Los disléxicos adultos cometen más errores de escritura

Spelling impairments in Spanish dyslexic adults

Spelling deficits have repeatedly been observed in children with dyslexia. However, the few studies addressing this issue in dyslexic adults have reported contradictory results. We investigated whether Spanish dyslexics show spelling deficits in adulthood and which components of the writing production process might be impaired in developmental dyslexia. In order to evaluate the involvement of the lexical and the sublexical routes of spelling as well as the graphemic buffer, lexical frequency, phonology-to-orthography consistency and word length were manipulated in two writing tasks: a direct copy transcoding task and a spelling-to-dictation task. Results revealed that adults with dyslexia produced longer written latencies, inter-letter intervals, writing durations and more errors than their peers without dyslexia. Moreover, the dyslexics were more affected by lexical frequency and word length than the controls, but both groups showed a similar effect of P-O consistency. Written latencies also revealed that while the dyslexics initiated the response later in the direct copy transcoding task than in the spelling-to-dictation task, the controls showed the opposite pattern. However, the dyslexics were slower than the controls in both tasks. Results were consistent with the hypothesis that spelling difficulties are present in adults with dyslexia, at least in a language with a transparent orthography such as Spanish. These difficulties seem to be associated with a deficit affecting both lexical processing and the ability to maintain information about the serial order of the letters in a word. However, the dyslexic group did not differ from the control group in the application of the P-O conversion procedures. The spelling impairment would be in addition to the reading deficit, leading to poorer performance in direct copy transcoding compared to spelling-to-dictation.

Introduction

It is well-documented that spelling deficits are present in most cases of developmental dyslexia, and it seems that these problems may persist into adulthood (Lefly and Pennington, 1991; Di Betta and Romani, 2006; Tops et al., 2012; Bogdanowicz et al., 2014). This impairment is likely to have negative consequences on academic and professional development, since dyslexics produce poor texts with multiple misspellings. Thus, writing difficulties are one of most frequent complaints among adults (Holmes and Castles, 2001; Di Betta and Romani, 2006). However, only a few studies have focused on the nature of the spelling impairment in dyslexia and its development across the life span, so many questions remain unanswered. This is in part due to the fact that research about the writing production process is relatively recent.

According to dual-process theories, spelling may be achieved through at least two different processing routes. The sublexical or assembled route makes use of knowledge about phonology-to-orthography (P-O) correspondences existing in the language, and provides phonologically plausible spellings for non-words or low-frequency words (Caramazza, 1988; Tainturier and Rapp, 2001). The so-called lexical route gives access to the spelling of whole-words from long-term memory, and thus would be used when spelling familiar words. The existence of both routes is almost undisputed, even in particularly shallow orthographies such as Spanish or Italian, in which spelling could be accomplished by resorting merely to the P-O conversion procedures (Valle, 1989; Cuetos, 1991, 1993; Barry and De Bastiani, 1997; Cuetos and Labos, 2001). Moreover, the independence of these routes is supported by the fact that they can be selectively damaged (Beauvois and Dérouesné, 1981; Shallice, 1988; Tainturier and Rapp, 2001). In surface dysgraphia, the lexical route is affected, so errors appear in words with non-predictable spellings (inconsistent and irregular words) as a result of the overreliance on the phoneme-to-grapheme conversion procedures. In the case of phonological dysgraphia, word spelling is spared, but non-word spelling is affected due to difficulties in correctly applying the P-O correspondences.

Growing evidence suggests that both routes interact during the spelling process. Specifically, some authors have proposed that lexical and sublexical information could integrate at a grapheme level (Tainturier and Rapp, 2001; Rapp et al., 2002; Bosse et al., 2003; Afonso and Álvarez, 2011). Both processes (lexical and sublexical) would “vote for” a candidate graphemic element, so when both systems produce the same output (the same grapheme is activated) the selection would be reinforced. As a result, sublexical phonological information would influence the writing of well-known words at the level of grapheme selection. Most of the evidence suggesting interaction of lexical and sublexical processes has revealed the influence of lexical knowledge on the sublexical route (Campbell, 1983; Barry and Seymour, 1988; Cuetos, 1991, 1993; Tainturier et al., 2000). However, some studies support that P-O correspondences also have an impact during word spelling (Bonin et al., 2001b; Folk and Rapp, 2004; Zhang and Damian, 2010; Afonso and Álvarez, 2011; Afonso et al., 2014).

