Elsevier

Brain and Language

Volume 98, Issue 2, August 2006, Pages 140-149
Brain and Language

The effect of word length on lexical decision in dyslexic and normal reading children

https://doi.org/10.1016/j.bandl.2006.04.003Get rights and content

Abstract

In the present study, the effect of word length on lexical decision in dyslexic and normal reading children was investigated. Dyslexics of 10-years old, chronological age controls, and reading age controls read words and pseudowords consisting of 3 to 6 letters in a lexical decision task. Length effects were much stronger in dyslexics and reading age controls than in chronological age controls. These results support the contention that dyslexics continue to rely on a predominantly sub-lexical reading procedure, whereas for normal readers the contribution of a lexical reading procedure increases. The relevance of the findings for current computational models of reading is discussed.

Introduction

A common finding in reading research is that, in skilled readers, length does not affect the reading speed for high frequent words, whereas longer pseudowords do take more time to recognize than short pseudowords (e.g., Juphard et al., 2004, Weekes, 1997). Recent research has shown, however, that in young readers the effect of length applies to both words and pseudowords, and that the influence of length on reading speed is even more pronounced in dyslexics (e.g., de Luca et al., 2002, Spinelli et al., 2005, van der Leij and van Daal, 1999, Ziegler et al., 2003, Zoccolotti et al., 2005).

Dyslexia refers to a deficient reading development, which often stems from problems in phonological processing (Vellutino, Fletcher, Snowling, & Scanlon, 2004). Due to phonological problems, a lot of errors are made in reading, and reading remains slow. However, in transparent orthographies, the highly regular grapheme-to-phoneme correspondences can help dyslexics to overcome phonological problems (Ziegler & Goswami, 2005). As a consequence, even dyslexic children can read accurately, by relying on a sub-lexical reading strategy in which letters are successively phonologically recoded. Therefore, in transparent orthographies, dyslexia is mainly characterized by laborious and slow reading (de Jong and van der Leij, 2003, Wimmer, 1993, Zoccolotti et al., 1999).

The influence of length on reading speed is often taken to suggest that word recognition is, at least partly, based on a sub-lexical reading strategy. The differential effect of length on words and pseudowords in skilled readers, young readers, and dyslexics further suggests that the nature of word recognition might change with the development of reading ability. In a word-naming study, Zoccolotti et al. (2005) observed that the effect of word length on vocal response times decreased dramatically from first grade to third grade in normal reading children, whereas in third grade dyslexics, the effect of word length was comparable to that of normal reading first graders. Zoccolotti et al. inferred that dyslexics ‘appear to fail in the transition from a sub-lexical to a lexical procedure’ (p. 372).

Until now, most of the studies investigating the effect of length in children have involved silent reading tasks in which eye-movements were registered (de Luca et al., 2002, Hutzler and Wimmer, 2004) or naming tasks (Spinelli et al., 2005, Ziegler et al., 2003, Zoccolotti et al., 2005). However, as words have to be read aloud in naming tasks, phonological processes may be assumed to contribute considerably to performance (e.g., Sprenger-Charolles, Siegel, Béchennec, & Serniclaes, 2003). Since phonological processing has often been shown to be problematic in dyslexics (see Vellutino et al., 2004), we aimed to investigate the effect of length in a reading task in which the contribution of phonology is less important: lexical decision. In this task, which requires a judgment on whether a string of letters constitutes a word or not, the focus is much more on orthography. As a lexical decision task specifically requires lexical knowledge, it may provide additional information about the nature of word recognition in dyslexic children. Given that lexical knowledge is crucial to perform a lexical decision task, but not (or less) for a naming task, this ‘extra’ lexical involvement might result in a processing strategy that relies to a larger extent on a lexical reading procedure.

There are at least two reading models that postulate an influence of length on word and pseudoword recognition: the Dual Route Cascaded model by Coltheart, Rastle, Perry, Langdon, and Ziegler (2001) and the connectionist network proposed by Ans, Carbonnel, and Valdois (1998). In some other connectionist models developed by Seidenberg and colleagues (e.g., Harm and Seidenberg, 1999, Plaut et al., 1996, Seidenberg and McClelland, 1989), the effect of word length is not explicitly modeled in word recognition, but is argued to be a consequence of neighbourhood effects. In the Dual Route Cascaded model (DRC, Coltheart et al., 2001), two reading routes are distinguished: the nonlexical route and the lexical route, which are activated at the same time. In the nonlexical route, the graphemes of a word are decoded into phonemes one-by-one, in a serial way. In the lexical route, all letters of a word are activated in parallel, and these letters activate a word’s entry in the orthographic lexicon. This word entry activates the corresponding word entry in the phonological lexicon, which activates the word’s phonemes (in parallel). From a developmental perspective, novel words will first be read through the nonlexical route. With the development of reading skill, an increasing number of words become represented in the orthographic lexicon, and then for these words the lexical route can be used (Jackson & Coltheart, 2001). The number of words readers can access directly in their orthographic lexicon depends on their reading skill, and consequently, so does the extent to which readers rely on the lexical and the nonlexical route.

