This article was downloaded by: [Fondren Library, Rice University ]
On: 11 November 2014, At: 14:12
Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer
House, 37-41 Mortimer Street, London W1T 3JH, UK

Neurocase: The Neural Basis of Cognition
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/nncs20

Knowing Its Gender Without Knowing Its Name:
Differential Access to Lexical Information in a
Jargonaphasic Patient
a

Joël Macoir & Renée Béland

b

a

Programme d'orthophonie and Unité de recherche en gériatrie, Laval University ,
Québec, Canada
b

École d'orthophonie et d'audiologie and Groupe de Recherche en Neuropsychologie
Expérimentale et Cognition, Université de Montréal , Québec, Canada
Published online: 17 Aug 2010.

To cite this article: Joël Macoir & Renée Béland (2004) Knowing Its Gender Without Knowing Its Name: Differential
Access to Lexical Information in a Jargonaphasic Patient, Neurocase: The Neural Basis of Cognition, 10:6, 471-482, DOI:
10.1080/13554790490900770
To link to this article: http://dx.doi.org/10.1080/13554790490900770

PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of
the Content. Any opinions and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied
upon and should be independently verified with primary sources of information. Taylor and Francis shall
not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other
liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or
arising out of the use of the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions

Neurocase, 10(6): 471–482, 2004
Copyright © Taylor & Francis Inc.
1355-4795/04/1006–471$16.00
Neurocase

Knowing Its Gender Without Knowing Its Name: Differential
Access to Lexical Information in a Jargonaphasic Patient
Joël Macoir1 and Renée Béland2
Programme d’orthophonie and Unité de recherche en gériatrie, Laval University, Québec, Canada, 2École d’orthophonie
et d’audiologie and Groupe de Recherche en Neuropsychologie Expérimentale et Cognition, Université de
Montréal, Québec, Canada,

Downloaded by [Fondren Library, Rice University ] at 14:12 11 November 2014

1

Abstract
According to recent models of word production, when we name a picture, we first retrieve the meaning of the object, and
then we independently retrieve the written or sound form of the word corresponding to the picture. In languages like
French, in which words have a gender, theoretical models disagree with respect to the moment at which this information is
retrieved. The lemma model (Levelt et al., 1999) posits that we access this information before the sound or written form of
the word is retrieved. In contrast, the “Independent network” (IN) model (Caramazza, 1997) model posits that we access
gender after retrieval of either the sound or written form of the word. This paper reports a single-case study of an aphasic
patient, BA, who showed deficits affecting spoken and written production in the presence of largely preserved
comprehension abilities. Experimental testing indicated that she presented with a deficit functionally localized in the
access to lexical representations. Results in picture naming and in gender identification also revealed that BA identified
the gender above chance level, whether she produced a correct response, a phonemic error, or a neologism. In contrast,
when she was unable to produce a spoken or written response, she could not identify the gender. This pattern of
performance is consistent with the lemma model in which access to lexical syntax is required before access to
phonological form can take place.

Introduction
According to recent models of speech production, words are
retrieved and produced through the activation of specialized
and inter-connected components. In these models, the production of words is conceived as a staged process, in which
the activation flow is initiated in a conceptual-semantic component and ends with the execution of articulation mechanisms.
For picture naming, lexical access involves two main steps.
In the first step, the non-verbal conceptual representation from
the semantic system maps into a lexical-semantic representation, specifying the semantic features of the word to produce.
In the second step, the lexeme, that is, the sound (phonological lexeme) or the written form (orthographic lexeme) of the
target word, including morphological, metrical and segmental
information, is retrieved. According to the “lemma” model,
(e.g. Levelt et al., 1999), the information from the lexicalsemantic level first addresses, in a one-to-one relationship, an
intermediate lexical level before the activation of the lexeme
level. This intermediate level, referred to as the lemma level,
contains information specifying the syntactic properties (e.g.
grammatical class, gender, and agreement in tense, number,
etc.) of the retrieved lexical-semantic representations (see
Fig. 1a). Although the question has not been clearly addressed

in the theoretical propositions, one can also assume that a
unique lemma level is activated, whether the concept is to be
spoken or written. Therefore, the output from the lemma level
is sent to the phonological or orthographic output lexicon if
the activated lexical-semantic representation is to be expressed
in the spoken or written modality, respectively (see Fig. 1a).
Evidence for the lemma level comes from the analysis of
tip-of-the-tongue states (TOT), a common phenomenon in
which normal subjects experience difficulty in retrieving a
specific word but still have access to some of its syntactic
(e.g. Vigliocco et al., 1997, 1999; Ferrand, 2001) or phonological (e.g. Brown, 1991) properties. In aphasia, DPI, the
patient recently reported by Bachoud-Lévi and Dupoux (2003)
as well as Dante, an Italian patient reported by Badecker et al.
(1995), showed largely intact abilities in identifying the
grammatical gender of words that they were unable to name or
for which they could not provide any phonological or orthographic information.
In contrast, the “independent network” (IN) model
(Caramazza, 1997) involves no lemma level. In this model,
syntactic information is accessed after activating either the
phonological or the orthographic lexemes, that is, at the final

Correspondence to: J. Macoir, Faculté de médecine, Département de réadaptation, Laval University, Pavillon F-Vandry, Québec (Qué) GIK 7P4, Canada.
Tel: 1-1-418-656-2131 (ext. 12190); Fax: +1-418-656-5476; e-mail: joel.macoir@rea.ulaval.ca
DOI: 10.1080/13554790490900770

472

J. Macoir and R. Béland

Semantic system

Semantic system

Lexical semantic
representation

A.

Lexical semantic
representation

B.

Lemma

Downloaded by [Fondren Library, Rice University ] at 14:12 11 November 2014

Phonolological
lexeme
Phonolological
lexeme

Orthographic
lexeme

Spoken output

Written output

Orthographic
lexeme

syntax

Spoken output

Written output

Fig. 1. Schematic depiction of the two accounts of the two-stage model of
lexical access. (A) The lemma mediation proposition based on Levelt et al.
(1999) (note that routes and components in grey are not implemented in Levelt
et al.); (B) The ‘Independent network model’.

stage of lexical retrieval (see Fig. 1b). Evidence in support of
the IN model arises from studies in which brain-damaged
patients showed correct access to lexemes without being able
to retrieve the information related to gender (e.g. Miceli and
Caramazza, 1988). In a recent study, Caramazza and Miozzo
(1997) also showed that normal subjects were able to retrieve
the grammatical gender and the initial phoneme of words in
TOT states above chance-level accuracy, but that the availability of the two types of information was not correlated.
According to the authors, these results demonstrate that the
retrieval of phonological information is independent of the
prior retrieval of syntactic information.
The lemma model and the IN model make different predictions with respect to normal and aphasic speech. In the
lemma model, normal speakers and aphasic patients should
be able to retrieve syntactic information without accessing
information at the lexeme level. In the IN model, if a

speaker retrieves syntactic information, he must also access
either phonological or orthographic information about the
word.
We report a single-case study of an aphasic patient, BA,
who showed deficits affecting spoken and written production
in the presence of largely preserved comprehension abilities.
We pursue two objectives in this case study. First, we investigate which of the two lexical access models, the lemma or the
IN model, better accounts for the patient’s abilities in picture
naming. Second, because the patient presents a clear dissociation in oral and picture naming, we investigate the lexeme
retrieval. Using a series of experiments, we attempt to distinguish instances where partial access to the phonological lexemes is possible from those in which this access is
completely blocked.

