Neologistic Jargon Sparing Numbers: A Category-specific ... · NeologisticJargonSparingNumbers:A Category-specificPhonologicalImpairment LaurentCohen Hôpital de la Salpêtrière,Paris,France

Neologistic Jargon Sparing Numbers ACategory-specific Phonological Impairment

Laurent CohenHocircpital de la Salpecirctriegravere Paris France

Patrick VerstichelHocircpital Beaujon Clichy France

Stanislas DehaeneINSERM Uniteacute 334 Orsay France

We report the case of a patient suffering from a severe neologistic jargon sparingnumber words Neologisms resulted from pervasive phoneme substitutions withfrequent preservation of the overall syllabic structure (eg revolver regreveltil) Word and nonword reading as well as picture naming were equallyaffected No significant influence of frequency imageability and grammaticalclass was found In striking contrast with this severe speech impairment thepatient made virtually no phonological errors when reading aloud arabic orspelled-out numerals but made frequent word selection errors (eg 250 reg ldquofourhundred and sixtyrdquo) This observation indicates that during speech planningdifferent categories of words are processed by separable brain systems down tothe level of phoneme selection a more peripheral level than was previouslyassumed Number words may be singled out during phonological processingeither because they constitute a particular semantic category or because theybenefit from special brain mechanisms devoted to the production of ldquoautomaticspeechrdquo or because they are the elementary building blocks of speech during theproduction of complex numerals

INTRODUCTION

Recent research with brain-damaged patients as well as functional brainimaging in normal subjects has brought increasing support to the idea that

Requests for reprints should be addressed to Dr Laurent Cohen Service de Neurologie 1Pavillon Paul Castaigne Hocircpital de la Salpecirctriegravere 57 Bd de lrsquoHocircpital 75651 Paris CEDEX 13France (Tel + 33 (1) 42 16 18 49+ 33 (1) 42 16 18 02 Fax +33 (1) 44 24 52 47 E-maillaurentcohen pslap-hop-parisfr)

Oacute 1998 Psychology Press Ltd

COGNITIVE NEUROPSYCHOLOGY 1997 14 (7) 1029ndash1061

knowledge relative to different categories of words rests in separable brainregions (H Damasio Grabowski Tranel Hichwa amp Damasio 1996 Dehaene1995) As yet such categorical organisation has been shown to prevail at twobroad levels of representation First at the semantic level fragments of knowl-edge that build up the meaning of words are stored in distinct cortical locationsfor different word categories For instance naming animals differing by subtleshape features strongly activates the early visual cortex whereas namingfamiliar tools or generating action words activates areas related to movementplanning or movement perception (Martin Haxby Lalonde Wiggs amp Unger-leider 1995 Martin Wiggs Ungerleider amp Haxby 1996) Second the ana-tomical layout of the connections that bind semantic knowledge on the onehand and word forms on the other hand also obeys a systematic categoricalorganisation Thus at least in the case of nouns the word forms and thecorresponding concepts activate one another through convergent two-wayprojections to the left inferotemporal region This temporal convergence zonecan be subdivided anatomically according to conceptual domains such asfamiliar people tools etc explaining why brain lesions can yield category-spe-cific word finding or word comprehension difficulties (Caramazza 1996 HDamasio et al 1996 Verstichel Cohen amp Crochet 1996)

We report the case of a patient with severe neologistic jargon resulting fromnumerous phoneme substitutions occurring during the planning of verbaloutput The impairment thus affected a relatively late stage of speech produc-tion which might be expected to be common to all types of words Howeverthe production of numberwords was almost entirely free of phonological errorsbut displayed word selection errors of a quite different type We will argue thatsuch a category-specific sparing suggests that the cerebral lexicon has acategorical organisation down to the phonological level

Following presentation of clinical data we will try to characterise the originof phonological errors through a detailed analysis of the jargon We will thenturn to the patientrsquos performance in number processing tasks Finally we willdiscuss the bearings that such a pattern may have on models of normal speechproduction and particularly on the categorical organisation of phonologicalplanning

CASE REPORT

Medical History

The patient was a 76-year-old right-handed retired college teacher He wasadmitted to hospital because his speech had suddenly turned into an incompre-hensible jargon Furthermore he seemed not to hear what was told to him Amild right brachio-facial motor deficit was noted which receded within a dayInitial CT scan revealed the sequelae of an old overlooked right-hemispheric

1030 COHEN VERSTICHEL DEHAENE

temporo-frontal infarct A second CT scan performed 1 week later showed anadditional recent infarct affecting the posterior part of the left superior andmiddle temporal gyri (Fig 1)

Language Assessment

General Description

The patientrsquos spontaneous speech had normal fluency prosody and articu-lation Many words contained one or several phonological transformations Forinstance when attempting to say ldquoquelques anneacuteesrdquo (a few years) he producedatildete instead of ane Other neologisms were impossible to relate conclusivelyto any specific word (eg jlsquoemitrematilde de dire quelque choserdquo [I emitrematilde tosay something]) Finally there were some syntax errors as well as rare wordsubstitutions that could not be accounted for by phonemic errors Additionallythe patient suffered from a major auditory deficit that included perception ofspeech as well as nonlinguistic sounds He complained that he could hear butnot understand what was told to him Actually he behaved almost as if he weredeaf he often did not orient to sudden noises nor react to the call of his name

The patient was given a French version of the Boston Diagnostic AphasiaExamination (Table 1) Due to the auditory deficit all subtests of repetition andauditory comprehension were performed at floor level The jargon affectedidentically all subtests requiring an oral response The relatively high scores inword reading semantic fluency and naming of body parts did not take intoaccount the phonemic transformations which affected about half of the re-sponses Written naming was partially preserved but the patient producedfrequent letter errors and written neologisms Comprehension of complexwritten sentences was impaired whereas comprehension of single writtenwords was good Additionally the patient was asked to name the 26 letters ofthe alphabet presented in random order He made only one error reading R asD

The patient seemed quite unaware of most of his speech errors He some-times complained that he had made an error although his response was actuallycorrect Conversely he was sometimes quite satisfied with grossly erroneousproductions For instance when asked to name a locker (ldquocadenasrdquo kadna inFrench) he said ldquothe fobœnwa ha I was looking for that word sinceyesterdayrdquo

We observed that a given word was sometimes produced correctly andsometimes subject to phonological transformations For instance when askedto read aloud the word ldquotortuerdquo torty (tortoise) the patient said ldquola torpi un amidZatilde qui va dans les jardins dans les jardins il y a des tortues (the torpian amidZatilde which goes in the gardens there are tortoises in the gardens)rdquo Inorder to document further this variability a set of 10 pictures was presented tothe patient 10 times for oral naming each time in a different order All 10 words

NEOLOGISTIC JARGON SPARINGNUMBERS 1031

FIG1

O

utli

neof

the

pati

entrsquos

lesi

ons

aspl

otte

don

Dam

asio

and

Dam

asio

rsquos(1

989)

tem

plat

esT

hele

fthe

mis

pher

eap

pear

son

the

righ

t-ha

ndsi

deof

axia

lsec

tion

s

1032

were produced correctly at least once but were also transformed at least onceFor instance the picture of a butterfly (ldquopapillonrdquo papijotilde in French) wassuccessively named as papijotilde paplotilde piSnotilde papijotilde patijotilde papijotildepapijotilde Satijotilde Sapiljotilde pipalotilde

Comprehension of Written Words

In order to test the comprehension of written words the patient was simul-taneously presented with 42 pictures selected from the line drawings inSnodgrass and Vanderwart (1980) He was then presented successively withthe written names corresponding to 21 out of the 42 pictures and was asked tomatch each name with the corresponding drawing He responded rapidly andflawlessly He would often produce naming errors at the same time as he was

TABLE 1Boston Diagnostic Aphasia Examination

Subtest Score

FluencyArticulation rating 77Phrase length 77Automatic speechReciting 12RepetitionWords 010High-probability sentences 08Low-probability sentences 08WritingMechanics 23Serial writing 4747Written confrontation naming 810Narrative writing 24Auditory comprehensionBody-part identification 020Word discrimination 072Commands 015Complex ideational material 012NamingNaming of body-parts 3030Confrontation naming 65105Semantic fluency (animals) 2023ReadingWord reading 3030Sentence reading 010Reading comprehensionWord-picture matching 1010Sentences and paragraphs 510Symbol discrimination 910


pointing to the correct picture For instance the patient said alugre instead ofldquoaraigneacuteerdquo arerege and zarkan instead of ldquoraquetterdquo raket while correctlypointing to the pictures of a spider and of a racquet respectively In a furthertest of word comprehension the patient could rapidly and flawlessly classify25 written words into 5 semantic categories (furnitures sports animals flow-ers trees) despite frequent naming errors These findings confirmed that thecomprehension of single written words was intact

Auditory Perception of Nonverbal Stimuli

In order to assess the auditory deficit beyond the domain of speech percep-tion the patient was presented with a series of 17 typical noises (eg ringingmooing etc) which he was asked to match with the corresponding pictures(eg a bell a cow etc) He succeeded on only 917 trials When presented withpopular tunes he perceived them as music but was unable to name them or tochoose among several written titles Thus as suggested by clinical examinationthe perceptual deficit extended to nonlinguistic sounds

Summary

In brief the patient suffered from Wernickersquos aphasia with neologistic jargonin good agreement with the localisation of his left posterior superior temporallesion (Naeser Helm-Estabroks Haas Auerbach amp Srinivasan 1987) Theassociation of right and left superior temporal lesions was probably responsiblefor the additional extension of the auditory deficit beyond the linguistic domainverging on cortical deafness In the following section we will try to locatewithin the speech production system the mechanisms underlying the jargonSince the patient was unable to process auditory stimuli but understood picturesand written words normally we used only reading aloud and picture namingtasks

EXPLORATION OF THE JARGON

Before presenting experimental data it may be useful to sketch an outline ofthe functional organisation of the speech production system (for reviews seeLevelt 1989 1992) Let us consider first the successive stages that lead fromthe perception of a picture to the production of the corresponding name (Fig2) Visual analysis of the picture of a cat results in the activation of the conceptof cat Access to this conceptual level consists in the retrieval of semanticinformation (animal chases mice etc) It is then possible to select the wordldquocatrdquo and to retrieve lexical information about this word There is evidence thatsuch lexical knowledge becomes available in two successive steps At a firststage activation of the so-called word lemma gives access to syntactic charac-teristics such as grammatical class and gender (Badecker Miozzo amp Zanuttini


