Cognitive processes in writing during pause and execution periods

Université de Poitiers Centre National de la Recherche Scientifique Centre de Recherche Sur la Cognition et l App e tissage Rapport technique : 2009/01/T.OLI Cognitive processes in writing during pause and execution periods Thierry Olive, Université de Poitiers, Poitiers, and Centre National de la Recherche Scientifique, France; Rui Alexandre Alves and São Luís Castro, Universidade do Porto, Portugal A paraître dans / To appear in : Olive, T., Alves R. A., & Castro, S. L.. Cognitive processes in writing during pause and execution periods. European Journal of Cognitive Psychology. Address correspondence to: Thierry Olive Laboratoire Langage, Mémoire & Développement Cognitif Maiso des S ie es de l Ho e et de la So iété 99 avenue du recteur Pineau 86000 Poitiers FRANCE Phone: + 33 (0)5 49 36 62 99 Fax: + 33 (0)5 49 45 46 16 e-mail: thierry.olive@univ-poitiers.fr 1 KEYWORDS: Writing; Writing processes; Pause; Execution periods; Cognitive effort. ABSTRACT The present study investigated how writing processes are activated during pause and execution periods. In two experiments, handwriting demands were manipulated by asking participants to compose with their familiar handwriting or with a highdemanding cursive uppercase calligraphy. Experiment 1 investigated narrative writing, a task with low planning demands. Experiment 2 addressed essay writing, a task with stronger planning demands. Occurrences of processes and their cognitive effort were analysed by asking participants to respond to random auditory probes and then to report their ongoing mental activity according to learned categories referring to the planning, translating, and revising writing processes. All together, the findings indicate that demands on planning did not affect how writing processes were activated during pauses and execution periods but automaticity of handwriting did. When handwriting was effortless, translating was mostly activated in parallel with motor execution, whereas revising and planning were mainly activated during pauses. However, none of the writing processes could be characterised as being typical of pauses, since translating was activated to a similar extent as the other two processes. By contrast, when handwriting was effortful, participants shifted to a more sequential functioning and activated translating mainly during pauses. INTRODUCTION A central issue in research on writing concerns how writers manage the cognitive processes they need to compose a text. Writing a text requires several high- and low-level processes (Hayes & Flower, 1980). Planning processes are required to set rhetorical goals, to generate ideas, and to organise them in a writing plan. Translating allows writers to formulate ideas into language. Revising includes reading to evaluate the text and editing it when problems are found. Finally, motor execution processes are needed to handwrite or to type the text in a written form. It has long been established that planning, translating, and revising, and, in children, motor execution, all require working memory resources (see Olive, 2004, for a review). In order to not exceed working memory capacity, writers have therefore to use and orchestrate these processes efficiently (Flower & Hayes, 1980). Management of the writing processes thus efle ts the ite s st ateg fo opi g with the demands of composition, and writing performance is considered to depend on how the writing processes are organised during writing (Breetvelt, van den Berg, & Rijlaarsdam, 1994; van den Berg & Rijlaarsdam, 2007). In that perspective, an important issue for writing research is to describe the functional ha a te isti s of the iti g p o esses a d ite s st ategies. Several studies have described how these writing processes are orchestrated during a writing session (Kellogg, 1987a, 1987b, 1988, 2001; Kellogg & Mueller, 1993; Levy & Ransdell, 1995; Penningroth & Rosenberg, 1995; Roussey & Piolat, 2008; see also Olive, Kellogg, & Piolat, 2002, or Piolat & Olive, 2000, for reviews). Translating takes 40!50% of the total writing time; planning and translating share the remaining time but, throughout a composition, the amount of time devoted to planning decreases, whereas time devoted to revising increases. Episodes of the writing processes generally do not exceed 20 s. Moreover, although the most familiar writing strategy that is taught advises writers to begin a composition by planning first, then to produce text, and finally to review the text already written, the writing processes are not activated linearly. Instead, they are recursive, with one process calling upon another, as when translating an idea into a sentence prompts the writer to engage in further planning. As Levy and Ransdell (1995, 1996) suggested, the pattern of transition between processes may i deed e the sig atu e of a ite s st ateg . The also o se ed that the writers who made numerous transitions between all the writing processes were also those that produced the best 2 quality texts. To go further in the o p ehe sio of ite s fu tio i g, it is e essa to ha a te ise ho the writing p o esses a e a ti ated a o di g to ite s a ti it . O e t writing behaviour indeed is made of pause and execution periods during which cognitive operations may be organised quite differently (Alves, Castro, Sousa, & Strömqvist, 2007; Hayes & Chenoweth, 2003, 2006). So, how can the orchestration of the writing processes during pauses and execution periods be portrayed ? One possibility is to conceive writing as strictly sequential. This conception implies that during motor execution writers only write down their text and that these periods of motor execution are separated by very long pauses dedicated to planning, translating, and reviewing. However, this does not correspond to the dynamics of writing, which is rather made of numerous short pauses between bursts of handwriting (Foulin, 1995; Hayes & Chenoweth, 2003, 2006; Schilperoord, 2002). Furthermore, if it were necessary to focus attention first on planning, second on generation, and third on motor execution, then this strict order would reduce the opportunity for rapid interactions between planning and translating. Yet, such an interaction is fundamental in skilled writing (McCutchen, 1988). Taken together, a cascade view of the organisation of the writing processes seems more appropriate. In such a view, when information has been processed by a cognitive component, this information is sent to the next component, while the first component can process another segment of information. For example, when an idea is constructed and organised, this idea is then formulated in language and next executed but, as soon as one segment of text is being written down, translating can begin to transform the following part of the text. More important, the high-level writing processes (planning, translating, and revising) can occur concurrently to execution making processing demands of pauses and execution periods quite different. Regarding pauses, writers spend at least half of the composition time pausing (Alamargot, Dansac, Chesnet, & Fayol, 2007; Alves et al., 2007; Strömqvist & Ahlsén, 1999). Pauses are interruptions of execution whose duration is a function of the complexity of the processes engaged in (Foulin, 1995). They can occur because of competition for limited capacity (Just & Carpenter, 1992; McCutchen, 1996), or for a common processing component (Pashler, 1994), or as consequence of memory decay and so are used to reinstate the intended message (Torrance & Galbraith, 2006) or even they may result from cross-talk between products and processing of ongoing activities (Navon & Miller, 1987). In either case, the function of pauses is poorly specified in writing research (see Torrance & Galbraith, 2006), and because during pauses all working memory capacity is freed from handwriting demands, pauses have generally been linked with the more effortful processes, namely planning and revising (Foulin, 1995; Schilperoord, 2002). Precise estimates of how much of planning, translating, and revising are carried out during pauses are however lacking. Regarding execution periods, skilled motor execution (handwriting or typing) frees working memory capacity. Therefore, working memory capacity available during handwriting is allocated to the high-level writing processes that can then be activated concurrently, at least in adults. For example, Bourdin and Fayol (1994, 2002) showed that performance of adults decreased when they had to recall series of digits or to compose sentences using cursive capital letters, a rarely practised and hence effortful calligraphy. Olive and Kellogg (2002) also observed that children were unable to activate high-level writing processes together with motor execution, and had to suspend handwriting to think over their texts. Conversely, adults were able to activate simultaneously motor execution and high-level writing processes. Chanquoy, Foulin, and Fayol (1990) observed increased fluency during the production of the last part of a sentence by contrast with