Case history
BA, an 81-year-old, right-handed woman, is a native speaker
of French. She has a grade thirteen education and worked as a
teacher. She suffered a cerebrovascular accident in August
1994 resulting in a severe right hemiplegia and language
impairments, characterized by a severe spoken and written
production deficit while comprehension was largely preserved.
The EEG performed during the acute period revealed abnormalities in the left fronto-temporal area. The cerebral perfusion brain scintigraphy performed at the same period showed
diminished activity in the left temporoparietal area. No CT
scan was done during the acute period. However, a CT scan
was performed 15 months after the stroke (November 1995).
This CT scan showed a large cortico-subcortical hypodense
lesion in the left sylvian area, involving the fronto-temporal
lobe (see Figs. 2 and 3). Two months after her stroke, BA was
admitted to a rehabilitation program and was referred to the
first author (JM). At that time she still showed language disturbances and right hemiplegia. The present investigation took

Figs. 2 and 3. Two axial CT scans depicting a large cortico-subcortical infarct in the left sylvian arca involving the fronto-temporal lobe.

Gender identification

place within a nine-month period during which BA’s performance remained stable.

Downloaded by [Fondren Library, Rice University ] at 14:12 11 November 2014

Neuropsychological evaluation
Neuropsychological testing showed no impairment in tasks
exploring orientation to time and space. BA showed good
face recognition and presented no clinical signs of visual
agnosia (Protocole Montréal-Toulouse d’évaluation des gnosies
visuelles, Agniel et al., 1992). She presented with a normal
performance (26/30, mean, score available for women between
ages 65–74 years is 25.8) on the line orientation judgment test
(Benton et al., 1978) as well as on the Bells cancellation test
(BA’s score = 34/35) (Gauthier et al., 1989), and there were
no signs of unilateral neglect. BA showed normal performance on the constructional apraxia subtest of the PENO
neuropsychological evaluation battery (Joanette et al., 1995).
Her performance on Raven’s Colored Progressive Matrices
was within the normal range. The patient’s short term memory was normal in the visuospatial modality (span of 5 on the
Corsi block tapping procedure; Milner, 1971) whereas she
presented with difficulties in the visuoverbal modality (span
of 2). BA presented with mild impairment on long term memory
tasks (picture recognition) on the BEM 144 (Signoret, 1991).

Language evaluation
BA was classified as mixed aphasia on an aphasia battery
including the MT-86β (Nespoulous et al., 1992) on the basis
of spontaneous speech, confrontation naming, oral reading,
repetition, comprehension and written expression.

Auditory and visual input
With respect to input components, BA’s abilities appeared to
be largely preserved. She was flawless in written letters (65/
65) and CV syllable (82/82) discrimination (same vs. different judgment task) and she only produced one error out of the
42 stimuli (98% correct) for auditory CV and V syllable discrimination. Her performance in lexical decision tasks (word
vs. nonword decision task) also pointed to the integrity of the
patient’s phonological (58/60, 97% correct) and orthographic
(60/60, 100% correct) input lexicons.

Semantics and syntactic comprehension
Comprehension abilities at the lexical-semantic level also
appeared to be largely preserved. Her performance was largely
preserved in a word-picture matching task adapted from
Caplan and Bub’s (1990) Psycholinguistic Assessment of
Language (PAL) in which the patient had to choose one of
two pictures as a match to a word. Paired pictures of animals,
fruits, vegetables and object stimuli were controlled for lexical frequency, visual similarity, word length and semantic
relationship (e.g. orignal ‘moose’-chevreuil ‘deer’). None of
the 65 word pairs share phonological similarities. The patient

473

performed flawlessly (65/65, 100% correct) on the written
version of the task and she only produced 8 errors out of the 65
stimuli (88% correct) when words were presented auditorily.
BA’s good performance in the written word categorization
task also indicates the large preservation of the semantic processing of words. She only produced 5 errors out of the 120
stimuli (96% correct) in the following categorization tasks:
fruit/vegetable (18/20); land vehicle/water vehicle (17/18);
bird/fish (20/20); clothes/body parts (18/20); beverage/food
(20/20); town/country (22/22). At the syntactico-semantic level,
she performed normally in sentence-picture matching tasks for
reversible sentences (16/18) but obtained lower scores for
embedded relative clauses and passive forms (8/16).

Spontaneous speech
Spontaneous speech was marked by the presence of abundant
neologisms. BA sometimes resorted efficiently to oral spelling to
compensate for word-finding difficulties during conversation
(e.g. neologistic jargon: “ché, zé, zomaso, sétromato, bouail,
souail, okro de la…” followed by the correct oral spelling
FROMAGERIE ‘cheese dairy’). [From now on, capital letters
are used for the transcription of the patient’s oral spelling productions.]

Repetition, reading aloud and oral spelling
to dictation of word and nonword stimuli
Method. We used two word stimuli sets in repetition, reading and oral spelling to dictation. The first set consisted in 90
stimuli, equally divided in one, two and three syllables in length.
Half the stimuli were high frequency (Baudot, 1992) and half
were low frequency. The second set consisted in 70 words with
high (40) and low (30) imageability ratings (Content et al.,
1990) matched for lexical frequency (30 low frequency and 40
high frequency). Both word stimulus sets were also controlled
for orthographic complexity. Words were categorized as regular
(74) or irregular (86) following criteria established for writing to
dictation by Beauvois and Derouesné (1981). A set composed of
60 nonword stimuli, two and three syllables in length, was used
in repetition and reading aloud. Since BA was very reluctant in
regard to oral spelling to dictation of nonword stimuli, she was
offered a shorter set of 25 stimuli controlled for length.
Results. As indicated in Table 1, lexical frequency had no
effect in repetition whereas a lexical frequency effect was
observed in reading aloud (χ2(1) = 3.69, p = 0.05) and oral
spelling to dictation (χ2(1) = 8.92, p = 0.01). With respect to
length, a better performance was recorded for longer words in
repetition (Kendall’s correlation coefficient = 1.000, p < 0.01)
and reading aloud (Kendall’s correlation coefficient = 1.000,
p < 0.01) while oral spelling appeared to be independent from
this variable, whether it was measured by the number of
syllables or by the number of letters (Kendall’s correlation coefficient = 0.60) (see Table 1). There was also no significant
difference when the performance on one and two syllable words
was compared to the performance on three syllable words.

474

J. Macoir and R. Béland

Table 1. Percent correct by length and lexical frequency in repetition, reading
aloud and oral spelling to dictation of word stimuli

Downloaded by [Fondren Library, Rice University ] at 14:12 11 November 2014

Syllable length

Low
frequency (%)

High
frequency (%)

Total
(%)

A. Repetition
1 syllable
2 syllables
3 syllables

80
87
100

80
100
100

80
93
100

Total

89

93

B. Reading aloud
1 syllable
2 syllables
3 syllables

27
47
67

67
73
67

Total

47

69

C. Oral spelling to dictation
1 syllable
2 syllables
3 syllables

40
60
27

87
73
60

Total

42

73

47
60
67

63
67
43

Summary

Table 2. Percent correct by imageability and lexical frequency in repetition,
reading aloud and oral spelling to dictation of word stimuli
Imageability

respect the syllable length (89%) and structure (74%) of the
intended target. Except for two lexicalization errors in reading aloud, all the errors produced on nonword stimuli in both
spoken transposition tasks consisted in phonemic errors.
Most of the errors produced by BA in oral spelling to dictation (58%) were non phonological plausible errors (NPPE) (i.e.
spelling errors including letter substitutions, deletions, transpositions and additions; e.g. élérat for éléphant ‘elephant’).
She also produced a smaller proportion (16%) of phonological
plausible errors (PPE) (i.e. errors that are explicable in terms
of the application of phoneme-grapheme correspondence rules
in generating the response; e.g. éléfan for éléphant ‘elephant’),
formal paragraphias (18%) and written neologisms (8%). Except
for one neologism, all the errors produced on nonword stimuli
in oral spelling to dictation consisted in NPPE.

