Professional Documents
Culture Documents
FROM 4 TO FOUR
A SUPPLEMENT TO 'FROM THREE TO 3'
INTRODUCTION
In a previous study we investigated the abilities of two groups of aphasic patients to
transcode integer quantities from alphabetically written forms (numerals) to Arabic
digit strings (numbers) (Deloche and Seron, 1982). The one-to-one correspondence
between the two coding systems, the small number of lexical primitives, the
simplicity of the grammars, and the frequent occurrence of the task in everyday life
for writing cheques, were why we chose this domain to investigate the pattern of
dissolution of aphasic subjects' performances in transcoding tasks theoretically
equivalent to translations from one language or modality to another. The results
indicated that Broca's (and more particularly those subjects with agrammatism) and
Wernicke's aphasics were differentially impaired in a way that paralleled the general
dissociation of linguistic abilities of the two types of aphasia along the syntactic/
semantic dimensions. It could thus be tentatively concluded that, despite its formal
autonomy with reference to language, the alphabetical coding system for quantities
and the psycholinguistic procedures of that system are not disrupted independently
of the other linguistic skills. This conclusion seems at variance with classic
observations according to which the reading of numbers is often preserved in alexic
syndromes (Hecaen and Kremin, 1976). As indicated by Coltheart (1980), the
'number preservation' effect found with ideographically written forms (numbers)
does not seem to have been systematically investigated in the alphabetical coding
system. Even Hecaen and Kremin (1976) did not actually compare their aphasic
patients' performances in the two scripts since only digital forms were used in their
study. The results of our previous study seemed to support the hypothesis of a
general disorder affecting the manipulation of hierarchical structures per se in
736 XAVIER SERON AND GERARD DELOCHE
Broca's aphasics, that is, the structures that underlie the arrangement of words in
sentences, the arrangement of lexical primitives in numeral forms, or the organiza-
tion of branches and nodes in trees as reported by Grossman (1980).
In the present study, we are concerned with the ability of Broca's and Wernicke's
aphasics to transcode integer quantities from the digital to the alphabetical system.
This task is the reverse of that used in our previous study but, given the peculiarities
of the coding systems and particularly the differences of serial ordering constraints,
the two tasks are far from being simple inverses of each other, as has been shown by
Power and Longuet-Higgins (1978), who developed transcoding algorithms in the
two directions. The present paper continues our previous study by analysing the
subjects' performance when handling a task that involves different psycholinguistic
procedures that cannot be simply inverted and that operate on essentially different
material, namely, ideographic versus alphabetical scripts (see Besner and Coltheart,
RESULTS
did not change the phonemic reading of the numeral lexicon elements, such as
'CINQUE' for 'CINQ' (5), were not registered as errors, nor was the incorrect use of
the letter'S' indicating the plural in some numeral lexicon elements. The omission of
the symbol '-' (hyphen) between some elements or the presence of monetary
indications ('Francs' or 'Frs') were also not considered errors.
With these conventions, 378 transcriptions were incorrect (mean frequency: 0.27).
The Broca's aphasics committed 246 errors and the Wernicke's only 132. Twelve
numbers were correctly transcoded by all the subjects, among them 10 lexical
primitives. The error frequency for the transcriptions of one-word numerals was
0.02 (2/126) for units, 0.13 (11/84) for simple tens (one-word tens names), but 0.26
(22/84) for particulars. It was 0.38 (16/42) for composed tens (two- or three-word
tens names). Since the two groups of patients were not equivalent in terms of a
criterion of severity of aphasia, the question of possible double dissociation of
Qualitative Analysis
We distinguish three sources of errors in relation to the cognitive processes: errors
committed during the recognition of the digits and/or the production of their
numeral name, serial ordering errors in parsing from left to right in the digit string,
and errors in the transcoding process itself.
Digit recognition and numeral lexicon production errors. Verbal, literal, phonemic,
morphemic and neologistic paragraphias. This set contains the 23 erroneous
transcriptions where one letter string at least was alien to the numeral lexicon.
The sequence of letters may be a word (verbal paragraphia) or not, and it may
or may not share most of its units with the correct form at some level (letters,
phonemes, morphemes, or neologisms). The Broca's aphasics produced 19 such
errors, and one Wernicke's aphasic was responsible for the remaining 4. In one case,
this subject made a verbal literal paragraphia: '200' (DEUX CENTS) -> 'DEUX
CENTRE' (two centre). In 3 other cases, the errors were neologistic as in '71'
(SOIXANTE ET ONZE) -+ 'SEPTON ONZE'. The small number of neologistic
productions in the corpus is probably due to two biases: first, low-level literal
paragraphia was a selection criterion; second, orthographic errors were ignored
when scoring the subjects' productions. This small number must not, consequently,
be viewed as a significant sample of a random set of numeral writing of aphasic
subjects. The erroneous transcriptions by the Broca's aphasics were of another
nature in that they were essentially morphemic or verbal paragraphias referring to
quantities. Verbal paragraphias were sometimes numeral adjectives such as
'NEUVIEME' (ninth) in '97' (QUATRE-VINGT-DIX-SEPT) -> 'NEUVIEME
738 XAVIER SERON AND GERARD DELOCHE
SEPT', but more frequently nouns such as 'DOUZAINE' (dozen) in '112' (CENT
DOUZE)->'CENT DOUZAINE'. Morphemic paragraphias generally resulted
from incorrect suffixation of digit names either because it was not the appropriate
bound morpheme or because it required a previous modification of the digit
name. For example, '45' (QUARANTE-CINQ) was erroneously transcribed as
'QUATRANTE CINQ' with 'QUATRE' (4) in the tens position becoming
'QUATR' + 'ANTE', which is the generalization of a procedure that leads to the
correct result with other tens names. We stress that the information relative to the
stack position was always preserved in all the errors by the Broca's aphasics. They
never produced verbal paragraphias on digit names.
