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.     

Unpredictability and complexity of print-to-speech correspondences increase reliance on lexical processes: more evidence for the orthographic depth hypothesis

The Orthographic Depth Hypothesis [Katz, L., & Frost, R. (1992). The reading process is different for different orthographies: The orthographic depth hypothesis. In R. Frost & L. Katz (Eds.), Orthography, phonology, morphology, and meaning (pp. 67–84). Amsterdam: Elsevier Science] proposes cross-linguistic differences in the involvement of lexical processing during reading. In orthographies with complex, inconsistent, and/or incomplete sublexical correspondences, decoding is more difficult and therefore slower. This gives more time to the lexical route to retrieve information, and leads to a greater ratio of lexical processing. We test whether this mechanism applies both for words with inconsistent (in English) and for words with complex (in French) correspondences. As complex correspondences are sufficient to derive a correct pronunciation, an increase in lexical processing may not occur. In a reading-aloud task, we used the frequency effect to measure lexical processing. The data showed stronger involvement of lexical processing for inconsistent compared to consistent words, and for complex compared to simple words. The results confirm that Katz and Frost’s proposed mechanism applies to different sources of orthographic depth.     

Reading and lexicalization in opaque and transparent orthographies: Word naming and word learning in English and Spanish

Do skilled readers of opaque and transparent orthographies make differential use of lexical and sublexical processes when converting words from print to sound? Two experiments are reported, which address that question, using effects of letter length on naming latencies as an index of the involvement of sublexical letter–sound conversion. Adult native speakers of English (Experiment 1) and Spanish (Experiment 2) read aloud four- and seven-letter high-frequency words, low-frequency words, and nonwords in their native language. The stimuli were interleaved and presented 10 times in a first testing session and 10 more times in a second session 28 days later. Effects of lexicality were observed in both languages, indicating the deployment of lexical representations in word naming. Naming latencies to both words and nonwords reduced across repetitions on Day 1, with those savings being retained to Day 28. Length effects were, however, greater for Spanish than English word naming. Reaction times to long and short nonwords converged with repeated presentations in both languages, but less in Spanish than in English. The results support the hypothesis that reading in opaque orthographies favours the rapid creation and use of lexical representations, while reading in transparent orthographies makes more use of a combination of lexical and sublexical processing.     

Children's and adults' reading of nonwords: effects of regularity and consistency.

The procedures used by novice readers to assemble pronunciations for nonwords were investigated. Children in Grades 1-3 read aloud consonant-vowel-consonant and longer monosyllabic nonwords. By the end of Grade 1, children displayed a good grasp of grapheme-phoneme (G-P) correspondences (e.g., ai, ow). Grade 2 and 3 readers increasingly used larger orthographic correspondences termed rimes (e.g., -ook, -ild). However, G-P correspondences determined most responses. Adults likewise used G-P rules when reading aloud nonwords and were more accurate at applying the rules. The strong reliance of Grade 1 and 2 readers on G-P rules was also demonstrated by their superior oral reading of regular words along with a tendency to regularize exception words (e.g., reading bull to rhyme with dull).     

Pathway control in visual word processing: Converging evidence from recognition memory

The extent to which readers can exert strategic control over oral reading processes is a matter of debate. According to the pathway control hypothesis, the relative contributions of the lexical and nonlexical pathways can be modulated by the characteristics of the context stimuli being read, but an alternative time criterion model is also a viable explanation of past results. In Experiment 1, subjects named high- and low-frequency regular words in the context of either low-frequency exception words (e.g., pint) or nonwords (e.g., flirp). Frequency effects (faster pronunciation latencies for high-frequency words) were attenuated in the nonword context, consistent with the notion that nonwords emphasize the characteristics of the frequency-insensitive nonlexical pathway. Importantly, we also assessed memory for targets, and a similar attenuation of the frequency effect in recognition memory was observed in the nonword condition. Converging evidence was obtained in a second experiment in which a variable that was more sensitive to the nonlexical pathway (orthographic neighborhood size) was manipulated. The results indicated that both speeded pronunciation performance and memory performance were relatively attenuated in the low-frequency exception word context in comparision with the nonword context. The opposing influences of list context type for word frequency and orthographic neighborhood size effects in speeded pronunciation and memory performance provide strong support for the pathway control model, as opposed to the time criterion model.     