FIG2

Sche

mat

icde

pict

ion

ofth

esp

eech

prod

ucti

onsy

stem

as

wel

las

inpu

tto

this

syst

emfr

ompi

ctur

ean

dw

ord

perc

epti

onW

esu

gges

ttha

tthe

pati

entrsquos

jarg

onre

sult

edfr

omim

pair

edpr

ojec

tions

from

the

lexe

me

leve

lto

the

phon

emic

leve

l

1035

1995 Schriefers Meyer amp Levelt 1990) The lemma may be viewed as themost abstract representation of the word and constitutes a convergence hub toand from stored word forms in various modalities auditory and visual inputlexicons phonological and orthographic output lexicons

Access to the phonological output lexicon constitutes the second stage oflexical access during word production It consists in activating the so-calledword lexeme that is the phonological features of the word that are necessaryfor oral output Lexemes may be themselves broken down into two distincttypes of information (1) the segmental composition of the word (eg thephonemes k a and t) and (2) its metrical structure including in particularthe number of syllables and possibly the consonant-vowel structure (Levelt1992 Sevald Dell amp Cole 1995) The idea that segments and the structure areboth stored although as separate records is supported by evidence from thestudy of spontaneous or elicited speech errors (eg Meyer 1992 Shattuck-Hufnagel 1979 Treiman 1983) as well as from recent on-line studies ofspeech production (eg Sevald et al 1995) Eventually the phonemic andstructural descriptions are combined giving access to the corresponding sylla-bles (eg kat) that constitute the building blocks of the articulatory program(Levelt amp Wheeldon 1994)

Written words can be read aloud in a similar fashion through a lexical routeFollowing processing of their constituent graphemes words are identified inthe visual input lexicon giving access to the corresponding lemma and thenceto the subsequent stages of verbal output Words can also be read aloud througha so-called surface nonlexical route This route maps any orthographicallylegal letter string into the corresponding sequence of phonemes based onlanguage-specific statistical regularities This surface route is essential fornaming nonwords which by definition have no stored semantic or lexicalrepresentations

Speech errors may result from dysfunctions affecting any one of these manyprocessing stages Leaving aside purely articulatory processes the translationfrom the conceptual level to oral output may be subject to two main sources oferror First faulty access to the lemma or lexeme levels may result in word-finding difficulties or in the selection of inappropriate words Second impairedprocessing at the phonemic or syllabic levels may result in phonologicaltransformations of the targets resulting in phonological jargon Needless tosay both mechanisms may be combined within a single trial resulting inphonological transformations of an incorrectly selected word (Butterworth1979)

Method

Over several testing sessions the patient was asked to name a total of 192pictures of simple objects from various semantic categories selected from the


line drawings inSnodgrass and Vanderwart (1980) and to read aloud 502 wordsand 192 nonwords Words were varied in frequency length imageability andgrammatical class

Frequency and imageability ratings were taken from the BRULEX Frenchlexical database (Content Mousty amp Radeau 1990) Frequency was measuredby the log10 of the number of occurrences by million The targets of the picturenaming task were all nouns Nouns also constituted a large majority of the wordreading material (840) whereas verbs adjectives and closed-class wordsrepresented 35 77 and 48 of targets respectively All nonwords wereorthographically and phonologically legal letter strings Targets are furtherdescribed in terms of frequency length and lexical neighbourhood in Table 2

The patientrsquos responses were transcribed by the examiner and tape-recordedfor subsequent checking of the transcription When the patient produced severalresponses to a given stimulus which occurred exceptionally only the first onewas considered The following analyses were run on phonetic transcriptions ofthe targets and responses

General Distribution of Errors

In agreement with clinical observations the patient produced only phonologi-cally and phonetically well-formed answers This allowed for a simple classi-fication of responses into four exclusive categories correct responses absence

TABLE 2Description of the Target Words and Nonwords Used in

Naming Tasks

Mean SD Min Max

Pictures (N = 192)Frequency

a(N = 178) 111 078 ndash11 295

Number of phonemes 456 174 20 1200Number of syllables 184 082 10 500Number of neighbours

b11180 1229 00 43400

Words (N = 502)Frequency

a(N = 480) 151 094 ndash11 381


b11000 1229 00 43400

Nonwords (N = 192)Number of phonemes 429 171 20 1100Number of syllables 172 086 10 400Number of neighbours

b10400 9260 00 43200

aLog10 of the frequency per millionbNumber of words of thesame length differing by one ortwo phonemes


of response verbal errors and neologisms Verbal errors and neologismsconsisted of inappropriate words and nonwords respectively

Overview of the Results

The patient made a total of 630886 (711) errors As shown in Table 3the error rate did not differ significantly across the three types of stimuli [ c 2(2)= 504 P = 08] In all three tasks neologisms were the most frequent type ofresponse (522886 589) and by far the most frequent type of error (522630829) Furthermore as illustrated in Table 4 a number of verbal errorsprobably resulted from phonemic errors accidentally producing real words (egthe picture of a fox was named as rœtar [delay] instead of rœnar) Thusphonemic errors were even more pervasive than the rate of neologisms would

TABLE 3Distribution of Error Types

Entire Corpusmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdash

Pictures Words Nonwords(N = 192) (N = 502) (N = 192)

Correct responses 66 (344) 131 (261) 59 (307)Errors 126 (656) 371 (739) 133 (693)

No response 9 (71) 6 (16) 1 (08)Verbal paraphasia 21 (167) 56 (151) 15 (113)Neologistic error 96 (762) 309 (833) 117 (880)

Matched Picturesand Words

mdashmdashmdashmdashmdashmdashmdashmdashmdashmdashndashPictures Words(N = 88) (N = 88)

Correct responses 31 (352) 22 (250)Errors 57 (648) 66 (750)

No response 1 (18) 0 (00)Verbal paraphasia 13 (228) 11 (167)Neologistic error 43 (754) 55 (833)

Matched Wordsand Nonwords

mdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashWords Nonwords

(N = 192) (N = 192)




suggest This issue will be further discussed in the section devoted to verbalerrors

Comparison of the patientrsquos behaviour with the different types of stimulimay have important bearings on the functional interpretation of the deficit Wetherefore performed comparisons restricted to matched subsets of the stimuliPicture naming and word naming were compared over a set of 88 nouns thatwere used in both tasks (Table 3) There was no difference in error rate [ c 2(1)= 173 P = 19] nor in the distribution of error types [ c 2(2) = 199 P = 37]We may note that with the 4688 words that were produced incorrectly in bothtasks the two errors were always different Word reading and nonword readingwere similarly compared over a set of 192 words and 192 nonwords matchedfor syllabic length (see Table 3) There was again no significant difference inerror rate [ c 2(1) = 195 P = 16] nor in the distribution of error types [ c 2(2) =518 P = 08]

Length

We expected a priori that the longer a target the more numerous thephonological transformations that it would suffer Indeed error rate wassignificantly correlated with length as measured by the number of phonemes[r(884) = 31 P lt 00001]1 This correlation prevailed across pictures [r(190)= 23 P = 0014] words [r(500) = 34 P lt 00001] and nonwords [r(190) =29 P = 00004]

Number of Lexical Neighbours

Phonological transformations can occasionally result in the production of areal word All targets are not equally exposed to this phenomenon For instancechanging one phoneme of the french word bal (or of the nonword nal) is

TABLE 4Examples of Verbal Errors Phonologically Close to

the Target

Stimulus Target Error

Picture fox rœnar rœtar (delay)Word boicircte (box) bwat pwE~t (tip)Nonword ulore ylor alor (then)

1The curve of error rate as a function of lengthwas well fittedby acumulativebinomial function

[ c2(12) = 104 P = 58] as may be expected assuming that the error probability on a given item

roughly reflected the probability of independent errors on each phoneme


likely to result in a real word such as cal dal gal mal pal sal andmany others whereas changing one phoneme of the word ldquomanivellerdquomanivel wonrsquot yield a single real word We therefore expected that the relativeproportion of errors resulting in a real word would be higher the larger thenumber of neighbours We computed the number of neighbours for all targetsin the corpus Neighbours were operationally defined as words comprising thesame number of phonemes as the target and differing from it by one or twophonemes As expected the relative proportion of verbal errors was signifi-cantly correlated with the number of neighbours [r(628) = 39 P lt 00001]This correlation prevailed across pictures [r(124) = 27 P = 0024] words[r(369) = 44 P lt 00001 and nonwords [r(131) = 33 P = 00011]

We thus found that length and lexical neighbourhood had the expectedinfluence on the overall error rate and on the relative rate of verbal errorsrespectively We then examined the influence of other possibly relevant pa-rameters for which we had no a priori expectancies frequency imageabilityand grammatical class

Frequency

We studied the influence of word frequency on the pattern of errors duringpicture naming and word reading This analysis was restricted to 178 picturetrials and 480 word trials for which a frequency rating was available (Contentet al 1990)

Error rate was significantly lower the higher the frequency [r(656) = ndash 15P = 00016] However frequency was strongly correlated with length thelonger a word the lower its frequency [r(656) = ndash 37 P lt 0001] The apparentfrequency effect could thus be an artefact of the length effect described earlierIn order to clarify this point we examined whether any frequency effect wouldpersist when keeping length constant The effect of frequency was thus studiedseparately for words of 3 4 5 and 6 phonemes (N = 155 173 102 and 66respectively) Words shorter than three or longer than six phonemes were fewand insufficiently distributed across the frequency range to be relevant in thisanalysis There was no significant correlation for any of the length classes (allP lt 13) Thus when keeping length constant the overall error rate did not varywith frequency

The relative proportion of verbal errors was also correlated with frequency[r(461) = 24 P lt 0001] However frequency was strongly correlated with thenumber of neighbours [r(461) = 37 P lt 0001] We therefore looked to see ifany frequency effect would persist when keeping the number of neighboursapproximately constant The 463 error trials were divided into 4 quartilesdepending on the number of lexical neighbours of the target word and the effectof frequency on the relative rate of verbal errors was studied separately for the4 quartiles There was no significant correlation for any of the neighbourhood


classes (all P gt 10) Thus when keeping neighbourhood size constant therelative rate of verbal errors did not vary with frequency

Imageability

We studied the influence of word imageability on the pattern of errors duringword reading This analysis was restricted to 214 trials for which an imageabil-ity rating was available (Content et al 1990)

There was no significant difference of overall error rate [ c 2(1) = 143 P =23] nor of relative proportion of verbal errors [ c 2(1) = 23 P = 13] betweenhigh- and low-imageability words (N = 100 and 114 respectively)

