Novelty, cOlllplexity. and hedonic value l

D. E. BERLYNE2UNIVERSITY OF TORONTO

Two experiment.~. in which Ss wereexposed to sequences of colored shapes.investigated effects on ratings of"pleasingness" and "interestingness" ofvariables that had previously been shownto affect ratings of "novelty. " The resultsindicate. on the whole. that bothpleasingness and interestingness increasewith novelty. These findings nm counter tothose of experiments indicating an inverserelation between novelty and verballyexpressed preference. Two furtherexperiments examined effects of somevariables that might account for thisapparen t discrepancy. Homogeneoussequences declined in judged"pleasantness" more than sequences inwhich several stimuli were interspersed,and simple stimuli became less pleasant asthey became less novel, while complexstimuli declined less or became morepleasant. The findings are related tohypotheses regarding mechanisms ofhedonic value. Two crucial predictiollSwere confirmed in a fifth experiment.

Two previous experiments (Berlyne &Parham, 1968) investigated determinantsof subjective novelty and revealed severalfactors that can cause a simple.nonrepresentational, visual stimulus (acolored patch of irregular shape) to berated more or less novel. The experimentsto be reported in this paper studied effectsof some of the same factors on rated"pleasingness" and "interestingness."

Novelty is one of the "collative"stimulus properties whose importance formotivation theory is becoming more andmore apparent, especially in such areas ofresearch as exploratory behavior,experimental aesthetics, developmentalpsychology, and personality study(Berlyne, 1960, 1963a, 1966). There isnow a sizable body of experimentalliterature (e.g., Berlyne, 1963b; Munsinger& Kessen, 1964; Day, 1965; Berlyne,

OgilVie, & Parham, 1968) on verbaljudgments applied to visual patterns ofdiffering complexity, another collativeproperty. Determinants of judgedcomplexity have been identified, and theinfluence of complexity on judgments ofpleasingness (liking, preference) and ofinterestingness have been studied. Theextension of this kind of investigation tonovelty seems warranted, and two of theexperiments to be reported studied hownovelty and complexity interact indetermining how pleasing a visual patternwill be rated.

The present state of knowledge withregard to effects of novelty on preferenceis rather puzzling. Zajonc (1968) reportssome experiments of his own, and reviewsa fair number of experiments by others,that indicate an inverse relation betweennovelty and verbally expressed preference.On the other hand, Cantor (1968) reportsan experiment with children in whichverbally expressed preference increasedwith novelty. Certainly, novelty tends inmany conditions to increase theprobability or duration of self-ex posure toa . stimulus object or manipulation of it,which many would regard as amanifestation of preference or liking. Ineveryday life, there are times whensomething is more attractive thansomething else because it is more familiar,just as there are times when something isfound particularly appealing because it isnovel. Some resolution of these apparentlydivergent findings is obviously called forand was one of the objectives of thepresent investigation.

EXPERIMENTS 1AND IISubjects

Eighty Ss took part in Experiment 1 and56 in Experiment II. All wereundergraduates taking elementarypsychology courses at the University ofToronto.

Table IRepresentation of Sequence Class*

Stimulus MaterialIn both experiments, colored patches of

irregular shape on a white background werepresented in six classes of sequences, andevery S went through two sequencesbelonging to different classes in turn. Thesequences were selected from those used inthe two experiments reported in the earlierarticle (Berlyne & Parham, 1968), andincorporated variables that thoseexperiments showed to have significanteffects on judged novelty.

The sequence classes can be representedas in Table I. In Sequence Classes I and 2,X differed from Y in color only, whereas Xdiffered from Y in both color and shape inSequence Classes l' and 2'. 'Stimulidenoted by different letters differed inboth color and shape in Sequence Classes 3and 4.

Experimental Design (see Table 2)The Ss of each experiment were divided

into eight equal groups, containing as far aspossible equal numbers of males andfemales. The key to the notation can befound in Tables I and 3 of the previousarticle (Berlyne & Parham, 1968). Groups5 and 6 underwent the prefamiJiarization(F) treatment, whereas the remaininggroups did not. Prefamiliarization consistedof exposing Ss, at the outset of theexperiment, to a slide showing all thestimuli to be used in the experiment andasking them to look at the slide until theyfelt "reasonably well acquainted with therange of pictures shown here" and able torecognize them if they saw them again.

ProcedureSs of Experiment I were first required to

learn the following numerical scale: 7-verypleasing, 6-quite pleasing, 5-sJightlypleasing, 4-neither pleasing nordispleasing, 3-slightly displeasing, 2-quitedispleasing, I-very displeasing. For Ss inExperiment II, the words "interesting" and

I and I'

2 and 2'

3

Yi Y2 Y3 Y4 Ys Y6 Y7 Ya Y9 YIO Yli Yi2 XI YI3 YI4 XL

Yi Yz Xi Y3 Y4 Ys Y6 Y7 Ya Y9 YlO Yll Y12 Yl3 Yi4 XL

Yi Y2 Y3 Y4 Ys Y6 Y7 Ya Y9 YIO Yii Yi2 Yi3 YI4 YIS XL

4 A B c D E F G H K L M N o

"Subscripts attached to the same letter represent successive appearances of the same stimulu~

Perception & Psychophysics, 1970, Vol. 8 (SA) Copyright 1970, Psychonomic Journals, Inc., Austin, Texas 279

Group

2

3

4

5 (F) and 7 (NF)

6 (F) and 8 (NF)

