Can “words” be processed as integrated units?

Three experiments examined the information processing of letters embedded within one-syllable words and similar unpronounceable sequences. A speeded discrimination task was used to detect processing differences between words and nonwords in a situation where both the identity and position of critical display information was known to subjects before stimulus presentation. Results indicated that word pairs differing by two letters were more quickly discriminated than two words differing in a single letter, while nonword pairs differing in two letters were discriminated no faster than two nonwords differing in a single letter. A further comparison showed a performance advantage for words over nonwords in a condensation task that forced a scan of stimulus letters for correct responding. These results suggest that familiarity affects information processing at a perceptual level, and are incompatible with theories suggesting that familiarity effects are due to inferential factors following letter feature analysis.     

Whole and part comparisons of words and nonwords

Under conditions of sequential presentation, two words are matched more quickly than are a single letter and the first letter of a word. An exception to this whole-word advantage was reported in 1980 by Umansky and Chambers, who used word pairs as stimuli, and asked subjects to compare the entire words or the words’ first letters. Experiment 1 showed that the stimulus lists used by Umansky and Chambers may not have constrained subjects to process the displays differently for wholistic and component comparisons. In those studies, the two words were identical onsame trials for both wholistic and first-letter comparisons, so that first-letter decisions could have been based on wholistic information. In the present study, lists were constructed so that first-letter decisions could not be determined correctly by wholistic information (e.g., BLAME/BEACH), and the whole-word advantage was replicated. Experiment 2 tested whether wholistic comparisons are generally superior to component comparisons. For consonant strings, first-letter comparisons were made more quickly than were whole-string comparisons. These results are interpreted as support for hierarchical models of visual word processing.     

Letters and words in word identification

Studies by Barron and Henderson (1977) and Johnson (1975) provide evidence that whole words may be the unit of identification in word perception, rather than single letters. Johnson found that words were matched faster than a letter to the first letter in a word. Barron and Henderson found faster matching times for words than for legal non-word items in a letter-matching task. These findings support the interpretation that words are identified before individual letters. If so, a word-frequency effect should be expected. Experiments 1 and 2 tested for word vs. first-letter-in-word differences, as well as for a word-frequency effect in simultaneous and delayed visual matching tasks. In the simultaneous task, first letters in words were matched faster than words. In the delayed task, there was no difference between matching words or matching the first letters in words. With both tasks there was a word-frequency effect for word matches but not for first-letter-in-word matches. In Experiment 3, first-letter matching time was unrelated to word frequency or lexical status, although it did vary with orthographic legality. These results, on the whole, are consistent with a race model in which identifications take place simultaneously at word, letter-cluster, and letter levels, rather than a sequential model in which the whole word is identified before the component letters.     

Detection of intraword and interword letter repetition: A test of the word unitization hypothesis

Do words, as familiar units or gestalts, tend to swallow up and conceal their letter components (Pillsbury, 1897)? Letters typically are detected faster and more accurately in words than in nonwords (i.e., scrambled collections of letters), and in more frequent words than in less frequent words. However, a word advantage at encoding, where the representation of the string is formed, might compensate for, and thus mask, a word disadvantage at decoding and comparison, where the component letters of the representation are accessed and compared with the target letter. To better reveal any such word disadvantage, a task was used in this study that increased the amount of letter processing. Subjects judged whether a letter was repeated within a six-letter word or a nonword (Experiment 1; intraword letter repetition) or was repeated between two adjacent unrelated six-letter words or nonwords (Experiment 2; interword letter repetition). Contrary to Pillsbury's word unitization hypothesis, both types of letter repetition (intraword and interword) were detected faster and just as accurately with words as with nonwords. In Experiment 2, however, interword letter repetition was detected less accurately on common words (but not on rare words or third-order pseudowords) than on the corresponding nonwords. Thus, although the familiar word does not deny access to its own component letters, it does make their comparison with letters from other words more difficult.     

Are letter codes always activated?