Strategies for visual word recognition and orthographical depth: a multilingual comparison

We investigated the psychological reality of the concept of orthographical depth and its influence on visual word recognition by examining naming performance in Hebrew, English, and Serbo-Croatian. We ran three sets of experiments in which we used native speakers and identical experimental methods in each language. Experiment 1 revealed that the lexical status of the stimulus (high-frequency words, low-frequency words, and nonwords) significantly affected naming in Hebrew (the deepest of the three orthographies). This effect was only moderate in English and nonsignificant in Serbo-Croatian (the shallowest of the three orthographies). Moreover, only in Hebrew did lexical status have similar effects on naming and lexical decision performance. Experiment 2 revealed that semantic priming effects in naming were larger in Hebrew than in English and completely absent in Serbo-Croatian. Experiment 3 revealed that a large proportion of nonlexical tokens (nonwords) in the stimulus list affects naming words in Hebrew and in English, but not in Serbo-Croatian. These results were interpreted as strong support for the orthographical depth hypothesis and suggest, in general, that in shallow orthographies phonology is generated directly from print, whereas in deep orthographies phonology is derived from the internal lexicon.     

Previewing the neighborhood: the role of orthographic neighbors as parafoveal previews in reading

In 2 experiments, a boundary technique was used with parafoveal previews that were identical to a target (e.g., sleet), a word orthographic neighbor (sweet), or an orthographically matched nonword (speet). In Experiment 1, low-frequency words in orthographic pairs were targets, and high-frequency words were previews. In Experiment 2, the roles were reversed. In Experiment 1, neighbor words provided as much preview benefit as identical words and greater benefit than nonwords, whereas in Experiment 2, neighbor words provided no greater preview benefit than nonwords. These results indicate that the frequency of a preview influences the extraction of letter information without setting up appreciable competition between previews and targets. This is consistent with a model of word recognition in which early stages largely depend on excitation of letter information, and competition between lexical candidates becomes important only in later stages.     

Orthographic learning at a glance: on the time course and developmental onset of self-teaching

Experiment 1 examined the time course of orthographic learning among Grade 3 children. A single encounter with a novel orthographic string was sufficient to produce reliable recall of orthographic detail. Moreover, newly acquired orthographic information was retained 1 month later. These data support the logistic learning functions featured in contemporary connectionist models of reading rather than a "threshold" model of orthographic learning. Experiments 2 and 3 examined self-teaching among novice readers. In contrast to the findings from less regular orthographies such as English and Dutch, beginning readers of a highly regular orthography (Hebrew) appear to be relatively insensitive to word-specific orthographic detail, reading in a nonlexical "surface" fashion. These results suggest fundamental differences between shallow and deep orthographies in the development of orthographic sensitivity.     

Orthographic structure versus morphological structure: principles of lexical organization in a given language

Most models of visual word recognition in alphabetic orthographies assume that words are lexically organized according to orthographic similarity. Support for this is provided by form-priming experiments that demonstrate robust facilitation when primes and targets share similar sequences of letters. The authors examined form-orthographic priming effects in Hebrew, Arabic, and English. Hebrew and Arabic have an alphabetic writing system but a Semitic morphological structure. Hebrew morphemic units are composed of noncontiguous phonemic (and letter) sequences in a given word. Results demonstrate that form-priming effects in Hebrew or Arabic are unreliable, whereas morphological priming effects with minimal letter overlap are robust. Hebrew bilingual subjects, by contrast, showed robust form-priming effects with English material, suggesting that Semitic words are lexically organized by morphological rather than orthographic principles. The authors conclude that morphology can constrain lexical organization even in alphabetic orthographies and that visual processing of words is first determined by morphological characteristics.     

The effects of spelling consistency on phonological awareness: a comparison of English and German

Within alphabetic languages, spelling-to-sound consistency can differ dramatically. For example, English and German are very similar in their phonological and orthographic structure but not in their consistency. In English the letter a is pronounced differently in the words bank, ball, and park, whereas in German the letter a always has the same pronunciation (e.g., Ball, Park, Bank). It is often argued that reading acquisition has a reciprocal effect on phonological awareness. As reading is acquired, therefore, spoken language representation may be affected differently for English and German children. Prior to literacy acquisition, however, phonological representation in English and German children should be similar due to the similar phonological structure of the two languages. We explored this hypothesis by comparing phonological awareness at the rime and phoneme levels in prereaders and beginning readers in English and German. Similar developmental effects were indeed observed in prereaders, but differential effects had emerged within the first year of reading instruction.     