Table 2Experiments 1 and II: Sequences*

SequencePhase Class Sequence

1 (1) R3 R3 R3 R3 R3 R3 R3 R3 R3 R3 R3 R3 Y3 R3 R3 Y3

II (2') G2 G2 l!l G2 G2 G2 G2 G2 G2 G2 G2 G2 G2 G2 G2 BI

1 (2') G2 G2 Y3 G2 G2 G2 G2 G2 G2 G2 G2 G2 G2 G2 G2 Y3

II (1) RI RI RI RI RI RI RI RI RI RI RI RI ~l RI . RI BI

1 (1') G2 G2 G2 G2 G2 G2 G2 G2 G2 G2 G2 G2 rl G2 G2 Y211 (2) RI RI l!l RI RI RI RI RI RI RI RI RI RI RI RI nI (2) R3 R3 Y3 R3 R3 R3 R3 R3 R3 R3 R3 R3 R3 R3 R3 ),lII (1') G2 G2 G2 G2 G2 G2 G2 G2 G2 G2 G2 G2 81 G2 G2 81

I (3) I I I 1 I 1 I I 1 I I 1 I I I Y3

II (4) 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 81

1 (4) I 4 5 6 7 8 9 10 11 12 13 14 15 16 17 Y3

II (3) 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 81

.. For key, see Tables 1 and 3, Berlyne & Parham, .1968F ; with prefamiliarization; NF = without prefamiliarizatlOn

"uninteresting" were substituted for"pleasing" and "displeasing." There wasthen a test for ability to supply the numbercorresponding to each judgment and viceversa. After both Sand E were satisfiedthat the scale had been mastered, the roomwas darkened, and the appropriate twosequences of stimuli were presented inturn. Each stimulus was projected for 9 secon the screen. A blank slide intervened for9 sec between the two phases, while Eannounced, "Another series will followimmediately after this." During theexposure of each stimulus, S had to assigna degree of pleasingness (Experiment I) orinterestingness (Experiment II) to it andutter the corresponding number.

ResultsThe curves in Fig. I show the results for

the different sequence classes, and Table 3sums up the statistical analyses. Included areorthogonal comparisons corresponding tovariables that were found to affectsubjective novelty in the previousinvestigation, analyses of linearcomponents of declining trends overmonotonous portions of sequences, andsign tests showing how judged pleasingnessor interestingness rose when a new stimulusappeared after a monotonouS run and sankagain when the stimulus figuring in themonotonous run reappeared.

Among the pleasingness data(Experiment I), two important interactionsinvolVing variables mentioned in Table 3were found. First, the ratings of the finalstimulus, XL, were subject to a significantinteraction between presence or absence ofprefamiJiarization (F vs NF) and thehomogeneity or heterogeneity of thesequence preceding XL (Sequence Classes 3

vs 4): F(1 ,36) = 4.97, P < .05. The meanfor Condition 4F (3.0) was markedly lowerthan the means for the other threecombinations of treatments, which wereclose together at 4.8-5.0. Secondly, therewas a significant interaction betweenphases and the prefamiliarization variablewith regard to the linear component of thetrend over the first IS stimuli in SequenceClass 3: F(1,504) = 5.75, p<.025. Thedecline was markedly less steep for the NFtreatment in Phase II than in the otherthree combinations.

In the trend analyses, the methodappropriate for equally spaced values ofthe independent variable was used. Theywere equally spaced numerically, but thequestion of whether they were equallyspaced psychologically may be raised. Inthis connection, it is worth noting Osgood,Suci, and Tannenbaum's (1957, p. 327)report of "... fairly satisfying evidencethat 7-step scales, defined by the linguisticquantifiers, 'extremely,' 'quite' and'slightly,' do yield nearly equalpsychological units in the process ofjudgment."

DiscussionIt will be seen that not all the variables

that were found to influence judgednovelty significantly in the previousinvestigation had significant effects onjudged pleasingness or interestingness.There was even a hint that interestingnessmay reach a maximum with a certaindegree of familiarity in the unexpectedfinding of a significantly higher meanrating (Experiment II) in Ss who had hadthe prefami1iarization treatment than inthose who had not (XL, Sequences 3 and4) had this treatment. Nevertheless, the

results tend strongly to favor one of thetwo divergent views mentioned earlier overthe other: Both pleasingness andinterestingness appear to increase withnovelty. There is a marked generalresemblance between the graphs in Fig. Iand those representing mean noveltyratings that were presented in the previousarticle (Berlyne & Parham, 1968). Inparticular, pleasingness and interestingness,like subjective novelty, decline withprolonged repetition of a particularstimulus, and they both rise temporarilywhen a monotonous sequence isinterrupted by a single appearance of acontrasting stimulus.

In experiments on complexity (Berlyne,1963; Day, 1965, 1967, 1968), conditionshave been found in which relatively highpleasingness coincides with relatively lowinterestingness, namely when complexity isparticularly low. But as far as novelty isconcerned, nothing of this sort seems tooccur. Pleasingness and interestingnessseem generally to go together, although thecurves in Fig. I suggest that interestingnessmay decline faster.

EXPERIMENT IIIAims

The results of Experiments I and II, likethose of Cantor's (1968) experiment,indicate a direct relation between noveltyand hedonic value, whereas theexperiments reported and reviewed byZajonc (1968) suggest an inverse relation.So we must look for differences in materialor procedure that might account for thesediscrepant outcomes.

