (2005) 333ndash341

[20] A J Berkhout D de Vries P Vogel Acoustic control bywave field synthesis J Acoust Soc Am 93 (1993) 2764ndash2778

[21] E A M Franssen J E F van Dongen J M H Ruys-broek H Vos R K Stellato Hinder door milieufactorenen de beoordeling van de leefomgeving in Nederland In-ventarisatie verstoringen 2003 RIVM report 815120 001National Institute for Public Health and the Environment(RIVM) Bilthoven The Netherlands 2004

[22] L E Marks D Algom Psychophysical scaling ndash In Mea-surement Judgment and Decision Making M H Birn-baum (ed) Academic Press New York 1998

[23] B Berglund Quality assurance in environmental psycho-physics ndash In Ratio Scaling of Psychological Magnitudes -In Honor of the Memory of S S Stevens S J BolanowskiG A Gescheider (eds) Erlbaum Hillsdale N J 1991

600

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

tribution to noise annoyance from qualitative content (egthe type of sound the time pattern and cues for speed anddistance) will hopefully be unchanged

Earlier research has shown that master scaling with ref-erences works well for loudness or annoyance of a one-occasion target exposure that is when repeated exposureis unfeasible (eg experiments with long duration expo-sures) or impossible (eg questionnaire surveys in fieldstudies) an example can be found in [23] The results ob-tained from the present field experiment are probably morereliable than the results that would have been obtained bycategory scaling The test-retest reliability of the panelistsrsquomagnitude estimates of annoyance of the reference soundswas found to be very good (above 08) compared to thereliability of 072 obtained in [8] for a group of 12 muchyounger subjects Considering that our panelists all werenaiumlve participants they also each produced high qualitypsychophysical functions for the reference as discussedin Section 26

43 Other possible explanations

There are several reasons why other investigators havefound a railway bonus (for a review see [3]) which wasnot found in this field experiment One of the reasons forfinding a railway bonus for short (one minute) noises inlistening experiments may be that the relation betweenloudness and LAeq is inherently different for train androad traffic noise Indeed some researchers have arguedthat noise annoyance evaluation in listening tests of shortsounds actually is close to a perceptual loudness evalua-tion (however see [26] on differences between loudness-based and quality-based perceived annoyance) If Zwickerloudness is a good first estimate of perceptual loudnessthe difference between train noise (of different types) andhighway noise would be seen in a Zwicker loudness ver-sus LAeq plot (Figure 11) Because the IC train noise usedin the present experiment was the noise of modern ratherquiet trains a few older and noisier IC train models wereadded in this acoustic analysis At levels above 65 dB(A)TGV and maglev trains seem to be a little louder thanhighway traffic or older IC trains However this effect onZwicker loudness is not significant and does therefore notsupport a railway bonus of 5 dB(A) stipulated in severalcountriesrsquo legislation Rather it seems to be a good actionto start to replace old IC trains by new ones The railwaybonus was originally based on studies with rather old low-speed trains and with much less dense traffic intensity thannowadaysThe intermittent character of railway noise could also be

an explanation for the railway bonus However this doesnot hold for aircraft noise which is also intermittent thiscan be explained by a difference in exposure In the case ofaircraft noise the exposure is on top of buildings and on allfaccedilades In the case of road traffic noise the probability ishigh that there are local roads also but there is a possibil-ity for a ldquoquiet siderdquo people are less annoyed if quiet sidesare available [27] In the case of railway noise there is alow probability for the presence of more than one track

Zw

ickerlo

udness

[sone]

