Serial processing in reading aloud: No challenge for a parallel model.

K. Rastle and M. Coltheart (see record 1999-10532-012) challenged parallel models of reading by showing that the cost of irregularity in low-frequency exception words was modulated by the position of the irregularity in the word. This position-of-irregularity effect was taken as strong evidence of serial processing in reading. This article refutes Rastle and Coltheart's theoretical conclusions in 3 ways: First, a parallel model, the connectionist dual process model (M. Zorzi, G. Houghton, & B. Butterworth, 1998b), produces a position-of-irregularity effect. Second, the supposed serial effect can be reduced to a position-specific grapheme-phoneme consistency effect. Third, the position-of-irregularity effect vanishes when the experimental data are reanalyzed using grapheme-phoneme consistency as the covariate. This demonstration has broader implications for studies aiming at adjudicating between models: Strong inferences should be avoided until the computational models are actually tested. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Serial processing in reading aloud: Reply to Zorzi (2000).

K. Rastle and M. Coltheart (1999; see also M. Coltheart & K. Rastle, 1994) reported data demonstrating that the cost of irregularity in reading aloud low-frequency exception words is modulated by the position of the irregularity in the word. They argued that these data implicated a serial process and falsified all models of reading aloud that operate solely in parallel, a conclusion that M. Zorzi (2000) challenged by successfully simulating the position of irregularity effect with such a model. Zorzi (2000) further claimed that a reanalysis of K. Rastle and A Coltheart's (1999) data demonstrates sensitivity to grapheme-phoneme consistency (which he claimed was confounded across the position of irregularity manipulation) rather than the use of a serial process. Here, the authors argue that M. Zorzi's (2000) reanalyses were inappropriate and reassert that K. Rastle and A Coltheart's (1999) findings are evidence for serial processing. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Serial and strategic effects in reading aloud.

M. Coltheart and K. Rastle (1994) reported that the size of the regularity effect on word-naming latency decreases across position of irregularity, implicating a serial process in reading aloud. In response to criticism by D. C. Plaut, J. L. McClelland, M. S. Seidenberg, and K. Patterson (1996), these results were replicated with monosyllabic words that had been controlled for consistency. In a second experiment, participants named nonword- or regular-word targets mixed with either first-position irregular fillers or third-position irregular fillers. Target naming was slowed when first-position irregular fillers were present, compared with target naming when third-position irregular fillers were present. These data suggest that participants can slow use of the nonlexical route if faced with very costly exception words. Simulations using the dual-route cascaded model (M. Coltheart, B. Curtis, P. Atkins, & M. Haller, 1993) are presented. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Contextual effects on reading aloud: Evidence for pathway control.

Recent evidence suggests that the processes responsible for generating a phonological code from print are flexible in skilled readers. An important goal, therefore, is to identify the conditions that lead to changes in how a phonological code is computed. Five experiments are reported that examine whether phonological processes change as predicted by the pathway control hypothesis when reading aloud words and nonwords. Changes in reading processes were assessed by measuring the effect of predictable switches between stimulus categories across trials. The results of the present experiments are argued to be consistent with the pathway control hypothesis. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Semantic involvement in reading aloud: Evidence from a nonword training study.

The current research uses a novel methodology to examine the role of semantics in reading aloud. Participants were trained to read aloud 2 sets of novel words (i.e., nonwords such as bink): some with meanings (semantic) and some without (nonsemantic). A comparison of reading aloud performance between these 2 sets of novel words was used to provide an indicator of the importance of semantic information in reading aloud. In Experiment 1, in contrast to expectations, reading aloud performance was not better for novel words in the semantic condition. In Experiment 2, the training of novel words was modified to reflect more realistic steps of lexical acquisition: Reading aloud performance became faster and more accurate for novel words in the semantic condition, but only for novel words with inconsistent pronunciations. This semantic advantage for inconsistent novel words was again observed when a subset of participants from Experiment 2 was retested 6-12 months later (in Experiment 3). These findings provide support for a limited but significant role for semantics in the reading aloud process. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Neighborhood effects in reading aloud: New findings and new challenges for computational models.

A word from a dense neighborhood is often read aloud faster than a word from a sparse neighborhood. This advantage is usually attributed to orthography, but orthographic and phonological neighbors are typically confounded. Two experiments investigated the effect of neighborhood density on reading aloud when phonological density was varied while orthographic density was held constant, and vice versa. A phonological neighborhood effect was observed, but not an orthographic one. These results are inconsistent with the predominant role ascribed to orthographic neighbors in accounts of visual word recognition and reading aloud. Consistent with this interpretation, 6 different computational models of reading aloud failed to simulate this pattern of results. The results of the present experiments thus provide a new understanding of some of the processes underlying reading aloud, and new challenges for computational models. (PsycINFO Database Record (c) 2011 APA, all rights reserved)     

Models of reading aloud: Dual-route and parallel-distributed-processing approaches.

