A developmental, interactive activation model of the word superiority effect

Parallel distributed processing (PDP) models of reading developed out of an appreciation of the role that context plays in letter and word perception. Adult readers can more accurately identify letters in a word than alone or in other random display contexts, a phenomenon known as the Word Superiority Effect (WSE). We examined the effects of orthographic context on the letter recognition skills of dyslexic children, comparing their performance to adults, and chronological- and reading-age matched groups. Consistent with previous studies, results showed adults better able to identify letters in the context of words and pseudowords than in random letter strings. Young normal readers demonstrated the WSE, but their pseudoword advantage was less than adults. The dyslexic children showed no WSE at all. PDP computer simulations for the experimental data using the Interactive Activation model (IA) suggested that the orthographic components of the lexical system of normal children are interactive and distributed as they are in adults but provide less bottom-up activation. In addition, top-down processing increases with age and reading skill, but may be absent for dyslexic readers.     

The effects of length and transposed‐letter similarity in lexical decision: Evidence with beginning,intermediate, and adult readers

Do length and transposed‐letter effects reflect developmental changes on reading acquisition in a transparent orthography? Can computational models of visual word recognition accommodate these changes? To answer these questions, we carried out a masked priming lexical decision experiment with Spanish beginning, intermediate, and adult readers (N=36, 44, and 39; average age: 7, 11, and 22 years, respectively). Target words were either short or long (6.5 vs. 8.5 letters), and transposed‐letter primes were formed by the transposition of two letters (e.g. aminal–ANIMAL) or by the substitution of two letters (orthographic control: arisal–ANIMAL). Children showed a robust length effect (i.e. long words were read slower than short words) that vanished in adults. In addition, both children and young adults showed a transposed‐letter priming effect relative to the control condition. A robust transposed‐letter priming effect was also observed in non‐word reading, which strongly suggests that this effect occurs at an early prelexical level. Taken together, the results reveal that children evolve from a letter‐by‐letter reading to a direct lexical access and that the lexical decision task successfully captures the changing strategies used by beginning, intermediate, and adult readers. We examine the implications of these findings for the recent models of visual word recognition.     

Letter information and orthographic context in word perception

Theories of word perception in reading can be categorized in terms of the assumption made about whether or not a word context modifies the feature analysis of its component letters. Independence theories assume that the visual information passed on by feature analysis is independent of word context. Nonindependence theories assume that a word context directly influences visual analysis. Some nonindependence theories have assumed that word context enhances feature analysis of letters, others have assumed that word context overrides feature analysis of letters, and some have assumed that word context directs which letters are analyzed. The present experiment provided a critical test between the two classes of theories by independently varying orthographic context and visual letter information in a letter recognition task. The results contradict the qualitative predictions of the class of nonindependence theories and are accurately described by a quantification of independence theory.     

Effects of parafoveal word length and orthographic features on initial fixation landing positions in reading

Previous research has demonstrated that readers use word length and word boundary information in targeting saccades into upcoming words while reading. Previous studies have also revealed that the initial landing positions for fixations on words are affected by parafoveal processing. In the present study, we examined the effects of word length and orthographic legality on targeting saccades into parafoveal words. Long (8?C9 letters) and short (4?C5 letters) target words, which were matched on lexical frequency and initial letter trigram, were paired and embedded into identical sentence frames. The gaze-contingent boundary paradigm (Rayner, 1975) was used to manipulate the parafoveal information available to the reader before direct fixation on the target word. The parafoveal preview was either identical to the target word or was a visually similar nonword. The nonword previews contained orthographically legal or orthographically illegal initial letters. The results showed that orthographic preprocessing of the word to the right of fixation affected eye movement targeting, regardless of word length. Additionally, the lexical status of an upcoming saccade target in the parafovea generally did not influence preprocessing.     

Not all visual features are created equal: early processing in letter and word recognition

In four experiments, we investigated the effect of deleting specific features of letters on letter and word recognition in the context of reading aloud. Experiments 1 and 2 assessed the relative importance of vertices versus midsegments in letter recognition. Experiments 3 and 4 tested the relative importance of vertices versus midsegments in word recognition. The results demonstrate that deleting vertices is more detrimental to letter and word identification than is deleting midsegments of letters. These results converge with those of previous research on the role of vertices in object identification. Theoretical implications for early processing in reading are noted.     

