Writing, working memory and dual-TASK
Writing is one of the most complex cognitive
activities that Human Beings accomplish during
their life. Writing indeed involves a large number
of cognitive components that operate at different
levels of representation. For instance, at a semantic
level, planning processes construct a pre-verbal
message that corresponds to the ideas a writer wants
to communicate. In this stage, ideas are retrieved
from long-term memory and (re-)organized if
necessary. Planning processes also allow scheduling
the unfolding of writing by preparing action plans
for composing (Hayes & Grawdol-Nash, 1996). A
second set of processes, the translating processes,
involves a linguistic level of representation in which
the pre-verbal message resulting from the planning
processes is transformed into a verbal message. The
conceptual structure elaborated during planning is
thus grammatically encoded by retrieving in the
mental lexicon the syntactic and morphological
properties of words (Bock & Levelt, 1994). A third
component allows comparing the text already
written to the writer's mental representation of its
own intended text by launching procedures aiming
at improving the text, both at the conceptual and
linguistic levels. A last component intervenes at a
motor level of representation (Kellogg, 1996).
Execution, or graphic transcription processes allow
writers transforming the linguistic message into a
sequence of motor programs for producing
handwriting. Writing being thus a strongly
integrative activity, understanding how writers
compose a text does not only consists in
descriptions of the processes underlying written
production but also in explaining how their
activation is orchestrated in the cognitive system,
whose main characteristic is its limited capacity in
simultaneously maintaining and processing
information. In other words, one objective of
writing research is to analyze the on-line
management of working memory.
Psychologists have developed several techniques
to analyze the management and cognitive demands
of mental activities. For example, they use mental
chronometry (Posner, 1978), visual search
(Schneider & Shiffrin, 1977), or event-related
potentials (Humphrey & Kramer, 1994). However,
the most widely used is the dual-tasks technique.
This is particularly true since working memory has
been identified as playing a crucial role in complex
cognitive activities. Numerous empirical evidences
supporting the existence of working memory and its
relationship with cognitive activities indeed came
from dual-task experiments (Baddeley, 2000).
In this framework, the present review aims at
showing that the dual-task technique largely
contributed to the understanding of the relationship
between writing and working memory. It begins by
presenting the general principle of the dual-task
technique and its potential limits. Then, the most
influential models of working memory are
described. The final section focuses on writing and
working memory. It begins by describing models of
writing that include working memory. Then, recent
findings on the role of the phonological loop, the
visuospatial sketchpad and the central executive
during writing are summarized. This review focuses
on the implication of working memory in writing at
the light of only dual-task experiments.
Consequently, other issues relevant to working
memory but not investigated with dual-tasks, such
as individual differences investigated with
correlational studies (for example, see Swanson &
Berninger, 1996) will not be examined.
Dual-TASKS
Requiring participants to simultaneously perform
two tasks (called on the one hand, the primary and,
on the other hand, the secondary), this technique
directly exploits the postulate that the cognitive
system has a limited pool of cognitive resources
(that must be defined before setting the secondary
task). When simultaneously performed, the primary
and secondary tasks share these cognitive resources.
No longer having the amount of resources that they
usually require, the primary or secondary tasks are
affected and their performance is presumed to
decrease. Consequently, comparing a dual-task
condition with a single-task condition allows
evaluating how resources are shared between the
primary and secondary tasks.
Numerous secondary tasks have been designed.
For example, in concurrent memory loads subjects
have to memorize a set of items (digits, characters,
matrices of dot) while performing the task under
investigation (the primary task). In this case, it is
postulated that the operations of the primary and
secondary tasks engage information of similar nature
and that they interfere. According to both the
number of items memorized and the nature of the
items, the secondary task has different impacts on
the primary task. Probe tasks are another kind of
secondary task. Signals (called probes) periodically
appear during the experiment (Kerr, 1973) and
participants are instructed to react as quickly as
possible as they detect the probe. Dual-task
performance at the secondary task is then compared
with the performance in single-task condition. The
latency of reaction to the probes assesses the amount
of cognitive resources that is devoted to the primary
task. Similarly, in tracking tasks, participants
follow a target and the deviations from the target are
interpreted as indicating resources consump-tion (for
an example, see Power, 1986). In this two latter
cases, it is postulated that the primary and secondary
tasks draw on the same pool of central resources.
Although findings of dual-task experiments
strongly contributed to a better comprehen-sion of
human cognition, many criticisms were raised
against this technique. Some authors even claimed
that the dual-task paradigm should no more be used.
Howe and Rabinowitz (1989) for instance argued
that the new interactions created for processing both
the primary and the secondary tasks disabled any
interpretation of performance in dual-task situation.
Without going so far, the dual-task technique