Stack errors. A stack error, such as '60' (SOIXANTE) -»• 'SIX' (6), is identified
in a numeral sequence that differs from the correct sequence only in that one element
of the erroneous production preserves the information regarding the position within
out of the 35 erroneous transcriptions produced by the two Broca's aphasics with
agrammatism. One transcribed the first digit '1' by 'CENT' (hundred) whatever the
digit string length (except particulars where the first digit T was transcoded by
'DIX' (10)). For this subject, the erroneous transcriptions may have resulted from
an overgeneralization of transcoding strategies correct in other contexts: ' 1 ' is in
fact 'CENT' in the hundreds position as in '153' (CENT CINQUANTE-TROIS)
and transcoding T in the tens position by the tens name 'DIX' (10) is correct when
followed by '0', '7', '8' and '9' but not in '112' (CENT DOUZE)^'CENT DIX
DEUX'. In the same way, the other agrammatic subject generalized the transcoding
of'1' in the thousands position by 'MILLE' to the case of six-digit numbers.
Erroneous lexical transcription of the digit T in the hundreds or the thousands
position. As a rule, the digits in the hundreds and thousands positions must be
transcribed by their digit name followed by 'CENT' (hundred) and 'MILLE'
(and). Seven such errors were produced by the Broca's aphasics, such as '51'
(CINQUANTE ET UN) -»'CINQUANTE UN' or '71' (SOIXANTE ET
ONZE) -• 'SOIXANTE ONZE'.
Mixed errors. There were 82 erroneous productions that could be interpreted
either as the result of the combination of two erroneous transcoding strategies
discussed above or in terms of one or the other strategy. Thus, the transcription
'TROIS DEUX' instead of 'TROIS CENT VINGT (320) may be either the result of
transcoding each digit into its digit name ('3' -> 'TROIS' and '2' -> 'DEUX') but
without violating the rule that '0' is never transcribed by 'ZERO', or by the omission
of 'CENT' (hundred) when a multiplicand and a stack error with '2' in the tens
position being transcribed as 'DEUX' (2) instead of'VINGT' (20). Such erroneous
transcriptions have been included in the 'mixed errors' category but, in most of the
cases, inspection of the most frequently used transcoding strategy by the par-
of the successive omissions of'QUATRE' (4), 'VINGT' (20), and 'DIX' (10) and the
substitution of'QUINZE' (15), 'UN' (1), and 'ZERO' (0), respectively. Such ad hoc
analysis is of little value since it is only a description of what is present and what is
absent in the erroneous transcriptions and has no relationship to the subprocesses
implied in the transcoding production task. Moreover, it cannot explain why
systematic erroneous transcriptions occur in particular contexts.
CONCLUSIONS
As indicated in the Table by the distributions of erroneous transcriptions
according to the different error sources, some factors clearly emerge that differ-
entially affected the performances of the two groups of aphasic subjects.
The group of 7 Broca's aphasics produced more erroneous transcriptions than the
battery may also have been biased in that types of items that present more difficulty
for one group than for the other may have been over-represented.
Most of the erroneous transcriptions of Broca's aphasics seemed to result from
grammatical (nonlexical) difficulties as indicated on the morphological and
syntactical levels. Their verbal or morphemic paragraphias belonging to the domain
of quantities, their stack errors being orientated toward the units names stack, and
the very few cases of position-within-stack errors clearly demonstrated that the
individual digits in the number string were correctly identified. These errors may
have been the result either of morphological difficulties in constructing composed
(root + bound morpheme) numeral names or of syntactical impairments in parsing
the digit string into hundreds, tens, and units. The nature of the errors may
tentatively favour thefirsthypothesis. The syntactical difficulties of Broca's aphasics
were evidenced at the production level in the numeral displacement errors where
ACKNOWLEDGEMENTS
We are indebted to an anonymous reviewer for helpful comments on an earlier expanded version of
this paper. We are also grateful to the speech therapists in the 'Centre de Reeducation du Langage', La
Salpetriere, Paris, and the 'Centre de Revalidation Neuropsychologique', Cliniques Universitaires St.
744 XAVIER SERON AND GERARD DELOCHE
Luc, UCL, Brussels, for their help in referring patients. We also thank Mrs N. Benoit, psychologist, for
her assistance in diagnoses. This research was supported by INSERM Grant no. 826023 and Grant no.
7, 81-82 from 'Les Echanges Scientifiques dans le Cadre des Accords Culturels Franco-Beiges'.
REFERENCES