Is a "Phoenician" reading style superior to a "Chinese" reading style? Evidence from fourth graders

This study compared normally achieving fourth-grade "Phoenician" readers, who identify nonwords significantly more accurately than they do exception words, with "Chinese" readers, who show the reverse pattern. Phoenician readers scored lower than Chinese readers on word identification, exception word reading, orthographic choice, spelling, reading comprehension, and verbal ability. When compared with normally achieving children who read nonwords and exception words equally well, Chinese readers scored as well as these children on word identification, regular word reading, orthographic choice, spelling, reading comprehension, phonological sensitivity, and verbal ability and scored better on exception word reading. Chinese readers also used rhyme-based analogies to read nonwords derived from high-frequency exception words just as often as did these children. As predicted, Phoenician and Chinese readers adopted somewhat different strategies in reading ambiguous nonwords constructed by analogy to high-frequency exception words. Phoenician readers were more likely than Chinese readers to read ambiguous monosyllabic nonwords via context-free grapheme-phoneme correspondences and were less likely to read disyllabic nonwords by analogy to high-frequency analogues. Although the Chinese reading style was more common than the Phoenician style in normally achieving fourth graders, there were similar numbers of poor readers with phonological dyslexia (identifying nonwords significantly more accurately than exception words) and surface dyslexia (showing the reverse pattern), although surface dyslexia was more common in the severely disabled readers. However, few of the poor readers showed pure patterns of phonological or surface dyslexia.     

Length, formats, neighbours, hemispheres, and the processing of words presented laterally or at fixation

It has long been known that the number of letters in a word has more of an effect on recognition speed and accuracy in the left visual field (LVF) than in the right visual field (RVF) provided that the word is presented in a standard, horizontal format. After considering the basis of the length by visual field interaction two further differences between the visual fields/hemispheres are discussed: (a) the greater impact of format distortion (including case alternation) in the RVF than in the LVF and (b) the greater facilitation of lexical decision by orthographic neighbourhood size (N) in the LVF than in the RVF. In the context of split fovea accounts of word recognition, evidence is summarised which indicates that the processing of words presented at fixation is affected by the number of letters to the left of fixation but not by the number of letters to the right and by the number of orthographic neighbours activated by letters to the left of fixation but not by the number of orthographic neighbours activated by letters to the right of fixation. A model of word recognition is presented which incorporates the notion that the left hemisphere has sole access to a mode of word recognition that involves parallel access from letter forms to the visual input lexicon, is disrupted by format distortion, and does not employ top-down support of the letter level by the word level.     

Orthographic redundancy in letter recognition: Orthographic neighbourhood or orthographic context?

Abstract

In explaining the word-superiority effect (i.e. the better detection of a letter in a word than in a nonword), the Interactive Activation Model (IAM) of McClelland and Rumelhart (1981) and the Fuzzy Logical Model of Perception (FLMP) of Massaro (1979) emphasise the importance of orthographic redundancy (i.e. the regularities of letters within words) in different ways. In the IAM, orthographic redundancy is defined by the number of “friends”; that is, words sharing the same letters except one with the word containing the target letter. Such friends constitute the orthographic “neighbourhood”. FLMP stresses the orthographic “context”; that is, the similarity of the word with a representation in the lexicon. The orthographic neighbourhood and context are manipulated independently in Experiment 1, and the findings are better understood in terms of the orthographic neighbourhood. By increasing the number of friends in nonwords, better letter detection is also obtained in nonwords as compared with letter detection in random letter strings (Experiment 2). These findings, together with the position effects obtained, are more clearly in agreement with the IAM than with the FLMP.     

Eye movements when reading: The importance of the word to the left of fixation

In reading, it is well established that word processing can begin in the parafovea while the eyes are fixating the previous word. However, much less is known about the processing of information to the left of fixation. In two experiments, this issue was explored by combining a gaze-contingent display procedure preventing parafoveal preview and a letter detection task. All words were displayed as a series of xs until the reader fixated them, thereby preventing forward parafoveal processing, yet enabling backward parafoveal or postview processing. Results from both experiments revealed that readers were able to detect a target letter embedded in a word that was skipped. In those cases, the letter could only have been identified in postview (to the left of fixation), and detection rate decreased as the distance between the target letter and the eyes' landing position increased. Most importantly, for those skipped words, the typical missing-letter effect was observed with more omissions for target letters embedded in function than in content words. This can be taken as evidence that readers can extract basic prelexical information, such as the presence of a letter, in the parafoveal area to the left of fixation. Implications of these results are discussed in relation to models of eye movement control in reading and also in relation to models of the missing-letter effect.     