Three possibilities suggest themselves,and some attempt was made inExperiment III to investigate all of them:

280 Perception & Psychophysics, 1970, Vol. 8 (SA)

One would expect the tedium factor to berelatively stronger when stimuli are simpleand low in information content, whereasthere will be more scope for theinformation processing presumed tounderlie the positive-habituation factorwhen stimuli are complex. Experiment III,therefore, used less complex and morecomplex stimuli in order to see if thehedonic value of the fonner would decline,and that of the latter would rise, as theylost their novelty.

(2) The stimuli used in the experimentssupporting a direct relation betweennovelty and hedonic value weremeaningless, whereas those used in theexperiments supporting inverse relationwere meaningful and symbolic, beingverbal or pictorial. In Experiment III, bothnonrepresentationl patterns andreproductions of representational paintingswere used as stimulus material. Theycan not, of course, be regarded asrepresentative samples of representationaland nonrepresentational visual patterns, sothat generalizations from any comparisonbetween their effects are hazardous. Theintention was simply to see whetherdifferent or even opposite trends mightappear with the two kinds of stimuli.

(3) In Experiments I and II (but not inCantor's experiment), one stimulus waspresented repeatedly 12 or 15 times. InZajonc's experiments. different stimuliwere exposed I, 2, 5, 10, and 25 timesbefore Ss rated them, and the stimulireceiving different numbers ofpreexposures were intenningled in thepreliminary sequence. One would expectmonotonous repetition to strengthen thetedium factor, while the appearance ofseveral stimuli in a sequence might wellmi tigate this factor and allow thepositive-habituation factor to gain theupper hand. So, in Experiment III, every Swe n t through both a homogeneoussequence, comprising 10 presentations ofone and the same stimulus, and aheterogeneous sequence of 40presentations, in which 10 presentations ofeach of four stimuli were interspersed.

PHASE H--....-.-.. .....PliASE 1

!5 6 (I ') 10, f2 '3MI!II6'7!} 2 1" !.I 16 r e 9 I() II 12 '3 t4 ~ Ifi!!

f---~~~--------i-~~-~~~--~

, • ~ 6 1 II 9 () " 12 13 14 ., 16 (,/'!} Z, 4 5 6 rIg 10 !J 12 13 14 r~ 16 !!

2' 4 - """"'----

~I-~ ------, >

I !G2~:?<:.1GO'~G'E~>G2::~'i.'C>.!~.<~;~ ?~!!GH02ijZlN!G2G2G2G2G2G2G2G2!!

~~._~._--

,

3F'~~~:

, , , , , I \ I , ; , I , ' , !? l 'I 2 2 2 'I 2 2 2 'I 'I 2 2 'I ~ !!

4F •

STIMULI

EXP II (INTERESTINGNESS)

22222< 22222 2222!.'j

f---------f- . ,

,

· PliASE I PliASE H SEQUENCE PHASE I PHASE n . ·.~',

. CLASS •

~, . , . .· I' :~ ...,~ .... ."

, ,,

"'fQlI3lQll1i1Olts~lItJiIOlt.1i1O"gfORJ~ ~~~~~~~I~~~~~~~~~

,~~~G2G2Q~G2QQQQgg.~ QQQ~QQ•••G2QG2~G2.~

, ,

· ... • .,.~

... ·- ,

~.~. -4 2' · .~., ,2 2

I IDM'glOlII5R3IQ"'ltJlt!liIOfQlOl'I:JfQ!} 'U'"t:!.1'I1lll'N'lljltlIt!ItIt'lI"""tlll"'!!,QG2~Q~G2G2QG2G2QQG2aG2~ QQ~QQaQQQG2G2QQG2Q~

, ,

· ~.• ·, . ,~.~·~ 3F ·, ,

, 2

,I I I I ," I " r' I I 1"3 2 22 2 2 Z 2: 2 2 :z 2 2 Z % ZI!

,I I I I I I J I I "" "~ 2222222221.2 'ill2,!!

, ,• ·

~~,

~~,

· 4F ·, ,, 2

, 4 ~ I'; I' I 9oQII12rJM~t'17YJ .. ,. :5 6 r I ') 10 I1I2I3Me" I ,. ,

14~'_

,~c~~1,~~6789~11~~~~~17~ 2'.~'rI9~1"21)~~~~

4

3

2

SEQUENCECLASS

EXP I (pI EASINGNESS )SEQUENCE'I~----:==-~~~r~~~~---=~~~~~;::::':::"-'-~~·_~~,--~~~~~~---,

CL ASS • PHASE I PHASE E S~~~~E.----.-

STIMULI

Fig. l. Experiments I and II: Mean ratings.

(1) The stimuli used in Experiments Iand II, like the black-and-white patternsfrom the Welsh Figure Preference Test usedby Cantor, were rather simple, whereasthose used in the experiments performedand cited by Zajonc (word!t, paralogs,Chinese ideograms, photographs of faces)were rather complex. If stimuli sometimesbecome less appealing and sometimesbecome more appealing as they becomefamiliar, this could be explained by the

interaction of two antagonistic processes,of which sometimes one prevails andsometimes the other. As a stimulus isrepeated or prolonged, its appeal maygradually succumb to a "tedium" factor.Preference may, on the other hand, risewith repetition or prolongation throughthe action of a "positive-habituation"factor. It may reflect increased liking for apattern as infonnation is absorbed from itand it becomes assimilated or organized.

