Elsevier

Acta Psychologica

Volume 124, Issue 3, March 2007, Pages 382-397
Acta Psychologica

Verbal, visual, and spatial working memory in written language production

https://doi.org/10.1016/j.actpsy.2006.02.005Get rights and content

Abstract

College students wrote definitions of either abstract or concrete nouns in longhand while performing a concurrent working memory (WM) task. They detected either a verbal (syllable), visual (shape), or spatial (location) stimulus and decided whether it matched the last one presented 15–45 s earlier. Writing definitions of both noun types elevated the response time to verbal targets above baseline. Such interference was observed for visual targets only when defining concrete nouns and was eliminated entirely with spatial targets. The interference effect for verbal targets was the same whether they were read or heard, implicating phonological storage. The findings suggest that language production requires phonological or verbal WM. Visual WM is selectively engaged when imaging the referents of concrete nouns.

Introduction

Language production requires the planning of semantic content and the grammatical encoding of this content (Bock & Levelt, 1994). Imaginal and propositional representations are translated into the ordered words of a phrase, clause, or sentence through grammatical encoding. Spoken output requires phonological encoding, whereas writing requires orthographic encoding either directly from an orthographic lexicon or via phonological mediation (Caramazza, 1991).1 Phonological encoding in writing may do more than provide an optional means of gaining access to orthography. Lexical entries that are encoded phonologically and heard as an inner voice allow the writer to review and edit a pre-text representation, prior to committing to handwritten or typed output (Chenoweth and Hayes, 2003, Witte, 1987).

Our interest was in examining the degree to which specific components of working memory (WM) support written language production. Because written production is slow compared with speech, it is feasible to measure performance on a concurrent task that uses executive attention plus either verbal, visual, or spatial storage (Jonides & Smith, 1997). Interference with the concurrent task would suggest that one or more processes involved in language production compete for the same WM component. For example, following Baddeley’s (1986) seminal model, a phonological loop allows the storage and maintenance of verbal representations. Interference with maintaining a syllable (ba or da) in phonological storage while concurrently writing (i.e., compared with baseline, single task performance) would suggest that writing involves storing phonological representations of the words of a sentence under construction. Understanding the WM requirements of writing is of interest in its own right and may shed light on language production in general to the extent that speaking and writing share processes and requirements.

If a writer encodes lexical entries phonologically during production, then the phonological loop may store these verbal representations either for review or for conversion to the orthographic codes needed in written motor output (Kellogg, 1996). Shelton and Caramazza (1999) noted that writing has traditionally been viewed as dependent on phonological encoding. Consistent with this view, concurrent tasks that make heavy demands on phonological or verbal WM shorten sentence length (Chenoweth and Hayes, 2003, Kellogg, 2004, Levy and Marek, 1998, Power, 1985, Ransdell et al., 2002) and can cause subject–verb agreement errors (Fayol et al., 1994, Largy and Fayol, 2001). For example, Chenoweth and Hayes (2003) found that repeatedly vocalizing the word “tap” suppresses the inner voice that often seems to accompany writing. They tracked pause durations in production and discovered that suppressing covert vocalization reduced the number of words that are produced in rapid succession prior to a long pause (2 s or longer).

Even so, two sources of neuropsychological evidence challenge the view that verbal working is necessary for holding lexical representations during language production. First, Vallar and Baddeley (1984) studied a patient with an impaired phonological loop whose memory span was limited to two items following auditory list presentation. This individual showed both normal spoken and written language. In reviewing cases of this type, Gathercole and Baddeley (1993) concluded that the language production process does not appear to require the verbal WM system. Second, writing does not necessarily require phonological processing based on clinical case studies (Shelton & Caramazza, 1999). For example, Shelton and Weinrich (1997) described a patient who could write real words successfully in 85% of the time but could not write a single non-word in which one must use phonological mediation to convert phonemes to graphemes. The patient’s ability to retrieve an orthographic lexical representation was intact, despite an impairment in the conversion mechanism.

The role of visual WM might be limited to planning the content of the message when the referents of concepts are imaged (Sadoski, Kealy, Goetz, & Paivio, 1997) as opposed to the mandatory stage of linguistic encoding. A concurrent visual–spatial task disrupts written production less than a verbal task (Kellogg, 2004), as one would expect if visual WM plays less of a role. On the other hand, perhaps visual WM at times stores the orthographic representations as an aid to catching spelling, punctuation, and other errors. Further, Hayes (1996) argued that spatial WM allows writers to represent the physical layout of an extended text. The empirical literature on how visual and spatial WM support writing is too limited to evaluate these hypotheses adequately.

The present research, therefore, examined again the assumption that verbal WM is required in written language production and expanded work on the roles played by the visual and spatial components of WM. We replicated the experiment of Sadoski et al. (1997) by making participants write definitions of either concrete or abstract nouns in longhand. They found that writers initiate production faster and compose more detailed, higher quality definitions for concrete compared with abstract nouns. Further, they reported using imagery more often in defining concrete relative to abstract words. We extended their work by having participants perform a concurrent task that required verbal, visual or spatial WM. We expected to replicate their definition findings and to find that only concrete nouns slow responses to a concurrent visual WM task. In contrast, both kinds of nouns were expected to interfere with a verbal WM task. Finally, we expected that the definition task would fail to prompt any demands on spatial WM, neither for concrete nor for abstract nouns, because the layout of an extended text was not developed.