For reading aloud, the DRC model predicts a smaller length effect on words than on pseudowords. When a word’s pronunciation is accessed via the lexical route, there should be no difference in reading speed for words of different lengths, all letters being processed in parallel. In contrast, when words are read via the nonlexical route, reading speed decreases with each additional letter. The larger the length effect, the more readers rely on sub-lexical decoding strategies.

For lexical decision, the predictions of the DRC model are different. The model makes lexical decisions on the basis of the contents of the orthographic lexicon, and is thus based on the lexical route only. The lexicon is searched in a serial way, in order of word frequency. As a consequence, a decision (‘yes’) on high frequent words will be reached faster than on low frequent ones. If the search on the lexicon is completed and no lexical match has been found, the model will produce ‘no’ for a decision. As a result, lexical decisions on words will be faster than lexical decisions on pseudowords. The model makes no explicit predictions for word length. Therefore, provided that word frequency is matched across words of different lengths, lexical decisions on long words may be expected to be as fast as on short ones.

The predictions of the DRC model with regard to lexical decision, however, bear upon normal adult readers. The model does not specify whether they also apply to children, or whether word length would affect lexical decision in children in a different way than in adults. If children base their lexical decisions solely on the contents of their orthographic lexicons, as outlined by the model, they will only make correct decisions on words they can read using a lexical reading strategy. However, words that are not yet in their orthographic lexicon would be erroneously rejected, possibly leading to high error rates. Even if these words could have been read correctly by a sub-lexical reading strategy, this would be of no avail, if lexical decision is exclusively based on a lexical reading strategy. As yet, however, it is unclear whether children exclusively rely on a lexical reading strategy in a lexical decision task, like adults do.

Another model that can account for the length effect on reading speed for words and pseudowords is provided by Ans et al. (1998). This model (for short, ACV98) is a connectionist network in which orthography is mapped to phonology by two reading procedures that work successively. First, a global procedure using knowledge about whole words is applied. In this procedure, the so-called ‘focal window’ or ‘visual attentional window’ spans all letters (or syllables) in a word at a time. Within this window, all letters are processed in parallel. If a word is not recognized by this global procedure, however, it will subsequently be processed by the analytic procedure. The analytic procedure is based on the activation of word syllabic segments or smaller segments. Processing is applied to the largest initial segment of the printed letter string that is recognized as familiar, and will proceed to the next familiar spelling pattern(s), up to the end of the string. Most often, familiar words are read by the global procedure, whereas unfamiliar words and pseudowords are read by the analytic procedure.

For words that are read aloud by the global procedure, no length effect is predicted, as all letters are processed in parallel. In contrast, in the analytic procedure, units are processed sequentially, resulting in slower reading speed for longer words. Lexical decision is exclusively performed by the global procedure in normal readers. Therefore, no effect of word length is expected in lexical decision in normal readers. For dyslexic readers, however, the model’s predictions are different (Juphard et al., 2004). Developmental dyslexia is assumed to be a consequence of either a problem in the analytic procedure, or a problem in the global procedure (Valdois et al., 2003). In regular orthographies, phonological decoding does not appear to be the major problem in dyslexia, but rather the development of direct print-to-sound connections. Thus, in these orthographies, the problem is seldom located in the analytic procedure, but is much more likely to be situated in the global procedure. Juphard et al. (2004) argued, that ‘when the size of the visual attentional window is reduced, then reading and lexical decision are based on analytic processing whatever the lexicality of the item to be read.’ (p. 333). They therefore expected a strong effect of word length in their dyslexic participant, both in reading aloud and lexical decision, which was indeed found. As in other studies (e.g., de Groot, Borgwaldt, Bos, & van den Eijnden, 2002), they did not find a length effect in a group of normal reading adults, which is in accordance with both the DRC model and the ACV98 model.

