French clinical data

[Brian MacWhinney]

Dear Info-CHILDES,
  I am happy to announce the addition to CHILDES of a new corpus of the
early sentences of six French children contributed by Marie-Therese Le
Normand. Two of the children had epilepsy and four were diagnosed with SLI.
The corpus includes about 200 short sentences from each child in a single
CHAT file linked to the audio productions in a .wav file.  These are not
conversational data, but are intended instead to represent the vocal
characteristics of speech in these children.
The data (audio and transcripts) can be downloaded from
http://childes.psy.cmu.edu/audio

Here is the documentation file for this fascinating new corpus:

Contact Address:

Dr. Marie-Thérèse Le Normand,
Laboratoire de Neuropsychopathologie du langage et de la cognition
INSERM 9609, Groupe Pitié-Salpétrière
47 Bld de l¹Hôpital
75651 Paris cedex 13
lenorman at chups.jussieu.fr
lenorman at infobiogen.fr

Le Normand audio corpora

Children with SLI and EPIlepsy (EPI)  involved in this corpora were part of
a larger project supported by L¹Institut National de la Santé et de la
Recherche Médicale  (INSERM), Paris,  investigating the language development
of children with neurogenic disorder. This project may contribute to central
issues in developmental psycholinguistics and neurolinguistics, particularly
in the understanding of the cerebral involvement in the acquisition of
language. 

Exclusion criteria for SLI and EPI children were premature birth, perinatal
complications, evidence  of obvious damage following a neurological
examination, a diagnosed behavior problem or other psychopathology.
Inclusion criteria were an Apgar score of 10 five minutes after birth, a
nonverbal intellect score greater than 90 (MacCarthy, 1976; range: 95-126),
intact hearing as assessed in an audiometric evaluation, native speaker of
French and uninterrupted participation in the study.

EEG evidenced for both EPI children a clear-cut left temporal focus. MRI
evidenced for Sofian a left temporal lesion, most probably dysplastic
without destruction of parenchyme. MRI evidenced no lesion for Benoît. PET
scan revealed left temporal hypometabolism for Sofian. No PET scan was
available for Benoît.

Comprehension tasks
The selection of language comprehension tasks included pointing (choosing
which items in a set of pictures are named by the experimenter), and
understanding of prepositions (positioning play figures; there were 18
requests to assess representation of space, quantity, partition and
localization). 


Table 1. z- scores of language comprehension tasks
    Pointing         Understanding of prepositions
    z-scores        z-scores
  
EPI
Benoît    1        0
Sofian    0        0
SLI
Beranger     1        -0.5
Hippolyte    +1        +2
Paul    -1          0.5
Samuel    -1.7                             -0.6
Sébastien      0          0
      

Production tasks

The linguistic production tasks included repetition of mono-, bi- and
polysyllabic words spoken by the experimenter; naming of pictures of
objects; lexical  and morphological categories diversity as measured by the
number of  type and tokens uttered in a 20 min controlled play session. A
JPEG photo of the setup is included with the data files. This is a form of
narrative where the subject is asked to verbalize all manipulations and
actions with dolls and objects in and around a doll house.

Normative data on the comprehension tasks are found in Chevrie-Muller, Simon
and Decante (1981) for ages between 3 years, 9 months and 8 years, and
normative data on the production tasks (lexical and morphological category
diversity) are found in Le Normand (1997) for ages between 2years  and 4
years.

Video and audio recordings

The video and audio recording sessions lasted approximately 20 minutes and
were conducted in the lab with mother or father using the same play
Fisher-Price toys materials. All sessions were transcribed in accordance
with the guidelines produced by the Codes for Human Analysis of Transcripts
(CHAT) which is part of the Child Language Data Exchange system (CHILDES)
(MacWhinney, 1995).  The computerized transcripts were then compared with
the original videotaped data by an independent transcriber in order to
verify their accuracy.  This process resulted in 97.0% inter-transcriber
reliability.  Any disagreements  concerning the transcription were resolved
by re-examination until consensus  was reached.