Low
frequency (%)

High
frequency (%)

Totals
(%)

A. Repetition
Low imageability
High imageability

90
90

95
90

93
90

Totals

90

92.5

B. Reading aloud
Low imageability
High imageability

40
60

40
80

Totals

53

60

C. Oral spelling to dictation
Low imageability
High imageability

50
80

55
85

Totals

70

70

40
70

53
82.5

Imageability had no effect in repetition whereas the difference between low and high imageability words was significant
in reading aloud (χ2(1) = 6.3, p < 0.05) and oral spelling to
dictation (χ2(1) = 6.9, p < 0.05) (see Table 2).
Orthographic complexity had no effect on BA’s performance
in oral spelling to dictation of word stimuli (regular words = 46/
74, 62% correct; irregular words = 55/86, 64% correct).
Finally, as compared to words, there was no difference in
BA’s abilities to repeat (χ2(1) = 1.92, p > 0.16), read aloud
(χ2(1) = 2.16, p > 0.14) and orally spell (χ2(1) = 0.41, p >
0.52) nonword stimuli. Moreover, BA’s performance in the
three tasks was not tied to stimulus length.
Error analysis. BA’s errors in repetition of the two word
stimulus sets consisted in phonemic (53%) and formal paraphasias (47%) in the same proportion. These errors largely
respect the syllable length (100%) and structure (87.5%) of
the intended target. Similar error types were also observed on
both stimulus sets in reading, with more phonemic (60%)
than formal (40%) paraphasias. As for repetition, these errors

In summary, BA presented with a spoken naming deficit in
spontaneous speech and picture naming tasks, marked by an
abundant production of neologisms. She produced phonemic
errors in reading aloud, whereas her performance was almost
perfect in repetition tasks, except for sentences. Written production was largely impaired and marked by the production
of non phonological plausible errors (NPPE). The patient’s
comprehension abilities are well preserved at the lexicosemantic level but are impaired at the syntactico-semantic
level. The spontaneous use of oral spelling to compensate for
word-finding difficulties in conversation was particularly fascinating and warranted a closer investigation.

Experiment 1. Picture naming
Method
BA’s ability to name pictures was assessed through a list of
stimuli composed of 373 line drawings of objects pertaining
to several living and nonliving semantic categories. The living
set comprised 145 drawings of animals, fruits, vegetables and
body parts; the nonliving set comprised 228 drawings of
vehicles, furniture, tools, kitchen utensils, buildings, clothing,
toys, musical instruments and miscellaneous. Stimuli were
distributed between low-frequency words (229) and highfrequency words (144) according to Baudot’s frequency tables
(Baudot, 1992). The mean frequency for the living and nonliving set was matched. The stimuli were presented in random
order. On each trial, BA was presented with a single line drawing of an object and was asked to produce its name orally or
by oral spelling. There was no time limit and the patient was
presented with approximately 35 items in each session.

Results
The distribution of BA’s spoken and orally spelled responses
in picture naming is presented in Table 3. In 7% (26/373) of
the stimuli, BA produced an orally spelled response only, and

Gender identification

475

Table 3. Distribution of BA’s spoken and orally spelled responses in picture naming (number and percent correct)

Downloaded by [Fondren Library, Rice University ] at 14:12 11 November 2014

Response type

Spoken

Orally spelled

Correct

Incorrect

Correct

Incorrect

Spoken response only
Orally spelled response only
Spoken response followed by oral spelling

52/145(35%)
–
26/202(13%)

93/145(64%)
–
176/202(87%)

–
12/26(46%)
111/202(55%)

–
14/26(54%)
91/202(45%)

Total

78/347(22%)

269/347(78%)

123/228(54%)

105/228(46%)

46% (12/26) of these responses were correct. In 39% of the
trials (145/373), she produced an oral response only, and was
correct in 36% (52/145) of these trials. For the remaining 202
picture stimuli, as reported in Table 3, BA produced first a
spoken response followed by an oral spelling response. In the
latter case, a new attempt to produce a spoken response was
observed 43 times. Her responses were correct in both spoken
and oral spelling modalities for 23 picture stimuli (23/202,
11%). For three pictures only (3/202, 1%), her correct spoken
production was followed by an incorrect orally spelled response;
the opposite pattern of response (i.e. incorrect spoken response followed by a correct orally spelled response) was
observed in 88 picture stimuli (88/202, 44%).
The overall difference in performance between spoken
(22%) and orally spelled (54%) naming was highly significant
(χ2(1) = 58.55, p < .001). The difference remained significant
when the analysis was restricted to the trials in which both a
spoken and an orally spelled response were produced (χ2(1) =
77.93, p < .001). There was no semantic category effect (spoken production: living = 29/127, 23% correct; nonliving = 49/
220, 22% correct/written output: living = 56/93, 60% correct;
nonliving = 67/135, 50% correct) while lexical frequency
appeared to influence the patient’s performance in both production modalities (spoken modality: low-frequency words =
41/218, 19% correct; high-frequency words = 37/129, 29%
correct) (χ2(1) = 3.99, p < .05)/oral spelling modality: lowfrequency words = 62/134, 46% correct; high-frequency
words = 60/94, 64% correct (χ2(1) = 6.16, p < .05). There was
no syllable length effect (Kendall’s correlation coefficient:
spoken modality: p = 0.1; written modality: p = 0.1) or phoneme/letter length effect (Kendall’s correlation coefficient:
spoken modality: p = 0.9; written modality: p = 0.2) on BA’s
performance.

Error analysis
Most of the errors (126/269, 47%) BA produced in spoken
output were neologisms for which the intended target was not
reliably recoverable (i.e. productions in which more than
50% of the phonemes differed from the target phonemes).
The syllable length of the intended target was respected in
only 57 of the 126 neologisms (45%), while 59/126 (47%)
differed by one syllable and 10/126 (8%) by two syllables [r =
.39, p < .0001]. The patient also produced a lower percentage of phonemic errors (26%), semantic paraphasias (10%),

Table 4. Distribution of spoken response types and corresponding orally
spelled response types in picture naming for the 202 trials in which a spoken
response was immediately followed by an orally spelled response
Spoken response

Orally spelled response
Correct NPPE PPE Neologism Semantic Formal
error
error

Correct
23
Neologism
34
Phonemic error
35
Semantic error
7
Formal error
8
Verbal unrelated error 4

2
30
19
3
7
5

1
–
–
–
1
–

–
2
–
1
–
2

–
8
–
3
–
2

–
3
–
–
2
–

formal paraphasias (9%) and verbal substitutions for which
the link with the intended target was not reliable (8%; e.g.
barbe “barb” ‘beard’ for drapeau “drapo” ‘flag’). Among the
105 errors produced by the patient in oral spelling modality,
77 (73%) were NPPEs. The remaining 28 errors were distributed as follows: semantic errors (18%), formal errors (5%),
neologisms (7%) and PPEs (2%).
Different response patterns were observed for the 202 trials
in which a spoken response was immediately followed by an
orally spelled response. The distribution of these response
patterns is presented in Table 4.