To date, most studies on lexical decision in dyslexics have involved adults (e.g., Milne et al., 2003, Milne et al., 2003) with acquired dyslexia (Arduino et al., 2003, Làdavas et al., 1997). The current study was concerned with the effects of length on the lexical decision performance of Dutch children with developmental dyslexia. The performance of these 10-year-old children was compared to the performance of a group of normal readers of the same age and to the performance of a group of younger normal readers, approximately 8 years of age, with the same level of reading ability. The lexical decision task induded words and pseudowords of 3 to 6 letters. Consistent with the results of previous studies on word naming (e.g., de Luca et al., 2002, Spinelli et al., 2005, Zoccolotti et al., 2005) and the performance of the adult dyslexic described by Juphard et al. (2004), we expected that the length effect in lexical decision would be stronger for dyslexic children than for normal reading children of the same age.

Section snippets

Participants

Sixty-six children took part in the study. These children came from 12 regular primary schools in the west and middle part of the Netherlands. Twenty-two children (13 boys, 9 girls) in Grade 4 had a reading lag of at least 15 months (mean 18.8; range 15–28). This group will be called the dyslexic group (DYS). The chronological age control group (CA) consisted of 22 normal readers (15 boys, 7 girls) in Grade 4. These children were individually matched in age, gender, vocabulary, and nonverbal

Results

The results are organized in two sections. First, we report the results on accuracy; second, we present the results concerning the response times.

Discussion

Length effects are often believed to indicate the use of a serial sub-lexical decoding strategy instead of a more parallel lexical reading procedure. The latter procedure might be less available to dyslexic readers because of a lack of orthographic knowledge (Ziegler et al., 2003, Zoccolotti et al., 2005). Indeed, in several studies length was found to affect the word- and pseudoword-naming speed of dyslexic readers more than the naming speed of normal readers. In the current study, we examined

References (41)

  • B. Ans et al.

    A connectionist multiple-trace memory model for polysyllabic word reading

    Psychological Review

    (1998)
  • N.C.M. Bakker

    TEL: een computerprogramma voor het berekenen van woord-, letter-, bigram-, trigram-, en tetragramfrequenties en versatiliteiten [TEL: a computer program to calculate word-, letter-, bigram-, trigram- and tetragram frequencies and versatility]

    (1990)
  • D.A. Balota et al.

    Visual word recognition of single-syllable words

    Journal of Experimental Psychology, General

    (2004)
  • N.D. Bleichrodt et al.

    Revisie Amsterdamse Kinder Intelligentie Test [Revised Amsterdam child intelligence test]

    (1987)
  • B.T. Brus et al.

    Een-minuut-test

    (1979)
  • M. Coltheart et al.

    DRC: a dual route cascaded model of visual word recognition and reading aloud

    Psychological Review

    (2001)
  • S.L. Crosbie et al.

    Auditory lexical decisions in children with specific language impairment

    British Journal of Developmental Psychology

    (2004)
  • P.F. de Jong et al.

    Developmental changes in the manifestation of a phonological deficit in dyslexic children learning to read a regular orthography

    Journal of Educational Psychology

    (2003)
  • M.A.R. Gijsel et al.

    Assessing reading skills by means of paper-and-pencil lexical decision: issues of reliability, repetition, and word-pseudoword ratio

    Reading and Writing: An Interdisciplinary Journal

    (2004)
  • M.W. Harm et al.

    Phonology, reading acquisition, and dyslexia: insights from connectionist models

    Psychological Review

    (1999)
  • Cited by (94)

    • Modelling the relationship of accurate and fluent word recognition in primary school

      2019, Learning and Individual Differences
      Citation Excerpt :

      We used orthographical decoding skills as a measure of accurate and fluent word recognition skills because of their importance for reading fluency and their strong relationship with text comprehension throughout primary school in German readers (Knoepke, Richter, Isberner, Naumann, & Neeb, 2014; Richter et al., 2012). Several studies suggest that poor reading skills in transparent orthographies, such as German, are associated with deficits in readers' orthographic decoding route (e.g., Richter et al., 2012; Martens & de Jong, 2006; Protopapas, Sideridis, Mouzaki, & Simos, 2007; Zoccolotti et al., 2005). In contrast, the phonological recoding is slow but reliable even in poor readers (e.g., Mayringer & Wimmer, 2000; Ziegler, Perry, Ma-Wyatt, Ladner, & Schulte-Körne, 2003).

    View all citing articles on Scopus
    View full text