Grammatical Class

The relatively small numbers of items in grammatical categories other thannouns make detailed comparisons difficult However post hoc analyses usingc 2 tests revealed no significant differences in error rate and no significantdifferences in the relative proportions of verbal errors and neologisms whencomparing closed-class with open-class words and when comparing the threetypes of open-class words (ie nouns adjectives and verbs) among them (allP gt 14) Similaranalyses were performed on a controlled subset of the materialcomprising 28 open-class and 28 closed-class words matched for length andfrequency (Segui Mehler Frauenfelder amp Morton 1982) There were againno differences between open- and closed-class words (all P gt 17)

Summary

Word reading nonword reading and picture naming yielded a similarpattern of errors with a massive predominance of neologisms Longer targetsproduced more errors whereas the number of lexical neighbours was a goodpredictor of the occurrence of verbal errors No significant influence of fre-quency imageability or grammatical class was found

Phonological Features of Neologisms

As mentioned in the clinical description neologisms covered the whole rangefrom single phoneme substitutions to neologisms very remote from the target(Table 5) However we observed that even the most abstruse neologisms oftenhad some similarity with the target word Even when a majority of phonemeswere transformed the similarity was still apparent in terms of overall lengthand syllabic structure This informal observation was evaluated by post hocanalyses of the corpus of 522 neologisms


Phonological Frame

Number of syllables Targets and neologisms were strongly correlated interms of number of syllables [r(520) = 87 P lt 0001] As many as 413522(791) neologisms had exactly the same number of syllables as thecorresponding targets while 100522 (192) differed by 1 syllable and only9522 (17) by 2 syllables Table 6 shows the distribution of neologisticerrors as a function of the length of the target and of the error A similarcorrelation between targets and neologisms prevailed with pictures wordsand nonwords (all P lt 0001) Analyses restricted to the matched subsets ofstimuli described earlier indicated that the proportion of neologisms with acorrect number of syllables was not different when naming pictures as com-pared to words [674 vs 818 c 2(1) = 270 P = 10] and when namingwords as compared to nonwords [804 vs 889 c 2(1) = 291 P = 09] Inbrief about 80 of neologisms had a correct number of syllables whateverthe type of stimuli

TABLE 5Examples of Neologisms


Phonologically closea

Picture necklace kolje kaljeWord eacutetoile (star) etwal etwanNonword tomu tomy comy

Phonologically remoteb

Picture accordion akordeotilde bormegatildeWord toiture (roof) twatyr patildepylNonword molitude molityd politotilded

aNeologisms phonologically close to the target

bNeologisms phonologically remote from the target

TABLE 6Preservation of the Number of Syllables in Neologisms

ResponseTarget 1 2 3 4 5 6 Total

1 99 32 3 1342 5 180 21 1 2073 9 84 11 2 1064 21 49 1 715 2 1 1 46 0Total 104 221 131 62 3 1 522


Syllabic structure In order to study the transformation of syllabic struc-tures we analysed the corpus of 912 syllables corresponding to the 413 trialson which target and neologism had the same number of syllables (see Table 7)Note that 224 of these neologisms (542) not only had the same number ofsyllables as the target but were entirely correct in terms of alternation ofconsonants and vowels Syllabic structure was preserved in 685912 (751)syllables The preservation of syllabic structure was not restricted to thesimplest and most frequent CV and CVC structures (eg a belt was namedfotildeter instead of sE~tyr [CV-CVC]) but also prevailed with complex syllableswith consonant clusters even when most phonemes were erroneous Forinstance the word ldquolivrerdquo (book) was read as pert instead of livr (CVCC)and the word ldquonormerdquo (norm) was read as diln instead of norm (CVCC)

Analyses restricted to the matched subsets of stimuli indicated that syllabicstructure was preserved somewhat more often for words as compared topictures [835 vs 679 c 2(1) = 488 P = 027] whereas there was nodifference between words and nonwords [745 vs 804 c 2(1) = 174 P =19]

We observed that whenever the patient produced an erroneous syllabicstructure he often selected syllabic structures that are frequent in French andmore rarely syllabic structures that are infrequent in French There was thus astrong correlation between the frequency of the various syllabic structures inthe language and in the patientrsquos errors [r(11) = 98 P lt 0001]2 For instanceCV and CVC syllables which are very frequent in French (501 and 195of syllables in our corpus of target words) constituted as much as 471 and241 of the erroneous structures produced by the patient

TABLE 7Preservation of Syllabic Structure in Neologisms

Responsemdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdash

Target CV CVC V CYV CCV CCVC CVCC Other Total

CV 393 34 4 16 7 3 457CVC 37 127 3 3 2 4 2 178V 16 58 1 3 78CYV 25 2 21 2 50CCV 20 1 2 24 2 49CCVC 2 1 18 1 1 23CVCC 4 1 16 2 23Others 9 11 3 1 2 28 54Total 500 181 65 42 38 22 23 41

2In this analysis the frequencies of the various syllabic structures in French were estimated on

the basis of the corpus of target words


Phonemic Content

In order to identify factors governing phoneme substitutions we analysedthe set of 224 neologistic errors on which the syllabic structure of the wholeword was preserved A total of 5611068 (525) phonemes were correct Thesame analysis performed with the matched subsets of stimuli showed that theproportion of correct phonemes did not differ between pictures and words(508 vs 507) nor between words and nonwords (500 vs 555)

There was no tendency for consonants to be substituted by consonants of thesame type (Table 8) When selecting an inappropriate consonant the patientshowed a decreasing preference for stops fricatives liquids and nasals irre-spective of the type of target consonant [ c 2(9) = 836 P = 50] This systematicbias closely followed the frequency distribution of consonant types in thelanguage as reflected in the composition of the corpus [r(2)= 972 P = 028]

As with consonants no systematic relationship was found between targetand response vowels The patient showed an idiosyncratic preference for nasalvowels (atilde otilde E~) which represented 186of target vowels but as much as403of vowel substitutions He also tended to avoid high vowels (i y u)which represented 309 of target vowels but only 141 of vowel substitu-tions

Finally we observed occasional perseverations of a given phoneme over aseries of consecutive trials For instance while the patient was reading aloud alist including words and nonwords the phoneme d was erroneously producedin 27 out of 38 consecutive trials associated with a variety of vowels andsyllabic structures

Summary

These analyses clearly confirmed the informal observation that neologismswere phonologically related to the corresponding targets In particular evenwhen the phonemic composition of neologisms was grossly incorrect theoverall phonological frame as measured by the number of syllables and theCVC structure was generally preserved We also noted that the selection of

TABLE 8Substitutions of Consonant Types in Neologisms

Responsemdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashmdashndash

Target Stops Fricat Liquids Nasals Total

stops 57 32 18 13 120fricatives 35 25 16 7 83liquids 26 20 9 13 68nasals 18 8 8 8 42Total 136 85 51 41 313


erroneous consonants and erroneous syllable structures followed the frequencydistributions of the French language This amounts to saying that the patientselected erroneous consonants and syllable structures randomly but still re-flecting French statistical regularities

Verbal Errors

In an attempt to characterise the patientrsquos phonological encoding impairmenton as well defined an empirical basis as possible neologisms were analysedseparately and verbal errors put aside However there is evidence that verbalerrors resulted from exactly the same phonological encoding deficit as neolo-gisms This hypothesis is already supported by the correlation between the rateof verbal errors and the number of lexical neighbours of the targets Further-more on informal inspection verbal errors were often phonologically close tothe targets (Table 7) In order to confirm this impression we resorted to moreformal analyses and showed that verbal errors had exactly the same degree ofphonological proximity to the targets as did neologisms

1 Targets and verbal errors were highly correlated in their number ofsyllables [r(90) = 71 P lt 0001] The percentage of errors with a correctnumber of syllables was as high with verbal errors as with neologisms [7892(848) vs 413522 (791) c 2(1) = 157 P = 21]

2 The percentage of errors in which the syllabic structure of the target wasfully preserved was not different with verbal errors and with neologisms [4592(489) vs 224522 (429) c 2(1) = 114 P = 28]

3 The percentage of correct phonemes in errors preserving syllabic struc-ture was not different with verbal errors and with neologisms [73151 (483)vs 5611068 (525) c 2(1) = 93 P = 34]

This suggests that the distinction we made between neologisms and verbalerrors was simply descriptive and did not reflect distinct pathophysiologicalmechanisms (for discussion of a contrasting case see Best 1996)

Discussion of the Jargon

The data clarify the functional localisation of the patientrsquos impairment withinthe speech production system The fact that there was no effect of imageabilitylexical status (word or nonword) frequency or grammatical word class on errorrates suggests a low-level impairment We have also seen that the pattern oferrors was similar with pictures words and nonwords This suggests that errorsoriginated downstream from the point of convergence of the three tasks Themost straightforward account would be a deficit in activating phonemes at arelatively late stage of speech planning common to all types of targets (seeGeneral Discussion)


The analysis did however reveal some subtle differences between picturenaming and reading the phonological similarity between targets and neolo-gisms was slightly weaker with pictures than with written words (674 vs818 of neologisms with a correct number of syllables 679 vs 824 ofsyllables with a correct structure) This finding can be readily explained by theexistence of two routes for word reading During word reading the phonemicoutput is conditioned both by the surface route that maps graphemes intophonemes and by the ldquodeeprdquo reading route in which phonemes are activatedfrom lexeme nodes In contrast the surface route is not involved in picturenaming The phonemic output is therefore less constrained and expected todiffer more widely from the expected target during picture naming than duringword reading

Did the patient also suffer from an additional impairment of word selectionAlthough it is possible that some neologisms resulted from phonologicaltransformations affecting inappropriate words we found little positive evi-dence for such word selection difficulties Word substitutions in aphasicpatients often result in semantically related responses (Kohn amp Goodglass1985) Here however most verbal errors could be explained by neighbour-hood effects in which the random substitution of a phoneme in a short word islikely to result in the production of another word rather than a neologism Aplausible meaning relationship between the target word and the patientrsquosresponse was discernible for only 1292 verbal errors Moreover these 12errors possibly included visual confusions between very similar objects inpicture naming (eg shirt reg coat orange reg apple cloud reg cake) as well asphonological transformations (eg a glove was named as dE~ [suede] insteadof gatilde) On the whole whatever the mechanism underlying such ambiguouserrors errors imputable to semantic proximity were exceptional contrary towhat could be expected if the patient suffered from a significant impairment ofword selection

EXPLORATION OF NUMBER PROCESSING

During informal clinical evaluation we observed that whenever the patientwas reading arabic numerals aloud his speech seemed free of phonemicerrors We felt that if confirmed this surprising observation could contributeto the understanding of the mechanism of the jargon as well as reveal disso-ciations between distinct word categories within the speech production sys-tem We will now turn to a study of the patientrsquos performance in numberprocessing tasks and particularly in simple reading tasks requiring oral outputof number names