Summary
BA presented with a severe naming impairment, more marked
in spoken modality than in written modality. The proportion,
as well as the nature of the errors, in orally spelled picture
naming was similar to what was observed in the oral spelling
to dictation task. These results confirmed the deficit in accessing orthographic lexemes associated with a deficit in the written
output (see Fig. 1). As compared to repetition and reading,
the patient’s performance in spoken picture naming was
much more affected. Moreover, while she only produced
phonologically related errors (i.e. phonemic and formal paraphasias) in repetition and reading, spoken picture naming was
marked by an abundant production of non phonologically
related errors, i.e. neologisms. According to the lemma model,
the patient may access the information at the lemma level even
when unable to name the picture. The next experiment tests
the ability of the patient to retrieve gender information from
the picture of an object.

476

J. Macoir and R. Béland

Downloaded by [Fondren Library, Rice University ] at 14:12 11 November 2014

Experiment 2: Retrieval of gender
According to the lemma model presented in the Introduction,
the syntactic properties of a word are recovered during
lemma retrieval, at the first step of lexical retrieval. For
example, the lemma of the word fauteuil ‘armchair’ specifies
all its syntactic parameters including word class (noun), number (singular), and gender (masculine). Indeed in French, the
grammatical gender is assigned arbitrarily to animate (i.e. all
nouns that denote males or females) as well as to inanimate
objects and abstract concepts. Nouns are assigned one of two
grammatical genders, commonly termed ‘masculine’ and
‘feminine’. Different elements, as determiners (e.g. the indefinite articles “un” and “une” are assigned to masculine and to
feminine nouns, respectively) and pronouns (e.g. the personal
pronouns “il” and “elle” are assigned to masculine and
feminine nouns respectively), are applied according to this
grammatical gender. The patient’s performance in gender
identification may differ according to the presentation of a
morphological (un/une) or a syntactic one (il/elle). Indeed,
BA could match the noun with un/une on the basis of memorized lexical associations without accessing the gender of the
noun at the syntactic level. In contrast, the matching of a
noun to il/elle involves obligatorily the access to the gender
information at the syntactic level, since in French as in
English, il or elle are most of the time not adjacent to the
noun to which they refer (the only exception are sentences
like: J’aime ce man-teau, il est magnifique ‘I like this coat, it
is splendid’).
In Experiment 2, we explore the access to the syntactic
properties of words from the semantic system by means of a
gender-decision task performed on nouns evoked by pictures.

the corresponding name. In the “UN/UNE” modality, she was
requested to point to one of two cards with the labels “masculine–UN” and “feminine–UNE”. In the “IL/ELLE” modality,
she had to point to one of two cards with the labels “masculine–IL” and “feminine–ELLE”.

Results
BA could not identify the grammatical gender without trying
to name the picture first. When she was discouraged from
producing any spoken or written response, she explicitly
stated her inability to perform the task. Spoken naming was
correct for 34 stimuli (34/90, 38% correct) in the “UN/UNE”
modality and for 29 stimuli (29/90, 32% correct) in the “IL/
ELLE” modality. For the remaining stimuli, BA was unable
to give any spoken response, produced neologisms, phonemic
errors, formal paraphasias or semantic paraphasias (in the
“IL/ELLE” modality only) (see Table 5a). In the “UN/UNE”
modality, she produced a written response for 22 stimuli
among the 56 failed in spoken modality and was correct for
10 of them (10/22, 45% correct). The proportion of correct
written responses (15/36, 42% correct) was similar in the “IL/
ELLE” modality. The remaining written responses consisted
in NPPE and formal paragraphias.
BA’s performance in gender identification was much better
than her ability to produce a spoken or a written response,
whether it was in the “UN/UNE” (72/90, 80% correct) or in
the “IL/ELLE” (73/90, 81% correct) modality. As shown in
Table 5b , a correct spoken response was followed by a correct gender identification in a large proportion (“UN/UNE” =
32/34, 94% correct; “IL/ELLE” = 28/29, 96.5% correct). When
naming was impossible or erroneous, gender identification was

Method
BA was presented with 90 drawings of objects which were as
unequivocal as possible in terms of name agreement. The corresponding names were unambiguous with regard to gender,
i.e. all items whose semantic gender was conveyed by two
different roots (e.g. coq ‘rooster’–poule ‘hen’) as well as
items whose grammatical gender was predictable from natural gender (e.g. curé ‘priest’) were excluded. The stimulus set
comprised 43 feminine and 47 masculine nouns matched for
lexical frequency (Baudot, 1992), phonological ending regularity (ending usually associated with masculine (32) or feminine (21) nouns; gender non predictable from the noun
ending = 37) and orthographic ending regularity (ending usually associated with masculine (42) or feminine (27) nouns;
gender non predictable from the noun ending = 21). In the
first session, half the stimuli (45) were presented for the “UN/
UNE” decision task, the other half (45) for the “IL/ELLE”
decision task. The stimuli that were presented in the “UN/
UNE” modality in the first session were given in the “IL/
ELLE” modality in the second session and vice versa. On
each of the trials, the patient was presented with a single picture and was requested to identify the grammatical gender of

Table 5. Number and percentage of correct responses in the “UN/UNE” and
“IL/ELLE” gender-decision tasks on pictures. (a) Spoken picture naming performance and distribution of errors; (b) gender identification performance for
successful and failed items in spoken naming
UN/UNE
N(%)

(a) Spoken naming response
Correct
Phonemic error
Formal paraphasia
Semantic paraphasia
Neologism
No response
(b) Gender identification
Correct
Phonemic error
Formal paraphasia
Semantic paraphasia
Neologism
No response

IL/ELLE
Critical
value

34/90(38)
13/90(14)
4/90(4)

N(%)

Critical
value

29/90(32)
19/90(21)
2/90(2)
4/90(4)
22/90(24)
14/90(15.5)

22/90(24)
17/90(19)
32/34(94)
10/13(77)
3/4*(100)

22
9
4

18/22(82)
8/17(47)

15
12

28/29(96.5)
18/19(95)
2/2(100)
3/4*(100)
17/22(77)
5/14(36)

19
13
2
4
15
10

*Gender-decision was performed with 100% (4/4) accuracy on the formal and
the semantic paraphasias produced.

Downloaded by [Fondren Library, Rice University ] at 14:12 11 November 2014

Gender identification

correct for 40 stimuli (40/56, 71% correct) in the “UN/UNE”
modality, and for 45 stimuli (45/61, 74% correct) in the “IL/
ELLE” modality, a performance much better than chance
(critical value = 34, p = 0.5, α = 0.95). When BA was unable
to produce any response, her performance in gender-decision
did not differ from chance (“UN/UNE” = 8/17, 47% correct;
“IL/ELLE” = 5/14, 36% correct; critical value = 12, p = 0.5,
α = 0.95). However, her ability to identify the grammatical
gender of nouns for which she produced phonemic errors,
formal paraphasias, semantic paraphasias or even neologisms
remained much better than chance (see Table 5b).
BA’s performance was not tied to lexical frequency (“UN/
UNE”: low frequency = 42/54, 78% correct, high frequency =
30/36, 83% correct; “IL/ELLE”: low frequency = 43/54, 80%
correct, high frequency = 30/36, 83% correct), gender (“UN/
UNE”: masculine = 35/47, 74% correct, feminine = 37/43,
86% correct, χ2(1) = 1.23, NS; “IL/ELLE”: masculine = 38/
47, 81% correct, feminine = 35/43, 81% correct), phonological ending regularity (“UN/UNE”: gender not predictable =
33/37, 89% correct; regular ending = 39/53, 74%; χ2(1) =
2.41, NS; “IL/ELLE”: gender not predictable = 32/37, 86%
correct; regular ending = 41/53, 77%, χ2(1) = 0.66, NS) or
orthographic ending regularity (“UN/UNE”: gender not
predictable = 17/21, 81% correct; regular ending = 55/69,
80% correct; “IL/ELLE”: gender not predictable = 16/21,
76% correct; regular ending = 57/69, 83% correct, χ2(1) =
0.53, NS).