This investigation assesses whether a word’s constituent-letter codes are activated when whole-word processing is encouraged, as well as when letter processing is encouraged. In Experiments 1 and 2, word primes were followed by a target item that had to be named. The target was a word, a constituent letter that had appeared in the prime, or a nonconstituent letter that had not appeared in the prime. The measure of constituent-letter activation was the difference in letter-naming latency between constituent and nonconstituent letters. Presumably, if constituent letters are named faster, letters are being activated by the words in which they appear. To encourage either whole-word or letter analysis during the processing of the priming word, the proportion of word versus letter targets was systematically varied. When the proportion heavily favored letter targets, constituent letters were named faster than their nonconstituent controls. However, a constituent-letter (vs. a nonconstituent-letter) advantage was not obtained when the proportion favored word targets. Experiment 3 replicated the effect with a priming task that required a discrimination response, instead of a naming response. Thus the results suggested that the activation of constituent-letter codes need not take place during the processing of words, but occurs only when letter analysis is stressed by the task.     

Familiarity effects and word unitization in visual comparison tasks

The decision that two words are identical is made more quickly than the decision that two non-words are identical. This familiarity effect was shown to be larger in a simultaneous matching task than in a sequential matching task. In the simultaneous task, two words were not matched as quickly as a single letter and a letter in a predesignated location within a word. The latter finding rules out a perceptual unitization account of the familiarity effect (Silverman, 1985). The familiarity effect was interpreted to be due to the facilitated encoding of a comparison item when a holistic cognitive unit representing the target is activated in memory.     

Familiarity affects visual processing of words.

While previous research has demonstrated that words can be processed more rapidly and/or more accurately than random strings of letters, it has not been convincingly demonstrated that the superior processing of words is a visual effect. In the present experiment, the cases of letters were manipulated in letter strings that were to be compared on the basis of physical identity. Mean response time was shorter for words than for nonwords even for pairs of letter strings that differed only in case (e.g., site-site). This finding implies that the advantage of words over nonwords (the familiarity effect) typically observed in the simultaneous matching task is not due solely to comparison of either the word names or the letter names and, thus, that at least part of the familiarity effect must be due to more rapid formation and/or comparison of visual representations of the two letter strings when they are words. Further analysis failed to reveal a significant involvement of phonemic or lexical codes in the comparison judgments.     

Multiletter Word Units in Visual Recognition: Direct Activation by Supraletter Visual Features

Multiletter priming effects have been interpreted as evidence for the representation of separable multiletter units in the visual word recognition system (Whiteley & Walker, 1994). The reported experiments examine whether the activation of such units is pre- or post-lexical. Experiments 2 and 3 employed priming in an alphabetic decision task in which subjects made a discrimination response to test stimuli which could be classed as either targets or foils. Targets were single letters, or consonant bigrams, present or absent in an immediately preceding word, or (Experiment 3 and 4) they were whole words semantically associated or not to a preceding word. Foils were single non-alphanumeric characters, a character plus a letter, or a word with one letter replaced by a character. Experiment 1 was a preliminary to determine the parameters of a sequential presentation manipulation. Experiment 2 compared conditions of simultaneous and sequential presentation where letters of prime words were presented together, or one at a time in rapid succession. With simultaneous presentation, responses to bigram targets were facilitated when these appeared in the prime word, while responses to individual constituent letters of those bigrams were not facilitated. Additionally, responses to primed bigram targets were faster than responses to primed single letter targets. The sequential presentation of prime words resulted in a qualitative change in the response pattern indicative of the disruption of multiletter unit activation. That change was replicated in Experiment 3 where semantic priming confirmed that the prime words were being processed to a level of meaning. The observations challenge a post-lexical account of the multiletter priming effects. Finally, Experiment 4 addressed the question of whether bigram priming reflects the intentional use of prime information to predict following targets. Strategic interpretations are undermined and it is argued that multiletter units are activated automatically as part of normal visual word recognition.     

Differential backward masking of words and letters by masks of varying orthographic structure

The effects of structural relationships between targets and masks were investigated using a backward-masking paradigm. Specifically, the masking of single letters, common fiveletter words, and five-letter pseudowords masked by a blank flash, strings of overlapped letters, pseudowords, and words was investigated. Target duration was varied from 2 to 32 msec, with mask duration held constant at 25 msec. The dependent measure was the critical interstimulus interval for correct target identification. Letters were more effectively masked than words and pseudowords. A blank mask caused the least amount of masking, followed by the overlapped letter strings, and then the word and pseudoword masks. In addition to the overall greater masking effectiveness for the three patterned masks, overlapped letter strings masked letters more effectively than they did words. The implications of current theories of masking for these results and the implications of these results for theories of word recognition were discussed.     