Regardless of the route followed to access the orthographic representation, this representation has been claimed to be held in a short-term memory store, the so-called graphemic buffer, in which the abstract graphemic units are kept for subsequent production (Cuetos, 1991; Tainturier and Rapp, 2003). The orthographic sequences computed by lexical or sublexical processes would be maintained by the buffer during the time needed for the sequential assignment of format-specific information depending on the output modality (for example, letter name in oral spelling or letter shape in handwriting). Patients with graphemic buffer deficits (Caramazza et al., 1987; Buchwald and Rapp, 2004; Tainturier and Rapp, 2004) make errors affecting individual graphemes (substitutions, deletions, additions, transpositions), and they show a significant effect of word length. Moreover, they perform comparably across modality of input (copying, spelling-to-dictation, written picture naming), across modality of output (typing, handwriting, oral spelling), and across type of stimulus (regular and irregular word, non-words). This pattern of errors seems to reflect a difficulty in maintaining serially-ordered abstract letter representations.

Some authors have claimed that the spelling impairment observed in dyslexia might be associated with phonological factors. Traditionally, phonological awareness has been associated with the spelling difficulties experienced by dyslexics (Goswami and Bryant, 1990; Caravolas et al., 2001; Shaywitz and Shaywitz, 2005; Friend and Olson, 2010). In contrast, some evidence suggests that dysgraphia in developmental dyslexia may emerge as a consequence of poor orthographic lexical representations (Goulandris and Snowling, 1991; Hanley et al., 1992; Angelelli et al., 2004; Di Betta and Romani, 2006).

One reason that might explain this lack of agreement in the literature is the fact that the studies addressing this issue have been conducted with samples of very different ages. It has been observed that spelling deficits are different in children and adults with dyslexia. In fact, different writing impairments have been described for children from different school grades. In a study with Italian dyslexic children attending Grades 3 and 5 (Angelelli et al., 2010), younger dyslexic children committed all type of errors while older children primarily made phonologically plausible errors. In another study, Di Betta and Romani (2006) asked dyslexic adults to learn spoken and written pseudowords and Dutch words in association to pictures. They found that the learning curve exhibited by adults with developmental dyslexia in reading and writing tasks was impaired from the first trial and that it remained impaired. Correlational analyses revealed that this difficulty in lexical learning was independent of the phonological deficits usually observed in dyslexic children. Altogether, these results are in line with the claim that spelling impairments in dyslexia may be a function of age (Wimmer, 1996; Bishop, 1997; Snowling and Nation, 1997; Landerl and Wimmer, 2000; Angelelli et al., 2010; Tops et al., 2012).

Finally, it has also been proposed that the core deficit in developmental dyslexia might affect the short-term memory system (Ramus and Szenkovits, 2008; Menghini et al., 2011) or the ability to process and/or access temporal order information (Beneventi et al., 2009; Szmalec et al., 2011). Szmalec et al. (2011) suggested that a difficulty to generate serial-ordered long-term representations from a sequence in short-term memory may be the cause of the spelling difficulties associated with developmental dyslexia. However, some findings have contradicted these hypotheses (Di Betta and Romani, 2006; Tops et al., 2012). In their study, Tops et al. (2012) observed that dyslexics did not make more transposition errors than controls. They interpreted this result as evidence of preservation of letter order information.

In the present study we aimed to shed light on the nature of spelling deficits in adults with developmental dyslexia. Specifically, we investigated how the main components of the writing production system (P-O conversion procedures, orthographic output lexicon, and graphemic buffer) are involved during written word production in both dyslexic and non-dyslexic adults in a language with a transparent orthography such as Spanish. In order to fulfill this objective, we examined the impact of P-O consistency, word frequency and word length in two different writing tasks. These variables are thought to reflect the use of the sublexical route, the lexical route and the graphemic buffer respectively during the spelling process.