Summary and comment
In summary, the data from Experiment 2 suggest that the
retrieval of the grammatical gender: (1) was largely preserved
in BA, (2) did not differ according to the gender agreement
markers, and (3) did not depend on the correct activation of
the corresponding phonological and/or orthographic forms.
BA’s ability to correctly identify the grammatical gender of
words for which she produced spoken errors was good, even
when the production was not target-related. However, the
access to gender information appeared to be completely
blocked when the patient was unable to give any spoken or
written response. BA produces a few semantic errors in the
IL/ELLE modality, suggesting a deficit in accessing phonological lexemes. When she produces phonemic errors, formal
paraphasias, or paragraphic errors, we assume that access to
the lexeme level is at least partial. However, when she produces no response, we cannot assume that she has no access
to any information at the lexeme level. The next experiments
attempt to probe word retrieval at the lexeme level.

477

Experiment 3a. Cued picture naming
Cueing is frequently used to determine the partial knowledge
of words that patients may still have for words they cannot
produce (e.g. Goodglass et al., 1976) or to improve their performance naming tasks (e.g. Bachy-Langedock and De Partz,
1989; Ward et al., 2000) by providing sufficient additional
activation to the phonological output lexicon to allow spoken
word production.

Method
The patient was presented with two lists of stimuli. List I was
composed of 55 drawings of objects pertaining to several living and nonliving semantic categories taken from the Batterie
d’examen des troubles en dénomination (Bachy-Langedock,
1989). The task was performed in two separate sessions. In
the first session, half the stimuli (27) were presented for spoken naming with a phonological cue (first phoneme) and the
other half (28) for written (oral spelling) naming with an
orthographic cue (First letter) (e.g. picture = cheminée ‘chimney’: phonological cue = “ch”; orthographic cue = C). Stimuli
for which the phonological cue was given in the first session
were presented in the second session with an orthographic cue
and vice versa. List 2 was composed of 51 drawings of objects
taken from the same testing battery (Bachy-Langedock, 1989)
and corresponding to bisyllabic or trisyllabic words. The procedure was identical to the one used in testing list 1, except
for the nature of the cue which corresponded either to the first
spoken syllable (e.g. picture = cheminée ‘chimney’: phonological cue = “che”) or the first orthographic syllable (e.g. picture = cheminée ‘chimney’, orthographic cue = CHE. When an
orthographic cue was presented, BA was instructed not to read
the syllables aloud.

Results
Table 6 shows the number and percentage of correct responses in the normal condition and two cued conditions of the
oral and written picture naming tasks. No significant cueing
effect was found, regardless of whether a phonemic or an orthographic cue was used. However, the superiority of written over
oral naming abilities was observed in the two cued picture
naming tasks. The difference between spoken and written
naming was significant, regardless of whether the cueing was
performed with the first letter (χ2(1) = 10.68, p < .01) or the
first syllable (χ2(1) = 5.07, p < .05).
Table 6. Number (%) of correct responses in the three different conditions
of picture naming

Experiment 3: Lexeme retrieval and word
production
We selected two tasks: the Cued picture naming task and the
Rhyming/orthographic ending word production task. The first
task probes the knowledge of the patient for word beginnings,
while the second probes her knowledge for word endings.

Normal condition
1st phoneme/letter cue
1st spoken/written syllable cue

Spoken naming

Written naming

χ2

78/347 (22%)
15/55 (27%)
13/51 (25%)

123/228 (54%)
33/55 (60%)
25/51 (49%)

***
**
*

Oral vs. written picture naming; ***p < 0.001; **p < .01; *p < .05.

Downloaded by [Fondren Library, Rice University ] at 14:12 11 November 2014

478

J. Macoir and R. Béland

Experiment 3b. Rhyming/orthographic ending
word production

Experiment 4: Lexeme retrieval without word
production

Method

In Experiment 4, BA was administered three different tasks
exploring access to lexemes without any involvement of
articulation mechanisms. In these tasks, her performance was
compared to the results of the three control subjects.

A list of 40 concrete word stimuli from the stimulus set used
in the picture naming task was presented to the patient. In the
spoken modality, she had to produce a spoken word, rhyming
with an auditorily presented stimulus (e.g. râteau “rato” ‘rake’→
photo “foto” ‘photo’). In the written modality, BA was asked
to spell aloud a word with a similar orthographic ending to a
written presented stimulus (e.g. RÂTEAU ‘rake’ → CHAPEAU
‘hat’). In the first session, half the stimuli (20) were presented
in written modality and the other half (20) in spoken modality.
The stimuli that were presented in written modality in the first
session were given in spoken modality in the second session
and vice versa. In this task, BA’s performance was compared
to the results of three female controls matched for age and
educational level (mean age = 76.3 years, mean education =
13 years).

Experiment 4a. Length judgment on pictures
According to the two-stage model of lexical retrieval, the
information related to word length (i.e. metrical information)
is available at the lexeme level. In experiment 5a, access to
this information from the semantic system was assessed through
a task in which the patient had to perform a length judgment
on words evoked by pictures.

Method

BA encountered serious difficulties when the stimulus was
presented auditorily and the task was suspended at her
request after 24 trials. In this condition, she only produced 8
rhyming words (8/24, 33% correct) which also respected the
orthographic ending of the target (e.g. échelle “échel” ‘ladder’
→ pelle “pel” ‘shovel’). For the remaining trials, she was unable
to produce any response for 13 of the stimuli and produced 2
non-rhyming words and a neologism. Her performance was
much better (33/40, 82.5% correct) in the written modality.
The seven failed trials consisted in no responses. Most of the
correctly orally spelled words (29/33, 88%) also respected the
phonological ending of the target (e.g. RIDEAU “rido”
‘curtain’ → CHAPEAU “chapo” ‘hat’) but in 4 of them (4/
33, 12%) the phonological form of the response differed from
the phonological ending of the target (e.g. OREILLER
“oreillé” ‘pillow’ → CHER “cher” ‘expensive’). The difference in correct performance for the two conditions was highly
significant (χ2(1) = 13.69, p < .001). Controls were flawless in
this task in both modalities.

BA was presented with 60 pairs of picture stimuli, vertically
aligned in the center of individual cards. For each trial, she
was asked to perform a length judgment on the spoken corresponding names (selected to be as unequivocal as possible
according to name agreement). The patient was instructed to
point to the picture for which the corresponding name takes
longer to pronounce or to point to both pictures when the
corresponding names have the same spoken length. The 60
pairs of stimuli were equally divided in four categories according to letter and syllable length similarity of the corresponding names (there were no final silent ‘e’s in the corresponding
names of the picture stimuli): (a) same number of syllables
but different number of letters (e.g. fauteuill “foteuil” ‘armchair’ –
canot “kano” ‘canoe’; expected response = ‘same’); (b) different number of syllables but same number of letters (e.g. fleur
“fler” ‘flower’ – balai “balai” ‘broom’; expected response =
‘different’); (c) same number of syllables and letters (e.g.
main “min” ‘hand’ – chat “cha” ‘cat’; expected response =
‘same’); and (d) different number of syllables and letters (e.g.
pain “pin” ‘bread’ – domino “domino” ‘domino’; expected
response = ‘different’). Stimuli were randomly presented to
the patient, who was discouraged from attempting to name
the corresponding words.