The role of configuration in the identification of visually degraded words

Two experiments were carried out in which reaction time for identification of words in sets of four was measured. The words were of three letters, with dichotomous variation of the first and third letters, in either upper- or lowercase, with letters chosen to give maximum variation in word configuration in lowercase. Four types of display were used, differing in size and masking. In Experiment 1, the words were easy to identify in uppercase as well as in lowercase, because the letters were easily discriminable in either case, and the results showed no advantage to lowercase words at any level of degradation. In Experiment 2, the letters were more difficult to discriminate in uppercase, and with the most severe degradation, the words in lowercase were identified more rapidly than those in uppercase. Thus the reaction time results showed an advantage of the configural properties available in lowercase only with words difficult to identify in uppercase and with extreme amounts of degradation. Error analyses of both experiments showed that errors were made to individual letters, not on the basis of the configuration, for words in both cases. It is concluded that letter processing occurs even when there is evidence for the use of configural information, thus implying a predominant role of letter perception in word identification. Some reservation is expressed whether whole-word configuration is the only pertinent configural property and whether the results from such simple identification tasks are appropriate to an understanding of natural reading.     

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.     

How to make the word-length effect disappear in letter-by-letter dyslexia: implications for an account of the disorder

The diagnosis of letter-by-letter (LBL) dyslexia is based on the observation of a substantial and monotonic increase of word naming latencies as the number of letters in the stimulus increases. This pattern of performance is typically interpreted as indicating that word recognition in LBL dyslexia depends on the sequential identification of individual letters. We show, in 7 LBL patients, that the word-length effect can be eliminated if words of different lengths are matched on the sum of the confusability (visual similarity between a letter and the remainder of the alphabet) of their constituent letters. Additional experiments demonstrate that this result is mediated by parallel letter processing and not by any compensatory serial processing strategy. These findings indicate that parallel processing contributes significantly to explicit word recognition in LBL dyslexia and that a letter-processing impairment is fundamental in causing the disorder.     

Part-whole relationships in the processing of small visual patterns

Subjects seem to react to a word faster than they react to a letter within a word. One interpretation is that words are processed holistically; another is that all visual stimuli are processed in terms of components, but that more stimulus information is available for use when the targets are words than when they are letters within words. The results of three experiments indicate that the word or pattern-level advantage occurs even when the stimulus information in the two situations is equated, but if the perceptual arrays cannot be unitized (e.g., consonant sequences), a pattern-level advantage does not occur. In addition, the experiments provide substantial evidence to indicate that if letter arrays cannot be unitized, then they are processed on a componentby-component basis, rather than holistically. Finally, the appropriate definition ofholistic processing is considered.     

Parafoveal processing of words and saccade computation during eye fixations in reading

Studied parafoveal word processing during eye fixations in reading to answer two questions: (a) Is the processing of parafoveally available words limited to the identification of beginning letters? (b) Does the parafoveal processing of words affect the following interword saccade? Reading afforded either no parafoveal preview, preview of beginning trigrams, preview of ending trigrams, or preview of the whole parafoveal word. Previews were controlled by replacing original letters either with X's or dissimilar letters. Preview benefits were larger for the whole word previews than for beginning or ending trigram previews. X-masks yielded preview benefits from intact beginning and ending trigrams but dissimilar letter masks yielded benefits from beginning trigrams only. Saccades were larger for whole word previews than for no previews. These results support Logogen-type models of word recognition and a model of saccade computation that posits a time-locked functional relation between the acquisition of parafoveal word information and the positioning of each fixation.     