SubjectsForty-eight students, for the most part

not students of psychology or of fine arts,were recruited in the corridors of theUniversity of Paris-Nanterre. Theyparticipated in the experiment in parties oftwo to six.

Stimulus MaterialFour nonrepresentational (N) patterns

were taken from a set (Berlyne, 1963b)that has been used over the years in anumber of experiments studying theeffects of complexity variables on

Perception & Psychophysics, 1970, Vol. 8 (SA) 281

B. Linear Components of Trends

Pleasingness Interestingness(Experiment I) (Experiment 11)

_ Stimuli Comparison F df p F df P

First 12 ofOverall Decline 31.09 1, 418 <.001 63.10 1,286 <.001

(1) and 0')

First 15 ofOverall Decline 195.04 I, 504 <.001 556.39 1,336 <.001

(3F) and (3NF)

First 15 ofOverall Decline 21.22 1, 504 <.001 28.29 1,336 <.001(4F) and (4NF)

Steeper Decline<.025 1,672 <.001First 15 in (3F, 3NF) than 5.13 1,1008 85.60

in (4F, 4NF)

C. Sign Tests

Pleasingness Interestingness(Experiment I) (Experiment 11)

Distribution Distribu tionSequence Comparison (Without Ties) p (Withou t Ties) p

(1, I') Xl > Preceding Y 27/8 <.01 23/3 <.010, 1') Xl> Succeeding Y 22/8 <.05 24/1 <.010, 1') XI, > Preceding Y 30/5 <.01 24/1 <.01(2, 2') Xl > Preceding Y 22/7 <.01 13/12 NS(2,2:) Xl> Succeeding Y 24/7 <.01 14/7 NS(2,2) XL> Preceding Y 35/1 <.01 25/2 <.01(3F,3NF) XL> Preceding Y 23/10 <.05 26/1 <.01

representational-complex (RC) items werecrowded canvases replete with a multitudeof human figures and other details.

One NS, one NC, one RS, and one RCitem were selected by tosses of a coin to bedesignated by the number 1 and to beelements of the homogeneous sequences.They consisted ofless complex members ofthe third pair in Category C and the morecomplex member of the first pair inCategory XC (Figs. 1 and 2, Berlyne,1963b), Ramsay's Norman McLeod, and

Table 3Experiments I and II

A. Orthogonal Dyadic Comparisons

4.97 1,36 <.05

ResultsSeparate analyses of variance were

carried out on the ratings for thehomogeneous sequence and heterogeneoussequence. The results are shown in Fig. 2.

Homogeneous sequence. The N itemswere rated higher than the R items-means= 3.7, 2.8, F(I ,40) =6.11, P < .025. Thisunexpected finding is peripheral to themain objectives of the experiment. It ispresumably connected with the fact thatcontemporary taste in painting,particularly in intellectual circles, favorsthe nonrepresentational. And such formsof representational painting as are now in

Rubens's La Kermesse. The remammgitems were designated by the number 2 andused for the heterogeneous sequence. Theyconsisted of the less complex member ofthe fourth pair in Category A, the morecomplex member of the second pair inCategory XA, Raeburn's Portrait of a Manand Rubens's Massacre of the Innocents.

Experimental Design and ProcedureSs were divided into eight groups of six.

Groups 1-4 went through a homogeneoussequence followed by the heterogeneoussequence, while Groups 5·8 had theheterogeneous sequence before thehomogeneous sequence. The homogeneoussequence comprised 10 successivepresentations of Item NS t for Groups 1and 5, of NC 1 for Groups 2 and 6, of RS 1for Groups 3 and 7, and of RC 1 for Groups4 and 8. The heterogeneous sequence wasthe same for all groups. It consisted of 10presentations each of Items NS2, NC2,RS2, and RC2, randomly intermingledwith the restriction that every item appearonce in each consecutive set of four.

Each presentation consisted of a 4-secprojection on a screen, and there was a4·sec interval between consecutivepresentations. An interval of 1 min 30 secintervened between the two phases, andthere was a break of 1 min between thefirst and the second halves of theheterogeneous sequence to allow the slidemagazine to be changed.

S had in front of him a booklet, eachpage of which bore a horizontal linedivided into seven compartments, with theword "Desagreable" (unpleasant) to its leftand the word "Agreable" (pleasant) on theright. The scale thus resembled those usedfor the semantic differential (Osgood, Suci,& Tannenbaum, 1957). After eachpresentation, S had to mark one of theseven compartments of the appropriatescale. He had been told that the scale wentfrom "extremely unpleasant" to"extremely pleasant" and that the centerrepresented the point of indifference "thatis to say, neither pleasant nor unpleasant."

4.48 1,24 < .05

17.76 1,24 <.001

Interestingness(Experiment II)

4.3

5.4

5.2

4.1

Means F df p

4.9NS

5.1

5.4NS

4.7

5.34.23 1,24 -0.5

4.6

4.9NS

5.0NS

p

NS

NS

NS

dfF

7.02 1,36 < .025

Pleasingness(Experiment I)

exploratory behavior and other processes.Two simple (NS) items and two complex(NC) items represented, respectively, thelower and upper reaches of thesubjective·complexity continuum, asscaling experiments (Day, 1965; Berlyne,Ogilvie, & Parham, 1968) have confirmed.The four representational (R) patternswere black·and·white reproductions ofpaintings. The representational·simple (RS)items were portraits of single figures on aplain background, and the