General Features

Reading Aloud Arabic Numerals

The patient was asked to read aloud 470 arabic numerals from 1 to 6 digitslong The most striking observation was that his production was essentially freeof all phonological errors (6 out of 470 trials eg 18 reg dez) in sharp contrastwith his usual speech production All phonological errors are listed in Table 9The patient made a total of 213470 (453) errors of which a large majority(190213 892) consisted of legal numerals Furthermore these legal re-sponses had generally the same length as the corresponding targets (145190763) single digits were often read as single digits two-digit numerals astwo-digit numerals etc The remaining 23 errors included the 6 phonologicalerrors 11 illegal numerals (eg 47 reg ldquofourteen sevenrdquo) some of which wereprobably self-corrections (eg 92 reg ldquoquatre quatre-vingt douzerdquo) 4 trans-formations into multiplication problems (eg 94 reg ldquonine times fourrdquo) and 2failures to respond

The mechanisms that allow one to transcode arabic numerals into strings ofwords may be broken down into two separate subprocesses (1) The generationof a word frame that specifies the sequence of word classes to produce Forexample the word frame for 235 would be stated informally ldquoa ones wordfollowed by lsquohundredrsquo then a tens word and finally a ones wordrdquo (2) Theretrieval of the appropriate words within the specified classes on the basis ofthe identity of digits (eg 2 reg laquotworaquo 3 reg laquothirtyraquo etc) There is goodneuropsychological evidence that the generation of the word frame and theselection of specific words can be impaired independently from one anotheryielding respectively frame (or syntactic) errors (eg 54 reg ldquofive hundred and

TABLE 9List of All Phonological Errors in Number Reading

Target Error

Arabic numerals 62 kuz deux(N = 470) 74 klE~fo

902 huit fatilde deux18 dez

601 six mel treize421 trois cent dotildez

Spelled-out numerals soixante quatorze six zatilde quatorze(N = 197) trois cent cinq quatre ka cinq

soixante seksatildetquarante twaratildetquatorze karotildeZsix mille treize six mel treize


fourrdquo) and lexical errors (54 reg ldquofifty threerdquo) (Cohen amp Dehaene 1991McCloskey Sokol amp Goodman 1986) The patientrsquos 190 legal errors includedboth frequent frame errors (eg 74 reg 704) and frequent lexical errors (eg 250reg 460) sometimes associated within a single trial (eg 687 reg 6714)attesting the disruption of multiple components of the transcodingmechanisms

Reading Aloud Spelled-out Numerals

The patient was asked to read aloud 197 spelled-out numerals from 1 to 7words long Just as in arabic number reading a striking feature of his behaviourwas the almost complete absence of phonological errors (6 out of 197 trials)For instance ldquoquaranterdquo was read as twaratildet instead of karatildet All phonologicalerrors are listed in Table 9

The patient made a total of 138197 (701) errors Almost all errors(130138 942) consisted of phonologically well-formed number words Theremaining eight errors included six phonological errors and two transforma-tions into multiplication problems (eg ldquoquarante sept mille treizerdquo reg ldquoseptfois cinq trenterdquo [seven times five thirty])

With numerals comprising several words the patient resorted to a word-by-word reading strategy and therefore most errors (104130 800) comprisedthe correct number of words However since he paid little attention to thegeneral structure of the target many responses formed illegal word strings Forinstance ldquomille neuf cent quatorzerdquo (1914) was read as ldquohuit quatre neufquatorzerdquo (eight four nine fourteen) French compound number words such asldquodix-neuf rdquo or ldquoquatre-vingtrdquo were printed with no dash and the patient treatedthe two components as independent words as evidenced by the occurrence oferrors such as ldquodix huitrdquo (18) reg ldquosept quinzerdquo (seven fifteen) Considering thecorpus of 104 error trials on which the number of words of the response wascorrect we noted that the patient did not substitute words with other words fromthe same class but showed a systematic response bias He had a decreasingpreference for ones words (139203 685) teens words (54203 266) andtens words (4203 20) This pattern prevailed whatever the type of targetincluding multipliers ldquohundredrdquo and ldquothousandrdquo

Thus when he was reading aloud arabic or spelled-out numerals and incontrast with his ordinary behaviour the patientrsquos speech was free ofphonological transformations Nonetheless he made frequent reading errorsIn order to determine whether these reading errors resulted from an impairedcomprehension of the stimuli or from an impaired visual-to-verbal mappingwe then assessed the patientrsquos comprehension of written numerals


Number Comprehension

Understanding meaningful arabic numerals The list of arabic stimuli thatthe patient was asked to name as reported before included numerals with aspecial meaning such as famous dates brands of cars etc (Cohen Dehaeneamp Verstichel 1994) When naming such numerals the patients would occa-sionally provide accurate semantic comments even when reading aloud waserroneous For instance when presented with the numeral ldquo1789rdquo he saidldquo1989 the French Revolutionrdquo With ldquo1939rdquo he said ldquo1839 itrsquos the warrdquo withldquo1515rdquo he said ldquo1815 Francis the Firstrdquo with ldquo1918rdquo he said ldquo1418 the endof World War Irdquo with ldquo13rdquo he said ldquo713 it brings luckrdquo He never producedinappropriate comments with numerals carrying no such special meaning

Magnitude comparison In order to assess further the patientrsquos comprehen-sion of numbers he was presented with 39 pairs of arabic numerals from 1 to6 digits long (the two numerals in each pair had the same number of digits)and with 37 pairs of spelled-out numerals (the two numerals in each pair hadthe same number of words) The patient was asked simply to point to the largernumeral in each pair but he often tried to name the stimuli at the same time

He made 139 (26) comparison errors with arabic numerals and 237(54) comparison errors with spelled-out numerals This excellent perform-ance contrasted with the numerous naming errors that the patient produced atthe same time as he was readily pointing to the larger number For instancewhen presented with the pair ldquo14 17rdquo the patient read ldquo29 18rdquo while correctlypointing to 17 Similarly ldquo14 13rdquo was read as ldquo14 15rdquo ldquo50 40rdquo as ldquo50 60rdquoldquo951 954rdquo as ldquo612 611rdquo while the patient always pointed correctly to the largernumeral The same phenomenon was observed with spelled-out numeralsThus the patient read ldquotrente (30) cinquante (50)rdquo as ldquo17 16rdquo while pointingto ldquocinquanterdquo With numerals comprising several words he occasionallyuttered illegal word strings as shown before while at the same time comparingthe two numerals correctly

Calculation The patient was presented visually with 24 arithmetic prob-lems involving 1- and 2-digit numerals (8 addition 8 subtraction and 8multiplication problems) and was asked to write down the result He made atotal of 524 errors He accurately solved relatively difficult problems such as8 times 7 50 ndash 34 8 + 27 or 31 + 12 It may be speculated that at least somecalculation errors actually resulted from errors in transcoding the arabic oper-ands into words (eg 2 times 2 reg 8 8 times 3 reg 12 8 ndash 3 reg 7) For instance thepatient may covertly have read 8 times 3 as ldquofour times threerdquo hence retrieving theerroneous result ldquotwelverdquo from arithmetic memory (Dehaene amp Cohen 1997)

In order to test calculation abilities without requiring the overt productionof any numerals the patient was presented visually with 35 simple arithmetic


problems (12 addition 11 subtraction and 12 multiplication problems) Theproposed result was erroneous in 23 problems (eg 7 times 6 = 40) and the correctresult in 12 problems (eg 6 times 8 = 48) The patient was asked to decide whethereach problem was correct or false He did not make a single error3

In conclusion the patientrsquos good performance in this group of tasks tappingnumber comprehension demonstrates that he could understand arabic andspelled-out numerals much more accurately than suggested by his poor readingperformance As a consequence we may conclude that naming errors resultedfrom an impaired selection of number words for verbal output on the basis ofa correct analysis of the input

Discussion of Number Processing

The Absence of Phonological Errors in NumberReading

The most striking feature of number reading was the almost completeabsence of phonological errors Could this dissociation be an artefact reflectingsome superficial and systematic difference between number words and otherwords One possibility is that number words are generally high-frequencywords a feature that might account for their sparing by phonological transfor-mations This hypothesis does not seem very likely given that frequency hadno influence on the rate of phonological errors in reading non-numerical wordsStill in order to evaluate it we selected for each number word all thenon-numerical trials with a target word that (1) had the same number ofphonemes as the considered number word and (2) had a strictly higherfrequency As is apparent from Table 10 those selected non-numerical wordswere subject to many neologistic errors whereas as shown before the corre-sponding number words were essentially free of such errors4 This conservativeanalysis allows us to conclude that number words were not spared simply byvirtue of their relatively high frequency In fact the only parameter that wasfound to have a significant influence on error rate during word reading is target

3The patientrsquos relatively good performance on the verification task may seem to be at odds

with our hypothesis (Dehaene amp Cohen 1995 1997) that the knowledge of arithmetic facts ispartly based on a verbal coding of the operands However we explicitly postulate that only alimited subset of arithmetic facts are solved on the basis of verbal automatisms (mostly smallmultiplications and some additions) In the present case we do not know exactly how the patientproceeded to verify the various types of problems Since he was not asked to read aloud theoperands and proposed result it is also quite possible that he occasionally selected the correctanswer by chance (50) or in spite of reading andor calculation errors At any rate this goodperformance does by no means undermine our conclusion that the patient understood numeralssatisfactorily

4This analysis was performed omitting the number word ldquounrdquo (one) which is identical to the

undefinite article and is the most frequent single-phoneme word in French


length Yet the length of number words ranges from one to seven phonemesso this variable obviously cannot account for the total sparing of numbers

Yet another possibility would be that the patient adopted a particularresponse strategy inthe case of numberwords due to the factthat numberwordsform a small closed set Whenever the phonological form of the target wordwould not be fully available the patient would elect to produce another moreaccessible word from the same category However several observations sug-gest that this hypothesis is probably not correct First as noted in the clinicalsection the patient was generally unaware of his phonological errors andseemed quite satisfied with gross transformations of the targets Why thenwould he always prefer to produce another number word rather than the correcttarget plus some phonological transformations Second the patientrsquos initialattempts at producing the correct target (ie before choosing to produce anothernumber word) should be detectable as initial phonological errors (something