L [dB(A)]Aeq45sec

0

10

20

30

40

50

40 50 60 70 80

Figure 11 Zwicker loudness versus LAeq45s for different types oftransportation sounds ( ) IC train ( ) TGV ( ) maglev train( ) highway traffic and ( ) some additional noisier IC trains(older type)

so the exposure will also be directed at only one faccediladeIn comparing road traffic and trains the faccedilade insulationwill be more effective in the case of train noise because ofthe smaller low-frequency proportion associated to trainnoise In comparing aircraft and trains which are both in-termittent the indoor exposure is certainly more intensivefor aircraft Considering these arguments it seems obviousthat aircraft is more annoying than road traffic which ismore annoying than train However faccedilade reduction wastaken into account in the present field experiment and stillthere was no clear railway bonus found Compared withthe field condition with closed windows and the faccediladefilter used in [8] a partially open window was used in thepresent field experiment which could explain this

In surveys questioning people at their home a lowerreported annoyance for train noise compared to highwaytraffic noise was observed in a particular range of noiselevels Most of the possible explanations proposed in liter-ature conflict with the fact that this railway bonus wouldbe observed in experiments based on single passages Wemention just a few The typical character of train noise andthe concentration of the sound energy in short time inter-vals may be advantageous with regard to activity distur-bance If the level is sufficiently low the probability ofnoticing the train noise is small compared to the prob-ability of noticing the sound of a continuous source Inaddition to physical differences in the sound the ldquogreenimagerdquo of trains as a means of transportation may add tothe acceptability of the source and thus increase the tol-erance to its noise that is as long as train passages arenot too frequent However a more recent hedonic pric-ing study found that householders in Birmingham placea greater value on reductions in railway noise than in roadtraffic noise [28] Cross-cultural studies (in field and lab-oratory context) have shown that a railway bonus is notuniversal [29 30] which would favor the argument aboveIt has further been shown that the bonus varies depend-ing on the (multiple) exposure situation [31] Based on theabove only part of the effect is supposed to be visible in

599

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

field experiments such as the one reported of in this paperPart of the effect is precisely what is observed

5 Conclusion

This study has shown that in an ldquoat homerdquo like contextnoise annoyance caused by different types of trains at thesame average outdoor faccedilade exposure level is not signifi-cantly different In particular magnetic levitation systemsare not more annoying than conventional high speed trainswhich is in agreement with earlier research Noise annoy-ance caused by conventional trains was not found to besignificantly lower than annoyance caused by TGVrsquos ormaglev trains at the same average faccedilade exposure Fieldsurveys have shown that for the same average sound levelrailway noise causes less annoyance or highly annoyedpersons than highway traffic noise Although our field ex-periment included several factors that may contribute tothis effect we could not observe it

More insight may be gained by taking into account thepsychoacoustic characteristics of the noise exposure andthe relevant personal factors of the panelists This paperhas focused on discussing the experimental methodologyin great detail and on presenting the results as a functionof the average outdoor faccedilade exposure level since thisis the main noise legislation indicator used in the Nether-lands Results of a detailed psychoacoustic analysis willbe reported in a future paper

Acknowledgments

This research was financed by the Project Group Zuider-zeelijn of the Ministry of Public Transport Public Worksand Water Management in the Netherlands The membersof the project steering committeemdashGilles Janssen (dBVi-sion) Annemarie Ruysbroek (RIVM) Martin van denBerg (VROM) and Pieter Jansse (Project Group Zuider-zeelijn) mdash are acknowledged for their valuable input Wealso appreciate the experimental assistance provided byMs Ingrid Decoster Finally we would like to thank thereviewers for their valuable comments and suggestions

References

[1] J M Fields J G Walker Comparing the relationships be-tween noise level and annoyance in different surveys a rail-way vs aircraft and road traffic comparison J Sound Vib81 (1982) 51ndash80

[2] U Moehler Community response to railway noise A re-view of social surveys J Sound Vib 120 (1988) 321ndash332

[3] H M E Miedema H Vos Exposure-response relation-ships for transportation noise J Acoust Soc Am 104(1998) 3432ndash3445

[4] H Fastl G Gottschling Subjective evaluation of noiseimmissions from transrapid Proceedings of the 25thInternational Congress on Noise Control Engineering(Intermiddotnoise) Liverpool UK Juli 1996 F A Hill R Law-rence (eds)

