JCPSLP Vol 22 No 1 2020

Error distribution within the word Another typical characteristic of graphemic buffer impairment is that letter errors often do not occur at an equal rate across the different positions in the word. This observation has been called a serial position effect . Individuals with buffer impairment are generally more accurate with letters at the start of the word than at other positions. However, this is in the context of two other patterns. For some individuals, errors are concentrated in the middle of the word with the beginning and end letters relatively less error-prone (a bow-shaped error distribution; Caramazza et al., 1987). This is thought to be because letters in the middle of the word have more neighbouring letters and this leads to more confusion. Other individuals with buffer impairment make more errors the further through the word they go, so that errors are concentrated at the end of the word (a linear error distribution, e.g., Schiller et al., 2001). This distribution is likely to reflect that information that has to be held active the longest (i.e., the final letters) may be the most vulnerable to impairment (and in its most extreme form this leads to fragment errors, see above). Non-linear spelling Usually in English, words are spelled from left to right – first letter to last letter. However, some individuals with graphemic buffer impairment instead write in a “non-linear” manner – with letters spelled out of order. Schubert and Nickels (2015) described a man with dysgraphia who wrote a large proportion of words in a non-linear order. For example, the word wish was written correctly, but the letters were not written from first to last, instead he first wrote W, followed by a gap, then S and H, with I being written last (but correctly placed between W and S). Guidelines for assessment To be able to detect any type of spelling impairment, including buffer impairment, and target intervention accordingly, it is important to investigate which components of spelling are intact and which are impaired by using a variety of targeted spelling assessments to investigate which spelling tasks are showing impairment and which type of errors are being made. In Table 1 we provide an overview of key tasks for testing spelling abilities and briefly describe assessment principles below. However, readers are referred to Whitworth et al. (2014) for more detailed guidance on assessment of spelling (see Chapter 7 on written word production). Comparison of different word types An important first step in spelling assessment is the comparison of writing to dictation of irregular words and nonwords to assess the lexical and sub-lexical pathways, 7 for example, using PALPA subtests 44 (Regularity & Spelling) and PALPA 45 (Nonword spelling). For example, poorer spelling of irregular words relative to regular words suggests an impairment along the lexical route. Similarly, impaired nonword spelling indicates impairment along the sub-lexical route. Impairment to the graphemic buffer (only) should result in “letter” errors (see earlier) across all of these different word types (although see footnote 6). Cross-modality testing Second, comparison of accuracy and error types in written picture naming and writing to dictation is also important to evaluate the locus of the spelling impairment. A graphemic buffer impairment should result in impairment and “letter”

output buffer , part of orthographic working memory . Information about the identity and order of graphemes is stored in this buffer and has to be held active before being translated into the shapes or names of letters for output. Based on the nature and function of this component within the spelling process we can expect certain patterns of impairment, as discussed below. How can graphemic buffer impairment be identified? Studies of acquired spelling impairment affecting the graphemic buffer in previously competent spellers have provided a picture of the symptoms of impairment to the graphemic output buffer (e.g., Caramazza, Miceli, Villa & Romani, 1987; Rapp, 2002; and see Barisic, Kohnen, & Nickels [2017] for a case of developmental graphemic buffer impairment). Similar performance across modalities, and words and nonwords The buffer is considered the final stage of the so-called “central” spelling processes before the more “peripheral” output processes (processes after the graphemic buffer in Figure 1). This means that the buffer holds information for spelling, no matter what the task (e.g., writing to dictation, written naming), modality of output (e.g., writing, typing, oral spelling) or type of word (regular, irregular or nonword). Consequently, impairment to the buffer results in poor spelling under all of these conditions. 6 Effect of length As the buffer is a working memory component, it is sensitive to the amount of information that is held. As there is more information (i.e., more letters, and more information about letter order) in a seven-letter word compared to a three-letter word, the seven-letter word will be more error prone. In line with this, the characteristic feature of buffer impairment is that individuals show a length effect on spelling accuracy with more errors on letters in long compared to short words. Letter errors In the buffer, information about the order and identity of graphemes is held active. Therefore, any disruption to the buffer will be reflected in a loss of this information and result in letter errors. For example, when writing the word yacht , single letters might be deleted ( yact ), transposed ( yahct ), added ( yiacht ) or substituted ( yackt ). Other errors, such as semantic errors (writing boat instead of yacht ) or phonologically plausible errors ( yot for yacht ) should not occur as frequently, and if they do, most likely reflect additional impairments. Furthermore, because the buffer processes information for all writing tasks and these processes are the same regardless of the output modality, error types should be similar in all writing tasks and modalities, i.e., in writing to dictation as well as written naming, and oral spelling as well as typing. Some individuals with dysgraphia are only able to write the first few letters of a word. This deletion of multiple letters towards the ends of words results in only a fragment of the target word (e.g., diamond ➝ dia ). While there exists debate whether these “fragment errors” instead reflect impaired retrieval of the lexical representation, it is likely that these errors also result from buffer impairment (Schiller, Greenhall, Shelton, & Caramazza, 2001).

17

JCPSLP Volume 22, Number 1 2020

www.speechpathologyaustralia.org.au