TABLE 10Neologisms with Non-numerical Words Matched with Number

Words

Number of Number ofNumber Words Frequency

aMatched Trials

bNeologisms ()

un 413 ndash ndashdeux 317 2 1 (500)trois 273 7 1 (143)quatre 244 20 7 (350)cinq 234 34 10 (294)six 215 44 14 (318)sept 195 66 23 (348)huit 211 47 16 (340)neuf 213 44 14 (318)dix 240 32 10 (313)onze 150 32 9 (281)douze 169 75 27 (360)treize 118 109 51 (468)quatorze 132 29 22 (759)quinze 188 69 26 (377)seize 129 104 37 (356)vingt 232 13 2 (154)trente 192 57 26 (456)quarante 175 34 25 (735)cinquante 181 32 24 (750)soixante 155 23 18 (783)cent 233 13 2 (154)mille 232 35 10 (286)

aLog10 of the frequency per million

bTrials with target words (1) matched in length with the number word

and (2) higher in frequency than the number word


like eg 3 reg bree glee four) At the very least assuming that theseputative initial attempts were covert the patientrsquos responses should have beensubstantially delayed Actually the patientrsquos errors were produced rapidly andwithout initial phonological errors Third it is not obvious why the patientwould not apply the same strategy to non-numerical targets also However thiswas not the case as seen from the small number of verbal errors and the almostcomplete absence of semantic errors A final argument against the existence ofsuch a response strategy would be the immunity of number words in connectedspeech when they are intermingled with other words a kind of data thatunfortunately we did not gather (see footnote 5 for relevant examples)

Thus it seems safe to conclude that the sparing of number words did notresult from some artefact but reflected their status as a particular category anissue that will be addressed in the General Discussion

The Origin of Number Reading Errors

Although the patient made numerous errors in reading aloud arabic andspelled-out numerals there was good evidence from number comparison andarithmetic verification tasks that he could comprehend these same numeralsaccurately Therefore reading errors resulted from a specific impairment inmapping arabic or spelled-out stimuli to the corresponding words With arabicnumerals errors affected both components of the mapping system framegeneration resulting in syntactic errors such as 74 reg ldquoseven hundred andfourrdquo and lexical retrieval resulting in lexical errors such as 66 reg ldquofifty sixrdquoWith spelled-out numerals all errors could be considered as lexical since thepatient merely read word by word seemingly building no word frame at all

GENERAL DISCUSSION

Summary

We have reported the case of a patient with Wernickersquos aphasia who sufferedfrom severe neologistic jargon (for reviews and related cases see Butterworth1979 1992 Ellis Miller amp Sin 1983 Lecours 1982) The neologismsalthough generally close to the target in terms of overall syllabic structure weremarred by pervasive phoneme substitutions Word and nonword reading aswell as picture naming were equally affected Performance was not influencedby lexical variables such as frequency imageability or grammatical classRemarkably phonological errors were almost never observed when the patientread numerals As mentioned in the Case Report this sparing extended to thenaming of isolated letters although this category was much less thoroughlystudied When reading aloud arabic or spelled-out numerals the patientrsquos errorsconsisted of substituting an entire numberword foranotherrather than selectinginappropriate phonemes


Number words and ordinary words thus displayed a double dissociationpattern On the one hand number words were subject to frequent word substi-tutions but escaped phonological transformations On the other hand ordinarywords were subject to pervasive phonological transformations but escapedword substitution errors5 The two major components of the patientrsquos deficitcould be accounted for in some detail each within the appropriate frameworkWe will now try to address the issues raised by the combination of the twodeficits and particularly by the intriguing fact that number words escapedphonological errors an observation that current models of speech planningwhich make no distinction between categories of words at the phonologicallevel are unable to account for

Origins of the Phonological Impairment

The simple fact that a category of words was unaffected by phonological errorsmay provide insights about the functional origin of the jargon Within currentmodels of speech production (eg Levelt 1989 1992) two broad alternativeorigins are conceivable for the patientrsquos phonological errors Errors could occurwhen retrieving the sequence of phonemes corresponding to a given word orthey could occur when selecting syllables on the basis of a given string ofphonemes For instance when trying to pronounce the word ldquoanneacuteerdquo an errorsuch as ane reg atildete could result either from the activation of phonemes atildeand t instead of phonemes a and n or at a later stage from the activationof syllables atilde and te instead of syllables a and ne (Levelt amp Wheeldon1994) Both accounts seem equally compatible with the good preservation of

5Following completion of this study we discovered that in his classical 1965 article Discon-

nexion syndromes in animals and man N Geschwind gave a brief but pithy description ofphenomena strikingly similar to those studied here occurring in patients labeled as conductionaphasics (Geschwind 1965 p 627) He noticed that number words were spared by phonologicalerrors while they could be erroneously substituted by one another ldquoAremarkable feature of manyof these cases is the frequently preserved ability of the patient to repeat polysyllabic numbers egseventy-eight while he fails to repeat even shorter words or repeats them paraphasically ( )Even when these patients fail to repeat numbers correctly their errors are different from thosewith words Thus the patient tends to substitute other numbers but rarely to produce grosslyparaphasic responses ( )rdquo Geschwind provided nice examples of this dissociation within singleutterances combining numerical and non-numerical words (eg fifty-five per cent reg fifty-fiveprogum three quarters reg three four one half reg fifty three quarters reg seventy five) Finallywhile his discussion is mostly based on speech repetition data Geschwind mentioned that ldquothisadvantage of numbers is not confined to the spoken modality since our first patient read printedwords paraphasically but read numbers aloud correctly whether printed as numerals or as wordsrdquoDue to his major perceptual impairment our patient was unable to repeat anything at all butGeschwindrsquos observations by extending the dissociation to a task with an auditory input furtherconfirm that the dissociation between word categories indeed reflects the structure of the speechproduction system


the phonological frame particularly of the number of syllables in the patientrsquosneologisms

Both hypotheses however are not equally compatible with the sparing ofnumber words In connected speech words of various syntactic and semanticcategories can be uttered in succession without phonological discontinuity Inparticular syllables may bridge over the boundaries of grammatical wordscontent words number words etc For instance the words ldquoquatrerdquo katr (four)and ldquoamisrdquo ami (friends) are syllabified as ka-tra-mi where the syllabletra combines phonemes tr from the number word ldquoquatrerdquo and phoneme afrom the noun ldquoamisrdquo This indicates that such ldquoresyllabificationrdquo cannot be acategory-specific process The retrieval of syllables from phonemes obviouslyspans across all word categories and does not take into account the particularwords to which the individual phonemes belong Hence an impairment at thesyllabic level should affect words of all types equally including number wordsThe sparing of number words therefore argues strongly against an impairmentof syllable retrieval (or of any subsequent process)

The hypothesis of an impaired activation of phonemes thus seems the onlyviable account of phonological errors It may be interesting to remember thatthe phonological perseverations observed during the production of series ofwords consisted of the iteration of a given phoneme and not of a given syllableThis observation further confirms that the phonological impairment affectedphoneme selection rather than syllable selection

Thus the main interest of the present observation is that it reveals that somecategorical organisation prevails in the speech production system down to theremarkably peripheral level of phoneme selection In our patient the activationof phonemes from the lexemes of number words was intact whereas theactivation of the same phonemes from the lexemes of otherwords was impaired

The dissociation of number words from other categories of words at thephonological level though unexpected is not without parallels in normalspeech production In French the phenomenon of liaison also provides indica-tions that number words and also letter names may be processed in a distinctway at the phoneme retrieval stage of speech production In many words thefinal consonant is not overtly realised in the speech output When the initialphoneme of the following word is a vowel however this final consonant mustbe uttered in order to provide a link (liaison in French) to the following wordFor instance ldquogrosrdquo (fat) is pronounced gro but ldquogros hommerdquo (fat man) ispronounced gro-zom Note that the added phoneme depends on the specificwords to be produced z in ldquogros hommerdquo but t in ldquogrand hommerdquo and nin ldquobon hommerdquo Hence liaison occurs at the processing stage most relevantto the present work namely the selection of phonemes appropriate to producea given lexeme A related phenomenon known as epenthesis consists of thesuppression of the final vowel of clitic words when the following word beginswith a vowel For instance one says ldquolrsquoamirdquo (the friend) and not le amirdquo


Remarkably there is no liaison and no epenthesis when the second word isa number word or a letter name For instance ldquoles huitresrdquo (the oysters) andldquoles onclesrdquo (the uncles) are pronounced with liaison as le-zegraveitr and le-zotildeklwhereas les huit rdquo (the eight ) and ldquoles onze rdquo (the eleven ) arepronounced as le-egraveit and le-otildez Similarly ldquoles airs les oeufs les ailesrdquo (theairs the eggs the wings) is pronounced as le-zer le-zœ le-zel but ldquoles R lesE les Lrdquo (the Rs the Es the Ls) as le-er le-œ le-el There is obligatoryepenthesis in ldquolrsquoonclerdquo (the uncle) and ldquolrsquohuissierrdquo (the usher) whereas onerather says ldquole onziegravemerdquo and ldquole huitiegravemerdquo Hence this observation indicatesthat in normal subjects phoneme selection may be influenced by the categoryof the words to be pronounced in agreement with the present observation thata deficit at this level can selectively spare the category of number words6

Our finding of a peripheral yet category-specific deficit is also strikinglysimilar in the spoken modality to a phenomenon reported by AndersonDamasio and Damasio (1990) in the domain of written language As a resultof a small left frontal lesion Anderson et alrsquos patient suffered from severealexia and agraphia Both of these deficits spared entirely arabic numerals Oneof the most remarkable features of this patient was the contrast between theillegible scribble resulting from her painful attempts at writing down lettersand her perfectly shaped and easily drawn arabic numerals Since the grapho-motor patterns required by letters and digits are of comparable complexity thisclear-cut dissociation suggested that the peripheral writing system may becontrolled by separate and dissociable systems devoted to arabic digits vsordinary words We draw a tentative parallel between such a sparing ofldquographological encodingrdquo of digits and the present sparing of phonologicalencoding of number words Taken together these results suggest that numbersactivate the elementary building blocks of the corresponding spoken and arabicoutput forms through pathways functionally and anatomically distinct fromother words

Why Are Numbers a Phonologically RelevantCategory

We have shown that number words were not spared by virtue of some superfi-cial difference from other words in terms of frequency length phonologicalcomplexity etc It also seems unlikely that numbers were spared because oftheir grammatical status From a syntactic point of view number words are akinboth to adjectives and to common nouns (Hurford 1987) two categories that