Variable Comparison Means

First or Second Xl 0, 1',2,2') 5.1

Appearance of X vsXL 0,1',2,2') 5.3

Xl 0, 1') 5.2NumbcrofYs vsBefore first X Xl (2, 2') 5.0

ReccncyofXL 0,1') 5.3

vsPrcvious X XL (2, 2') 5.3

Degree of Change XL O,2) 5.1from Y to X vs

XL 0', 2') 5.4

Homogeneous or XL (3F, 3NF) 4.9Heterogeneous vsSequence Before X XL (4F, 4NF) 4.0

Prefamiliarization XL (3F, 4F) 4.0vs

XL (3NF, 4NF) 4.9

282 Perception & Psychophysics, 1970, Vol. 8 (SA)

HOMOGENEOUS SEQUENCE

PRESENTATIONS

EXP m

AimsThe results of the three experiments

reported so far certainly cast doubt on thevalidity for all kinds of stimulus material ofZajonc's conclusion that hedonic valueincreases with familiarity. There is,however, one more factor that must beexamined. In Zajonc's experiments, Shadsimply to look at the stimulus patternsduring the familiarization phase and didnot have to record judgments until a laterphase. In our experiments, a rating had tobe made after every presentation. Thismight very well have made the proceduremore laborious and therefore more tedious.If no ratings had been required duringfamiliarization, it is possible that positivehabituation would have come to the fore,producing a rise in preference even for thesimpler stimuli as they became morefamiliar. Experiment IV was carried out tocheck this possibility.

The necessity for a judgment after everyrepresentation might have affected theratings by making Ss pay more attention tothe stimuli. To see if this was an operativefactor, half of the Ss in Experiment IVwere given instructions designed to induceclose attention, while the other half werenot, but repeated judgments were requiredof neither group during familiarization.

EXPERIMENT IV

sequences, the representational stimulimade for a steeper decline than thenonrepresentational stimuli.

SubjectsForty-eight students were recruited as in

Experiment HI. They all went through theheterogeneous sequence alone and had torate stimuli only during the first fourpresentations of the sequence (dUringwhich each of the four stimuli appearedonce) and during the last fourpresentations (during which again eachstimulus appeared once). After the firstfour presentations had been concluded,24 Ss, comprising the attention-instruction(AI) group, were told that they would seethe pictures again a number of times andthat they should look at them closelybecause they would later be questionedabout them. The next 32 presentations (8of each stimulus) then followed, and Sswere then instructed about the second setof ratings just before the last fourpresentations were given. The 24 Ss of theno-attention instruction (NAI) group weresimply told after the first fourpresentations that they would see thestimuli again a number of times.

ResultsNeither the main effect nor any of the

interactions involving the presence or

DiscussionOur hypothesis predicting a decrease in

the hedonic value of simpler patterns andan increase in that of more complexpatterns as repetition reduces noveltyreceives partial confirmation. This isprecisely what happened with the Npatterns of the heterogeneous sequence inPhase I. And in both sequences, there was asignificantly steeper decline with the Sstimuli than with the C stimuli.Nevertheless, ratings of both Sand Cstimuli declined in the heterogeneoussequence when it came second and in thehomogeneou s sequence. We must,therefore, suppose that thehomogeneity-heterogeneity variable playedsome part and that monotony favors thetedium factor. Furthermore, Ss wereapparently more susceptible to the tediumfactor during the second phase of theexperiment. This is shown by the fact thatratings declined for all stimuli of theheterogeneous sequence in Phase H,including the NC stimulus that produced arise in Phase I.

The N/R variable also proved to hilVesome influence, but it took the oppositeform to that suggested by the observationthat experiments indicating an inverserelation between novelty and hedonic valuehave used meaningful, symbolic stimuli,while those indicating a direct relation haveused meaningless stimuli. In both thehomogeneous and the heterogeneous

having the heterogeneous sequencesecond)-F = 15.35, P < .001.

There was, however, a significant tripleinteraction involving the R/N, SIC, andphase variables-F = 12.91, P < .001. Forthis reason, the results for the two phasesare plotted separately in Fig. 2. Inspectionof the curves shows that the tripleinteraction resulted from the greaterdissimilarity of trends in Phase I than inPhase II. So the data for this phase weresubjected to a separate trend analysis. InGroups 5·8, which had the heterogeneoussequence first, the overall linearcomponent was not significant. Theinteraction between the linear componentand the R/N and SIC variables wassignificant-F(I,864) = 9.65, P < .005.Data for the Rand N items were thereforeexamined separately. The linearcomponents for the RS and RC items didnot differ significantly, and there was asignificant decline over both R items takentogether-F = 7.70, P < .01. Linearcomponents differed significantly for theNS and NC items-F = 29.22, p < .00 I.The NS item produced a significant decline-F = 10.01, P < .005, while the NC itemsproduced a rise with a significant linearcomponent-F = 19.20, P < .00 I.

6 7 e g 10, 2 J •

HETEIlOGEN€OUS SEQUENCE

A-_~~=~-~--o-...o NC~~ ................ - ~~N5a

:;:

Fig. 2. Experiment III: Mean ratinp.

vogue (e.g., Pop Art, Nouvelle Figuration)are certainly very different from ourBaroque and Neoclassical R items.

A trend analysis showed the linearcomponent of the overall declining trendto be significant: F(I,432)= 52.51,P< .001. The linear component wassteeper for N items than f'" Ritems-F( 1,432) = 4.49, P < .0:' ~nd

steeper for S items than f~,r Citems-F( 1,432) = 11.61, p < ,005. Thedeclines were, however, significant for theN items alone (F = 42.94, P < .001), the Ritems alone (F = 13.03, P < .ool), the Sitems alone (F = 55.81, P < .001), and theC items alone (F = 7.25, P < .01).