Orthographic processing in visual word identification

A series of experiments is reported examining orthographic priming effects between briefly presented pairs of letter strings. The experiments investigate the effects of the number and position of letters shared by primes and targets, and the effects of prime-target length. Priming effects increase nonlinearly as a function of both the number and the position of shared letters, and they are dependent on the positions of letters relative to both the end positions in the string and to the identities of their nearest neighbours. There is little effect of absolute string length on priming. These priming effects can be distinguished from intrusion errors where letters from primes are reported in response to targets. An account of orthographic processing is outlined which attributes priming to cooperative interactions between coarse relative-position coded letter cluster representations activated by primes and targets. The implications of the findings for understanding other effects in word recognition and reading are discussed.     

Reduced orthographic learning in dyslexic adult readers: Evidence from patterns of letter search

Visual letter search performance was investigated in a group of dyslexic adult readers using a task that required detection of a cued letter target embedded within a random five-letter string. Compared to a group of skilled readers, dyslexic readers were significantly slower at correctly identifying targets located in the first and second string position, illustrating significantly reduced leftward facilitation than is typically observed. Furthermore, compared to skilled readers, dyslexic readers showed reduced sensitivity to positional letter frequency. They failed to exhibit significantly faster response times to correctly detect target letters appearing in the most, compared to least, frequent letter position within five-letter words, and response times correlated with positional letter frequency only for the initial, and not the final, letter position. These results are compatible with the SERIOL (sequential encoding regulated by inputs to oscillations within letter units) model of orthographic processing proposed by Whitney and Cornelissen (2005). Furthermore, they suggest that dyslexic readers are less efficient than skilled readers at learning to extract statistical regularities from orthographic input.     

Reduced orthographic learning in dyslexic adult readers: evidence from patterns of letter search

Visual letter search performance was investigated in a group of dyslexic adult readers using a task that required detection of a cued letter target embedded within a random five-letter string. Compared to a group of skilled readers, dyslexic readers were significantly slower at correctly identifying targets located in the first and second string position, illustrating significantly reduced leftward facilitation than is typically observed. Furthermore, compared to skilled readers, dyslexic readers showed reduced sensitivity to positional letter frequency. They failed to exhibit significantly faster response times to correctly detect target letters appearing in the most, compared to least, frequent letter position within five-letter words, and response times correlated with positional letter frequency only for the initial, and not the final, letter position. These results are compatible with the SERIOL (sequential encoding regulated by inputs to oscillations within letter units) model of orthographic processing proposed by Whitney and Cornelissen (2005). Furthermore, they suggest that dyslexic readers are less efficient than skilled readers at learning to extract statistical regularities from orthographic input.     

Integrating information across eye fixations in reading: the role of letter and word units

Eye movements and eye fixations were recorded to study the integration of letter/word information across interword fixations in reading. Two hypotheses were examined. One hypothesis posits that readers obtain effective information from the beginning two or three letters of a parafoveal word. This information facilitates the recognition of the word when it is being fixated. The alternative posits that effective information is obtained from the complete parafoveal word. The results of the present study showed faster reading rates when parafoveal previews comprised complete words than when they comprised beginning letters alone. Furthermore, the usability of parafoveally available partial word information from beginning and ending letters was not affected by small variations in retinal eccentricity. Both findings were taken as evidence that readers gain useful information from all letters of the parafoveally available word and that whole word information, rather than specific letter information, is integrated across interword fixations in reading.     

The independence of letter identity and letter doubling in reading

The ability to read requires processing the letter identities in the word and their order, but it is by no means obvious that our long-term memory representations of words spellings consist of only these dimensions of information. The current investigation focuses on whether we process information about another dimension—letter doubling (i.e., that there is a double letter in WEED)—independently of the identity of the letter being doubled. Two experiments that use the illusory word paradigm are reported to test this question. In both experiments, participants are more likely to misperceive a target word with only singleton letters (e.g., WED) as a word with a double (e.g., WEED) when the target is presented with a distractor that contains a different double letter (e.g., WOOD) than when the distractor does not contain a double letter (e.g., WORD). This pattern of results is not predicted by existing computational models of word reading but is consistent with the hypothesis that written language separately represents letter identity and letter doubling information, as previously shown in written language production. These results support a view that the orthographic representations that underlie our ability to read are internally complex and suggest that reading and writing rely on a common level of orthographic representation.     