6There are other French words beyond numbers and letters that idiosyncratically escape

epenthesis We simply propose the hypothesis that numbers and letters escape epenthesis becausethey belong to phonologically particular categories and draw a parallel between this fact and thepresent neuropsychological dissociation


were heavily affected by phonological errors We now consider other potentialaccounts of why number words should constitute a relevant category at thephonological level At least three peculiar features of numbers may be respon-sible for their special status within the word production system the specificsemantic domain they refer to (quantities) the fact that they form an over-learned series of words and the special combinatorial syntax that enables themto compose multi-digit numerals The present study alone is not sufficient todetermine which of these properties if any is responsible for the selectivesparing phenomenon that we have reported However we try to suggest howeach of these alternative theories could be empirically tested in furtherresearch

Number Words and Lexicosemantic Categories

A first specific property of number words is that they refer to a narrowcategory-specific semantic domain that can be selectively spared or impairedby a brain lesion (Cipolotti Butterworth amp Denes 1991 Dehaene amp Cohen1997 Goodglass Klein Carey amp James 1966) It is now well established thatword finding and word comprehension difficulties can affect differentiallycategories such as nouns vs verbs tools vs animals or familiar people etcSuch evidence along with functional brain imaging studies suggests thatconsidering only nouns word forms and concepts activate one another throughconvergent two-way projections to the left inferotemporal region (H Damasioet al 1996) These convergence points are the neural substrate of word lemmasas defined in the functional framework outlined earlier (Caramazza 1996)Furthermore the temporal convergence zone can be subdivided anatomicallyaccording to conceptual domains (AR Damasio amp Tranel 1993 H Damasioet al 1996) It has been proposed that in the case of action verbs the bindingof word forms and concepts is achieved by the frontal rather than by thetemporal lobe (AR Damasio amp Tranel 1993 Denes Meneghello Vallese ampVanelli 1996) In summary the cerebral layout of word lemmas is organisedaccording to semantic (and possibly grammatical) categories It is possible thatat least some of these topographical distinctions are further propagated alongthe speech production pathway down to the level of word forms or lexemesOn this account the lexemes of a given category of words such as numerals inthe present case could be selectively impaired or spared yielding category-specific phonological errors

However it should be noted that aside from the present case evidence forsuch a categorical organisation at the lexeme level is as yet extremely limitedA few case reports suggest that closed-class words may occasionally escapephonological distortions (Butterworth 1979 Coslett Gonzalez-Rothi amp Heil-man 1984 Friederici amp Schoenle 1980) Friederici and Schoenle studied apatient with Wernickersquos aphasia (case 1) who when reading aloud lists ofwords made 88 errors (including 76 neologisms) with open-class words


but only 29 errors (including 24 neologisms) with closed-class words Nospecific information was provided concerning the patientrsquos performance withnumber words We are not aware of reports of selective sparing or impairmentof other word categories at the phonological level As yet therefore there islittle reason to generalise from numbers and closed-class words to the wholelexicon or to propose that phonological encoding relies on dissociable sub-systems for different word categories beyond these two special cases

Number Words and Automatic Speech

Beyond their semantic specificity an original feature of number words isthat they are commonly recited as an ordered verbal series (one two three ) for instance for the purpose of counting objects Young children learnnumber words by reciting the series even before comprehending the correspon-dence with abstract quantities (Wynn 1990) The overpractised series ofnumber words is part of what is often called ldquoautomaticrdquo or ldquononpropositionalrdquospeech along with other series such as the alphabet the days of the weekelementary arithmetic facts (eg ldquotwo times three sixrdquo) etc A relativesparing of automatic speech in otherwise severely aphasic patients has repeat-edly been observed (eg Henschen 1926 Lum amp Ellis 1994) even in patientswith a severe jargon (Cummings Benson Walsh amp Levine 1979) andstandard language batteries include an assessment of this particular speechmodality As an illustration of the ability to access number words by a pathwaydedicated to automatic speech aphasic patients may be able to name arabicdigits or spelled-out numerals only through serial counting (Cohen et al 1994Dehaene amp Cohen 1991 Gazzaniga amp Hillyard 1971) Similarly patients withGerstmannrsquos acalculia are unable to solve elementary problems such as 7-3but can still provide the result of problems overlearned as verbal automatismssuch as ldquothree times six eighteenrdquo (Dehaene amp Cohen 1997)

It may therefore be hypothesised that number words were spared by thejargon because in addition to their use in normal propositional speech theycan be uttered in the context of automatic speech We do not claim that ourpatient resorted to a counting strategy when reading aloud numerals There wasindeed no clinical indication that he did so We only suggest that words thathave been learned as automatic speech might benefit from particular accessmechanisms even when produced out of the context of overlearned word seriesThis hypothesis has the advantage of readily accounting for the parallel sparingof letter reading Whereas letters can be named individually just like any othervisual object children also learn their names by rote in a fixed order Adultsoften recite this series to determine the relative position of two letters in thealphabet for instance when looking up a word in a dictionary As reportedearlier our patient did not make a single phonological error when namingletters His only error in 26 trials consisted in naming letter R (er) as D (de)


Although letter names may seem trivially simple they are phonologicallycomparable and sometimes identical to ordinary monosyllabic words (eg Cand sea B and bee) The patient made as many as 2743 (628) errors withmonosyllabic words with a CV or VC structure comparable to that of letternames (613 with pictures and 2130 with written words) Thus letter namesjust like number words significantly escaped the errors affecting ordinarywords of a matched complexity [ c 2(1) = 241 P lt 0001]

The associated sparing of letters and numbers suggests that perhaps numberwords were not spared as such but only because the neural systems underlyingthe production of automatic speech were intact (Cummings et al 1979)Unfortunately we have no data concerning the patientrsquos behaviour with otherinstances of automatic speech such as days of the week months of the yearnursery rhymes etc The relatively preserved automatic speech in aphasics hasbeen often ascribed to the intact right hemisphere (eg Graves amp Landis 1985Henschen 1926 Kinsbourne 1971) Alternatively it has also been proposedthat left cortico-subcortical loops play a role in verbal automatisms amidst otherfamiliar motor sequences (Dehaene amp Cohen 1997 Houk amp Wise 1995)Whatever the correct account the present data are compatible with the idea thatthe output word forms of ldquoautomatic wordsrdquo rest on cerebral systems anatomi-cally distinct from the systems that underlie ordinary propositional speech

Number Words Letter Names and Phonemes TheBuilding Blocks of Speech

Number words and letter names share another interesting feature beyondtheir role in automatic speech They can both be combined to form complexwords Letter names can be combined to form acronyms such as ju-es-es-ar (USSR) Similarly number words can be combined to form complexnumerals such as twenti- q ri (23) When involved in such combinationsindividual number words and letter names behave as complex phonologicalentities intermediate between the word and the phoneme levels In some sensewhile phonemes are the building blocks of most words one may say thatindividual letter names are the building blocks of acronyms and individualnumber words the building blocks of complex numerals

One may then speculate that the speech production system may haveintegrated the combinatorial particularities of numerals and acronyms as afunctional regularity of the mapping between words and articulation patternsWhereas most words are programmed as a sequence of phonemes it is at leastpossible that the units of speech production in the case of numerals are entirenumber words The entire sequence of phonemes making up the pronunciationof a number word (for) or a letter word (ju) might be retrieved as a distinctunit during speech production This hypothesis would provide an explanation


for the above-mentioned special status of number words and letters with respectto the rules of liaison and epenthesis

In this perspective the patientrsquos deficit might tentatively be interpreted asaffecting a general stage of processing during which the appropriate phonologi-cal units making up the word to be produced would be retrieved Noise duringthis unit selection stage would result in random substitutions of inappropriateunits Entire individual number words would be substituted during numberproduction whereas in the case of other types of words individual phonemeswould be subject to errors

Although clearly speculative the theory outlined here has the advantage ofaccounting simultaneously for all the main features of the patientrsquos deficitnamely the word substitution errors in number reading and the phonemesubstitution errors in normal word production All other accounts of thecategory-specificity of the jargon need to make the ad hoc hypothesis that twofunctionally unrelated deficits were independently responsible for the word andnumber reading impairments The postulation of two functionally independentdeficits seems implausible however because it would be a remarkable coinci-dence if the phoneme selection deficit spared precisely the same lexicalcategory (number words) that was exclusively affected by word selectionerrors Further work should tell whether the more economical concept of asingle combinatorial process occurring at multiple levels during word produc-tion (single phonemes letter names or individual number words) is a viabletheoretical construct

Manuscript first received 9 September 1996Revised manuscript received 15 September 1997

Revised manuscript accepted 16 October 1997

REFERENCES

Anderson SW Damasio AR amp Damasio H (1990) Troubled letters but not numbersDomain specific cognitive impairment following focal damage in frontal cortex Brain 113749ndash766

Badecker W Miozzo M amp Zanuttini R (1995) The two-stage model of lexical retrievalEvidence from a case of anomia with selective preservation of grammatical gender Cognition57 193ndash216

Best W (1996) When rackets are baskets but baskets are biscuits where do the words comefrom A single case study of formal paraphasic errors in aphasia Cognitive Neuropsychology13 443ndash480

Butterworth B (1979) Hesitation and production of verbal paraphasias and neologisms injargon aphasia Brain and Language 8 133ndash161

Butterworth B (1992) Disorders of phonological encoding Cognition 42 261ndash286Campbell JID amp Clark JM (1988) An encoding-complex view of cognitive number proc-

essing Comment on McCloskey Sokol amp Goodman (1986) Journal of ExperimentalPsychology General 117 204ndash214

Caramazza A (1996) The brainrsquos dictionary Nature 380 485ndash486


Cipolotti L Butterworth B amp Denes G (1991) A specific deficit for numbers in a case ofdense acalculia Brain 114 2619ndash2637

Cohen L amp Dehaene S (1991) Neglect dyslexia for numbers A case report CognitiveNeuropsychology 8 39ndash58

Cohen L Dehaene S amp Verstichel P (1994) Number words and number nonwords A caseof deep dyslexia extending to arabic numerals Brain 117 267ndash279

Content A Mousty P amp Radeau M (1990) Brulex une base de donneacutees lexicales informa-tiseacutee pour le franccedilais eacutecrit et parleacute Lrsquoanneacutee Psychologique 90 551ndash566

Coslett HB Gonzalez-Rothi LG amp Heilman KM (1984) Reading Selective sparing ofclosed-class words in Wernickersquos aphasia Neurology 34 1038ndash1045

Cummings JL Benson DF Walsh MJ amp Levine HL (1979) Left-to-right transfer oflanguage dominance a case study Neurology 29 1547ndash1550

Damasio H amp Damasio AR (1989) Lesion analysis in neuropsychology Oxford OxfordUniversity Press

