A deficit in post-graphemic writing processes: Evidence for a graphomotor buffer

ABSTRACT

Post-graphemic writing processes transform abstract letter representations into representations of writing movements. We describe an individual with an acquired post-graphemic writing deficit. NGN is normal in spelling words aloud, but impaired in writing words to dictation, with most errors involving letter substitutions (e.g., RUMOR written as BUMOR). NGN’s deficit affects graphic motor plans, which specify the writing strokes for producing letters. Analyses of writing speed, fluency, and stroke patterns suggest that NGN’s errors result from incomplete motor-plan activation. NGN’s error rate is high for the first letter in a word, and declines across subsequent positions. On the basis of this serial position effect and other results, we propose that post-graphemic writing mechanisms include a graphomotor buffer, a writing-specific working memory that holds activated graphic motor plans bound to specific serial positions. We suggest that NGN’s graphomotor buffer is damaged such that early serial positions are affected most severely. Finally, we present results speaking to the roles and capabilities of the graphomotor buffer, and the structure of graphic motor plans.

KEYWORDS:

• Acquired dysgraphia
• peripheral dysgraphia
• writing

Acknowledgements

We are grateful to NGN and his wife for their cooperation and patience over many sessions of testing. Correspondence should be addressed to Michael McCloskey, Department of Cognitive Science, Krieger Hall, Johns Hopkins University, Baltimore, MD 21218 USA (Email: [email protected]).

Disclosure statement

No potential conflict of interest was reported by the authors.

ORCID

Teresa Schubert http://orcid.org/0000-0002-1627-3893

Notes

1 The graphemic buffer should not be confused with the graphomotor buffer mentioned in the introduction, and discussed at length in later sections. The graphemic buffer stores abstract letter representations, and plays a role in both writing and oral spelling. We argue that the graphomotor buffer (not shown in Figure 3) maintains graphic motor plans, and is involved only in writing.

2 Note that the terms allograph and allographic are used somewhat inconsistently in the literature. Whereas some authors use these terms as we do, to refer to letter shape representations with no motoric content, others use the terms in referring to what we have called graphic motor plans.

3 Several studies with neurotypical participants have found that elimination of visual feedback leads to stroke and letter duplication and omission errors in cursive writing (e.g., Cubelli & Lupi, 1999; Lebrun, 1976; Smyth & Silvers, 1987). Hence, the failure to find poorer eyes-closed than eyes-open performance for NGN might be taken to suggest an inability to use visual feedback (i.e., the deficit assumed to underlie afferent dysgraphia). However, as we have already noted, NGN’s error pattern does not conform to that expected from an inability to use feedback. NGN’s equivalent accuracy for eyes-open and –closed writing probably reflects simply that upper-case print writing (the form produced by NGN) is less dependent on visual feedback than cursive writing (used in the studies of neurotypical participants). The continuous nature of cursive writing, and the frequent occurrence of consecutive identical strokes (e.g., in n or m) create the need for careful monitoring of which strokes have and have not been produced, with monitoring failures leading to duplications and omissions. We also note that NGN’s writing was neater with eyes open than with eyes closed (e.g., spacing between letters was more even, and strokes were positioned more accurately within letters), indicating that NGN used visual feedback when writing with eyes open.

4 Although our focus in this article is on NGN’s writing in upper-case print, in Appendix 3 we present results showing that his graphic motor plan impairment also affects writing in cursive and in lower-case print.

5 The mean values were computed by first calculating, for each letter of the alphabet (e.g., K), the median writing time for the productions of the letter as a correct response (e.g., the K in KEEP written correctly), and the median for productions of the letter as an error (e.g., the K in YEAR → KEAR). This method avoids over-representing extreme values for individual responses (through the use of medians), and controls for differences in frequency of various letters as correct and incorrect responses (by averaging letter medians rather than times for individual responses). The velocity minima results reported in the next paragraph were computed with the same method.

6 We are uncertain to what extent NGN’s attempts to finish letters after reaching an impasse were under conscious control. On the one hand he may have applied knowledge of letter syntax and/or actual letter shapes as a deliberate strategy. On the other hand, these forms of knowledge may impose constraints on written characters (not only for NGN but also for neurotypical writers) via non-conscious processes. We note in this context that NGN typically seemed unaware of his struggles and errors in writing, even when queried directly.

7 The error rates plotted in the figure are slightly different from those reported in the text for initial and non-initial serial positions (e.g., for position 1: 17.1% in the figure, 19.0% in the text), due to differences in how the values were calculated. Appendix 4 explains the two calculation methods and our rationale for applying one or the other. The appendix also describes how we evaluated the statistical reliability of differences between initial and non-initial serial positions.

8 The graphomotor buffer hypothesis attributes NGN’s serial position effect to non-uniform disruption of position-specific representations. An alternative possibility is that the serial position effect arises from non-uniform disruption of position-specific processes that convert abstract letter representations to graphic motor plans. In a later section we present evidence against this alternative account.

9 To be precise, the stimuli were initialisms (non-pronounceable strings, such as FBI, created from the initial letters of a word sequence and produced in spoken form by naming each letter) rather than acronyms (pronounceable words, such as NATO, created from initial letters and produced in spoken form by pronouncing the letter sequence as a word).

10 This account differs from the entirely letter-by-letter production process we rejected in light of NGN’s serial position effect (see Figure 14), because the strictly letter-by-letter process included no advance planning at motor levels.

11 Note that these errors are entirely consistent with the conclusion that NGN’s central spelling processes are intact. Even normal spellers make occasional errors (e.g., when attempting to write a word they have never learned to spell).

Additional information

Funding

This research was supported by a grant from the Johns Hopkins Science of Learning Institute to the first author.