Phonological recoding and rapid orthographic learning in third-graders' silent reading: a critical test of the self-teaching hypothesis

This study examined rapid orthographic learning following silent reading in third-grade children as a function of number of target nonword repetitions and test delay. In each of two test sessions at least 6 days apart, children read a series of short stories, with each story containing a different nonword repeated either four or eight times. In the second session, after the stories had been read, children were asked to read short lists of target nonwords or homophonic alternatives. Children read target nonwords faster than homophones, indicating that they had formed functional orthographic representations of the target nonwords through phonologically recoding them during silent story reading. They also preferred target nonwords to homophones in an orthographic choice task in which the alternatives included the target, the homophone, and a visually similar foil, although here orthographic learning was stronger for items encountered eight times within stories and stronger for items tested immediately. These findings provide critical evidence in support of Share's self-teaching through phonological recoding hypothesis.     

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.     

Differential Effects of Number of Letters on Word and Nonword Naming Latency

The issue addressed in this study is whether there are differential effects of number of letters on word and nonword naming latency. Experiment 1 examined the effect of number of letters on latency for naming high-frequency words, low-frequency words, and nonwords. Number of letters affected latency for low-frequency words and nonwords but did not affect latency for high-frequency words. Number of letters was also negatively correlated with number of orthographic neighbours, number of friends, and average grapheme frequency. Number of letters continued to affect nonword naming latency, but not low-frequency word naming latency, after the effects of orthographic neighbourhood size, number of friends, and average grapheme frequency had been accounted for. Experiment 2 found that number of letters had no effect on the latency of delayed naming of the same words and nonwords. It is concluded that the effect of number of letters on nonword naming reflects a sequential, non-lexical reading mechanism.     

Reading strategies in orthographies of intermediate depth are flexible: Modulation of length effects in Portuguese

This paper examines the role of grapheme–phoneme conversion for skilled reading in an orthography of intermediate depth, Portuguese. The effects of word length in number of letters were determined in two studies. Mixed lists of five- and six-letter words and nonwords were presented to young adults in lexical decision and reading aloud tasks in the first study; in the second one, the length range was increased from four to six letters and an extra condition was added where words and nonwords were presented in separate, or blocked, lists. Reaction times were larger for longer words and nonwords in lexical decision, and in reading aloud mixed lists, but no effect of length was observed when reading words in blocked lists. The effect of word length is thus modulated by list composition. This is evidence that grapheme–phoneme conversion is not as predominant for phonological recoding in intermediate orthographies as it is in shallow ones, and suggests that skilled reading in those orthographies is highly responsive to tasks conditions because readers may switch from smaller segment-by-segment decoding to larger unit or lexicon-related processing.     

Foundation literacy acquisition in European orthographies

Several previous studies have suggested that basic decoding skills may develop less effectively in English than in some other European orthographies. The origins of this effect in the early (foundation) phase of reading acquisition are investigated through assessments of letter knowledge, familiar word reading, and simple nonword reading in English and 12 other orthographies. The results confirm that children from a majority of European countries become accurate and fluent in foundation level reading before the end of the first school year. There are some exceptions, notably in French, Portuguese, Danish, and, particularly, in English. The effects appear not to be attributable to differences in age of starting or letter knowledge. It is argued that fundamental linguistic differences in syllabic complexity and orthographic depth are responsible. Syllabic complexity selectively affects decoding, whereas orthographic depth affects both word reading and nonword reading. The rate of development in English is more than twice as slow as in the shallow orthographies. It is hypothesized that the deeper orthographies induce the implementation of a dual (logographic + alphabetic) foundation which takes more than twice as long to establish as the single foundation required for the learning of a shallow orthography.     