Summary and comments

Results

These results confirmed the lexical localization of BA’s
spoken production deficit as well as the better preservation of
lexical access to orthographic than to phonological lexemes.
However, it still remains possible that: (1) the actual word
production, requiring the activation of the output phonological buffer and articulation mechanisms, played a role in the
patient’s performance, and (2) the short term memory deficit
could influence her performance in the spoken modality as
compared to the written modality, in which the stimulus
remained visible during the task. These possible influences
were controlled in the next experiment.

As shown in Table 7, BA’s performance appeared to be
greatly influenced by the orthographic length of the stimuli.
Whereas BA’s performance was comparable to the controls
for the trials in which the letter and syllable length were the
same (categories c and d), large deficits were observed when
she had to perform the spoken length judgment on the trials in
which the letter and syllable length were different (categories
a and b) (see Table 7).
The patient resorted to an orthographically-oriented strategy to perform the length judgment on pictures. She was at a
control level when spoken and written lexical forms were the

Results

Gender identification
Table 7. Performance of BA (number (%) of correct responses) and controls
(mean number (%) of correct responses and range) in the length judgement on
picture tasks according to letter and syllable length concordance
Stimulus categories Expected response BA

Downloaded by [Fondren Library, Rice University ] at 14:12 11 November 2014

(a) Same # of syll./
diff. # of letters
(b) Diff. # of syll./
same # of letters
(c) Same # of syll./
same # of letters
(d) Diff. # of syll./
diff. # of letters

Controls

Same

7/15(47%) 14/15(93%) (13–15)

Different

9/15(60%) 13/15(87%) (12–13)

Same

14/15(93%) 14/15(93%) (13–15)

Different

13/15(87%) 15/15(100%)

same length. For items in which there was a discrepancy between the written and spoken form, BA’s strategy was ineffective and led to substantial difficulties.

Experiment 4b. Letter/phoneme identification
on pictures
In this experiment, access to lexemes from the semantic system
was assessed through a task in which the patient had to judge
the presence or absence of a particular letter or phoneme in
words evoked by pictures.

Method
BA was presented with 80 drawings of objects pertaining to
several living and nonliving semantic categories which were
as unequivocal as possible according to name agreement. For
each stimulus, she was asked to determine if the name of the
pictured object comprised a specific letter or phoneme1 (e.g.
“If you had to write/pronounce the name corresponding to
this picture, would you use the letter T/would you say the
sound /t/?”). The stimuli were equally divided among the following four categories: a) the corresponding name comprises
the phoneme but not the letter (e.g. /o/, ‘O’ in chapeau ‘hat’/
chapo/); (b) the corresponding name comprises the letter but
not the phoneme (e.g. /m/, ‘M’ in trompette ‘trumpet’/trompette/); (c) the corresponding name comprises both the phoneme
and the letter (e.g. /t/, T in bateau ‘boat’/bato/); and (d) the
corresponding name comprises neither the phoneme nor the
letter (e.g. /i/, I in vache ‘cow’/vache/). Stimuli were randomly
presented within each modality. In the first session, half the

479

stimuli (40) in each category were presented in the written
modality (question probing the presence or absence of a letter)
and the other half (40) in the spoken modality (question probing the presence or absence of a phoneme). Stimuli presented
in the written modality in the first session were given in the
spoken modality in the second session and vice versa. The
patient was discouraged from attempting to name the picture.

Results
Table 8 shows the number and percentage of correct responses in phoneme judgement and letter judgment. Results
indicate that BA was better at judging the presence or
absence of a particular letter in the name of a pictured object
(74/80, 92.5% correct; mean score for the controls = 78.4/80,
98%; SD = 1.1, range = 77–80) than the presence or absence
of a particular phoneme (47/80, 59% correct; mean score for
the controls = 76.8/80, 96%; SD = 2.2, range = 75–80) and the
difference was highly significant (χ2(1) = 22.92, p < .001).
The results in the written modality (letter judgment) were
not tied to the degree of transparency in the spelling-to-sound
correspondence since BA’s performance was 80% or better in
all four categories (see Table 8). Results in the phoneme
judgment showed that, except for category c in which the
name of the pictured object comprises both the letter and the
phoneme, BA’s scores were lower than the control subjects’
scores, especially in category b (see Table 8).

Experiment 4c. Pseudohomophone/picture
matching
This experiment required a comparison between the phonological form activated through the lexical/semantic route and
the phonological form activated through the non-lexical reading route.

Method
BA was given a task drawn from the case studies reported by
Bub and Kertesz (1982) and by Caramazza et al. (1983). In
this task, from two pictured objects the patient had to choose
the one corresponding to a written pseudohomophone, Pseudohomophones are orthographically legal pseudowords that
would “sound like” French words when pronounced using

Table 8. Performance of BA (number (%) of correct responses) and controls (mean number (%) of correct responses) in the letter/phoneme identification
on picture task (+ stands for present, − for absent)
Stimulus categories

BA

Controls

Phon. identif.

Let. identif.

Phon. Identif.

Let. identif.

(a) Phoneme+/letter−
(b) Phoneme−/letter+
(c) Phoneme+/letter+
(d) Phoneme−/letter−

14/20(70%)
4/20(20%)
17/20(85%)
12/20(60%)

19/20(95%)
16/20(80%)
19/20(95%)
20/20(100%)

18.7/20(93.5%)
18.7/20(93.5%)
19.7/20(98.5%)
19.7/20(98.5%)

18.7/20(93.5%)
19.7/20(98.5%)
20/20(100%)
20/20(100%)

Total

47/80(59%)

74/80(92.5%)

76.8/80(96%)

78.4/80(98%)

Downloaded by [Fondren Library, Rice University ] at 14:12 11 November 2014

480

J. Macoir and R. Béland

the application of spelling-to-sound conversion rules (e.g.
LAPEINT pronounced “lapin” lapin ‘rabbit’). Pseudowords
differed from words by only one grapheme (e.g. EINT in
LAPEINT) and this grapheme was always of a lower sublexical frequency than the corresponding grapheme of the target
word (e.g. in LAPEINT, EINT is less frequent than IN in
LAPIN). BA was presented with 60 stimuli equally divided in
three categories. In each category, the pseudohomophone and
the name of the target picture differed by two letters while the
distance, measured by the number of letters that were changed,
between the pseudohomophone and the name corresponding
to the foil picture was manipulated. In category a, the
pseudohomophone and the name of the foil picture differed
by one letter (underlined in the example) (e.g. for the pseudohomophone BAON “ban” (homophone of BANC ‘bench’)
the name of the foil picture was PAON “pan” ‘peacock’); in
category b, the name of the foil picture differed by two letters
from the pseudohomophone (e.g. for the pseudohomophone
SOURSID “soursi” (homophone to SOURCIL ‘eyebrow’), the
name of the foil picture was SOURIS “souri” ‘mouse’); and
finally in category c, the pseudohomophone and the name of
the foil picture differed by three letters (e.g. for the pseudohomophone TOUPPIS “toupi” (homophone of TOUPIE
‘spinning top’), the name of the foil picture was TAPIS “tapi”
‘carpet’). The lexical frequency of the target name and the
foil name was controlled within each category. According to
the phonological form, the target name and the foil name differed by either one (53 trials) or two phonemes (7 trials).
Stimuli were randomly presented in two testing sessions. The
patient was discouraged from attempting to read the pseudohomophones aloud.