It has often been argued that various facts about skilled reading aloud cannot be explained by any model unless that model possesses a dual-route architecture (lexical and nonlexical routes from print to speech). This broad claim has been challenged by M. S. Seidenberg and J. L. McClelland (1989, 1990). Their model has but a single route from print to speech, yet, they contend, it can account for major facts about reading that have hitherto been claimed to require a dual-route architecture. The authors identify 6 of these major facts about reading. The 1-route model proposed by Seidenberg and McClelland can account for the 1st of these but not the remaining 5. Because models with dual-route architectures can explain all 6 of these basic facts about reading, the authors suggest that this remains the viable architecture for any tenable model of skilled reading and learning to read. The dual-route cascaded model, a computational version of the dual-route model, is described. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

A quantitative multiple-levels model of reading aloud.

A new computational model of reading aloud is described that gives a detailed quantitative account of response latency and accuracy in word naming studies. The model implements a simple multiple-levels system (T. Shallice, E. K. Warrington, & R. McCarthy, 1983) in which different levels of spelling-to-sound correspondence (e.g., phoneme, rime, word) are combined in an interactive-activation network to produce a final pronunciation of the input. Naming latency is modeled as a function of the number of network cycles required to exceed the response criteria. The model's parameters are set by fitting the model to 2 sets of data. The model then makes accurate predictions of response time and error rates in studies of word frequency, regularity, and consistency. Depending on one's perspective, the model can equally well be viewed as a dual-route theory, a single-route theory, or an analogy theory. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Lexical and nonlexical phonological priming in reading aloud.

Five homophone priming experiments were reported in which the lexicality of primes and targets were varied, so that primes and targets were either nonword homophones (keff-keph), word homophones (brake-break), pseudohomophones (brayk-braik), or of mixed lexicality (brake-brayk and brayk-break). Results showed that naming of targets was facilitated by a phonologically identical prime only when a word was in the prime-target pairing. Simulations of these data using the dual-route cascaded model of reading (e.g., M. Coltheart, B. Curtis, P. Atkins, & M. Haller, 1993) were also reported. These results are evidence against the view that there is a critical early stage in the process of visual word recognition in which words are represented in purely phonological form, and they are evidence for the view that knowledge of orthography and phonology is represented locally in the reading system. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Vowel processing during silent reading: Evidence from eye movements.

Two eye movement experiments examined whether skilled readers include vowels in the early phonological representations used in word recognition during silent reading. Target words were presented in sentences preceded by parafoveal previews in which the vowel phoneme was concordant or discordant with the vowel phoneme in the target word. In Experiment 1, the orthographic vowel differed from the target in both the concordant and discordant preview conditions. In Experiment 2, the vowel letters in the preview were identical to those in the target word. The phonological vowel was ambiguous, however, and the final consonants of the previews biased the vowel phoneme either toward or away from the target's vowel phoneme. In both experiments, shorter reading times were observed for targets preceded by concordant previews than by discordant previews. Implications for models of word recognition are discussed. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Reading aloud: Evidence for contextual control over lexical activation.

Can readers exert control (albeit unconsciously) over activation at particular loci in the reading system? The authors addressed this issue in 4 experiments in which participants read target words aloud and the factors of prime–target relation (semantic, repetition), context (related, unrelated), stimulus quality (bright, dim), and relatedness proportion (RP; high, low) were manipulated. In the high RP condition (RP = .5), an interaction between semantic context and stimulus quality was observed in which low stimulus quality slowed unrelated targets more than related ones, replicating previous work. In contrast, the low RP condition (RP = .25) yielded additive effects of semantic context and stimulus quality. However, when low RP was examined within the context of repetition priming, context and stimulus quality once again interacted. These results are discussed in the context of a widely endorsed framework with the addition of the central assumption that there is control over feedback between various levels. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Cross-language computational investigation of the length effect in reading aloud.

The authors examined whether 2 computational models of reading, the dual-route cascaded model (M. Coltheart, K. Rastle, C. Perry, R. Langdon, and J. C. Ziegler, 2001) and the connectionist 2-layer model (M. Zorzi, G. Houghton, and B. Butterworth, 1998), were able to predict the pattern that the length effect found in reading aloud is larger in German than in English (J. C. Ziegler, C. Perry, A. M. Jacobs, and M. Braun, 2001). The results showed that the dual-route cascaded model, which uses a serial mechanism for assembling phonology, successfully predicted this cross-language difference. In contrast, the connectionist model of Zorzi et al. (1998) predicted the opposite: a larger length effect in English than in German. Both the success of one model and the failure of the other highlight fundamental differences between 2 major classes of computational models. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Code-specific processes in word naming: Evidence supporting a dual-route model of word recognition.