[5] D Neugebauer J Ortscheid Geraumluschbewertung des Tran-srapid Ist der Transrapid wie die Bahn zu beurteilenBerlin Umweltbundesamt Texte 25 1997

[6] D Neugebauer J Ortscheid Geraumluschbewertung des Tran-srapid Fortschritte der Akustik mdash DAGA 97 Kiel Ger-many 1997

[7] J Vos Annoyance caused by magnetic levitation trainTransrapid 08 mdash a laboratory study TNO report TM-03-C001 The Netherlands Organisation for Applied ScientificResearch (TNO) Soesterberg The Netherlands Jan 2003

[8] J Vos Annoyance caused by the sounds of a magnetic lev-itation train J Acoust Soc Am 115 (2004) 1597ndash1608

[9] B Berglund A Preis Is perceived annoyance more sub-ject-dependent than perceived loudness Acta Acusticaunited with Acustica 83 (1997) 313ndash319

[10] M Ishibashi K Ueno H Tachibana M Watanabe Sub-jective experiment on annoying sounds in living environ-ments Proceedings of the 18th International Congress onAcoustics (ICA) Kyoto Japan 2004

[11] J Maillard J Martin P Champelovier J Lambert Percep-tive evaluation of road traffic noise inside buildings using acombined image and wave field synthesis system Proceed-ings of CFADAGArsquo04 Strasbourg France Maumlrz 2004

[12] R F S Job Community response to noise A review offactors influencing the relationship between noise exposureand reaction J Acoust Soc Am 83 (1988) 991ndash1001

[13] H M E Miedema H Vos Demographic and attitudinalfactors that modify annoyance from transportation noise JAcoust Soc Am 105 (1999) 3336ndash3344

[14] W Ellermeier M Eigenstetter K Zimmer Psychoacousticcorrelates of individual noise sensitivity J Acoust SocAm 109 (2001) 1464ndash1473

[15] D Vaumlstfjaumlll M Kleiner T Gaumlrling Affective reactions tointerior aircraft sounds Acta Acustica united with Acustica89 (2003) 693ndash701

[16] T Feng C Xiao-hong L Tan-feng A social survey onthe effects of high-speed maglev noise on residents inShangai Proceedings of The 2005 Congress and Expo-sition on Noise Control Engineering (Intermiddotnoise) Rio deJaneiro Brazil Aug 2005

[17] J J Gibson The ecological approach to visual perceptionLawrence Erlbaum Associates Inc Hillsdale New JerseyUSA 1979

[18] J G Neuhoff Ecological psychoacoustics Introductionand history ndash In Ecological Psychoacoustics J G Neuhoff(ed) Elsevier Academic Press London UK 2004

[19] C Guastavino B F G Katz J-D Polack D J LevitinD Dubois Ecological validity of soundscape reproductionActa Acustica united with Acustica 91 (2005) 333ndash341

[20] A J Berkhout D de Vries P Vogel Acoustic control bywave field synthesis J Acoust Soc Am 93 (1993) 2764ndash2778

[21] E A M Franssen J E F van Dongen J M H Ruys-broek H Vos R K Stellato Hinder door milieufactorenen de beoordeling van de leefomgeving in Nederland In-ventarisatie verstoringen 2003 RIVM report 815120 001National Institute for Public Health and the Environment(RIVM) Bilthoven The Netherlands 2004

[22] L E Marks D Algom Psychophysical scaling ndash In Mea-surement Judgment and Decision Making M H Birn-baum (ed) Academic Press New York 1998

[23] B Berglund Quality assurance in environmental psycho-physics ndash In Ratio Scaling of Psychological Magnitudes -In Honor of the Memory of S S Stevens S J BolanowskiG A Gescheider (eds) Erlbaum Hillsdale N J 1991