fluency during the first part of the sentence. According to the authors, this suggests that during the first part of a sentence, adult writers begin to plan or to translate the ending part of the sentence, which is written down without any concurrent process to motor execution. Moreover, as eye movements indicate, adult writers often read the text already produced to either create new content, or to evaluate what has been produced so far (Alamargot, Chesnet, Dansac, & Ros, 2006; Alamargot et al., 2007). In sum, concurrent activation of highlevel writing processes and motor execution is now well documented. The nature of the writing processes that are activated concurrently to handwriting has been highlighted in a recent study. Alves, Castro, and Olive (in press) analysed the writing processes occurring during pauses and execution periods. To discriminate between pauses and execution periods, Al es et al. e o ded all ite s ke st okes he o posi g a 3 narrative on a keyboard. To identify the writing processes that occurred during pauses and execution periods, before writing, participants learned to categorise examples of introspective thoughts as different types of activities related to writing (planning, translating, or revising). Next, while composing, writers had to react as quickly as possible to randomly distributed auditory probes, and immediately after each probe they had to report their ongoing activity according to the learned categories (for description of the technique, see Kellogg, 1987b; Olive et al., 2002; Piolat, Olive, Roussey, Thunin, & Ziegler, 1999). Occurrences of the writing processes were then analysed according to the ite s a ti it . T a slati g as ostl a ti ated du i g oto execution, whereas revising and planning were mainly activated during pauses; however, none of the writing processes could be characterised as being typical of pauses, since translating was activated to a similar extent as the other two processes. Of major interest for writing research is whether these findings can be generalised to writing situations that involve different processing demands. Indeed, because all writing processes share a common working memory capacity (Kellogg, 2001; McCutchen, 2000), changes in processing demands of the composition task can affect functioning of the writing processes. For example, top-down and bottom-up interactions have been observed, with changes in processing demands of the high-level writing processes affecting motor execution and vice versa (Brown, McDonald, Brown, & Carr 1988; Fayol, 1999; Kellogg, 2001). PRESENTATION OF THE EXPERIMENTS The fi st goal of the p ese t e pe i e ts as to test hethe Al es et al. s (in press) findings, obtained in typing a narrative, can be generalised to different writing tools and type of texts. Specifically, in the present experiments writers composed their text, a narrative in Experiment 1 and an essay in Experiment 2, with pen and paper rather than by typing. Several findings indicate that writing processes are activated differently when writing by hand or with a computer. Composing texts on a computer may increase ite s og iti e effo t e ause it e ui es the use of a keyboard to produce the text. Although typewriting is becoming more and more widespread due to the increased use of computers, few people have a degree of proficiency needed to attain the level of automatisation of handwriting achieved by adults (Bosman, 1993; Rieger, 2004). Furthermore, Goldberg, Russell, and Cook (2003) confirmed that when composing with keyboard revision occurs more frequently, but mainly to edit the text at local level (see also, Kellogg & Mueller, 1993). If planning and revision are more demanding when composing by hand because they operate at nonlocal levels, then they should mostly be carried out during pauses, and not concurrently with execution. Only translating should occur during motor execution since there is no reason to believe that translating operations are affected by the writing medium or tool. This should be particularly true in Experiment 2 where writers composed by hand an essay, a particularly demanding task, unlikely to narrative composition. The major difference between these two types of text relates to their planning demands. The narrative schema is learnt very early and so planning a narrative requires filling in the different parts of the schema with knowledge retrieved from long-term memory without major structuring work. By contrast, there is no such schema for an argumentative text where the organisation of content is self-sustained (Favart & Coirier, 2006). The management of an essay involves logical and coherent reasoning, rational and abstract reflection, and skills in semantic organisation and given-new linearisation. Consequently, planning demands of essay composition being extremely high, this would lead writers to plan only during pauses. Such a pattern of result should i di ate that Al es et al. s i p ess fi di gs a e specific to typing. The second goal of the experiments was to examine whether changes in processing demands of motor execution i pa t o ite s st ateg . I deed, when no sufficient working memory capacity is available, writers would shift from a concurrent activation of the writing processes to a more sequential one. This was not confirmed by Alves et al. (in press) when they split participants into fast and slow typing groups. Even if the demands of execution were greater in the slow than in the fast typing group, strategies for activating the writing processes did not differ. It is possible that the distinction between fast and slow typists was not sufficiently extreme to affe t ite s st ateg . I o posi g ha d, 4 Olive and Kellogg (2002) have shown that adults could be forced to adopt a more sequential strategy when they had to write down their text using an unpractised uppercase cursive calligraphy. Accordingly, in Experiments 1 and 2, half of the participants composed a text using their usual and familiar calligraphy while the other half used a cursive uppercase calligraphy which has been shown to increase demands of handwriting (Bourdin & Fayol, 1994; Olive & Kellogg, 2002). It is expected that when processing demands of motor execution are high, writers shift to a more sequential strategy. Specifically, writers composing with an unfamiliar handwriting should translate their text mainly during pauses. Moreover, planning and revising should also occur only during pauses as their high processing demands prevent writers from activating these processes while handwriting. For writers using their familiar handwriting, motor execution should be coordinated concurrently to all writing processes, with translating being more activated while handwriting, and planning and revising being more activated during pauses. To test these hypotheses, all participants performed the triple task technique, which is designed to answer questions about the time course and cost of the writing processes (Kellogg, 1987a, 1988; Olive et al., 2002; Piolat & Olive, 2000; Piolat et al., 1999). The method involves measuring the allocation of working memory resources using reaction times (RTs) to auditory probes and coupling these measurements to specific writing processes engaged in during composition. The procedure calls for participants to focus attention on composing a text and to respond as rapidly as possible to auditory probes (see Method section, Exp. 1). This dual-task technique is based on the assumption that the primary and secondary tasks compete for the limited capacity of working memory (Engle, 2002). Therefore, performance in the secondary task decreases as the demands of the primary task increase. The greater the performance degradation in RTs to the auditory probes during composition compared with baseline single task RTs to the same stimuli, the greater the cognitive effort required by the writing processes. To couple the RTs with specific writing processes, after each response to an auditory probe, the writers are asked to categorise their thoughts at the moment the probe occurred with a directed and immediate verbalisation. For the verbalisation task, participants are requested to choose among response categories referring to particular writing processes. Thus, the extra time it takes to detect the auditory signal can be taken as a measure of the processing demands associated with the writing process interrupted by the signal. By analysing the mean frequency with which the writing processes are reported, the directed verbalisations provide an estimate of how much planning, translating, and reviewing are carried out during the composition. The triple task technique raises several methodological questions concerning particularly the reactivity and validity of directed verbalisations. Several studies have been conducted to assess the validity of this technique (Kellogg, 1987b; Piolat, Kellogg, & Farioli, 2001; Piolat et al., 1999; Piolat, Roussey, Olive, & Farioli, 1996; Ransdell, 1995). These studies indicate that the triple task does not disrupt the writing process. Neither the functional characteristics of writers