Results
BA only made 9/20 (45%) correct matches in category a, in
which the orthographic form of the pseudohomophone was
more similar to the foil name than to the target name. Her performance was very impaired relative to the controls (mean
score = 18.3/20, 92%; SD = 1.6; range = 16–20). In category b,
her score was 15/20 (75%), which was lower than the mean
score of the controls (mean score = 19/20, 95%; SD = 1; range
= 18–20). Finally, BA performed flawlessly (20/20, 100% correct) in category c, a performance similar to the controls, who
all obtained a perfect score of 20/20, 100% in this category.
BA’s performance in this experiment thus appeared to be
largely tied to the orthographic form of the stimuli since she
tended to perform the matching task by choosing the picture
whose orthographic form was closest to the pseudohomophone in both categories a and b. Her score in category b,
although slightly higher, was still lower than the mean score
of the controls.

Summary and comments
To summarize, the results from Experiment 4: (1) confirmed
that BA presented with a deficit in accessing the lexeme level

and that the deficit was more marked in the spoken than in the
written modality; (2) showed that this dissociation was not
tied to the nature of the input code or to the differential task
demands; and (3) confirmed that this dissociation was not
tied to post-lexical deficits since it was also observed in tasks
requiring no spoken or written word production.

Discussion
BA’s ability to correctly identify the grammatical gender of
words was largely preserved; Moreover, this preservation of
access to syntactic information did not seem to depend on the
correct activation of their corresponding spoken or orthographic forms. Except in the case where she could not produce
any spoken or written response, BA identified the gender of
nouns, even when she could not retrieve any phonological
information (i.e. neologisms) or when this retrieved information was degraded (i.e. phonemic and formal paraphasias).
Her performance in gender decision was very similar after the
morphological (un/un) and after the syntactic (il/elle) cue.
Given that her performance in gender decision was independent of her performance in spoken naming of the pictures,
this indicates that she was not relying on the memorized lexical
associations after the un/une cue. Otherwise, she would have
shown a better performance in picture naming than in the un/
une gender identification task. Moreover, the difficulty in
accessing phonological representations sometimes resulted in
the production of semantic paraphasias because of the concurrent activation of a phonologically related item. This pattern
of performance provides additional evidence for the independence of access to syntactic and phonological information in
the output lexicon. We believe that it also clearly supports the
two-stage model of lexical access and could readily be
accounted for by resorting to the lemma proposition (Levelt
et al., 1999) in which access to lexical syntax is required
before access to phonological form can take place. In contrast, the “IN model” (Caramazza, 1997) hardly accounts for
this performance since, in this model, gender identification
should be dependent on the efficiency of lexeme retrieval.
Both models predict a possible dissociation in the access to
the orthographic and phonological lexeme. As a matter of
fact, data from all the experiments we conducted showed that
BA privileged access to the orthographic lexeme. A similar
dissociation was previously documented for English (Patterson
and Shewell, 1987; Hillis et al., 1999), Italian (Semenza et al.,
1992) as well as French patients (Lhermitte and Derouesné,
1974; Michel, 1979). Furthermore, in the present study, we
showed that the dissociation held in tasks that do not involve
a spoken output, thus confirming that the dissociation is
located at the lexeme level. Indeed, if BA was suffering from
a disconnection between the phonological lexeme level and
the spoken output (see Fig. 1), then she should be unimpaired
in tasks which tap the phonological lexeme level only (Experiments 4 a, b, and c).
In the lemma model, we suggest that this pattern of performance mainly resulted from an impairment in the activation

Downloaded by [Fondren Library, Rice University ] at 14:12 11 November 2014

Gender identification

flow of lexical retrieval. More specifically, BA presented with
a partial disconnection between the lexical semantic level and
the phonological output lexicon. This impairment leads to the
production of different error types, depending on the input
code and the source of lexical activation. Her repetition abilities are largely preserved because they benefit from a direct
intra-code mapping between input and output phonological
representations and from a partial activation of the whole-word
lexical semantic route. However, BA produced neologisms in
sentence repetition probably because the acoustic trace decayed
more rapidly than in word repetition. The phonological representations also benefit from non-lexical and lexical-semantic
activation in reading, but this ability appeared to be more
affected because of the ambiguous mapping between orthography and phonology. In BA, the summation of both sources
of activation is insufficient to achieve the complete retrieval
of the phonological lexemes in the output lexicon, giving rise
to phonemic and formal para-lexias. This lexical origin was
confirmed by lexical frequency and imageability effects on
the patient’s performance in reading. As already suggested by
Best (1995) for picture naming, we also suggest that the reverse
length effect found in BA’s reading aloud of word stimuli is
consistent with a deficit in accessing phonological lexical
representations. According to this interpretation, such a deficit is more likely to affect the lexical retrieval of short words
as compared to longer words. Short words may have a large
number of phonological neighbours in the lexicon whereas
long words frequently have none. Therefore, a deficit in
retrieving lexical spoken forms is more likely to result in the
production of a phonologically related word for short words
than for long words. Neologisms were found in spontaneous
speech and in picture naming whereas they are completely
absent from the sample in word repetition and reading aloud,
two tasks in which the patient benefits from an oral or written
input to retrieve the phonological lexemes. In spontaneous
speech and in picture naming, the patient receives no oral or
written input and thus must retrieve the phonological lexeme
without linguistic cues. Thus it appears that BA produced
neologisms whenever access to the phonological representations was blocked, an interpretation already proposed by
Buckingham (1993) and Butterworth (1992). However, the
experiment in gender identification revealed that BA’s performance was better when she produced a neologism than
when she produced no response. This result calls for a distinction between a blocked access resulting in the production
of a neologism and a blocked access resulting in a no
response. The interaction we found between gender and phonological activation suggests that neologisms result from a
noisy activation coming from the semantic level. This noisy
activation, as shown in the error analysis in the picture naming task, contained no information with respect to the nature
of the phonemes or to the number of syllables of the target
word. Therefore, while this activation is sufficient to activate
gender information, it seemed to be insufficient to activate
the corresponding phonological lexeme, giving rise to the
production of a neologism. In some trials, access to the corre-

481

sponding orthographic lexeme is better preserved and BA
produced a written response to compensate for her difficulties
in spoken production. In the case of no responses, the activation is too weak to reach the lemma level and, consequently,
the phonological or orthographic lexeme levels. Consistent
with other cases reported in the literature, the disconnection
observed in BA between the lexical semantic level and the
phonological output lexicon leads to the production of semantic errors in picture naming (Caramazza and Hillis, 1990; Rapp
and Goldrick, 2000; Béland and Mimouni, 2001). However,
the proportion of these errors in BA’s performance decreased
in the gender identification task in which she both named the
picture and identified the gender. A possible explanation for
this decrease in the production of semantic errors is that the
additional activation of the lemma level required for gender
identification has the effect of reducing the noisy activation
coming from the semantic level. Further investigation is needed
to understand how the activation of lemma level mediates the
mapping between the semantic and the phonological representations.