The relative benefit of word context is a constant proportion of letter identification time

Letter identification is reduced when the target letter is surrounded by other, flanking letters. This visual crowding is known to be impacted by physical changes to the target and flanks, such as spatial frequency content, polarity, and interletter spacing. There is also evidence that visual crowding is reduced when the flanking letters and the target letter form a word. The research reported here investigated whether these two phenomena are independent of each other or whether the degree of visual crowding impacts the benefit of word context. Stimulus duration thresholds for letters presented alone and for the middle letters of 3-letter words and nonwords were determined for stimuli presented at the fovea and at the periphery. In Experiment 1, the benefit of word context was found to be the same at the fovea, where visual crowding is minimal, and at the periphery, where visual crowding is substantial. In Experiment 2, visual crowding was manipulated by changing the interletter spacing. Here, too, the benefit of word context was fairly constant for the two retinal locations (fovea or periphery), as well as with changes in interletter spacing. These data call into question both the idea that the benefit of word context is greater when stimulus quality is reduced (as is the case with visual crowding) and the idea that words are processed more effectively when they are presented at the fovea.     

Reading Ability and the Use of Context in Orthographic Information Processing

Reading ability and context use in orthographic processing during silent reading were investigated. Poor readers and reading-level matched controls were presented target words containing a letter substitution in two contrasting context conditions. The hypothesis was that presenting a word in a highly predictable context would induce readers to proceed through the text without completely processing orthographic units at lower processing levels in the hierarchy (e.g., constituent letters). This general context effect was found. Normal and poor readers did not differ in context dependency. Poor readers more often missed substitutions, regardless of context. Poor readers also processed orthographic information less accurately. Target letters located in final word position were missed most often. Substitutions located in high bigram-frequency letter clusters were more often missed, and this effect was independent of intraword location. The implications of these results for understanding poor readers' basic reading problems are discussed.     

Direct assessments of the processing time hypothesis for the missing-letter effect

When participants search for a target letter while reading, they make more omissions if the target letter is embedded in frequently used words or in the most frequent meaning of a polysemic word. According to the processing time hypothesis, this occurs because familiar words and meanings are identified faster, leaving less time for letter identification. Contrary to the predictions of the processing time hypothesis, with a rapid serial visual presentation procedure, participants were slower at detecting target letters for more frequent words or the most frequent meaning of a word (Experiments 1 and 2) or at detecting the word itself instead of a target letter (Experiment 3). In Experiments 4 and 5, participants self-initiated the presentation of each word, and the same pattern of results was observed as in Experiments 1 and 3. Positive correlations were also found between omission rate and response latencies.     

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.     

Interpretation of orthographic uniqueness point effects in visual word recognition

The orthographic uniqueness point (OUP) of a word is the position of the first letter from the left that distinguishes a word from all other words. In 2 recent studies (P. J. Kwantes & D. J. K. Mewhort, 1999a; A. K. Lindell, M. E. R. Nicholls, & A. E. Castles, 2003), it has been observed that words with an early OUP were processed more quickly than words with a late OUP. This has been taken to suggest that observers process the letters of words sequentially in a left-to-right order. In this article, it is shown that the OUP results do not provide selective evidence for left-to-right sequential processing in visual word recognition because the data are also compatible with an account in which letter processing occurs in random order.     

Visual detection of multi-letter patterns

This study addressed two basic questions about the detection of multi-letter patterns: (a) How is the detection of a multi-letter pattern related to the detection of its individual components? (b) How is the detection of a sequence of letters influenced by the observer's familiarity with that sequence? In three experiments observers searched for one-, two-, or three letter patterns embedded in a rapid series of multiple six-letter frames. In Experiment 1, unfamiliar two-letter patterns were detected more accurately than their one-letter components. This two-letter advantage reflects the fact that in an array of fixed size, larger target stimuli contain more information and are easier to discriminate from nontarget alternatives. Quantitative analyses indicated that observers combine information not decisions, about the component letters in a pattern. In Experiment 2, with statistical and physical properties equated, a familiar three-letter pattern (i.e., CAT) was detected more accurately than its unfamiliar anagram (i.e., TCA). This word advantage in word (not letter) detection persisted even after extensive practice and was uninfluenced by the lexical character of distractor items. In Experiment 3, words (e.g., FIB), pronounceable non words (e.g., FIF(, and familiar acronyms (e.g., FBI) were detected more readily than unfamiliar items (e.g., IBF). Thus both orthographic knowledge and familiarity with specific sequences can facilitate perceptual processing in "word" detection.