nor the quality of the text produced are influenced. Moreover, directed verbalisations provide valid information about the processes underlying the primary task (Ericsson & Simon, 1993) and do ot efle t ite s eta og itio s a out ho they compose (Levy & Ransdell, 1995). Of course, directed verbalisations, given their discrete nature, provide only an approximation of the writing process. Furthermore, they do not allow studying the automatic processes involved in composition. EXPERIMENT 1 : NARRATIVE COMPOSITION Experiment 1 investigated how planning, translating, and revising are activated during pauses and execution periods when composing a narrative with pen and paper. Half of the participants composed their text using a familiar calligraphy (their own calligraphy) and the other half used an unfamiliar calligraphy (cursive uppercase). To analyse which high-level writing processes are activated during pauses or during handwriting, and at which extent, participants composed their text using a digitising tablet (see Method section). To identify the writing processes, writers learned to categorise their ongoing mental activity. Then, during composition, auditory probes were distributed, and, each time writers heard a probe, they were asked to label their ongoing mental activity as being either planning, translating, or revising. It was thus possible to analyse how many times the various writing processes were labelled during pauses and during handwriting. 5 It was also necessary to check whether our manipulation of the processing demands of handwriting through the calligraphy requirements succeeded. So, after composing their text, writers were asked to copy it, again on a digitising tablet, and to respond as quickly as possible to auditory probes. By analysing the reaction times (RTs) associated to handwriting only during the copying task, it was possible to assess the cognitive effort of motor execution in the two groups of familiar and unfamiliar calligraphy. Moreover, because participants responded to the probes also during composition, we were able to analyse the probable shift from a parallel strategy in the familiar calligraphy condition to a more sequential strategy in the unfamiliar calligraphy condition by comparing RTs in the copy task associated with handwriting with RTs in the composition task associated with pause and with handwriting (for the rationale, see Olive & Kellogg, 2002). Finally, because after each reaction to a probe writers labelled their ongoing mental activities, it was possible to associate each reaction time to a writing processes and hence to analyse the cognitive effort of the writing processes. METHOD PARTICIPANTS Forty undergraduate psychology students participated in this experiment (mean age = 22.2 years, SD = 3.2). Half of the students composed and copied their texts using their usual handwriting (standard condition). The other half composed and copied their texts using cursive uppercase calligraphy (uppercase condition). In that condition, before composing their text, participants were shown a sample of a cursive uppercase calligraphy and were asked to write the alphabet twice. All participants were tested individually and were treated according to adequate ethical standards. TASKS AND MATERIAL C OMPOSITION TASK . To collect information about write s a ti it i.e., handwriting or pausing), participants composed their narrative using a Wacom Intuos2 digitising tablet. Two kinds of pauses can be distinguished in writing: pauses during which high-level writing processes occur, and pauses due to handwriting operations such as writing down the dot o a i o the a of a t . The latte pauses a e i deed too short to involve highlevel writing processes, and so functionally they do not differ from handwriting. For the present research, it is indispensable to distinguish between pauses during which handwriting has stopped long enough so that high-level writing processes do occur, from pauses where the pen is, for example, just moving between words. For that purpose, defining a threshold affording such a distinction is necessary. Very different thresholds varying between 130 ms and more than 5 s have been used (in handwriting: 130 ms by Alamargot et al., 2007; 200 ms by Passerault, 1991; 250 ms by Olive & Kellogg, 2002; in typing: 1 s by Alves et al., in press; 2 s by Alves et al., 2007; Levy & Ransdell, 1995; Schilperoord, 2002; Wengelin, 2007; 5 s by Jansen, van Waes, & van den Berg, 1996). A recent finding suggests that the highlevel writing processes can occur during pauses shorter than a quarter of second (Alamargot et al., 2007). In order to be able to compare our RT findings with the ones previously observed (Olive & Kellogg, 2002), a 250 ms criterion was set. Therefore, pauses were defined as an interruption of handwriting longer than 250 ms; handwriting included pauses shorter than 250 ms and actual handwriting. The topic of the narrative was prompted with a sequence of seven coloured pictures printed on an A4 page. This sequence shows a boy going for a walk; he meets a balloon seller, and gets a red balloon; very pleased, he continues strolling. Suddenly a blast of wind takes the balloon away; the boy breaks into tears. Instructions asked participants to use that story as a prompt for writing a story for children and to compose a well-structured and well-formulated text. The elicitation pictures were in front of the participants during the entire composition. While composing, participants were required to react as fast as possible to auditory beeps and then to indicate the ongoing mental activity that was interrupted by the beep (see later for details about these two tasks). In order to be able to collect sufficient responses to the probes, writing time was set to 30 min, with a minimum of 20 min. Two judges, blind to the objectives and method of the study, independently judged the quality of the texts. Final texts were typed with the errors they contained to control for handwriting legibility in the assessment of quality and were 6 then assessed with two dimensions related to (1) the number and structure of information and (2) style of language, with a scale ranging from 1 (very low quality) to 7 (very high quality). Reliable agreement between the judges (all correlations significant at p<.01) was found for the two scales (rs >.76). C OPYING TASK . After having composed their text, participants were asked to read it twice and then to copy it without editing during 10 min. They copied their text using the digitising tablet in order to be able to distinguish when they were handwriting the text or when they were pausing. During that task, participants were required to perform only the secondary reaction time task. To verify that the uppercase calligraphy did increase the demands of handwriting, we compared reaction times that occurred during handwriting only (i.e., when the electronic pen was on the tablet or up less than 250 ms) in the standard and uppercase conditions. S ECONDARY RT S AND DIRECTED VERBALISATION . While composing participants had to react as fast as possible to auditory beeps by pressing on the spacebar of a computer keyboard with their nondominant hand. After each reaction to a probe, they had to indicate the mental activity that was ongoing when the probe occurred. Reaction times and directed verbalisations were collected with a modified version of ScriptKell (Piolat et al., 1999) that can be connected to a digitisi g ta let a d that s a s ite s a ti it . Mo e p e isel , S iptKell lau hed a do eeps, olle ted ‘Ts under single or dual task conditions, and gathered the verbalisation responses, which were associated with RTs to the ite s a ti it pausi g o ha d iti g . Fo the verbalisation task, writers were required to indicate whether they were pla i g, t a slati g, e isi g, o doi g a thi g else othe la el . Planning referred to finding and organising ideas, translating to language formulation, and revising to reading and editing text already written. The othe la el accounted for thoughts unrelated to the task (e.g., day dreaming). Orchestration and cognitive effort of the writing processes were analysed with number of occurrences of each writing process and with length of RTs. PROCEDURE Data were collected in individual sessions that lasted for around 60 min. The experimental procedure involved the following steps: the directed retrospection training, the collection of baseline RTs, the composition with concurrent RT probes followed by directed verbalisations, and a copy task with RT probes only. In the first step of the experiment, participants were trained in the directed verbalisation categories. The experimenter began with instructions that defined the writing processes (planning, translating, and revising) in nontechnical terms. As explained before, a fourth category unrelated to the iti g p o esses as added othe , fo all thoughts that did not fit the defined writing processes. The instructions then continued with examples of thinking-aloud protocols to illustrate each writing process (for example, for pla i g: I thi ki g a out hat to sa . . . , I ha