TEXT REVISING STRATEGIES AND PROCEDURES
159
- has constructed a representation of the to-be-attained product that conforms
to the product expected by the researcher;
- has properly diagnosed all of the linguistic problems;
- corrects in an economical manner.
Natural situations of text modification during composition are obviously more complex
than the one imposed on the revisers here. Nevertheless, if we find out that the theoretical
strategies were in fact employed, it will be possible to deduce that the functional principles
upon which they are based are essential to the overall revision of texts.
Method
Forty-eight children whose mean age was 10.6 years (ranging from 9 years and 11 months
to 12 years and 1 month) and forty-eight adults, classified as expert or nonexpert on the
basis of their degree of expertise in writing, revised one of two experimental texts. The experts
and nonexperts in each group were not comparable, since classification was not based on
the same criteria in each case. The expert/nonexpert distinction was designed simply to take
into account prior experimental results providing evidence of significant variations in
performance within same age groups (Fayol & Gombert, 1987). The children's degree of
expertise was assessed by their teacher (each child in the class was labelled as expert or
nonexpert and then names were drawn in each of the two groups). The adult experts were
high school graduates who had frequent contact with writing; the nonexperts had not had
much education, and did not work in a profession that involved writing on a daily basis.
The texts (a narrative and a description) each consisted of eight statements, six of which
had been modified (errors were inserted) so as to encourage the subjects to make three local
corrections (within-statement) in order to reestablish proper syntax and semantics (insert,
delete, move nouns) and three global corrections (between-statement) in order to reestablish
text coherency (insert, delete, move statements; see Appendix).
The computer-assisted revising system (SCRIPREV) required the subjects to modify the
text by using only the verbal material displayed on the screen. This system operates like
a highly simplified word processor. The reviser used an optical pencil to point on the screen
to the linguistic segment (from 1 to N words) he or she wished to move, and to the location
he or she wished to move it to (in the text or outside the text). After each move, the computer
justified the text and recorded all manipulations.
To determine what revising strategy was used by the subjects, each of the predefined
model procedures was coded in time series format. A time series is a set of data organized
into a sequence of temporal instants. Two events were recorded at each instant: the exact
identity of the linguistic item involved in the manipulation, and the location in the text where
the subject was revising.
Whenever a subject was found to have revised the text in the expected manner (52 subjects
out of 96; see Roussey, Piolat, & Guercin, 1990), his/her procedure (actual times series)
was compared to the model time series by calculating the distance between structures according
to a method developed by Guercin (1986) and Guercin, Roussey, and Piolat (1990). The strategy
corresponding to the model time series closest to the actual time series was attributed to
that subject. Only a small number of subjects, mostly experts, could be attributed a strategy
(28 out of the 52 subjects who revised the text as expected).
Results
The actual procedures varied by age and type of text.