Silent letters and phonological priming

Many written English words contain silent letters. Omitting them produces nonwords pronounced identically to the original words, for example, SALM for PSALM and COLUM for COLUMN. We report two naming and two lexical decision experiments in which targets of 4–11 letters followed primes exposed for 100 ms in mask-prime-mask-target sequences. Priming in SALM-psalm and COLUM-column pairs exceeded priming in orthographic control pairs such as ASTA-pasta and COUSI-cousin, pairs in which pronounced letters are omitted to form the primes. SALM and COLUM, however, were less effective primes than PSALM and COLUMN. Results were discussed in terms of the phonological coherence hypothesis, the role of orthographic codes in filtering phonologically activated representations, and graphemes as reading units.     

Phonology and orthography in reading aloud

This study investigated the orthographic and phonological contribution of visually masked primes to reading aloud in Dutch. Although there is a relatively clear mapping between the spelling and sound of words in Dutch, words starting with the letter c are ambiguous as to whether they begin with the phoneme /S/ (e.g., citroen, “lemon”) or with the phoneme /k/ (e.g., complot, “conspiracy”). Therefore, using words of this type, one can tease apart the contributions of orthographic and phonological activation in reading aloud. Dutch participants read aloud bisyllabic c-initial target words, which were preceded by visually masked, bisyllabic prime words that either shared the initial phoneme with the target (phonologically related) or the first grapheme (orthographically related) or both (phonologically and orthographically related). Unrelated primes did not share the first segment with the target. Response latencies in the phonologically related conditions were shorter than those in the unrelated condition. However, primes that were orthographically related did not speed up responses. One may conclude that the nature of the onset effect in reading aloud is phonological and not orthographic.     

Cambridge University versus Hebrew University: The impact of letter transposition on reading English and Hebrew

Is reading similarly affected by letter transposition in all alphabetic orthographies? “The Cambridge University effect,” demonstrating that jumbled letters have little effect on reading, was examined using rapid serial visual presentation (RSVP) in English and in Hebrew. Hebrew-English bilinguals were presented sentences in both languages containing words with transposed letters. Sentences were presented rapidly on the screen word by word, and participants had to reproduce the sequence of words perceived. We found a marked difference in the effect of transpositions in the two languages. In English, transpositions had little effect on performance, whereas in Hebrew, performance deteriorated dramatically for words with transposed letters. The differential effects of transposition are accounted for by the difference in lexical organization in Hebrew and in English, suggesting that models of reading in alphabetic orthographies may be language specific.     

Lexical and sublexical processes in the perception of transposed-letter anagrams

Evidence from priming and lexical decision tasks suggests that nonwords created by transposing adjacent letter pairs (TL nonwords) are very effective in activating lexical representations of their base words, because the process of orthographic matching tolerates minor changes in letter position. However, this account disregards the possible role of sublexical processing in reading. TL nonwords are perceptually ambiguous, with lexical and sublexical processing giving rise to conflicting interpretations. The consequences of this ambiguity were investigated in a lexical decision experiment with primes that were either high or low bigram frequency TL versions of target words. Priming effects were much larger for low BF primes (e.g., pucnh-PUNCH) than for high BF primes (e.g., panit-PAINT). This finding is interpreted as evidence that lexical activation can be inhibited by competing output resulting from sublexical processing of TL letter string. We conclude that phonological processing is an important determinant of responses to TL stimuli, and we consider how this interpretation might be accommodated within the dual-route cascaded (DRC) model of word recognition.     

Differences between Chinese morphosyllabic and German alphabetic readers in the Stroop interference effect

The goal of our study was to localize the source of the stronger Stroop interference effect found in morphosyllabic readers as compared with alphabetic readers. Twenty-three Chinese and 24 German undergraduate students were tested in a Stroop paradigm with the following stimuli: color patches, colorneutral words (e.g.,friend printed in yellow), incongruent color-associated words (e.g.,blood printed in blue), and incongruent color words (e.g.,yellow printed in blue). Results revealed no differences in German and Chinese students’ response times to color patches. Chinese participants, however, showed longer color naming latencies for neutral words as well as for color words and color-related words. No differences between German and Chinese participants were found when print color latencies for neutral words were subtracted from print color latencies for color words and color-related words. This result does not support theories which suggest that for morphosyllabic readers there is a direct route from orthography to the semantics of a word. We rather argue, with reference to dual route models of reading, that access from print to phonology is faster for morphosyllabic than for alphabetic readers, and therefore interference caused by conflicting phonologies of color name and written word will be stronger in Chinese readers than in German readers.