e to ake a threepart te t ; fo t a slati g: I sea hi g a o d , This se te e has to e o st u ted like this ; fo e isi g: The se o d se te e of this pa ag aph is too lo g a d o ple , I he ki g spelli g . Ne t, pa ti ipants were required to categorise several examples of thinking-aloud protocols (for e a ple, fo pla i g: I sea hi g hat to sa a out this . . . , I ha e to sa this idea fi st a d that o e afte ; fo t a slati g: This se te e is appropriate , Does this o d takes a s at the e d ; fo e isi g: This pa ag aph should e displa ed at the e d , I eadi g te t . Ea h time an error occurred, the experimenter provided feedback to correct the pa ti ipa t s u de sta di g of the meaning of the different categories. During the second step of the experiment, the reaction time task was introduced. Participants were informed that during the composition they ould o asio all hea a audito sig al eep . The e e asked to react as quickly as possible to these beeps by pressing the spacebar of a keyboard with their nondominant hand while handwriting with the dominant one. After delivering these instructions, a series of 25 single-task RTs were collected. The first five trials were treated as warm-up signals and the mean baseline RT was calculated from the 20 remaining RTs. The probes were distributed in a random interval with a mean interval of once every 10 s and a range of 5 s to 15 s. During the composition, probes were distributed with a mean interval of once every 30 s and a range of 15-45 s. Directly after the collection of the baseline RTs, the experimenter read the writing assignment and gave the topic of composition to the 7 participant who then began composing. In the final phase of the experiment, participants received instructions regarding the copying task. They were informed that while copying, they would have to continue responding as fast as possible to probes (in an interval ranging from 15 to 45 s), but that no labelling would be required. RESULTS REACTION TIMES Baseline reaction times were not reliably different between the standard (M = 414 ms, SD = 89 ms) and uppercase (M = 423 ms, SD = 83 ms) conditions, t(38)= - 0.321. For all other analyses, we calculated interference in reaction time (RT baseline RT). To test whether amount of effort devoted to handwriting increased in the uppercase calligraphy relative to the standard calligraphy, we first compared interference RT scores (iRT) associated to handwriting in the copying task in the two calligraphy conditions. As expected, iRTs were higher in the uppercase condition (M = 383 ms, SD = 90 ms) than in the standard condition (M = 208 ms, SD = 74 ms), t(38) = - 6.68, p <.0001. Then, to analyse whether the increase in demands of handwriting in the uppercase condition affected coordination of the writing processes, an ANOVA was conducted with calligraphy as a between-participants factor and activity (copy transcription, composition pause, composition transcription) as a within-participant factor. Figure 1 shows the iRTs for the standard and uppercase conditions while handwriting during copying, and pausing and handwriting during composition. As predicted, the Calligraphy x Activity interaction was significant, F(2, 76)= 5.144, MSE = 4948.64, p <.001. As a consequence, we conducted separate analyses for the standard and uppercase conditions. Figure 1. Interference in RTs while transcribing during dictation, and while executing or pausing during composition of a narrative for writers using a standard or uppercase calligraphy (Experiment 1). Error bars indicate standard deviations. 8 For participants using the uppercase calligraphy, the iRTs were not reliably different, F(2, 38) = 1.221, MSE = 5474.82. By contrast, for participants using their standard calligraphy, RT interferences scores were reliably different, F(2, 38) = 28.810, MSE = 4830.41, p <.0001. Scheffé post hoc analyses showed that iRTs associated with transcription or pausing in composition (M = 364 ms, SD = 72 ms, and M = 337 ms, SD = 91 ms, respectively) were both higher than when transcribing during the copying task. This esult, hi h epli ates Oli e a d Kellogg s fi di gs, indicates that participants that used their familiar handwriting were engaged in high-level writing processes concurrently with transcription. Regarding the cognitive effort associated with planning, translating and revising during pauses or while handwriting (see Table 1), iRTs associated to the writing processes were entered in a mixed ANOVA with calligraphy, processes, and activity as factors. The analysis revealed a reliable main effect of the calligraphy condition, F(1, 29) = 14.662, MSE = 30460.5, p < .001. Reaction time interference scores were longer in the uppercase condition (M = 457 ms, SD = 94 ms) than in the standard condition (M = 358 ms, SD = 97 ms), indicating that the unfamiliar handwriting had an impact on available resources for high-level writing processes. A reliable main effect of writing processes was also observed, F(2, 58) = 31.147, MSE = 3300.54, p < .0001. Post hoc analyses showed that translating was significantly (p < .0001) less demanding than planning and revising. No significant interaction was found. Table 1 Cognitive effort of the writing processes (RT interference scores in ms) according to transcribing) and type of calligraphy (standard vs. uppercase) ite s a ti it pausi g o Standard deviations are indicated in parentheses. OCCURRENCES OF WRITING PROCESSES Percentages of occurrences of the writing processes (planning, translating, and revising) during pauses or while handwriting were calculated for each participant and then averaged across subjects. A mixed ANOVA with calligraphy, processes, and activity as factors was conducted. Percentages of occurrences are presented in Figure 2. The analysis revealed a main effect of activity, F(1, 38) = 4.791, MSE = 58.93, p < .05. There were more processes designated while pausing (M = 17.8%, SD = 9.8%) than while handwriting (M = 15.5%, SD = 12%). However, this difference happened because when composing a text writers spend more time on pauses than on handwriting (Alamargot et al., 2007; Alves et al., 2007). A main effect of processes was also observed, F(2, 76) = 14.668, MSE = 160.58, p < .0001. Post hoc 9 comparisons (p < .001) showed that translating was more frequent (M = 21.9%, SD = 11%) than revising (M = 11.1%, SD = 8.7%) but not than planning (M = 17%, SD =9.7%), the two latter being reliably different. The interaction between processes and activity was significant, F(2, 76) = 11.924, MSE = 74.04, p < .0001, as was the Processes x Activity x Calligraphy interaction, F(2, 76) = 5.135, MSE = 74.04, p < .01. Accordingly, we conducted separate ANOVAs in the standard and uppercase calligraphy conditions with processes and activity as within-participant factors. Figure 2. Occurrences of writing processes during pauses or transcription for writers composing a narrative with a standard or uppercase calligraphy (Experiment 1). With uppercase calligraphy, only a main effect of processes was observed, F(2, 19) = 5.089, MSE = 149.23, p < .01. Post hoc comparisons indicated that translating (M = 20.1%, SD = 9.6%) was more frequent than revising (M = 12.1%, SD = 8.6%) but not than planning (M = 17.1%, SD = 8.7%), the two latter being no different. No other reliable effect was observed. In the standard calligraphy condition, the writing processes were also differently activated, F(2, 38) = 9.831, MSE = 171.94, p < .001. Post hoc comparisons revealed that revising (M = 10.1%, SD = 8.9%) was less frequent than translating (M = 23.1%, SD = 12.3%) but not than planning (M = 16.8%, SD = 10.1%). By contrast with the uppercase condition, the Processes x Activity interaction was significant, F(2, 38) = 12.401, MSE = 97.412, p < .0001. As shown in Figure 2, translating was more activated during handwriting than during pauses (Scheffé, p < .05). Revising and planning showed the reverse pattern: These processes were more frequent during pauses than during handwriting (p < .01). During handwriting, revising and planning were activated but less so than translating (p < .01), and during pauses, translating occurred as often as planning and revising. WRITING PERFORMANCE F LUENCY . During the copying task, the number of words per minute differed significantly between the two calligraphy conditions, t(38)= -10.64, p < .0001 (see Table 2). Writers copying in uppercase produced fewer words per minute than 10 writers using standard calligraphy. During composition, fluency was also significantly reduced in the uppercase than in the standard handwriting condition, t(38) = -1.349, p <.0001 (see Table 2). Table 2 Writing performance in the standard and uppercase handwriting conditions Standard deviations are indicated in parentheses. S ENTENCE LENGTH . Writers in the uppercase condition composed sentences that were reliably shorter than in the standard handwriting condition, t(38) = - 6.707, p < .05 (see Table 2). R EVISIONS . The percentage of words that were revised did not significantly differ between the standard and uppercase conditions, t(38) = 0.187 (see Table 2). N