600

ACTA ACUSTICA UNITED WITH ACUSTICA De Coensel et al Annoyance caused by high-speed trainsVol 93 (2007)

field experiments such as the one reported of in this paperPart of the effect is precisely what is observed

5 Conclusion

This study has shown that in an ldquoat homerdquo like contextnoise annoyance caused by different types of trains at thesame average outdoor faccedilade exposure level is not signifi-cantly different In particular magnetic levitation systemsare not more annoying than conventional high speed trainswhich is in agreement with earlier research Noise annoy-ance caused by conventional trains was not found to besignificantly lower than annoyance caused by TGVrsquos ormaglev trains at the same average faccedilade exposure Fieldsurveys have shown that for the same average sound levelrailway noise causes less annoyance or highly annoyedpersons than highway traffic noise Although our field ex-periment included several factors that may contribute tothis effect we could not observe it

More insight may be gained by taking into account thepsychoacoustic characteristics of the noise exposure andthe relevant personal factors of the panelists This paperhas focused on discussing the experimental methodologyin great detail and on presenting the results as a functionof the average outdoor faccedilade exposure level since thisis the main noise legislation indicator used in the Nether-lands Results of a detailed psychoacoustic analysis willbe reported in a future paper

Acknowledgments

This research was financed by the Project Group Zuider-zeelijn of the Ministry of Public Transport Public Worksand Water Management in the Netherlands The membersof the project steering committeemdashGilles Janssen (dBVi-sion) Annemarie Ruysbroek (RIVM) Martin van denBerg (VROM) and Pieter Jansse (Project Group Zuider-zeelijn) mdash are acknowledged for their valuable input Wealso appreciate the experimental assistance provided byMs Ingrid Decoster Finally we would like to thank thereviewers for their valuable comments and suggestions

References

[1] J M Fields J G Walker Comparing the relationships be-tween noise level and annoyance in different surveys a rail-way vs aircraft and road traffic comparison J Sound Vib81 (1982) 51ndash80

[2] U Moehler Community response to railway noise A re-view of social surveys J Sound Vib 120 (1988) 321ndash332

[3] H M E Miedema H Vos Exposure-response relation-ships for transportation noise J Acoust Soc Am 104(1998) 3432ndash3445

[4] H Fastl G Gottschling Subjective evaluation of noiseimmissions from transrapid Proceedings of the 25thInternational Congress on Noise Control Engineering(Intermiddotnoise) Liverpool UK Juli 1996 F A Hill R Law-rence (eds)

[5] D Neugebauer J Ortscheid Geraumluschbewertung des Tran-srapid Ist der Transrapid wie die Bahn zu beurteilenBerlin Umweltbundesamt Texte 25 1997

[6] D Neugebauer J Ortscheid Geraumluschbewertung des Tran-srapid Fortschritte der Akustik mdash DAGA 97 Kiel Ger-many 1997

[7] J Vos Annoyance caused by magnetic levitation trainTransrapid 08 mdash a laboratory study TNO report TM-03-C001 The Netherlands Organisation for Applied ScientificResearch (TNO) Soesterberg The Netherlands Jan 2003

[8] J Vos Annoyance caused by the sounds of a magnetic lev-itation train J Acoust Soc Am 115 (2004) 1597ndash1608

[9] B Berglund A Preis Is perceived annoyance more sub-ject-dependent than perceived loudness Acta Acusticaunited with Acustica 83 (1997) 313ndash319

[10] M Ishibashi K Ueno H Tachibana M Watanabe Sub-jective experiment on annoying sounds in living environ-ments Proceedings of the 18th International Congress onAcoustics (ICA) Kyoto Japan 2004

[11] J Maillard J Martin P Champelovier J Lambert Percep-tive evaluation of road traffic noise inside buildings using acombined image and wave field synthesis system Proceed-ings of CFADAGArsquo04 Strasbourg France Maumlrz 2004

