Conrad Perry

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

This article describes the Dual Route Cascaded (DRC) model, a computational model of visual word recognition and reading aloud. The DRC is a computational realization of the dual-route theory of reading, and is the only computational model of reading that can perform the 2 tasks most commonly used to study reading: lexical decision and reading aloud. For both tasks, the authors show that a wide variety of variables that influence human latencies influence the DRC models latencies in exactly the same way. The DRC model simulates a number of such effects that other computational models of reading do not, but there appear to be no effects that any other current computational model of reading can simulate but that the DRC model cannot. The authors conclude that the DRC model is the most successful of the existing computational models of reading.

Nested Incremental Modeling in the Development of Computational Theories: The CDP+ Model of Reading Aloud.

At least 3 different types of computational model have been shown to account for various facets of both normal and impaired single word reading: (a) the connectionist triangle model, (b) the dual-route cascaded model, and (c) the connectionist dual process model. Major strengths and weaknesses of these models are identified. In the spirit of nested incremental modeling, a new connectionist dual process model (the CDP+ model) is presented. This model builds on the strengths of 2 of the previous models while eliminating their weaknesses. Contrary to the dual-route cascaded model, CDP+ is able to learn and produce graded consistency effects. Contrary to the triangle and the connectionist dual process models, CDP+ accounts for serial effects and has more accurate nonword reading performance. CDP+ also beats all previous models by an order of magnitude when predicting individual item-level variance on large databases. Thus, the authors show that building on existing theories by combining the best features of previous models--a nested modeling strategy that is commonly used in other areas of science but often neglected in psychology--results in better and more powerful computational models.

Developmental dyslexia in different languages: Language-specific or universal?

Most of the research on developmental dyslexia comes from English-speaking countries. However, there is accumulating evidence that learning to read English is harder than learning to read other European orthographies (Seymour, Aro, & Erskine, 2003). These findings therefore suggest the need to determine whether the main English findings concerning dyslexia can be generalized to other European orthographies, all of which have less irregular spelling-to-sound correspondences than English. To do this, we conducted a study with German- and English-speaking children (n=149) in which we investigated a number of theoretically important marker effects of the reading process. The results clearly show that the similarities between dyslexic readers using different orthographies are far bigger than their differences. That is, dyslexics in both countries exhibit a reading speed deficit, a nonword reading deficit that is greater than their word reading deficit, and an extremely slow and serial phonological decoding mechanism. These problems were of similar size across orthographies and persisted even with respect to younger readers that were at the same reading level. Both groups showed that they could process larger orthographic units. However, the use of this information to supplement grapheme-phoneme decoding was not fully efficient for the English dyslexics.

Beyond single syllables: large-scale modeling of reading aloud with the Connectionist Dual Process (CDP++) model.

Most words in English have more than one syllable, yet the most influential computational models of reading aloud are restricted to processing monosyllabic words. Here, we present CDP++, a new version of the Connectionist Dual Process model (Perry, Ziegler, & Zorzi, 2007). CDP++ is able to simulate the reading aloud of mono- and disyllabic words and nonwords, and learns to assign stress in exactly the same way as it learns to associate graphemes with phonemes. CDP++ is able to simulate the monosyllabic benchmark effects its predecessor could, and therefore shows full backwards compatibility. CDP++ also accounts for a number of novel effects specific to disyllabic words, including the effects of stress regularity and syllable number. In terms of database performance, CDP++ accounts for over 49% of the reaction time variance on items selected from the English Lexicon Project, a very large database of several thousand of words. With its lexicon of over 32,000 words, CDP++ is therefore a notable example of the successful scaling-up of a connectionist model to a size that more realistically approximates the human lexical system.

Identical Words are Read Differently in Different Languages

It is hypothesized that written languages differ in the preferred grain size of units that emerge during reading acquisition. Smaller units (graphemes, phonemes) are thought to play a dominant role in relatively consistent orthographies (e.g., German), whereas larger units (bodies, rhymes) are thought to be more important in relatively inconsistent orthographies (e.g., English). This hypothesis was tested by having native English and German speakers read identical words and nonwords in their respective languages (zoo-Zoo, sand-Sand, etc.). Although the English participants exhibited stronger body-rhyme effects, the German participants exhibited a stronger length effect for words and nonwords. Thus, identical items were processed differently in different orthographies. These results suggest that orthographic consistency determines not only the relative contribution of orthographic versus phonological codes within a given orthography, but also the preferred grain size of units that are likely to be functional during reading.

No more problems in Coltheart's neighborhood: resolving neighborhood conflicts in the lexical decision task.