UMBER OF GRAMMATICAL ERRORS . We calculated the number of spelling and syntactic errors for 100 words (see Table 2). A mixed 2 (errors: spelling, syntactic) x 2 (calligraphy: standard, uppercase) ANOVA was carried out. No reliable difference was observed between the two handwriting conditions, F(1, 38) = 1.32, MSE = 0.41. A main effect of errors was observed, F(1, 38) = 4.45, MSE = 0.21, p < .05. Writers made more syntactic than spelling errors. There was no significant interaction, F(1, 38) < 1. T EXT QUALITY . Mean scores of quality are presented in Table 2. A mixed 2 (calligraphy: standard, uppercase) x 2 (quality: style, content) ANOVA was carried out on scores of text quality. The handwriting conditions significantly affected holistic quality, F(1, 38) = 32.88, MSE = 0.578, p < .0001. Texts produced by writers using their standard handwriting were judged of higher quality than texts produced by the writers using an uppercase handwriting. The style and content subscales did not differ, F(1, 38) < 1. The Calligraphy x Quality interaction was not reliable, F(1, 38) < 1. DISCUSSION The findings of this experiment showed that planning, translating, and revising were mostly activated during pauses (between 20% and 15%). Thus, none of these writing processes can be said to be typical of pauses. During execution periods, translating dominated (30%), but planning and revising also co-occurred, even if to a limited extent (13% and 6%, respectively). So, this a age e t of the iti g p o esses a o di g to o e t ite s behaviour does not seem to be specific to the execution mode since the esults ith ite s usi g thei fa ilia ha d iti g epli ate Al es et al. s (in press) findings obtained with typists. The findings also indicate that when resources are available, any high-level writing processes can be managed concurrently with execution, but the large standard deviations of occurrences of the writing processes indicate a great variability i ite s st ateg and suggest that some writers used a more sequential strategies while others presumably always activated the high-level writing processes concurrently to handwriting. It is also important to notice the close link between translating and handwriting. This finding is in line with cascade models 11 of speech production in which the linguistic formulation of speech closely precedes articulation (Kempen & Hoenkamp, 1987; Rapp & Goldrick, 2000). I o t ast ith Al es et al. s i p ess fi di gs, however, increased handwriting demands affected how planning, translating, and revising are activated. Writers shifted to a more sequential strategy by activating translating in pauses as much as while motor execution. This change can be explained by the fact that in this condition fewer resources were available to writers; so, they were not able to activate at the same time motor execution and high-level writing processes. Unfamiliar motor execution also affected writing fluency. Indeed, when handwriting their text with a cursive uppercase calligraphy, writers composed more slowly. Slowness of uppercase handwriting of course results from uppercase calligraphy itself, which implies longer drawings of letters, and as indicated by the reduction of fluency in the copying task with an uppercase calligraphy. Interestingly, the cognitive effort of the writing processes was higher in the uppercase condition, suggesting an impact of the unfamiliar handwriting on high-level writing processes. Sentences were shorter in uppercase than when writers used their standard handwriting. This sentence reduction has often been observed in writing when composition is most difficult (Kellogg, Olive, & Piolat, 2007a, 2007b; Ransdell, Levy, & Kellogg, 2002), when motor execution is slowed, but also when producing sentences in high demanding o te ts Po e , . Follo i g Po e s suggestio , the sho te sentences probably indicate that writers reduced the amount of semantic work to manage the high processing demands of handwriting with an unfamiliar calligraphy. Finally, text quality was judged lower with uppercase handwriting indicating that uppercase handwriting has drawn resources away from high-level writing. This is also evidence of interaction between motor execution and high-level writing processes. EXPERIMENT 2 : ESSAY COMPOSITION Experiment 1 gave evidence that writing processes are activated differently during pauses and execution periods, and that the observed online management of the writing processes are similar when composing by typing (Alves et al., in press) or by writing by hand (Experiment 1). Moreover, when fewer resources were available because of an increase in amount of resources devoted to motor execution, writers shifted to a more sequential strategy for managing the writing processes. Experiment 2 tested the generality of this online management of the writing processes in a more demanding writing task, namely essay composition. As in Experiment 1, half of the participants composed their text using an uppercase calligraphy while the other half composed using their standard handwriting. They all composed their text while reacting as quickly as possible to auditory probes and categorising their ongoing mental activity. METHOD PARTICIPANTS Forty undergraduate psychology students participated in this experiment (mean age = 23 years, SD = 3.7). Half of the participants composed and copied their text using their usual handwriting (standard condition). The other half composed and copied their text using cursive uppercase calligraphy (uppercase condition). All participants were tested individually and were treated according to currently accepted ethical standards. TASKS AND PROCEDURE Except for the topic of the composition, the tasks and procedure were identical to Experiment 1. The composition task asked participants to compose an essay about a debate that was occurring in France at the time of the experiment related to prohibition of smoking inside offices, administrations, and public places. Instructions asked participants to articulate pro and con arguments about this topic. Again, composition time was set to 30 min (with a minimum of 20 min), participants were required to react as fast as possible to auditory beeps, and then to indicate their ongoing mental activity interrupted by the beep. They also performed the copying task. 12 RESULTS AND DISCUSSION REACTION TIMES Baseline reaction times did not reliably differ between the standard (M = 585 ms, SD = 99 ms) and uppercase conditions (M = 577 ms, SD = 111 ms), t(38) = 0.241. Again, in the uppercase condition, the mean iRT associated to handwriting during the copying task was reliably higher (M = 464 ms, SD = 109 ms) than in the standard condition (M = 277 ms, SD = 88 ms), t(38) = - 5.963, p < .0001. This finding, which replicates the result observed in Experiment 1, indicates that handwriting with cursive uppercase calligraphy increases the amount of resources devoted to motor execution. Consequently, in that condition, less cognitive capacity was available for writers to activate high-level planning, translating, and revising processes concurrently with handwriting. It is thus highly probable that writers in the uppercase condition changed how they activated writing processes during handwriting. Changes in coordination of the writing processes were analysed with a 2 (calligraphy: standard, uppercase) x 3 (activity: copy- transcription, composition pause, composition transcription) mixed ANOVA. Figure 3 shows the iRTs for the standard and uppercase conditions while handwriting under dictation, or handwriting and pausing under composition. As predicted, the Calligraphy$Activity interaction was significant, F(2, 76) = 13.707, MSE = 6028.44, p < .0001. So, we conducted separate analyses for the standard and uppercase conditions. Figure 3. Interference in RTs while transcribing during dictation, and while executing or pausing during composition of an essay for writers using a standard or uppercase calligraphy (Experiment 2). Error bars indicate standard deviations. 