[12] R F S Job Community response to noise A review offactors influencing the relationship between noise exposureand reaction J Acoust Soc Am 83 (1988) 991ndash1001

[13] H M E Miedema H Vos Demographic and attitudinalfactors that modify annoyance from transportation noise JAcoust Soc Am 105 (1999) 3336ndash3344

[14] W Ellermeier M Eigenstetter K Zimmer Psychoacousticcorrelates of individual noise sensitivity J Acoust SocAm 109 (2001) 1464ndash1473

[15] D Vaumlstfjaumlll M Kleiner T Gaumlrling Affective reactions tointerior aircraft sounds Acta Acustica united with Acustica89 (2003) 693ndash701

[16] T Feng C Xiao-hong L Tan-feng A social survey onthe effects of high-speed maglev noise on residents inShangai Proceedings of The 2005 Congress and Expo-sition on Noise Control Engineering (Intermiddotnoise) Rio deJaneiro Brazil Aug 2005

[17] J J Gibson The ecological approach to visual perceptionLawrence Erlbaum Associates Inc Hillsdale New JerseyUSA 1979

[18] J G Neuhoff Ecological psychoacoustics Introductionand history ndash In Ecological Psychoacoustics J G Neuhoff(ed) Elsevier Academic Press London UK 2004

[19] C Guastavino B F G Katz J-D Polack D J LevitinD Dubois Ecological validity of soundscape reproductionActa Acustica united with Acustica 91 (2005) 333ndash341

[20] A J Berkhout D de Vries P Vogel Acoustic control bywave field synthesis J Acoust Soc Am 93 (1993) 2764ndash2778

[21] E A M Franssen J E F van Dongen J M H Ruys-broek H Vos R K Stellato Hinder door milieufactorenen de beoordeling van de leefomgeving in Nederland In-ventarisatie verstoringen 2003 RIVM report 815120 001National Institute for Public Health and the Environment(RIVM) Bilthoven The Netherlands 2004

[22] L E Marks D Algom Psychophysical scaling ndash In Mea-surement Judgment and Decision Making M H Birn-baum (ed) Academic Press New York 1998

[23] B Berglund Quality assurance in environmental psycho-physics ndash In Ratio Scaling of Psychological Magnitudes -In Honor of the Memory of S S Stevens S J BolanowskiG A Gescheider (eds) Erlbaum Hillsdale N J 1991

600

De Coensel et al Annoyance caused by high-speed trains ACTA ACUSTICA UNITED WITH ACUSTICAVol 93 (2007)

[24] M Nilsson B Berglund Effects of noise from combina-tions of traffic sources Archives of the Center for SensoryResearch 6 (2001) 1ndash59

[25] B Berglund U Berglund T Lindvall A study of responsecriteria in populations exposed to aircraft noise J SoundVib 41 (1975) 33ndash39

[26] B Berglund The role of loudness as guide for commu-nity noise Proceedings of the 20th International Congresson Noise Control Engineering (Intermiddotnoise) Sydney Aus-tralia 1991

[27] E Oumlhrstroumlm A Skaringnberg H Svensson A Gidloumlf-Gun-narsson Effects of road traffic noise and the benefit of ac-cess to quietness J Sound Vib 295 (2006) 40ndash59

[28] I J Bateman B H Day L I The valuation of transport-related noise in Birmingham Technical Report Departmentof Transport London UK Sep 2004

[29] T Yano T Morihara T Sato Community response toShinkansen noise and vibration a survey in areas along theSanyo Shinkansen Line Proceedings of Forum AcusticumBudapest Hungary Aug 2005

[30] H Ma T Yano Railway bonus for noise disturbance inlaboratory settings Acoustical Science and Technology 26(2005) 258ndash266

[31] C Cremezi P E Gautier J Lambert P Champelovier An-noyance due to combined noise sources mdash advanced re-sults Proceedings of The 17th International Congress onAcoustics (ICA) Rome Italy Sep 2001

601