Acknowledgements
We are grateful to Johanne Bernier and Lise Gagnon for their
contribution to BA’s neuropsychological assessment. We
thank Dr Lucie Morissette for her help in the interpretation of
radiological data. We are also grateful to two anonymous
reviewers for helpful comments on a previous version of this
paper. The first author was supported by a Chercheur-clinicien
Boursier award from the Fonds de la Recherche en Santé du
Québec (FRSQ–003186). The second author was supported
by the FRSQ (01ER66426) and by the Social Sciences and
Humanities Research Council of Canada (410-2002-0752).

Note
1

French like English has a deep orthography, that is, there is
no one-to-one mapping in sound-to-spelling correspondences
or spelling-to-sound correspondences. For instance, the phoneme/o/has different graphemic correspondences (O, EAU, AU,
OS, etc.) and the grapheme “EN” has many phonemic correspondences (e.g. “an” in enfant “anfan” ‘child’ and “in” in
moyen “moyin” ‘mean’)

References
Agniel A, Joanette Y, Doyon B, Duchein C, Protocole Montréal-Toulouse
d’évaluation des gnosies visuelles – PEGV. Isbergues: 1: Ortho-Édition, 1992.
Bachoud-Lévi AC, Dupoux E. An influence of syntactic and semantic variables
on word form retrieval. Cognitive Neuropsychology 2003; 20: 163–88.
Bachy-Langedock N. Batterie d’examen des troubles en dénomination. Brussels: Editest, 1989.
Bachy-Langedock N, De Partz M-P. Coordination of two rcorganization therapies in a deep dyslexic patient with oral naming disorder. In: Seron X,
Deloche G, editors. Cognitive approaches in neuropsychological rehabilitation. Hillsdale: Lawrence Erlbaum Associates, 1989: 211–47.
Badecker W, Miozzo M, Zanuttini R. The two-stage model of lexical retrieval:
Evidence from a case of anomia with selective preservation of grammatical
gender. Cognition 1995; 57: 193–216.

Downloaded by [Fondren Library, Rice University ] at 14:12 11 November 2014

482

J. Macoir and R. Béland

Baudot J. Fréquences d’utilisation des mots en français écrit contemporain.
Montréal: Les Presses de l’université de Montréal, 1992.
Beauvois MF, Demuesné J. Lexical or orthographic agraphia. Brain 1981; 104:
21–49.
Béland R, Mimouni Z. Deep dyslexia in the two languages of an arabic-french
bilingual patient. Cognition 2001; 82: 77–126.
Benton AL, Varney NR, Hamsher K, de S. Visuo-spatial judgement: A clinical
test. Archives of Neurology 1978; 35: 364–7.
Best WMA. Reverse length effect in dysphasic naming: When elephant is
casier than ant. Cortex 1995; 31: 637–52.
Brown AS. A review of the tip-of-the-tongue experience, Psychological Bulletin
1991; 109; 204–23.
Bub D, Kertesz A. Evidence for lexicographic processing in a patient with preserved written over oral single word naming. Brain 1982; 105: 697–717.
Buckingham HW. Disorders of word-form processing in aphasia. In: Blanken G,
Dittmann J, Grimm H, Marshall JC, Wallesch C-W, editors. Linguistic disorders and pathologies – An international handbook. New York: W. de Gruyter,
1993: 187–97.
Butterworth B. Disorders of phonological encoding. Cognition 1992; 42: 261–86.
Caplan D, Bub D. Psycholinguistic Asscssment of Aphasia. Paper presented
at the American Speech and Ilearing Association Conference, Seattle,
1990.
Caramazza A. How many levels of processing are there in lexical access?
Cognitive Neuropsychology 1997; 14: 177–208.
Caramazza A, Berndt RS, Basili AG. The selective impairment of phonological processing: A case study. Brain and Language 1983; 18: 128–74.
Caramazza A, Hillis AE. Where do semantic errors come from? Cortex 1990;
26: 95–122.
Caramazza A, Miozzo M. The relation between syntactic and phonological
knowledge in lexical access: Evidence from the ‘tip-of-the-tongue’ phenomenon. Cognition 1997; 64: 309–43.
Content A, Mousty P, Radeau M. Brulex, une base de données lexicales informatisées pour le français écrlt et parlé. L’Année Psychologique 1990; 90:
551–66.
Ferrand L. Grammatical gender is also on the tip of French tongues. Current
Psychology Letters: Behavior, Brain and Cognition 2001; 5: 7–20.
Gauthier L, DeHaut F, Joanette Y. The Bells Test: A quantitative and qualitative test for visual neglect. International Journal of Clinical Neuropsychology 1989; 11: 49–54.

Goodglass II, Kaplan E, Weintraub S, Ackerman N. The “tip-of-the-tonguc”
phenomenon in aphasia. Cortex 1976; 12: 145–53.
Hillis AE, Boatman D, Hart J, Gordon B, Muking sense out of jargon: A neurolinguistic and computational account of jargon aphasia. Neurology 1999;
53: 1813–24.
Joanette Y, Ska B, Belleville S, Lecours AR, Peretz I, Poissant A. Évaluation
neuropsychologique dans la démence de type Alzheimer: un compromis
optimal. L’année gérontologique 1995; 3: 69–83.
Levelt WJM, Roelofs A, Meyer AS. A theory of lexical access in speech production. Behavioral and Brain Sciences 1999; 22: 1–75.
Lhermitte F, Deroucsné J. Paraphasje et jargonaphasie dans le langage oral avec
conservation du langage écrit. Revue Neurologique 1974; 130: 21–38.
Miceli G, Caramazza A. Dissociation of inflectional and derivational morphology. Brain and Language 1988; 35: 24–65.
Michel F. Préservation du langage écrit malgré un déficit majeur du langage
oral, Lyon Médical 1979; 241: 141–9.
Milner B. Interhemispheric differences and psychological processes. British
Medical Bulletin 1971; 27: 273–7.
Nespoulous J-L, Lecours AR, Lafond D, Lemay A, Puel M, Joanette Y, Cot F,
Rascol A. Protocole Montréal-Toujouse d’examen linguistique de l’aphasie.
MT-86 Module Standard Initial: Mlâ. 2nd edition revised by Renée Béland
& Francine Gironx, Isbergues: L’Ortho-Édition, 1992.
Patterson K, Shewell C. Speak and spell: Dissociations and word-class effects.
In: Coltheart M. Sartori G, Job R, editors. The cognitive neuropsychology of
language. London: Lawrence Erlbaum Associates, 1987: 273–94.
Rapp B, Goldrick M. Discreteness and interactivity in spoken word production,
Psychological Review 2000; 107: 460–99.
Semenza C, Cipolotti J., Denes G. Reading aloud in jargonasphasia: An
unusual dissociation in speech output. Journal of Neurology, Neurosurgery,
and Psychiatry 1992; 55: 205–8.
Signoret JL. Batterle d’efficience mnésique BEM 144. Paris: Éditions scientifiques, Elsevier, 1991.
Vigliocco G, Antonini T, Garrett MF. Grammatical gender is on the tip of Italian
tongues. Psychological Sciences 1997; 8: 314–7.
Vigliocco G, Vinson DP, Martin RC, Garrett MF. Is ‘count’ and ‘mass’ information available when the noun is not? An investigation of the tip-of-the-tongue
states and anomia. Journal of Memory and Language 1999; 40: 534–58.
Ward J, Stolt R, Parkin AJ. The role of semantics in reading and spelling: Evidence for the ‘summation hypothesis’. Neuropsychologia 2000; 38: 1643–53.