13 In the uppercase condition, the three iRTs did not reliably differ, F(2, 38) = 1.832, MSE = 6474.84. In contrast, for participants using their standard calligraphy, iRTs were reliably different, F(2, 38) = 47.593, MSE = 5582.05, p < .0001. As expected, Scheffé post hoc analyses showed that iRTs associated with transcription or pausing in composition (M = 454 ms, SD = 103 ms, and M = 494 ms, SD = 72 ms, respectively) were both higher than when handwriting during the copying task (M = 277 ms, SD = 88 ms). This finding indicates that participants that used their familiar handwriting engaged the high-level writing processes concurrently with transcription. In other words, having less cognitive resources at their disposal when handwriting with an unfamiliar calligraphy, writers shifted from a parallel strategy to a more sequential one. Then, iRTs associated to the writing processes were entered in a mixed ANOVA with calligraphy, processes, and activity factors. Data are presented in Table 1. The analysis revealed a reliable main effect of calligraphy, F(1, 33) = 7.476, MSE = 34036.28, p < .01. Global iRT was longer in the uppercase condition (M = 546 ms, SD = 127 ms) than in the standard condition (M = 476 ms, SD = 136 ms), indicating that demands of handwriting had an impact on demands of high-level writing processes. A reliable main effect of writing processes was also observed, F(2, 66) = 95.473, MSE = 6186.7, p < .0001. Post hoc analyses showed that iRTs associated to the writing processes reliably (p < .01) differed from each other, with the shortest RT associated to translating. No significant interaction was found. This result again replicates the findings of Experiment 1. It shows that, as already documented in the literature, translating is less demanding than planning or translating. More interesting, although calligraphy affected the global cognitive effort of composition it did not affect specific writing processes, as indicated by the absence of an interaction. OCCURRENCES OF WRITING PROCESSES The percentage of occurrence of each writing process was computed and entered into a mixed ANOVA with the calligraphy, processes, and activity factors. A main effect of processes was observed, F(2, 76) = 21.16, MSE = 91.53, p < .0001. Post hoc comparisons (p < .001) revealed that translating was more frequent (M = 21.8%, SD = 10.6%) than revising (M = 12.8%, SD = 7.5%) and planning (M = 15.5%, SD = 8.5%), the latter two being reliably different. That translating is more often activated than planning and revising during a composition is now well established in writing research. As expected, the Processes$Activity interaction was significant, F(2, 76) = 9.657, MSE = 83.47, p = .0002, as was the Processes x Activity x Calligraphy interaction, F(2, 76) = 8.055, MSE = 83.47, p < .001 (see Figure 4). Accordingly, we conducted separate ANOVAs in the standard and uppercase calligraphy conditions with processes and activity as withinparticipant factors. In the uppercase calligraphy condition, as expected, only a main effect of processes was observed, F(2, 38) = 5.629, MSE = 87.41, p = .007. Post hoc comparisons indicated that translating (M = 20.44%, SD = 8%) was more frequent than revising (M = 13.5%, SD = 7%) and planning (M = 16%, SD = 8.8%), the latter two being no different. No other reliable effect was observed. With the standard calligraphy, writing processes were also differently activated, F(2, 38) = 17.766, MSE = 75.64, p < .0001. As in the uppercase condition, translating (M = 23.2%, SD = 12.8%) was more frequent than revising (M = 12%, SD = 8.1%) and planning (M = 14.9%, SD = 8.2%), the latter two being no different. As predicted, the Processes x Activity interaction was reliable, F(2, 38) = 15.411, MSE = 95.6, p < .0001. As shown in Figure 2, translating was more activated during handwriting than during pauses (Scheffé, p < .01). Revising and planning showed the reverse pattern: Revising and planning were more frequent during pauses than during handwriting (p < .01). Finally, during handwriting, revising and planning were activated less than translating (p < .0001), and during pauses, translating occurred as often as planning and revising. These findings replicate the ones of Experiment 1: We again found evidence that the demands of handwriting affected how planning, translating, and revising are coordinated with handwriting. More precisely, especially translating was less activated concurrently with handwriting when execution demands where increased. This is mainly due to lack of available resources when handwriting. 14 Figure 4. Occurrences of writing processes during pauses or transcription for writers composing an essay with a standard or uppercase calligraphy (Experiment 2). WRITING PERFORMANCE F LUENCY . During the copying task, the number of words per minute significantly differed between the two handwriting conditions, t(38) = - 19.48, p < .0001 (see Table 2). Writers copying with an uppercase calligraphy produced fewer words per minute than writers using standard handwriting. During composition, fluency was also reliably slower in the uppercase condition than in the standard handwriting condition, t(38) = 17.31, p < .0001 (see Table 2). These results replicate findings of Experiment 1. S ENTENCE LENGTH . A Student t-test showed a significant difference between the two handwriting conditions, t(38) = 2.44, p < .05. Writers in the uppercase condition composed shorter sentences than writers in the standard handwriting condition (see Table 2). R EVISIONS . The percentage of words that were revised was significantly affected by the handwriting conditions, t(38) = 2.78, p < .01. Writers in the standard condition revised more words than writers in the uppercase condition (see Table 2). N UMBER OF GRAMMATICAL ERRORS . Number of spelling and syntactic errors for 100 words (see Table 2) was entered into a mixed 2 (errors: spelling, syntactic) x 2 (calligraphy: standard, uppercase) ANOVA. No reliable difference was observed between the two handwriting conditions, F(1, 38) = 2.78, MSE = 1.32. A main effect of errors was observed, F(1, 38) = 7.78, MSE = 0.52, p < .01.Writers made more grammatical errors than spelling errors. There was no significant interaction, F(1, 38) < 1. T EXT QUALITY . Mean scores of quality are presented in Table 2. Reliable agreement between the judges (all correlations significant at p < .01) was found for the two scales (rs > .80). A mixed 2 (calligraphy: standard, uppercase) x 2 (quality: style, content) ANOVA was carried out on scores of text quality. The calligraphy conditions significantly 15 affected holistic quality, F(1, 38) = 19.57, MSE = 1.26, p < .0001. Texts produced by the writers using their standard handwriting were judged of higher quality than texts produced by the writers using an uppercase handwriting. Style was evaluated as better than content, F(1, 38) = 1.178, MSE = 0.59. The Calligraphy x Quality interaction was reliable, F(1, 38) = 7.96, MSE = 0.59, p < .01. Post hoc comparisons showed that the quality of style tended to decrease in the uppercase condition relative to the standard condition (p = .053), while the score of content decreased more largely in the uppercase condition (p < .0001). In sum, handwriting with an unfamiliar calligraphy impacted on the cognitive effort of writing, on writing strategies that shifted from parallel to more sequential, and on writing performance. Writers were less fluent, produced texts with shorter sentences, of lower quality, and with more errors. It must be noticed, however, that the decrease in the number of revisions carried out may not result from the low availability of resources, but may athe efle t the ite s difficulty in reading their text written in cursive uppercase calligraphy. It is difficult to interpret the reduction of sentence length and in holistic quality in the same line. Even if the difficulty to read the text written in uppercase calligraphy may have prevented writers to make extensive revision, global cognitive effort was higher in the uppercase condition indicating that writers were probably near to a cognitive overload. This may explain why writers were less able to perform more demanding planning and revising operations. GENERAL DISCUSSION The two experiments presented in this paper studied how writers manage writing processes when composing a text. Online management of the writing processes has received strong attention because several findings suggest a elatio ship et ee te t ualit a d ite s skill to deal ith the li ited apa it of iti g e o Le & ‘a sdell, 1995, 1996). As all writing processes are resource-consuming and share working memory capacity, writers have to use writing strategies that optimise the involvement of working memory (Kellogg, 1996; McCutchen, 1996, 2000; Olive et al., 2002). This suggests that pauses are ideal moments for writers to carry out the more demanding operations (structuring their text, revising its meaning, etc.; Schilperoord, 2002). The present findings indicate that none of the writing processes is typical of pauses: Planning, translating, and revising occur at the same extent during pauses. However, writers still have the opportunity to coordinate some of the high-level writing processes concurrently to handwriting si e adults ha d iti g is effo tless Oli e & Kellogg, . So, which processes occur when writers are actually writing down their text? The present findings indicate that all types of writing processes can cooccur with handwriting, but that translating dominates (planning: between 10% and 15%; translating: around 30%; revising: between 5% and 10%). The finding that translating is closely linked to motor execution is coherent with a cascade view of the flow of information in writing. Why is translating most probable to be executed during handwriting than other processes? First, planning requires several central cognitive operations, such as decision making, inference, and goal setting, all of which require executive resources of working memory (Hayes, 1996; Hayes & Nash, 1996; Kellogg, 1996). Second, revision requires that the writer reads his/her text not only at the linguistic but also at the conceptual level. Such reading is also very effortful (Roussey & Piolat, 2008). By contrast, translating is the less demanding process. In a capacity view of writing in which working memory capacity is shared between the writing processes, the process most likely to cooccur with another is the less effortful one, in the present case translating. Of course, this does not mean that planning and revising can occur only during pauses. As the findings of Experiments 1 and 2 indicate, some planning and revising activities are carried out concurrently to motor execution. Because planning and revising are cognitive components made of several subprocesses, it is possible that less demanding subprocesses such as idea retrieval for planning or checking spelling errors for revising can be activated in parallel with motor execution. This pattern of online management of the writing process during pauses and execution periods does not seem to be specific either of genre or of output modality since it was observed both in narrative and essays composition and with typing (Alves et al., in press) and handwriting (this paper) as motor execution modalities. Rather, this management of high-level writing processes in parallel with motor execution is a general tendency of writers, it is not situation specific. 