The concurrent WM task was designed to require either the verbal, visual, or spatial component, plus the executive functions demanded in juggling the secondary task with writing (D’Esposito et al., 1995). The verbal WM task required detecting visually presented syllables (ba or da) on a 30 s variable interval schedule and deciding rapidly if it matched the last one presented. In reading the syllables, both phonological and orthographic representations are activated (Massaro & Cohen, 1994). It was desirable to equate the verbal and visual tasks with respect to presentation modality, varying only the kind of materials used. Our aim was to ensure that the two tasks were equally difficult to perform under baseline conditions. Thus, the visual WM task was identical except that visual objects instead of syllables were presented. In both the verbal and visual tasks, the syllables and objects were presented either on the left or right side of the computer screen and the participants ignored the location in making their responses. In the spatial task, the same objects were presented as in the visual task, but the participants ignored their shapes and instead responded when the location was novel compared to the last stimulus presentation.

In Experiment 1, we tested the prediction that both concrete and abstract nouns would reduce the speed of responding to targets in the verbal task, whereas only concrete nouns should interfere with the visual task. In Experiment 2, we examined again the selective interference effect for the visual task and tested our prediction of no interference for a spatial task. Whether the interference observed in the verbal task was phonological in nature was assessed in Experiment 3 by comparing aural with visual presentation of the syllables.

Section snippets

Methods

College students (N = 60) were assigned in equal numbers to one of four groups defined by the factorial combination of materials (concrete versus abstract nouns) and task (verbal versus visual). In each condition, participants wrote definitions of 10 nouns while concurrently performing a WM task that required the detection of a visually presented target and a speeded decision regarding whether to respond. They were instructed to respond by clicking a mouse button whenever the target was different

Experiment 2

The primary purpose of Experiment 2 was to replicate the selective interference found for visual targets using a larger sample size than that used in Experiment 1. One might also argue that the triangles and circles were perhaps coded verbally as well as visually in Experiment 1, implying that part of the RT interference observed could have reflected the dependence of writing on verbal WM. This interpretation cannot account for the selective interference in writing definitions of concrete but

Experiment 3

The orthographic encoding of graphemes can be accomplished by direct retrieval of word spelling from an orthographic output lexicon without phonological mediation (Caramazza, 1991). Computing orthography does not necessarily require a conversion of syllables to graphemes. Direct orthographic encoding, therefore, could have slowed RT on the verbal task used here because the phonological segments were visually presented. That is to say, the visual presentation of ba and da allowed for the

General discussion

It was hypothesized that verbal WM supports necessary processes in written language production, whereas visual WM supports optional processes associated with the planning of image-based conceptual content. Here, writing a definition to either a concrete or an abstract noun slowed the responses made to a concurrent verbal WM task. This outcome is consistent with the notion that grammatical, phonological, or orthographic encoding required the use of verbal WM.

As expected, only concrete nouns

Acknowledgments

The authors thank Adele Handley, Julie Lee, Kevin Kelley, Bridgett Gregg, Mike Donovan, Heather Mertz, Mark Morgan, Zane Price, and Mike Cahill for their assistance in preparing materials, collecting data, and assigning ratings to the definitions. This research was supported in part by NATO Collaborative Research Grant No. LST.CLG 974939.

References (28)

  • D.Z. Hambrick et al.

    The role of working memory in higher-level cognition

  • J.R. Hayes

    A new framework for understanding cognition and affect in writing

  • J. Jonides et al.

    The architecture of working memory

  • R.T. Kellogg

    A model of working memory in writing

  • Cited by (82)

    • Comparing factual recall of tapped vs. handwritten text

      2021, Acta Psychologica
      Citation Excerpt :

      Sometimes called the generation effect, it has been suggested that information that has been produced is more likely to be remembered, due to the notion that production necessarily requires recall, planning, and encoding (e.g., DeWinstanley & Bjork, 2004; Mulligan, 2004; Rosner et al., 2013). In the case of notetaking, a student must keep information in phonological or verbal working memory (e.g., Kellogg et al., 2006), at least for the time it takes to write it down, which is believed to lead to superior long-term memory formation. In fact, advantages for note-takers were found even in cases where they were not allowed to review their notes before assessment (e.g., Faber et al., 2000).

    • Working memory, executive functions, and emotional intelligence in second language writing

      2020, Journal of Second Language Writing
      Citation Excerpt :

      In order for these elements to become available, at least some updating of the information is required. Regarding accuracy, the results are in line with empirical evidence suggesting that grammatical, phonological, and orthographic encoding rely on WM resources (Fayol et al., 1994; Hupet et al., 1998; Kellogg et al., 2007; Olive et al., 2008). Again, updating ability contributed substantially to the variability observed in linguistic accuracy, followed by shifting ability.

    • Working memory and second language writing: A systematic review

      2023, Studies in Second Language Acquisition
    View all citing articles on Scopus
    View full text