Effects of P-O consistency during spelling have been reported in the literature (Ziegler et al., 1996; Bonin et al., 2001b, 2014). Along with effects of P-O regularity (Kandel and Valdois, 2005; Delattre et al., 2006; Roux et al., 2013), P-O consistency effects have been interpreted as evidence of the application of the conversion procedures linking phonemes and graphemes. More errors would be made in inconsistent than in consistent words because in the former case successful spelling cannot be achieved by resorting only to knowledge of the P-O conversion rules. Thus, if the correct orthographic representation is not stored in the lexicon, the sublexical route would be likely to produce an incorrect form. Even when the orthographic word-form is available, P-O consistency effects are expected to be observed in written latencies. As mentioned above, both routes are thought to be involved during the normal spelling process, and the graphemic units will be reinforced when both routes produce the same output. When spelling P-O consistent words, the sublexical route would activate the same graphemes activated by the lexical route; however, when there is a P-O inconsistent segment in the word, the sublexical route would produce more than one potential spelling. Therefore, graphemes corresponding to P-O consistent segments would be strengthened in comparison to P-O inconsistent segments. Regarding word frequency, this variable has not been as thoroughly studied in handwriting research as in, for example, visual word recognition. However, some evidence suggests that orthographic representations are stored in the lexicon weighted according to their frequency, leading to shorter written latencies for high-frequency than for low-frequency words (Bonin et al., 2001a; Bonin and Fayol, 2002). Finally, marked word length effects observed in dysgraphic patients have been linked to the short-term memory system (Tainturier and Caramazza, 1996; Miceli et al., 1997; Tainturier and Rapp, 2003), but this variable has not been systematically studied in non-dysgraphic populations (although see Van der Plaats and Van Galen, 1990). We believe that the study of the effects of word frequency, P-O consistency and word length allow us to identify the component or components of the writing production system that might be deficient in adults with dyslexia.

Recruitment of the graphemic buffer seems to be required during any spelling task. However, it is widely admitted that the impact of the lexical and the sublexical routes during writing depends, among other things, on the type of task being performed (Cuetos, 1991; Bonin et al., 2001b, 2014). For example, Bonin et al. (2001b) observed that the presence of inconsistencies in the P-O mappings affected written latencies during a written picture naming task only when inconsistencies appeared at initial positions. However, during spelling-to-dictation written latencies were affected by the presence of initial, middle and final inconsistencies. The authors concluded that the semantic influence on orthographic encoding when writing words from pictures allowed final inconsistencies to be solved before production. Nevertheless, analyses of errors made by dysgraphic patients have revealed effects of P-O consistency at all positions (Rapp et al., 2002; Tainturier and Rapp, 2003) even in written picture naming (Folk et al., 2002). Thus, it is yet unclear to what extent each particular process influences spelling depending on the demands of the task.

In the present study we manipulated P-O consistency, word frequency and word length and tested the same words in a direct copy transcoding task and in a spelling-to-dictation task with a three-fold aim. First, we attempted to compare the impact of these variables on each task. P-O consistency may have an effect only in the spelling-to-dictation task, since access to the phonological representation is not required to perform the direct copy transcoding task. Second, we sought to establish whether or not the dyslexic group has impaired spelling abilities and how their deficit affects different tasks. For example, dyslexic adults may not show impaired spelling abilities per se, but only reading difficulties. Thus, differences between the dyslexic and the non-dyslexic group might reveal only differences in the copying task due to the potential involvement of reading processes, but not in spelling-to-dictation when the input is auditory. However, if the dyslexic group has an independent deficit affecting the writing production system, then differences between groups should be observed in both tasks. Finally, and if the dyslexics show spelling impairments, we aim to determine which specific processes might underlie those difficulties. Effects of P-O consistency, word frequency or word length in the dyslexic group differing from those obtained for the control group may suggest the existence of differences in the use of sublexical information, lexical information or the graphemic buffer respectively.

We collected written latencies, inter-letter interval (ILI) durations and whole-word writing durations from a group of adults with dyslexia and a control group. While written latencies are thought to reflect access to the orthographic representation (central processes), writing and ILI durations are considered to tap into more peripheral processes. However, recent evidence suggests that central processing may also affect writing and ILI durations (Roux et al., 2013). It has been recently observed that sublexical phonological information can produce an effect in ILI durations in the direct copy transcoding task without affecting written latencies (Afonso, 2014; Afonso et al., 2014). These findings may reflect the fact that sublexical phonological information can be used to reinforce the orthographic representations at the grapheme level, producing an effect only when the specific grapheme has to be actually produced. Moreover, besides written latencies (Van Galen et al., 1986), ILI durations should be strikingly affected by word length effects. Graphemic buffer deficits affect the correct retrieval of the individual graphemes, and patients with this type of deficits made fewer errors at initial than at middle positions (Tainturier and Rapp, 2003, 2004; Buchwald and Rapp, 2004). This evidence suggests that the involvement of the buffer is less relevant at the beginning of the written response. The analysis of written latencies and ILI durations in addition to the analysis of errors allows us to obtain a more detailed picture of the writing process, and it also makes it possible to detect more subtle differences between dyslexic and control participants. To our knowledge, this is the first study addressing the spelling abilities of individuals with dyslexia that provides chronometric measures of the written response.