16 One important difference between Alves et al. s stud a d the t o p ese t e pe i e ts ust e oted. I iti g research, pauses of interest are pauses during which high-level writing processes occur. Pauses are an interruption of handwriting. However, as indicated in the Method section, some short pauses are the result of mere handwriting actions. So, usually, pauses of interest are pauses that exceed a predefined length. In the present experiments, a 250 ms threshold was used while Alves et al. used a 1 s threshold. Interestingly, despite some minor changes in amount of occurrence of the writing processes, this difference did not affect how writing processes were orchestrated in the both studies. This again provides strong support for the idea that the management of the writing processes, as it is observed in the experiments reported here, is a very general way to organise the writing processes given the limits of working memory capacity. The role of available working memory capacity is also confirmed by the change that was observed when writers wrote down their text with an unfamiliar handwriting. In that case, cooccurrence of translating and of motor execution dropped. However, the increase in processing demands of execution did not result in a strict sequential organisation of the writing processes. Even when handwriting with an unfamiliar calligraphy, planning, translating, and revising sometimes occurred concurrently with motor execution. As indicated previously, planning, translating, and revising are cognitive components made of several subprocesses. It is this possible that less demanding subprocesses can cooccur with motor execution, even if this process is very effortful. It must be noted that, at some times in the production (with a familiar calligraphy), cumulated demands of handwriting and of the high-level writing processes may exceed available capacity. When faced with such demands, to avoid decrement in performance, writers have to shift to a more sequential strategy (Brown & Carr, 1989; Fayol, 1999). Such a shift in functioning could announce a pause to come, but they could also signal the end of the processing of a planning unit. Future research should investigate these shifts in order to better understand the dynamics of the writing processes. The present findings have implications for the comprehension of writing development. As shown by McCutchen (1996, 2000) and Swanson and Berninger (1994, 1996), one important constraint in writing development is related to the huge demands or handwriting in beginning writers. For instance, difficulties with motor transcription in children are known to correlate with poor writing skills (Graham, Berninger, Abbot, Abbot, & Whitake , . Child e diffi ult i managing the writing processes may be quite similar to that encountered by adult writers using an unfamiliar calligraphy. As Olive and Kellogg (2002) found, third-graders are not able to activate the high-level writing processes concurrently to handwriting because handwriting is not yet automatised. Thus, the high demands and difficulties of motor transcription in children not only preclude concurrent activation of processes, they may explain in part why young writers fail to engage in highlevel processes (McCutchen, 1996). To better understand how children acquiring skills in composition succeed in orchestrating the writing processes, it is now necessary to evaluate when beginning writers become able to concurrently activate different types of writing processes, and which processes are more likely to cooccur with motor execution. According to the developmental model of Berninger and Swanson (1994), as translating is the first process (after handwriting and spelling) that is functionally efficient, it should also be the first to be coordinated with motor execution. However, since planning and revising are fundamental for the producing text of quality, the moment in writing acquisition when these processes are sufficiently mastered to be coordinated with handwriting may be of i po ta e i p edi ti g hild e s iti g a hie e e t. 17 REFERENCES Alamargot, D., Chesnet, D., Dansac, C., & Ros, C. (2006). Eye and pen: A new device to study reading during writing. Behavior Research Methods, Instruments, and Computers, 38, 287-299. Alamargot, D., Dansac, C., Chesnet, D., & Fayol, M. (2007). Parallel processing before and after pauses: A combined analysis of graphomotor and eye movements during procedural text production. In G. Rijlaarsdam (Series Ed.), M. Torrance, L. van Waes, & D. Galbraith (Eds.), Studies in writing: Vol. 20. Writing and cognition: Research and applications (pp. 13-29). Amsterdam: Elsevier. Alves, R. A., Castro, S. L., & Olive, T. (in press). Execution and pauses in writing narratives: Processing time, cognitive effort and typing skill. International Journal of Psychology. Alves, R. A., Castro, S. L., Sousa, L., & Stro¨mqvist, S. (2007). Typing skill and pause-execution cycles in written composition. In G. Rijlaarsdam (Series Ed.), M. Torrance, L. van Waes, & D. Galbraith (Eds.), Studies in writing: Vol. 20. Writing and cognition: Research and applications (pp. 55-66). Amsterdam: Elsevier. Be i ge , V. W., & S a so , H. L. 4 . Modif i g Ha es a d Flo e s odel of skilled iti g to e plai egi i g and developing writing. In E. C. Butterfield (Ed.), Ad ances in cognition and educational practice: Vol 2. Children’s riting: To ard a process theory of the development of skilled writing (pp. 57-81). Greenwich, CT: JAI Press. Bosman, E. A. (1993). Age-related differences in the motoric aspects of transcription typing skill. Psychology of Aging, 7, 87-102. Bourdin, B., & Fayol, M. (1994). Is written language production more difficult than oral language production? A working memory approach. International Journal of Psychology, 29, 591-620. Bourdin, B., & Fayol, M. (2002). Even in adults, written production is still more costly than oral production. International Journal of Psychology, 37, 219-227. Breetvelt, I., van Den Bergh, H., & Rijlaarsdam, G. (1994). Relations between writing processes and text quality: When and how? Cognition and Instruction, 12(2), 103-123. Brown, T. L., & Carr, T. H. (1989). Automaticity in skill acquisition: Mechanisms for reducing interference in concurrent performance. Journal of Experimental Psychology: Human Perception and Performance, 15, 686-700. Brown, J. S., McDonald, J. L., Brown, T. L., & Carr, T. H. (1988). Adapting to processing demands in discourse production: The case of handwriting. Journal of Experimental Psychology: Human Perception and Performance, 14, 45-59. Chanquoy, L., Foulin, J. N., & Fayol, M. (1990). Temporal management of short text writing by children and adults. European Bulletin of Cognitive Psychology, 10, 513-540. Engle, R. W. (2002). Working memory capacity as executive attention. Current Directions in Psychological Science, 11, 19-23. Ericsson, K. A., & Simon, H. A. (1993). Protocol analysis: Verbal reports as data (Rev. ed.). Cambridge, MA: MIT Press. Favart, M., & Coirier, P. (2006). Acquisition of the linearization process in text composition in third to ninth graders: Effects of textual superstructure and macrostructural organization. Journal of Psycholinguistic Research, 35(4), 305-328. Fayol, M. (1999). From on-line management problems to strategies in written composition. In M. Torrance & G. Jeffery (Eds.), The cognitive demands of writing: Processing capacity and working memory effects in text production (pp. 13!23). Amsterdam: Amsterdam University Press. Flower, L. S., & Hayes, J. R. (1980). The dynamics of composing: Making plans and juggling constraints. In L. W. Gregg & E. R. Steinberg (Eds.), Cognitive processes in writing (pp. 31-49). Hillsdale, NJ: Lawrence Erlbaum Associates, Inc. Foulin, J. M. (1995). Pauses et de´bits: Les indicateurs temporels de la production