Word recognition cues for beginning and skilled readers

Three experiments are reported in which subjects chose from an array the response alternative that most resembled a nonsense stimulus trigram or quingram. Subjects could respond on the basis of individual letter positions or overall word shape. However, word shape was not manipulated independently of the individual letters that comprise the word as has been the case in previous studies. Kindergarten children and first graders who were poor readers showed a more varied pattern of responses than first graders who were good readers. With simple stimuli (trigrams), beginning and skilled readers chose alternatives on the basis of overall shape and the initial letter. With more complex stimuli (quingrams), the skilled readers continued choosing on the basis of word shape and initial letters, but beginning readers only chose on the basis of the first letter.     

Parafoveal letter-position coding in reading

The masked-priming lexical decision task has been the paradigm of choice for investigating how readers code for letter identity and position. Insight into the temporal integration of information between prime and target words has pointed out, among other things, that readers do not code for the absolute position of letters. This conception has spurred various accounts of the word recognition process, but the results at present do not favor one account in particular. Thus, employing a new strategy, the present study moves out of the arena of temporal- and into the arena of spatial information integration. We present two lexical decision experiments that tested how the processing of six-letter target words is influenced by simultaneously presented flanking stimuli (each stimulus was presented for 150 ms). We manipulated the orthographic relatedness between the targets and flankers, in terms of both letter identity (same/different letters based on the target’s outer/inner letters) and letter position (intact/reversed order of letters and of flankers, contiguous/noncontiguous flankers). Target processing was strongly facilitated by same-letter flankers, and this facilitatory effect was modulated by both letter/flanker order and contiguity. However, when the flankers consisted of the target’s inner-positioned letters alone, letter order no longer mattered. These findings suggest that readers may code for the relative position of letters using words’ edges as spatial points of reference. We conclude that the flanker paradigm provides a fruitful means to investigate letter-position coding in the fovea and parafovea.     

Letter position dyslexia in Arabic: from form to position

This study reports the reading of 11 Arabic-speaking individuals with letter position dyslexia (LPD), and the effect of letter form on their reading errors. LPD is a peripheral dyslexia caused by a selective deficit to letter position encoding in the orthographic-visual analyzer, which results in migration of letters within words, primarily of middle letters. The Arabic orthography is especially interesting for the study of LPD because Arabic letters have different forms in different positions in the word. As a result, some letter position errors require letter form change. We compared the rate of letter migrations that change letter form with migrations that do not change letter form in 10 Arabic-speaking individuals with developmental LPD, and one bilingual Arabic and Hebrew-speaking individual with acquired LPD. The results indicated that the participants made 40% letter position errors in migratable words when the resulting word included the letters in the same form, whereas migrations that changed letter form almost never occurred. The error rate of the Arabic-Hebrew bilingual reader was smaller in Arabic than in Hebrew. However, when only words in which migrations do not change letter form were counted, the rate was similar in Arabic and Hebrew. Hence, whereas orthographies with multiple letter forms for each letter might seem more difficult in some respects, these orthographies are in fact easier to read in some forms of dyslexia. Thus, the diagnosis of LPD in Arabic should consider the effect of letter forms on migration errors, and use only migratable words that do not require letter-form change. The theoretical implications for the reading model are that letter form (of the position-dependent type found in Arabic) is part of the information encoded in the abstract letter identity, and thus affects further word recognition processes, and that there might be a pre-lexical graphemic buffer in which the checking of orthographic well-formedness takes place.     

Orthographic effects in the word substitutions of aphasic patients: an epidemic of right neglect dyslexia?

Aphasic patients with reading impairments frequently substitute incorrect real words for target words when reading aloud. Many of these word substitutions have substantial orthographic overlap with their targets and are classified as "visual errors" (i.e., sharing 50% of targets' letters in the same relative position). Fifteen chronic aphasic patients read a battery of words and non-words; non-word reading was poor for all patients, and more than 50% of errors on words involved the substitution of a non-target word. An investigation of the factors conditioning these word substitutions, as well as the production of visual errors, identified a number of similarities to patterns previously reported for patients with right neglect dyslexia, which has been said to occur relatively rarely. These included a strong tendency for errors to overlap targets in initial letter positions, maintenance of target length in errors, and the substitution of words higher in imageability than targets. It is proposed that left hemisphere damage frequently leads to disruption of a level of representation for written words in which letter position is ordinally coded, resulting in exacerbation of a normal processing advantage for early letter positions. A computational model is discussed that incorporates this level of representation and successfully simulates relevant normal and patient data.