In the area of visual word recognition, there is considerable disagreement as to whether neighborhood effects for words in the lexical decision task are facilitatory or inhibitory: While they seem to be mostly facilitatory in English, they tend to be absent or inhibitory in French or Spanish. The present study investigated the possibility that the facilitatory neighborhood effect obtained in English is due to the fact that most neighbors in English are body neighbors (i.e. they share the same orthographic rime). Our results showed that when words were matched for orthographic neighborhood (N), the effects of body neighbors (BN) were facilitatory (i.e. shorter reaction times for words with many body neighbors than for words with few body neighbors). In contrast, when words are matched for BN, the effects of N are unreliable with a tendency towards inhibition. In conclusion, it appears that research conducted in English has always found neighborhood effects to be facilitatory because of the dominant role of body neighbors in English. In contrast, neighborhood effects in French and Spanish may have been more ambiguous because these languages either do not confound N and BN, or they do not require a greater sensitivity to the body/rime unit.

The DRC model of visual word recognition and reading aloud : an extension to German

The dual-route cascaded (DRC) model of word recognition and reading aloud was implemented in German. In this paper, we describe crosslinguistic differences and similarities between the German and the English DRC. The German DRC was evaluated with respect to its ability to correctly pronounce all German monosyllabic words and to simulate the loan word (regularity) effect. Furthermore, we obtained DRC predictions concerning a number of benchmark effects previously investigated in English, namely effects of word frequency, word length, and neighbourhood size.

Do current connectionist learning models account for reading development in different languages

Learning to read a relatively irregular orthography, such as English, is harder and takes longer than learning to read a relatively regular orthography, such as German. At the end of grade 1, the difference in reading performance on a simple set of words and nonwords is quite dramatic. Whereas children using regular orthographies are already close to ceiling, English children read only about 40% of the words and nonwords correctly. It takes almost 4 years for English children to come close to the reading level of their German peers. In the present study, we investigated to what extent recent connectionist learning models are capable of simulating this cross-language learning rate effect as measured by nonword decoding accuracy. We implemented German and English versions of two major connectionist reading models, Plaut et al.s (Plaut, D. C., McClelland, J. L., Seidenberg, M. S., & Patterson, K. (1996). Understanding normal and impaired word reading: computational principles in quasi-regular domains. Psychological Review, 103, 56-115) parallel distributed model and Zorzi et al.s (Zorzi, M., Houghton, G., & Butterworth, B. (1998a). Two routes or one in reading aloud? A connectionist dual-process model. Journal of Experimental Psychology: Human Perception and Performance, 24, 1131-1161); two-layer associative network. While both models predicted an overall advantage for the more regular orthography (i.e. German over English), they failed to predict that the difference between children learning to read regular versus irregular orthographies is larger earlier on. Further investigations showed that the two-layer network could be brought to simulate the cross-language learning rate effect when cross-language differences in teaching methods (phonics versus whole-word approach) were taken into account. The present work thus shows that in order to adequately capture the pattern of reading acquisition displayed by children, current connectionist models must not only be sensitive to the statistical structure of spelling-to-sound relations but also to the way reading is taught in different countries.

Modelling reading development through phonological decoding and self-teaching: implications for dyslexia

The most influential theory of learning to read is based on the idea that children rely on phonological decoding skills to learn novel words. According to the self-teaching hypothesis, each successful decoding encounter with an unfamiliar word provides an opportunity to acquire word-specific orthographic information that is the foundation of skilled word recognition. Therefore, phonological decoding acts as a self-teaching mechanism or ‘built-in teacher’. However, all previous connectionist models have learned the task of reading aloud through exposure to a very large corpus of spelling–sound pairs, where an ‘external’ teacher supplies the pronunciation of all words that should be learnt. Such a supervised training regimen is highly implausible. Here, we implement and test the developmentally plausible phonological decoding self-teaching hypothesis in the context of the connectionist dual process model. In a series of simulations, we provide a proof of concept that this mechanism works. The model was able to acquire word-specific orthographic representations for more than 25 000 words even though it started with only a small number of grapheme–phoneme correspondences. We then show how visual and phoneme deficits that are present at the outset of reading development can cause dyslexia in the course of reading development.

A POSITION-SENSITIVE STROOP EFFECT : FURTHER EVIDENCE FOR A LEFT-TO-RIGHT COMPONENT IN PRINT-TO-SPEECH CONVERSION

In the classical Stroop effect, response times for naming the color in which a word is printed are affected by the presence of semantic, phonological, or orthographic relationships between the stimulus word and the response word. We show that color naming responses are faster when the printed word shares a phoneme with the color name to be produced than when it does not, in conditions where there is no semantic relationship between the printed word and the color name. This result is compatible with a variety of computational models of reading. However, we also found that these effects are much larger when it is the first phoneme that the stimulus and response share than when it is the last. Our data are incompatible with computational models of reading in which the computation of phonology from print is purely parallel. The dual route cascaded model computational model of reading, which has a lexical route that operates in parallel and a nonlexical route that operates serially letter by letter, successfully simulates this position-sensitive Stroop effect. The model also successfully simulates the “onset effect” in masked priming (Forster & Davis, 1991) and the interaction between the regularity effect and the position in a word of a grapheme-phoneme irregularity (Rastle & Coltheart, 1999b)-effects which, we argue, arise for the same reason as the position-sensitive Stroop effect we report.