The orthographic representation of consonant–vowel status: Evidence from two cases of acquired dysgraphia

Abstract

Department of Cognitive Science, Johns Hopkins University 3400 N. Charles St., Baltimore, MD 21218, USAIntroductionMost theories of the cognitive architecture of the spelling systeminclude a post-lexical working memory process called the ‘‘graphemicbuffer.’’ The buffer maintains orthographic representations (retrievedfrom long-term memory or assembled via the sublexical system) duringthe selection of output forms (letter names or shapes) for spelling(Caramazza & Miceli, 1990). Individuals with a selective graphemicbuffer impairment exhibit a characteristic profile, including: a wordlength effect, letter errors such as substitutions, deletions, insertions,exchanges, and shifts; and the absence of a frequency effect, semanticerrors, or phonologically plausible errors.One striking pattern in certain individuals with graphemic bufferdamage is a very high rate of preservation of consonant–vowel (CV)status in substitution errors (CLAW fi CRAW). For example, LB(Caramazza & Miceli, 1990) preserved CV status on 99% of lettersubstitution errors. This pattern is difficult to explain if orthographicrepresentations include only letter identity and position information.The high CV preservation rate (among other things) led Caramazzaand Miceli to claim that orthographic representations are multi-tiered, containing a separate representation of orthographic CV sta-tus. We refer to this as the Orthographic CV proposal (see alsoMcCloskey, Badecker, Goodman-Schulman, & Aliminosa, 1994).However, Johnsdottir, Shallice, and Wise (1996) argued that it maynot be necessary to posit that orthographic representations containCV information. They suggested that basic phonological informationabout a word can constrain the output in cases of damage, yieldingthese high CV preservation rates. They did not, however, formalizethis Phonological CV claim, creating a problem for testing it. Weattempt to address this problem and test the alternative hypotheses inthis investigation.Fig. 1a depicts the representation of CHAIN under the Ortho-graphic CV account with Column 1 depicting an intact representation,Column 2 one that is damaged, and Column 3 a repaired representa-tion. The figure illustrates that the CV status of the substituted letter isconstrained to be a vowel in the output, because the ‘‘damaged’’ letter(I) is associated with a vowel on the orthographic CV tier. Fig. 1bdepicts the corresponding representations under the Phonological CVaccount according to which each letter is associated with its corre-sponding phonologically defined CV identity. Under conditions ofdamage, the underspecified letter is represented as a vowel on thephonological CV tier, constraining the substituted letter to also be avowel.As Fig. 1a and b indicate, both accounts predict that the CV statusof substituted letters will be preserved in words whose letters matchtheir corresponding phonemes in CV status (consistent polygraphs). Toempirically distinguish between these accounts, we must considerwords with a mismatch in phoneme and grapheme CV status (e.g.,THIGH) (mixed polygraphs). As shown in Fig. 1c and d, for thesewords, the two accounts make different predictions. The orthographicaccount predicts that the G will be substituted by a consonant, whilethe phonological CV account predicts that the G will be substituted bya vowel, matching the phonological CV status of the G (part of thediphthong /aI/).In the investigation reported here, we evaluate the spelling per-formance of two brain-damaged individuals, examining the substitu-tion errors produced on mixed versus consistent polygraphs. Accordingto the orthographic CV account, CV preservation in mixed polygraphsshould be no different from that in consistent polygraphs, whereasaccording to the phonological CV account, CV preservation in mixedpolygraphs should be significantly reduced.Case studiesRSB, a right-handed man, suffered a CVA at the age of 54, fiveyears prior to the onset of this investigation. BWN, also a right-handedman, suffered a CVA at the age of 72, seven years prior to the onset ofthis investigation. Both hold PhDs, and neither had premorbid spellingdifficulties. In both cases, MRI and CT scans reveal a left parietal le-sion that produced moderate difficulties in spoken language produc-tion, and significant difficulties in spelling and number processing.Analysis 1: Locus of spelling impairmentRSB and BWN exhibited the characteristics of a graphemic bufferdeficit: (1) word length effects: RSB, 79% correct on four-letter words,14% correct on seven-letter words (v