None of the comparisons or interactionsinvolving the phase difference (whether thehomogeneous series occurred first orsecond) reached significance.

Heterogeneous sequence. In theheterogeneous sequence, the overall meanrating was once again higher for N itemsthan for R items-means = 4.23, 3.24;F(1,1728) =481.44, P < .00 I-and higherfor C items than for S items-means = 4.03,3.44; F = 173.82, P < .001. There was,however, a significant interaction betweenthese two dichotomies: F =97.32,P< .001. It turned out that the NC itemwas rated significantly higher than the NSitem-means = 4.7, 3.7; F = 26.56,P < .00l-but the means for the RC andRS items (3.3, 3.2) were close together.

Turning to the trend analysis, we findboth the overall linearcomponent-F(I,1794) = 31.21,p < .00 I-and the quadraticcomponent-F = 9.29, P < .005-to besignificant. The linear component wassignificantly steeper for the S items thanfor the C items-F = 10.60, P < .005. Itwas also steeper in Phase II (i.e., in groups

Perception & Psychophysics, 1970, Vol. 8 (SA) 283

oL.------I.-----------'-----'

Fig. 3. Experiment IV: Mean ratings.

EXP jy6.--------=------,

w::>...J

:i!u

~ f----t:::y==1c:=y==fc~h==t_!J:!I

AROUSAL POTENTIAL

Fig. 4. The Wundt curve.

moderate increase in arousal will berewarding and pleasant. The manner inwhich the second or arousal-increasemechanism works can be summed up by areinterpretation of the Wundt curve(Fig. 4). Here, the abscissa is taken torepresent "arousal potential," a termmeant to cover all the stimulus propertiesthat tend to raise arousal, including noveltyand complexity. Positive hedonic valuereaches a maximum with moderate arousalpotential (stimulation producing amoderate arousal increment) and then, asarousal potential increases further, hedonicvalue takes on lower and lower positivevalues and finally becomes negative.

A high degree of novelty means a highdegree of arousal potential, so that, as astimulus becomes familiar and loses itsnovelty, we must imagine ourselves movingalong the horizontal axis of the Wundtcurve from right to left.

The decreasing appeal of a simplestimulus with familiarization is easy toaccount for in terms of the curve. Astimulus that is high in novelty but low incomplexity will have medium arousalpotential (somewhere around Region B inFig. 4). Progressive loss of novelty throughrepeated or prolonged exposure wouldcorrespond to a leftward movement alongthe horizontal axis from Region B toRegion A accompanied by falling hedonicvalue. An additional mechanism is,however, apt to intervene eventually.Protracted exposure to stimulation of lowinformation content apparently leads to anincrease in at least the noncerebral indicesof arousal and to negative hedonic value(Berlyne, 1960; Schultz, 1965). Aboredom factor, conceived in this way,could explain why ratings of simple stimulisometimes end up negative.

The positive-habituation factor thatseems to preponderate with complexstimuli in at least some circumstancescould have at least two explanations, and itis difficult at this stage to know the relativeimportance of either. First, the initialimpact of a complex stimulus pattern canbe expected to engender uncertain ty,conflict, disorientation (Berlyne, 1960,1963a, 1966, in press). Opportunities for

interaction involving these variables andthe experiments variable are of in terest inview of the different results that wereobtained with Nand R patterns when theheterogeneous sequence came first inExperiment III. However, neither of theseinteractions reached significance.

GENERAL DISCUSSIONOur findings support our hypothesis that

the hedonic value of complex stimuli tendsto rise as they become less novel while theopposite holds for simple stimuli. We have,however, some evidence that monotonous,homogeneous sequences are moreconducive than varied sequences to adecline in hedonic value afterfamiliarization.

In our experiments, there p re sen t ational-n on represen t a ti onalvariable influenced ratings bu t in ways thatare rather complicated and may very wellbe peculiar to the specific kinds of stimulusused. So it would be hazardous to drawany general conclusions from these effects,except that we have no support for theconjecture stated at the beginning of thisarticle that a rise in hedonic value withfamiliarization is characteristic ofmeaningful symbolic material, while adeeline is characteristic of nonsensematerial. We have also no indication thatchanges in hedonic value take a differentcourse when Ss simply look at stimuli asthey appear from when they have to recordjudgments during familiarization or lookwith special attentiveness.

It was suggested earlier in this articlethat the apparent discrepancies amongexperimental findings might be resolved ifwe supposed that changes in hedonic valuedepended on the interaction of twoantagonistic factors, a tedium factor thatmight plausibly preponderate when asimple stimulus was encounteredrepeatedly and a positive-habituationfactor that might preponderate duringrepetition of a complex stimulus.

