Writing, working memory and dual-TASK
result therefore demonstrates that the phonological
loop is also engaged during planning.
These different experiments highlights the role of
the phonological loop and the visuospatial
sketchpad in writing. Even if the majority of
studies that have been conducted lead to findings
generally consistent with the initial proposals of
Kellogg about the relationship between the sub-
systems of working memory and the writing
processes, some evidences stem the predictions of
the model. In particular, the question of the
relationships between the phonological loop and the
planning process still remains. Finally, with respect
to the dual-task issue, it is important to notice that
in the different dual-task experiments presented
above, the "unintrusiveness" criterion suggested by
Fisk et al. (1986-87) cannot be respected. Indeed, in
order to decide whether or not a component of
working memory is required for performing the
primary task, the effect of the secondary task can be
assessed by comparing performance in single- versus
dual-task condition but of either the primary task or
the secondary task, or both. As it is shown in the
next section, performance has to be analyzed
differently when one focuses on attentional and
executive functions of working memory.
Some findings on attentional and executive
functioning in writing
As indicated in the introduction of this paper,
writing is a highly integrative activity. Furthermore,
all writing processes, excepted graphic transcription
in adults, place high demands on the central
executive of working memory (Kellogg, 1996).
Composing a text thus requires shifting between --
and presumably concurrently activating-- several
writing processes. Research on the role of the
central executive in writing has therefore to analyze
how the different writing processes are activated in
the course of writing. In other words, it means
exploring the strategies writers activate for managing
the writing processes in the course of writing
(Fayol, 1999). On another hand, it has to study
how writers succeed (or not) in activating the
different writing processes without exceeding the
limits of working memory (for a review, see Brown
& Carr, 1989). This means also evaluating the
amount of resources that is devoted to each of these
processes.
The demands and temporal organization of the
writing processes. The triple task method (initially
proposed by Kellogg, 1987) had proved fruitful for
addressing these issues (for a review, see Olive,
Kellogg & Piolat, 2001). The general principle of
the triple-task method requires participants to
perform a primary task (in the present case the
composition of a text), a probe task and to verbalize
about the primary task. In practice, two variants of
the triple task have been used. In the first variant,
the verbalization is directed. After each reaction to a
probe the writer categorizes his thought according to
categories referring to the writing processes
(Kellogg, 1987). In the second variant, participants
think aloud throughout the primary task and the
writing processes are identified by mean of a
classical verbal protocol analysis (Levy & Ransdell,
1995). Two variables are then analyzed. First, the
verbalization data provide information about how
the writing processes are activated. Second, the
reaction times to the probes allows studying
allocation of cognitive resources to these processes.
Indeed, each reaction time (RT) to a probe is
associated with a verbalization data and the latency
of these RTs (in term of a difference with a baseline
RT) can be interpreted as estimating the amount of
cognitive resources (or processing demands) that is
devoted to each writing process. Latency of
secondary reaction times has indeed been considered
reflecting allocation of cognitive resources in studies
of mental chronometry.
The experiments that were carried out with the
triple task thus analyzed processing demands and
activation of the writing processes. They attempted
to examine whether one process dominates the
others at particular points in time and whether these
relationship changes as writing progresses, in other
words the goal was to determine the general
functioning characteristics of writing. They also
tried to assess the demands in central resources of
each writing process. The impact of several factors
on both the processing demands and activation of
the writing processes was also examined to further
understand how these general functioning
characteristics interacted with writer's inter-
individual differences (domain knowledge, verbal
skill, working memory capacity....) and with the
composition situations (writing strategies, writing
in first or second language, medium of
production,.; for a review, see Olive, Kellogg &
Piolat, 2001, and Piolat & Olive, 2000).
Regarding the general functioning characteristics
of writing (i.e., the processing demands and the
activation of the writing processes during a
composition), the results do not always converge.
This is due to the fact that both the processing
demands and the activation of the writing processes
vary with the specific demands of the task and the
knowledge and skills of the writers. Nevertheless,
without taking into account the factors that were
manipulated, some stable findings can be drawn
from these experiments. First, planning, but also
revision with a less extent, place more demands on
working memory than translating processes do. The
lighter demands on working memory of translating
can be explained by noting that translating sub-
processes are also involved in the production of
spoken language. Second, translating is the most
often activated process but planning is more
activated than revision, although the reverse is
observed in a few studies for these two last
processes (more detailed information can be found in
Olive, Kellogg & Piolat, 2001, and in Piolat &
Olive, 2000). Moreover, throughout a writing
session, activation of translating remains constant
while that of planning decreases and that of revision