Damasio AR amp Tranel D (1993) Nouns and verbs are retrieved with differently distributedneural systems Proceedings of the National Academy of Sciences of the USA 90 4957ndash4960

Damasio H Grabowski TJ Tranel D Hichwa RD amp Damasio A (1996) A neural basisfor lexical retrieval Nature 380 499ndash505

Dehaene S (1995) Electrophysiological evidence for category-specific word processing in thenormal human brain NeuroReport 6 2153ndash2157

Dehaene S amp Cohen L (1991) Two mental calculation systems A case study of severeacalculia with preserved approximation Neuropsychologia 29 1045ndash1074

Dehaene S amp Cohen L (1995) Towards an anatomical and functional model of numberprocessing Mathematical Cognition 1 83ndash120

Dehaene S amp Cohen L (1997) Cerebral pathways for calculation Double dissociationsbetween Gerstmannrsquos acalculia and subcortical acalculia Cortex 33 219ndash250

Deloche G amp Seron X (1987) Numerical transcoding A general production model In GDeloche amp X Seron (Eds) Mathematical disabilities A cognitive neuropsychologicalperspective (pp 137ndash170) Hillsdale NJ Lawrence Erlbaum Associates Inc

Denes G Meneghello F Vallese F amp Vanelli L (1996) Task-dependent noun retrievaldeficit An experimental case study Neurocase 2 35ndash43

Ellis AW Miller D amp Sin G (1983) Wernickersquos aphasia and normal language processingA case study in cognitive neuropsychology Cognition 15 111ndash114

Friederici AD amp Schoenle PW (1980) Computational dissociation of two vocabulary typesEvidence from aphasia Neuropsychologia 18 11ndash20

Gazzaniga MS amp Hillyard SA (1971) Language and speech capacity of the right hemi-sphere Neuropsychologia 9 273ndash280

Geschwind N (1965) Disconnexion syndromes in animals and man Part II Brain 88585ndash644

Goodglass H Klein B Carey P amp James KJ (1966) Specific semantic word categories inaphasia Cortex 2 74ndash89

Graves T amp Landis T (1985) Hemispheric control of speech expression in aphasia Archivesof Neurology 42 249ndash251

Henschen SE (1926) On the function of the right hemisphere of the brain in relation to the leftin speech music and calculation Brain 49 110ndash123

Houk JC amp Wise SP (1995) Distributed modular architectures linking basal ganglia cere-bellum and cerebral cortex Their role in planning and controlling action Cerebral Cortex 295ndash110

Hurford JR (1987) Language and number Oxford Basil BlackwellKinsbourne M (1971) The minor cerebral hemisphere as a source of aphasic speech Archives

of Neurology 25 302ndash306


Kohn SE amp Goodglass H (1985) Picture naming in aphasia Brain and Language 24266ndash283

Lecours AR (1982) On neologisms In J Mehler ECT Walker amp MF Garrett (Eds)Perspectives on mental representation Hillsdale NJ Lawrence Erlbaum Associates Inc

Levelt WJM (1989) Speaking From intention to articulation Cambridge MA MIT PressLevelt WJM (1992) Accessing words in speech production Stages processes and repre-

sentations Cognition 42 1ndash22Levelt WJM amp Wheeldon L (1994) Do speakers have access to a mental syllabary

Cognition 50 239ndash269Lum CC amp Ellis AE (1994) Is ldquononpropositionalrdquo speech preserved in aphasia Brain and

Language 46 368ndash391Martin A Haxby JV Lalonde FM Wiggs CL amp Ungerleider LG (1995) Discrete

cortical regions associated with knowledge of color and knowledge of action Science 270102ndash105

Martin A Wiggs CL Ungerleider LG amp Haxby JV (1996) Neural correlates of category-specific knowledge Nature 379 649ndash652

McCloskey M Sokol SM amp Goodman RA (1986) Cognitive processes in verbal numberproduction Inferences from the performance of brain-damaged subjects Journal of Experi-mental Psychology General 115 307ndash330

Meyer AS (1992) Investigation of phonological encoding through speech error analysesAchievements limitations and alternatives Cognition 42 181ndash211

Naeser MA Helm-Estabrooks N Haas G Auerbach S amp Srinivasan M (1987) Relation-ship between lesion extent in lsquoWernickersquos arealsquo on computed tomographic scan and predictingrecovery of comprehension in Wernickersquos aphasia Archives of Neurology 44 73ndash82

Schriefers H Meyer AS amp Levelt WJM (1990) Exploring the time course of lexical accessin language production Picture-word interference studies Journal of Memory and Language29 86ndash102

Segui J Mehler J Frauenfelder U amp Morton J (1982) The word frequency effect andlexical access Neuropsychologia 20 615ndash627

Sevald CA Dell GS amp Cole JS (1995) Syllable structure in speech production Aresyllables chunks or schemas Journal of Memory and Language 34 807ndash820

Shattuck-Hufnagel S (1979) Speech errors as evidence for serial-order mechanism in sentenceproduction In WE Cooper amp EC Walker (Eds) Psycholinguistic studies presented toMerrill Garrett Hillsdale NJ Lawrence Erlbaum Associates Inc

Snodgrass JG amp Vanderwart M (1980) A standardized set of 260 pictures Norms for nameagreement image agreement familiarity and visual complexity Journal of ExperimentalPsychology Human Learning 6 174ndash215

Treiman R (1983) The structure of spoken syllables Evidence from novel word gamesCognition 15 49ndash74

Verstichel P Cohen L amp Crochet G (1996) Associated production and comprehensiondeficits for names of people following left temporal lesion Neurocase 2 221ndash234

Wynn K (1990) Childrenrsquos understanding of counting Cognition 36 155ndash193


knowledge relative to different categories of words rests in separable brainregions (H Damasio Grabowski Tranel Hichwa amp Damasio 1996 Dehaene1995) As yet such categorical organisation has been shown to prevail at twobroad levels of representation First at the semantic level fragments of knowl-edge that build up the meaning of words are stored in distinct cortical locationsfor different word categories For instance naming animals differing by subtleshape features strongly activates the early visual cortex whereas namingfamiliar tools or generating action words activates areas related to movementplanning or movement perception (Martin Haxby Lalonde Wiggs amp Unger-leider 1995 Martin Wiggs Ungerleider amp Haxby 1996) Second the ana-tomical layout of the connections that bind semantic knowledge on the onehand and word forms on the other hand also obeys a systematic categoricalorganisation Thus at least in the case of nouns the word forms and thecorresponding concepts activate one another through convergent two-wayprojections to the left inferotemporal region This temporal convergence zonecan be subdivided anatomically according to conceptual domains such asfamiliar people tools etc explaining why brain lesions can yield category-spe-cific word finding or word comprehension difficulties (Caramazza 1996 HDamasio et al 1996 Verstichel Cohen amp Crochet 1996)

We report the case of a patient with severe neologistic jargon resulting fromnumerous phoneme substitutions occurring during the planning of verbaloutput The impairment thus affected a relatively late stage of speech produc-tion which might be expected to be common to all types of words Howeverthe production of numberwords was almost entirely free of phonological errorsbut displayed word selection errors of a quite different type We will argue thatsuch a category-specific sparing suggests that the cerebral lexicon has acategorical organisation down to the phonological level

Following presentation of clinical data we will try to characterise the originof phonological errors through a detailed analysis of the jargon We will thenturn to the patientrsquos performance in number processing tasks Finally we willdiscuss the bearings that such a pattern may have on models of normal speechproduction and particularly on the categorical organisation of phonologicalplanning

CASE REPORT

Medical History

The patient was a 76-year-old right-handed retired college teacher He wasadmitted to hospital because his speech had suddenly turned into an incompre-hensible jargon Furthermore he seemed not to hear what was told to him Amild right brachio-facial motor deficit was noted which receded within a dayInitial CT scan revealed the sequelae of an old overlooked right-hemispheric



Language Assessment

General Description






FIG1

O

utli

neof

the

pati

entrsquos

lesi

ons

aspl

otte

don

Dam

asio

and

Dam

asio

rsquos(1

989)

tem

plat

esT

hele

fthe

mis

pher

eap

pear

son

the

righ

t-ha

ndsi

deof

axia

lsec

tion

s

1032





Subtest Score






Summary





FIG2

Sche

mat

icde

pict

ion

ofth

esp

eech

prod

ucti

onsy

stem

as

wel

las

inpu

tto

this

syst

emfr

ompi

ctur

ean

dw

ord

perc

epti

onW

esu

gges

ttha

tthe

pati

entrsquos

jarg

onre

sult

edfr

omim

pair

edpr

ojec

tions

from

the

lexe

me

leve

lto

the

phon

emic

leve

l

1035





Method









Naming Tasks

Mean SD Min Max


a(N = 178) 111 078 ndash11 295


b11180 1229 00 43400


a(N = 480) 151 094 ndash11 381


b11000 1229 00 43400


b10400 9260 00 43200

















(N = 192) (N = 192)






Length





the Target









Frequency






Imageability



Grammatical Class


Summary





Phonological Frame












1 99 32 3 1342 5 180 21 1 2073 9 84 11 2 1064 21 49 1 715 2 1 1 46 0Total 104 221 131 62 3 1 522












Phonemic Content





Summary








Verbal Errors














General Features





Target Error


































Words


aMatched Trials

bNeologisms ()










GENERAL DISCUSSION

Summary











































REFERENCES
























































Language Assessment

General Description






FIG1

O

utli

neof

the

pati

entrsquos

lesi

ons

aspl

otte

don

Dam

asio

and

Dam

asio

rsquos(1

989)

tem

plat

esT

hele

fthe

mis

pher

eap

pear

son

the

righ

t-ha

ndsi

deof

axia

lsec

tion

s

1032





Subtest Score






Summary





FIG2

Sche

mat

icde

pict

ion

ofth

esp

eech

prod

ucti

onsy

stem

as

wel

las

inpu

tto

this

syst

emfr

ompi

ctur

ean

dw

ord

perc

epti

onW

esu

gges

ttha

tthe

pati

entrsquos

jarg

onre

sult

edfr

omim

pair

edpr

ojec

tions

from

the

lexe

me

leve

lto

the

phon

emic

leve

l

1035





Method









Naming Tasks

Mean SD Min Max


a(N = 178) 111 078 ndash11 295


b11180 1229 00 43400


a(N = 480) 151 094 ndash11 381


b11000 1229 00 43400


b10400 9260 00 43200

















(N = 192) (N = 192)






Length





the Target









Frequency






Imageability



Grammatical Class


Summary





Phonological Frame












1 99 32 3 1342 5 180 21 1 2073 9 84 11 2 1064 21 49 1 715 2 1 1 46 0Total 104 221 131 62 3 1 522