This research examined visual and phonological coding in visual word recognition, Participants named words while performing 1 of 3 memory tasks. The difficulty of the memory tasks was varied. In Experiment 1, increasing the difficulty of a digit-memory load resulted in slower naming of low-frequency regular (e.g., wink) words but faster naming of low-frequency exception (e.g., pint) words. In Experiment 2, increasing the difficulty of a dot-memory load slowed naming of low-frequency exception words more than naming of low-frequency regular words. In Experiment 3, increasing the difficulty of a tone-memory load resulted in slower naming of both low-frequency regular and exception words. The results are consistent with dual-route assumptions concerning code-specific processes in word recognition. (PsycINFO Database Record (c) 2011 APA, all rights reserved)     

Reading aloud: Evidence for the use of a whole word nonsemantic pathway.

[Correction Notice: An erratum for this article was reported in Vol 48(1) of Canadian Journal of Experimental Psychology (see record 2007-10228-001). In Table 1, an author's error caused the digits in the Related-Unfamiliar condition to be transposed. The correct table is presented in this erratum.] Tested the assumptions that (1) only whole word orthographic knowledge can produce priming and (2) it is automatic. Two experiments with 20 adult Japanese readers were conducted in the context of reading Japanese Kana. Results show that, taken alone, neither the presence nor the absence of priming effects in oral reading permits an inference as to whether the addressed or assembled routine is used. Converging operations that do permit such an inference are reported. The data support the view that (1) components of the word recognition system operate interactively such that use of the assembled routine yields priming under certain conditions and (2) normal readers of a shallow orthography use a nonsemantic, whole-word pathway to name words. (French abstract) (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Localizing the lexicon for reading aloud:replication of a PET study using fMRI

Functional magnetic resonance imaging (fMRI) was used to investigate the neural basis of written word recognition in two normal subjects. With a 1.5T scanner and temporal surface coil, T2 gradient echo images were obtained while subjects read words aloud. As a control condition, subjects visualized false font strings and said the word "range' each time such a string appeared. These two conditions were presented in an oscillatory pattern, alternating 30 s of each condition for a total of 4 min. Comparison of the two conditions using cross-correlation demonstrated strong activation in both subjects in the left posterior superior temporal gyrus, near the site predicted for the visual input lexicon by Déjerine and recently demonstrated by positron emission tomography.     

Two routes or one in reading aloud? A connectionist dual-process model.

A connectionist study of word reading is described that emphasizes the computational demands of the spelling–sound mapping in determining the properties of the reading system. It is shown that the phonological assembly process can be implemented by a two-layer network, which easily extracts the regularities in the spelling–sound mapping for English from training data containing many exception words. It is argued that productive knowledge about spelling–sound relationships is more easily acquired and used if it is separated from case-specific knowledge of the pronunciation of known words. It is then shown how the interaction of assembled and retrieved phonologies can account for the combined effects of frequency and regularity–consistency and for the reading performance of dyslexic patients. It is concluded that the organization of the reading system reflects the demands of the task and that the pronunciations of nonwords and exception words are computed by different processes. (PsycINFO Database Record (c) 2011 APA, all rights reserved)     

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 model's 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. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Neighbourhood density effects in reading aloud: New insights from simulations with the DRC model.

A series of nine simulations with the Dual Route Cascaded (DRC) model (M. Coltheart, K. Rastle, C. Perry, R. Langdon, & J. Ziegler, 2001) investigated neighbourhood density (N) effects in nonword and word naming. Two main findings emerged from this work. First, when naming nonwords there are two loci for the effect of N in the model, contrary to M. Coltheart et al.'s single locus explanation of what the model is doing. The early N effect involves interactive activation between the orthographic lexicon and the letter units such that high N facilitates letter identification, which in turn affects the nonlexical route. The late N effect arises from activation in the orthographic lexicon that feeds forward to the phonological lexicon and primes phonemes in the phoneme system. Second, when naming words the presence/absence of an effect of N on the Letter Units through feedback from the lexical level depends on the parameter settings. Implications and suggestions for future directions are made. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Submorphemic processing in reading Chinese.

Three experiments are reported here to address the question of whether submorphemic information is involved in the recognition of Chinese characters. A character decision task was used in which frequency characteristics of the radicals within a character and radical status of the components within a noncharacter were manipulated. The frequency of the right-hand radical affected responses to 2-radical characters, whereas the radical status of both left and right components affected noncharacter responses. Furthermore, the impact of radical frequency was shown to be sensitive to radical position. For 3-radical characters, it appeared that the frequency of a compound radical (composed of 2 subradicals) had no effect on responses, whereas the frequency of the subradicals did. It is concluded that all simple radicals are independently activated in the process of character recognition. Compound radicals are not activated in this way despite their common occurrence. The results are considered within a multilevel interactive-activation framework incorporating position sensitivity. (PsycINFO Database Record (c) 2010 APA, all rights reserved)