Writing, working memory and dual-TASK
consist in investigating the demands of writing in
short-term storage, namely the relationship between
the writing processes and the slave systems.
Secondly, taking into account the great number of
processes required to compose a text and the limited
capacity of working memory, the role of working
memory in written production can be investigated
by studying the attentional and management
functions of the central executive.
Some findings on verbal and visuospatial short-
term working memory in writing
Kellogg (1996) has argued that spatial working
memory supports planning conceptual content and
that verbal working memory supports translation of
this content into sentences. Kellogg, Olive and
Piolat (in press) tested this assumption in a sentence
generation task. Participants wrote definitions of
either abstract or councrete nouns while concurrently
performing a secondary task that required the
detection of visually presented targets that were
either verbal (ba or da) or visual (a triangle or a
circle). Participants had to detect whether the target
was different from the last one presented. Sentence
length, secondary task accuracy and secondary
reaction times were collected and were disrupted by
the verbal and visual secondary tasks. More
precisely, the verbal task disrupted the production of
definitions of both the concrete and abstract nouns
whereas the visual task disrupted the production of
definitions of only the concrete nouns. This
findings is consistent with Kellogg's (1996) claims
that verbal working memory is needed for the
linguistic encoding of sentences and that visual
working memory is needed for planning image-
based conceptual content.
The role of the visuospatial sketchpad in writing
has been mainly conducted with tasks asking
subjects to memorize a spatial or visual material
(Levy & Ransdell, 2001). For example, arguing
that the composition of a descriptive text implies
more mental imagery than that of argumentative
text, Passerault and Dinet (2001) hypothesized that
overloading the visuospatial sketchpad should have
more impact with descriptive texts than with
argumentative texts. As expected, writers' fluency
was more slowed down when composing descriptive
texts.
By contrast, studies of the phonological loop
generally do not involve any memorization. The
goal of such phonological secondary tasks used is to
interfere with either the processing, the temporary
storage or the refreshment of information in the
phonological loop. For example, in the unattended
speech task, while they compose a text, participants
hear a speech to which they do not have to pay
attention. In the articulatory suppression task,
participants continually repeat unmeaning syllables.
Madigan, Johnson and Linton (1994) observed that
an unattended speech task affected writer's fluency
but had no effect on the quality of the texts.
Exploring further why an unattended speech affects
writing, Marek and Levy (1999) observed any
differences between an unattended speech task made
by a coherent text or an incoherent one. They thus
demonstrated that the effect of this unattended
speech is mainly related to a phonological
interference rather than to its semantic integration.
Marek and Levy (1999) extended their findings by
exploring the relation of the phonological loop with
each writing process. They used three different
writing tasks, each one being designed to engage
mainly one particular writing process. In the first
task where participants had to generate a sentence
from several words, formulation processes were
assumed to be mainly engaged. In the second task,
only execution processes where supposed to be
involved because participants copied a text. Finally
in the third task, editing a text, revision processes
were mainly engaged. Simultaneously to these three
tasks, participants were submitted to a secondary
task of unattended speech. According to Kellogg
(1996), because only formulation would engage the
phonological loop, the unattended speech task
should have no effect on the copying and revision
tasks. The speed and errors of typing during the
copying task were not affected by the unattended
speech. In the revision task, the unattended speech
had no effect on the number of errors detected. By
contrast, in the sentence generation task writer’s
forgot to included source words and subjective
quality of the sentences was judged lower when it
was performed with an unattended speech than
without an unattended speech. These data lead
Marek and Levy (1999) to conclude that the
formulation processes need stronger access to the
phonological loop than the execution and revision
processes did.
In another experiment, Levy and its collaborators
(see Levy & Ransdell, 2001) instructed their
participants to perform secondary tasks that aimed at
overloading each sub-system of working memory
(i.e., the phonological loop, the central executive
and the visuospatial sketchpad). All secondary
stimuli were characters. The single char-acteristic
that allowed differentiating these secondary tasks
was the dimension of the characters that participants
had to process to make a response. Indeed, about all
two seconds, the character, its color, its font or its
location changed. Thus, in the phonological
condition, writers had to reply when two
consecutive characters shared the same phoneme (for
instance [i], e.g., in English: b, c, d). In the
visuospatial task, writers had to detect when two
consecutive characters were at the same place, or
were similar in color. In the task aiming at
overloading the central executive, writers had to
reply when two consecutive characters formed a two-
letter English word. Consistent with Kellogg's
model (1996), the visuospatial task produced the
strongest decrease of the initial planning time (pre-
handwriting pause). However, contrarily to
Kellogg’s propositions, the phonological task
produced a larger effect on this initial planning time
than when the central executive was required. This