Phonemic Content





Summary








Verbal Errors














General Features





Target Error


































Words


aMatched Trials

bNeologisms ()










GENERAL DISCUSSION

Summary











































REFERENCES























































FIG1

O

utli

neof

the

pati

entrsquos

lesi

ons

aspl

otte

don

Dam

asio

and

Dam

asio

rsquos(1

989)

tem

plat

esT

hele

fthe

mis

pher

eap

pear

son

the

righ

t-ha

ndsi

deof

axia

lsec

tion

s

1032





Subtest Score






Summary





FIG2

Sche

mat

icde

pict

ion

ofth

esp

eech

prod

ucti

onsy

stem

as

wel

las

inpu

tto

this

syst

emfr

ompi

ctur

ean

dw

ord

perc

epti

onW

esu

gges

ttha

tthe

pati

entrsquos

jarg

onre

sult

edfr

omim

pair

edpr

ojec

tions

from

the

lexe

me

leve

lto

the

phon

emic

leve

l

1035





Method









Naming Tasks

Mean SD Min Max


a(N = 178) 111 078 ndash11 295


b11180 1229 00 43400


a(N = 480) 151 094 ndash11 381


b11000 1229 00 43400


b10400 9260 00 43200

















(N = 192) (N = 192)






Length





the Target









Frequency






Imageability



Grammatical Class


Summary





Phonological Frame












1 99 32 3 1342 5 180 21 1 2073 9 84 11 2 1064 21 49 1 715 2 1 1 46 0Total 104 221 131 62 3 1 522












Phonemic Content





Summary








Verbal Errors














General Features





Target Error


































Words


aMatched Trials

bNeologisms ()










GENERAL DISCUSSION

Summary











































REFERENCES



























































Subtest Score






Summary





FIG2

Sche

mat

icde

pict

ion

ofth

esp

eech

prod

ucti

onsy

stem

as

wel

las

inpu

tto

this

syst

emfr

ompi

ctur

ean

dw

ord

perc

epti

onW

esu

gges

ttha

tthe

pati

entrsquos

jarg

onre

sult

edfr

omim

pair

edpr

ojec

tions

from

the

lexe

me

leve

lto

the

phon

emic

leve

l

1035





Method









Naming Tasks

Mean SD Min Max


a(N = 178) 111 078 ndash11 295


b11180 1229 00 43400


a(N = 480) 151 094 ndash11 381


b11000 1229 00 43400


b10400 9260 00 43200

















(N = 192) (N = 192)






Length





the Target









Frequency






Imageability



Grammatical Class


Summary





Phonological Frame












1 99 32 3 1342 5 180 21 1 2073 9 84 11 2 1064 21 49 1 715 2 1 1 46 0Total 104 221 131 62 3 1 522












Phonemic Content





Summary








Verbal Errors














General Features





Target Error


































Words


aMatched Trials

bNeologisms ()










GENERAL DISCUSSION

Summary











































REFERENCES


























































Summary





FIG2

Sche

mat

icde

pict

ion

ofth

esp

eech

prod

ucti

onsy

stem

as

wel

las

inpu

tto

this

syst

emfr

ompi

ctur

ean

dw

ord

perc

epti

onW

esu

gges

ttha

tthe

pati

entrsquos

jarg

onre

sult

edfr

omim

pair

edpr

ojec

tions

from

the

lexe

me

leve

lto

the

phon

emic

leve

l

1035





Method









Naming Tasks

Mean SD Min Max


a(N = 178) 111 078 ndash11 295


b11180 1229 00 43400


a(N = 480) 151 094 ndash11 381


b11000 1229 00 43400


b10400 9260 00 43200

















(N = 192) (N = 192)






Length





the Target









Frequency






Imageability



Grammatical Class


Summary





Phonological Frame












1 99 32 3 1342 5 180 21 1 2073 9 84 11 2 1064 21 49 1 715 2 1 1 46 0Total 104 221 131 62 3 1 522












Phonemic Content





Summary








Verbal Errors














General Features





Target Error


































Words


aMatched Trials

bNeologisms ()










GENERAL DISCUSSION

Summary











































REFERENCES























































FIG2

Sche

mat

icde

pict

ion

ofth

esp

eech

prod

ucti

onsy

stem

as

wel

las

inpu

tto

this

syst

emfr

ompi

ctur

ean

dw

ord

perc

epti

onW

esu

gges

ttha

tthe

pati

entrsquos

jarg

onre

sult

edfr

omim

pair

edpr

ojec

tions

from

the

lexe

me

leve

lto

the

phon

emic

leve

l

1035





Method









Naming Tasks

Mean SD Min Max


a(N = 178) 111 078 ndash11 295


b11180 1229 00 43400


a(N = 480) 151 094 ndash11 381


b11000 1229 00 43400


b10400 9260 00 43200

















(N = 192) (N = 192)






Length





the Target









Frequency






Imageability



Grammatical Class


Summary





Phonological Frame












1 99 32 3 1342 5 180 21 1 2073 9 84 11 2 1064 21 49 1 715 2 1 1 46 0Total 104 221 131 62 3 1 522












Phonemic Content





Summary








Verbal Errors














General Features





Target Error


































Words


aMatched Trials

bNeologisms ()










GENERAL DISCUSSION

Summary











































REFERENCES



























































Method









Naming Tasks

Mean SD Min Max


a(N = 178) 111 078 ndash11 295


b11180 1229 00 43400


a(N = 480) 151 094 ndash11 381


b11000 1229 00 43400


b10400 9260 00 43200

















(N = 192) (N = 192)






Length





the Target









Frequency






Imageability



Grammatical Class


Summary





Phonological Frame












1 99 32 3 1342 5 180 21 1 2073 9 84 11 2 1064 21 49 1 715 2 1 1 46 0Total 104 221 131 62 3 1 522












Phonemic Content





Summary








Verbal Errors














General Features





Target Error


































Words


aMatched Trials

bNeologisms ()










GENERAL DISCUSSION

Summary











































REFERENCES





























































Naming Tasks

Mean SD Min Max


a(N = 178) 111 078 ndash11 295


b11180 1229 00 43400


a(N = 480) 151 094 ndash11 381


b11000 1229 00 43400


b10400 9260 00 43200

















(N = 192) (N = 192)






Length





the Target









Frequency






Imageability



Grammatical Class


Summary





Phonological Frame












1 99 32 3 1342 5 180 21 1 2073 9 84 11 2 1064 21 49 1 715 2 1 1 46 0Total 104 221 131 62 3 1 522












Phonemic Content





Summary








Verbal Errors














General Features





Target Error


































Words


aMatched Trials

bNeologisms ()










GENERAL DISCUSSION

Summary











































REFERENCES





































































(N = 192) (N = 192)






Length





the Target









Frequency






Imageability



Grammatical Class


Summary





Phonological Frame












1 99 32 3 1342 5 180 21 1 2073 9 84 11 2 1064 21 49 1 715 2 1 1 46 0Total 104 221 131 62 3 1 522












Phonemic Content





Summary








Verbal Errors














General Features





Target Error


































Words


aMatched Trials

bNeologisms ()










GENERAL DISCUSSION

Summary











































REFERENCES

























































Length





the Target









Frequency






Imageability



Grammatical Class


Summary





Phonological Frame












1 99 32 3 1342 5 180 21 1 2073 9 84 11 2 1064 21 49 1 715 2 1 1 46 0Total 104 221 131 62 3 1 522












Phonemic Content





Summary








Verbal Errors














General Features





Target Error


































Words


aMatched Trials

bNeologisms ()










GENERAL DISCUSSION

Summary











































REFERENCES

























































Frequency






Imageability



Grammatical Class


Summary





Phonological Frame












1 99 32 3 1342 5 180 21 1 2073 9 84 11 2 1064 21 49 1 715 2 1 1 46 0Total 104 221 131 62 3 1 522












Phonemic Content





Summary








Verbal Errors














General Features





Target Error


































Words


aMatched Trials

bNeologisms ()










GENERAL DISCUSSION

Summary











































REFERENCES
























































Imageability



Grammatical Class


Summary





Phonological Frame












1 99 32 3 1342 5 180 21 1 2073 9 84 11 2 1064 21 49 1 715 2 1 1 46 0Total 104 221 131 62 3 1 522












Phonemic Content





Summary








Verbal Errors














General Features





Target Error


































Words


aMatched Trials

bNeologisms ()










GENERAL DISCUSSION

Summary











































REFERENCES























































Phonological Frame












1 99 32 3 1342 5 180 21 1 2073 9 84 11 2 1064 21 49 1 715 2 1 1 46 0Total 104 221 131 62 3 1 522












Phonemic Content





Summary








Verbal Errors














General Features





Target Error


































Words


aMatched Trials

bNeologisms ()










GENERAL DISCUSSION

Summary











































REFERENCES

































































Phonemic Content





Summary








Verbal Errors














General Features





Target Error


































Words


aMatched Trials

bNeologisms ()










GENERAL DISCUSSION

Summary











































REFERENCES























































Phonemic Content





Summary








Verbal Errors














General Features





Target Error


































Words


aMatched Trials

bNeologisms ()










GENERAL DISCUSSION

Summary











































REFERENCES
























































Verbal Errors














General Features





Target Error


































Words


aMatched Trials

bNeologisms ()










GENERAL DISCUSSION

Summary











































REFERENCES




























































General Features





Target Error


































Words


aMatched Trials

bNeologisms ()










GENERAL DISCUSSION

Summary











































REFERENCES























































General Features





Target Error


































Words


aMatched Trials

bNeologisms ()










GENERAL DISCUSSION

Summary











































REFERENCES


















































































Words


aMatched Trials

bNeologisms ()










GENERAL DISCUSSION

Summary











































REFERENCES











































































Words


aMatched Trials

bNeologisms ()










GENERAL DISCUSSION

Summary











































REFERENCES




































































Words


aMatched Trials

bNeologisms ()










GENERAL DISCUSSION

Summary











































REFERENCES


























































Words


aMatched Trials

bNeologisms ()










GENERAL DISCUSSION

Summary











































REFERENCES



























































GENERAL DISCUSSION

Summary











































REFERENCES































































































REFERENCES

























































































REFERENCES



















































































REFERENCES












































































REFERENCES







































































REFERENCES


































































REFERENCES




























































REFERENCES

























































































































Documents

Neologistic Jargon Sparing Numbers: A Category-specific ... · NeologisticJargonSparingNumbers:A Category-specificPhonologicalImpairment LaurentCohen Hôpital de la Salpêtrière,Paris,France