TEXT REVISING STRATEGIES AND PROCEDURES
161
able to process both linguistic levels at once. They revised the overall structure of the text
after having done the necessary within-statement ((clean up».
These results are interpreted in another article (Roussey, Piolat, & Guercin, 1990). The
interpretation concerns three groups of revisers: those who managed to produce the expected
text by employing one of the model strategies, those who proceeded in a different manner,
and finally, those who were unable to produce the required revised text.
Discussion
Our goal here was to propose an initial account of the utility of the notion of revising
strategy.
Our method of analysis of the step-by-step progression of the text revising activity enabled
us to point out that certain model procedures were used especially by child experts, while
others were implemented by adult experts. (The adults were able to simultaneously handle
problems at two linguistic levels). However, the adults did not use the same strategy to revise
the narrative and the description.
A parallel may be drawn between these findings and the proposals made by Hayes et
al. (1987).
Accordingly, revisers who use a simultaneous processing strategy may have defined the
task outright by assessing both the micro- and macro-linguistic levels of the text. Revisers
using a consecutive processing strategy, on the other hand, may have had that same linguistic
goal, but have decided to reach it via two consecutive passes through the text. Or, they may
have had a more limited definition of the task at the onset, only later to discover as they
revised at the local level that they needed to pass through the text a second time to improve
the global level, thus changing their definition of the task (adaptive «flexibility»).
As with the description, the expert adults may, at task definition time, have decided
to improve both linguistic levels but in a consecutive fashion in order to handle the more
delicate revision of this more difficult type of text; or they may have modified their task
definition, incomplete at first, since problems are more difficult to anticipate for this more
complex text structure.
The main limitation of our approach is that in our definition of model revising procedures,
only one final state of the text was considered (the state obtained when the expected corrections
were made). This meant rejecting a relatively large number of subjects from our protocol
analysis. However, the defined models did enable us to characterize the functioning of a
third of all subjects, and over half of all experts.
Furthermore, the final state of the text could be obtained only by using the linguistic
items displayed on the screen. In a spontaneous revising situation where the writer can insert
new items, it would only be possible to predict the solution for problems involving «rules»
or «maxims» (see Hayes et al., 1987). In this case, the problem definition would lead to
diagnostics calling for modifications in the text to which anyone would agree. On the other
hand, problems pertaining to «communicative intentions» can be solved in many different
ways, all equally valid and thus unpredictable.
The main advantage of our approach lies in the fact that in this experimental situation,
which is constraining for the revisers but highly controlled, it is possible to compare the
procedures actually used to model procedures, while varying different factors (characteristics
of the subjects, type of errors introduced in the texts, text difficulty, etc.). Several new research
goals may now be set, particularly in the area of development (Fayol & Gombert, 1987),
such as:
- Determine whether the strategies employed can be ordered by difficulty (work
load, simultaneous use of several knowledge bases, etc).