It may be worth our while to considerthese two processes a little further and, inparticular, to relate them to a tentativeview of the mechanisms underlyinghedonic value for which the evidence hasbeen reviewed elsewhere (Berlyne, 1967).According to this view, hedonic value (aterm meant provisionally to cover bothreward value, as judged by the capacity ofa stimulus to reinforce an instrumentalresponse, and preference or pleasure, whichis reflected in verbal evaluations) can beproduced in either of two ways. First, anycondition that drives arousal up to a highlevel is aversive, so that a stimulus thatsubsequently reduces arousal will berewarding and pleasant. Secondly, anystimulus condition that produces a

-s

FINAL

PRESENTATION

=_--_.------ C

INITIAL

PRESENTATION

~ 5

~Cl:

If)

~ 4Z.....Z<lIf)

<l 3....,...Ja..Z<l...., 2::.

absence of attention instructions reachedsignificance. So the results of the AI andNAI groups will be reported together. Themain effects of the initial-rating/final-ratingand N/R and SiC dichotomies were notsignificant. The only significantcomparisons in the analysis-of-variancetable were the N/R by S/Cinteraction-F(I,322) = 51,65,P < .001-and the comparison pertainingto the primary aim of the experiment,namely the Initial-Final by SiCinteraction-F(I ,322) = 5.84, P < .025.

The former interaction reflects the factthat the RS stimulus was rated higher thanthe RC stimulus-F(I ,322) = 15.53,P < .00l-whereas the NC stimulus wasrated higher than the NSstimulus-F =38.90, P < .001. As for thesecond of these in teractions, the mean finalrating of the C stimuli was significantlyhigher than their initialrating-F( I ,322) = 4.49, P< .05-while, inthe case of the S patterns, the final ratingwas lower than the initial rating but notsignificantly so (see Fig. 3).

Finally, the results of Experiment IVwere examined in conjunction with thosefor the heterogeneous sequence in Groups5, 6, 7, and 8 of Experiment III, Le., thegroups that had the heterogeneoussequence first. The double Experiments bySiC interaction and the triple Experimentsby Initial/Final by SiC interaction were farfrom significant. Over both experiments, theI ni tial-Final by SiC interaction wassignificant-F( I ,483) = 10.42, p < .005.The mean final rating for the S stimuli(3.3) was significantly lower than the meanin i tial rating for those stimuli(4.2)-F(I,483) =6.00, P < .025. For theC stimuli, on the other hand, the meanfinal rating (4.6) was significantly higherthan the mean initial rating(4.1)-F =6.19, p < .025.

Furthermore, the triple Initial-Final bySiC by N/R interaction and the quadruple

284 Perception & Psychophysics, 1970, Vol. 8 (SA)

TESTS

Fig. 5. Experiment V: Mean ratings.

further acquaintance with the pattern willprovide scope for absorption of additionalinformation and for the perceptual andideational processing that enablesuncertainty and conflict to be resolved aselements are discriminated, classified,recognized, and grouped together assub·wholes. This can evidently be a sourceof pleasure, presumably dependent on thearousal-reduction mechanism. However,the arousal-increase mechanism to whichthe Wundt curve is relevant is also likely tocome into play. A stimulus that is high inboth novelty and complexity will be highin arousal value and thus correspond to apoint on the horizontal axis in Region C.As it lost its novelty, the point representingits arousal potential would shift towardsRegion B accompanied by an increase inhedonic value.

Skaife's (1967) findings with auditorystimulus material are consonant with oursand open to a similar interpretation. Sheused musical sequences varying in what shecalled "complexity," but, since it involvedvarying degrees of deviation from normalmelodies, it might better be designated"surprisingness" or "uncertainty."Whatever name is applied to it, it can berecognized as a constituent of arousalpotential. When repeated over 20 days,low·"complexity" sequences that wereinitially rated pleasant gradually lost theirpleasantness, whereas high·"complexity"sequences that were initially given lowratings underwent a significant rise injudged pleasantness.

Our interpetation in terms of the Wundtcurve implies that, if familiarization hadproceeded further, ratings of the simplepatterns would have continued to decline,whereas those of the complex patternswould have climbed to a peak and thendropped. Skaife (1967) recorded such arise followed by a fall with some of hermusical sequences and some of her Ss. Sodid Alpert (I953) with repetitions of anunfamiliar rhythmic sound pattern. It

47

46

<? 45Z

~ 44Cl:(/) 43(/)W 42Z<?Z 4'i7i<1 4.0w...J0- 39Z

38<1w::< 37

0

EXP Y.

4 5 6

seemed desirable, however, to verify thesepredictions with our visual material.Experiment V was undertaken for thispurpose.

EXPERIMENT VSubjects

Forty undergraduates from elementarypsychology classes at the University ofToronto took part, two at a time.

ProcedureThe procedure was the same as for

Experiment IV, except that it wasprolonged. S went through six tests, duringeach of which the four patterns werepresented in a random order and had to berated on a 7-point Osgood scale from"displeasing" to "pleasing." Between eachtest and the next, the four patterns wereeach presented eight times in the randomorder that had been used between theinitial and final presentations inExperiment IV, but no judgments wererecorded.

ResultsThe overall mean was higher for the R

patterns (4.4) than for the N patterns(3.8): F( I,897) = 30.12, p < .00 I. Thiswas opposite to the preference recorded inExperiment m. This difference might havereflected differences in taste betweenFrench and Canadian students. It might, onthe other hand, have been due to the factthat the projector used in Toronto (aKodak Carousel) produced brighter imagesthan the projector used in Paris (aLiesegang nonautomatic), since the N slidesconsisted mostly of white background.There was also a significant interactionbetween the N/R and CISdichotomies-F(I,897) = 15.26, P < .001.The means for the SN, SR, CN, and CRpatterns were, respectively, 3.5, 4.6, 4.0,and 4.2.