écrite. L. Année Psychologique, 95, 483-504. Goldberg, A., Russell, M., & Cook, A. (2003). The effect of computers on student writing: A metaanalysis of studies from 1992 to 2002. Journal of Technology, Learning and Assessment, 2(1). Retrieved 10 September 2005, from http://www.bc.edu/research/intasc/jtla.html Graham, S., Berninger, V. W., Abbott, R. D., Abbott, S. P., & Whitaker, D. (1997). Role of mechanics in composing of elementary school students: A new methodological approach. Journal of Educational Psychology, 89, 170-182. Hayes, J. R. (1996). A new framework for understanding cognition and affect in writing. In C. M. Levy & S. E. Ransdell (Eds.), The science of writing: Theories, methods, individual differences, and applications (pp. 1-27). Mahwah, NJ: Lawrence Erlbaum Associates, Inc. Hayes, J. R., & Chenoweth, N. A. (2003). The inner voice in writing. Written Communication, 23(2), 135-149. Hayes, J. R., & Chenoweth, N. A. (2006). Is working memory involved in the transcribing and editing of texts? Written Communication, 20(1), 99-118. Hayes, J. R., & Flower, L. S. (1980). Identifying the organization of writing processes. In L. W. Gregg & E. R. Steinberg (Eds.), Cognitive processes in writing (pp. 3-30). Hillsdale, NJ: Lawrence Erlbaum Associates, Inc. Hayes, J. R., & Nash, J. G. (1996). On the nature of planning in writing. In C. M. Levy & S. E. Ransdell (Eds.), The science of writing: Theories, methods, individual differences, and applications (pp. 29-55). Mahwah, NJ: Lawrence Erlbaum Associates, Inc. Janssen, D., van Waes, L., & van den Bergh, H. (1996). Effects of thinking aloud on writing processes. In C. M. Levy & S. E. Ransdell (Eds.), The science of writing: Theories, methods, individual differences, and applications (pp. 233-250). Mahwah, NJ: Lawrence Erlbaum Associates, Inc. Just, M. A., & Carpenter, P. A. (1992). A capacity theory of comprehension: Individual differences in working memory. Psychological Review, 99, 122-149. Kellogg, R. T. (1987a). Effects of topic knowledge on the allocation of processing time and cognitive effort to writing processes. Memory and Cognition, 15, 256-266. Kellogg, R. T. (1987b). Writing performance: Effect of cognitive strategies. Written Communication, 4, 269-298. 18 Kellogg, R. T. (1988). Attentional overload and writing performance: Effects of rough draft and outline strategies. Journal of Experimental Psychology: Learning, Memory, and Cognition, 14, 355-365. Kellogg, R. T. (1996). A model of working memory in writing. In C. M. Levy & S. E. Ransdell (Eds.), The science of writing: Theories, methods, individual differences, and applications (pp. 57-71). Mahwah, NJ: Lawrence Erlbaum Associates, Inc. Kellogg, R. T. (2001). Competition for working memory among writing processes. American Journal of Psychology, 114, 175-191. Kellogg, R. T., & Mueller, S. (1993). Performance amplification and process restructuring in computer-based writing. Journal of Man!Machine Studies, 39, 33-49. Kellogg, R. T., Olive, T., & Piolat, A. (2007a). Verbal, visual and spatial working memory in written language production. Acta Psychologica, 124, 382-397. Kellogg, R. T., Olive, T., & Piolat, A. (2007b). Verbal and visual working memory during sentence production. In G. Rijlaarsdam (Series Ed.), M. Torrance, L. van Waes, & D. Galbraith (Eds.), Studies in writing: Vol. 20. Writing and cognition: Research and applications (pp. 97-108). Amsterdam: Elsevier. Kempen, G., & Hoenkamp, E. (1987). An incremental procedural grammar for sentence formulation. Cognitive Science, 15, 173-218. Levy, C. M., & Ransdell, S. E. (1995). Is writing as difficult as it seems? Memory and Cognition, 23(6), 767-779. Levy, C. M., & Ransdell, S. E. (1996). Writing signatures. In C. M. Levy & S. E. Ransdell (Eds.), The science of writing: Theories, methods, individual differences, and applications (pp. 149-161). Mah ah, NJ: La e e E l au Asso iates, I . M Cut he , D. . Fu tio al auto ati it i hild e s iti g: A p o lem of metacognitive control. Written Communication, 5, 306-324. McCutchen, D. (1996). A capacity theory of writing: Working memory in composition. Educational Psychology Review, 8, 299-325. McCutchen, D. (2000). Knowledge, processing, and working memory: Implications for a theory of writing. The Educational Psychologist, 35, 13-23. Navon, D., & Miller, J. (1987). Role of outcome conflict in dual task interference. Journal of Experimental Psychology: Human Perception and Performance, 13, 435-448. Olive, T. (2004).Working memory in writing: Empirical evidence from the dual-task technique. The European Psychologist, 9, 32-42. Olive, T., & Kellogg, R. T. (2002). Concurrent activation of high- and low-level production processes in written composition. Memory and Cognition, 30, 594-600. Olive, T., Kellogg, R. T., & Piolat, A. (2002). The triple task technique for studying the process of writing. In T. Olive & C. M. Levy (Eds.), Contemporary tools and techniques for studying writing (Vol. 10, pp. 31!59). Dordrecht, the Netherlands: Kluwer Academic Publishers. Pashler, H. (1994). Dual-task interference in simple tasks: Data and theory. Psychological Bulletin, 116, 220-244. Passerault, J. M. . L a al se e te ps éel de l a ti ité de ésu é: u e étude des temps de pause. In M. Charolles & A. Petitjean (Eds.), Le résumé de texte. Aspects linguistiques, sémiotiques, psycholinguistiques et automatiques (pp. 207-219). Paris: Klincksieck. Penningroth, S. L., & Rosenberg, S. (1995). Effects of a high information-processing load on the writing process and the story written. Applied Psycholinguistics, 16, 189-210. Piolat, A., Kellogg, R. T., & Farioli, F. (2001). The triple task technique for studying writing processes: On which task is attention focused? Current Psychology Letters: Brain Behavior and Cognition, 4, 67-83. Piolat, A., & Olive, T. (2000). Comment étudier le coût et le déroulement de la rédaction de textes? La méthode de triple-tâche: un bilan méthodologique. L’Année Psychologique, 100, 465-502. Piolat, A., Olive, T., Roussey, J.-Y., Thunin, O., & Ziegler, J. C. (1999). ScriptKell: A computer assisted tool for measuring the distribution of time and cognitive effort in writing and other complex cognitive activities. Behavior Research Methods, Instruments and Computers, 31, 113-121. Piolat, A., Roussey, J.-Y., Olive, T., & Farioli, F. (1996). Charge mentale et mobilisation des processus rédactionnels: examen de la procédure de Kellogg. Psychologie française, 41, 339-354. Power, M. J. (1986). A technique for measuring processing load during speech production. Journal of Psycholinguistic Research, 15, 371-382. Ransdell, S. (1995). Generating thinking aloud protocols: Impact on the narrative writing of college students. American Journal of Psychology, 108, 89-98. Ransdell, S., Levy, C. M., & Kellogg, R. T. (2002). The structure of writing processes as revealed by secondary task demands. L1Educational Studies in Language and Literature, 2, 141-163. Rapp, B., & Goldrick, M. (2000). Discreteness and interactivity in spoken word production. Psychological Review, 107, 460-499. Rieger, M. (2004). Automatic keypress activation in skilled typing. Journal of Experimental Psychology: Human Perception and Performance, 30, 555-565. Roussey, J.-Y., & Piolat, A. (2008). Critical reading effort during text revision. European Journal of Cognitive Psychology, 20, 765-792. Schilperoord, J. (2002). On the cognitive status of pauses in discourse production. In T. Olive & C. M. Levy (Eds.), Contemporary tools and techniques for studying writing (pp. 61-87). Dordrecht, The Netherlands: Kluwer Academic Publishers. Strömqvist, S., & Ahlsén, E. (Eds.).(1999). The process of writing: A progress report (Gothenburg Papers in Theoretical Linguistics No. 83). Gothenburg, Sweden: University of Göteborg, Department of Linguistics. Swanson, H. L., & Berninger, V. W. (1994). Working memory as a source of individual differences i hild e s iti g. I J. S. Ca lso (Series Ed.), E. C. Butterfield (Ed.), Advances in cognition and educational practice: Vol. 2. Children’s riting: To ard a process theory of the development of skilled writing (pp. 31-56). Greenwich, CT: JAI Press. 19 Swanson, H. L., & Berninger, V. W. (1996). Individual differences i hild e s o ki g e o and writing skill. Journal of Educational Psychology, 63(2), 358-385. Torrance, M., & Galbraith, D. (2006). The processing demands of writing. In C. A. MacArthur, S. Graham, & J. Fitzgerald (Eds.), Handbook of writing research (pp. 67-80). New York: Guilford Publications. Van Den Bergh, H., & Rijlaarsdam, G. (2007). The dynamics of idea generation during writing: An on-line study. In G. Rijlaarsdam (Series Ed.), M. Torrance, L. vanWaes, & D. Galbraith (Eds.), Studies in writing: Vol. 20. Writing and cognition: Research and applications (pp. 125-150). Amsterdam: Elsevier. Wengelin, A. (2007). The word-level focus in text production by adults with reading and writing difficulties. In G. Rijlaarsdam (Series Ed.), M. Torrance, L. van Waes, & D. Galbraith (Eds.), Studies in writing: Vol. 20. Writing and cognition: Research and applications (pp. 67-82). Amsterdam: Elsevier. 20 21