To turn to the comparisons that aregermane to the chief purpose of theexperiment, the Tests by CiS interactionwas significant-F(5,897) =2.65, p < .025.But neither the Tests by N/R nor the Testsby CiS by N/R interaction approachedsignificance. Consequently, curves for thetwo S patterns and the two C patterns arepresented in Fig. 5.

Ferguson's (1965) nonparametric trendanalysis was used. Since the hypothesesunder investigation implied a decline for Spatterns and a rise followed by a declinefor C patterns, one-tailed p values wererelevant. With regard to C patterns, therewas significant bitonicity-z = 2.19,p < .015. The monotonic and tritoniccomponents were not significant. As forthe S patterns, there was both a significantmonotonic component-z = 4.30,

p < .001-and a significant bitoniccomponent-z =2.83, p < .005.

Discussio11

The wo predictions from theinterpret:- cion in terms of the Wundt curvethus received confirmation, and thisexplanation of the positive-habituation andtedium factors is therefore tenable. The Ccurve reached a maximum at the third test.After that, it dropped slowly anderratically. This may have been becauserewarding effects of perceptual processingcarne into play. The bitonic component inthe predominantly downward trend of theS curve seems to represent a flattening outthat is compatible with the shape of theWundt curve (see Fig.4 and Berlyne,1967), if we make the assumption thatarousal potential drops linearly as numberof presentations increases. The rise in the Scurve over the last three tests was far fromsignificant-z =0.52. Since the S curvesank only slightly below the indifferencepoint (4.0), boredom seems to have playedno more than a minor role.

REFERENCESALPERT, R. Perceptual determinants of affect.

Unpublished MA thesis, Wesleyan University,1953.

BERLYNE, D. E. Conflict. arousol and curiosity,New York: McGraw-Hili, 1960.

BERLYNE, D. E. Motivational problems raisedby exploratory and epistemic behavior. In S.Koch (Ed.), Psychology-A study ofa science.Vol. 5. New York: McGraw-Hili. 1963a.Pp. 284-364.

BERLYNE, D. E. Complexity and incongruityvariables as detenninants of exploratory choiceand evaluative ratings. Canadian Journal ofPsychology, 1963b, 17,274-290.

BERLYNE, D. E. Curiosity and exploration.Science, 1966, 153,25-33.

BERLYNE, D. E. Arousal and reinforcement. InD. Levine (Ed.), Nebraska Symposium onMotivation. 1967. Lincoln, Nebr: University ofNebraska Press, 1967. Pp. I-IIO.

BERLYNE, D. E. Psychobiology and aesthetics.New York: Appleton-Century-Crofts, in press.

BERLYNE, D. E., OGILVIE, J. c., & PARHAM,L. C. C. The dimensionality of visualcomplexity, interestingness and pleasingness.Canadian Journal of Psychology, 1968. 22,376-387.

BERLYNE, D. E., & PARHAM. L. C. C.Detenninants of SUbjective novelty. Perception& Psychophysics, 1968.3.415-423.

CANTOR, G. N. Children's "like-dislike" ratingsof familiarized and unfamiliarized visualstimuli. Journal of Experimental ChildPsychology. 1968,6,651-657.

DAY, H. Exploratory behavior as a function ofindividual differences and level of arousal.Unpublished PhD thesis, University ofToronto, 1965.

DAY. H. Evaluations of subjective complexity.pleasingness and interestingness for a series ofrandom polygons varying in complexity.Perception & Psychophysics, 1967, 2,281-286.

DAY, H. The importance of symmetry andcomplexity in the evaluation of com plexity,interest and pleasingness. PsychonomicScience, 1968, 10, 339-340.

Perception & Psychophysics, 1970, Vol. 8 (SA) 285

FERGUSON, G. A. Nonparametric tremfanalysis. Montreal: McGill University Press,1965.

MUNSINGER, H., & KESSEN, W. Unct:rtainty,structure and preference. PsychologicalMonographs, 1964,78, No.9.

OSGOOD, C. E., SUCI, G. J., & TANNENBAUM.P. H. The measurement of meaning. Urbana:University of Illinois Press, 1957.

SCHULTZ, D. P. Sensory restriction: E1fects onbehavior. New York: Academic Press, 1965.

SKAIFE, A. M. The role of complexity anddeviation in changing musical taste.Unpublished PhD thesis, University of Oregon.1967.

ZAJONC, R. B. Attitudinal effects of mere

exposure. Joumal of Personality & SocialPsychology Monograph Supplement, 1968,9,No. 2, Part 2.

NOTES1. Th is investigation was supported by

Research Grant APB-73 from the NationalResearch CounCIl of Canada. Experiments tlJ andIV were carried out while the author held aVisiting Professorship at the Institut d'Estbetiqueet des Sciences de I'Art, University of Paris,awarded jointly by the Scientific Affairs Divisionof NATO and the Minna-Tames-HeinemanJ'oundation. while on sabbatical leave from theUniversity of Toronto. The data of Ellperiments I

and JJ were collected by Jerry Colglazier, thoseof Experiments III and IV by Lydie Boudet, andthose of Experiment V by Mary Louise King andKathryn Davies. The author-is indebted to themfor their abTe collaboration. Experiments Ifl andIV were carried out in the Laboratory ofExperimental Psychology of the University ofParis·N3Ilterre. thanks to the cooperation ofProfessor R-. "'ranees.

2. Address: Department of Psychology,University of Toronto, Toronto. Ontario.Canada.

(Accepted for publication January 15,1970.;

286 Perception It Psychophysics, 1970, Vol. 8 (SA)