Semantic errors or statistical flukes? a note on allport's “on knowing the meaning of words we are unable to report”

Allport (1977) has claimed that normal subjects who view briefly a multi-word array, which is then pattern-masked, produce semantic errors analogous to those made by subjects with “deep dyslexia”. We claim that the rate at which normal subjects make such errors is consistent with their arising purely by chance.     

Red is read but eye is blue: a case study of developmental dyslexia and follow-up report

Recently, analogies have been drawn between the developmental and acquired dyslexias, but there has been no unequivocal report of developmental deep dyslexia. The case is reported here of a 9-year-old child whose reading performance resembles deep dyslexia in several ways. The incidence of errors with a semantic component is shown to be significantly above chance. The pattern of errors is discussed in relation to recent data on error patterns in normal beginning readers. Text reading, spelling, naming, and repetition are described. In contrast to previous case reports, follow-up data are presented which chart the qualitative changes in performance over time despite relatively little quantitative change. The child described has several specific handicaps. Developmental deep dyslexia may not be a prevalent subtype because the multiple impairments necessary to produce the disorder seldom co-occur and may tend to preclude reading development altogether.     

Are semantic errors actually semantic?: Evidence from alzheimer's disease

Patients with Alzheimer's disease (AD) produce a high rate of semantic errors when naming to confrontation. This is considered to be one of the many consequences of their semantic memory deficit. However, it has been shown, in aphasic patients with focal lesions, that semantic errors could arise from impairment to any one of the levels in the naming process. To check this hypothesis in AD, we assessed in 15 patients the capacity to name and access semantic knowledge (by multiple-choice probe questions) about 14 objects presented successively in the visual, tactile, auditory, and verbal modalities. In the visual naming task, 33 errors were recorded: 26 (78.8%) were semantic and 7 (21.2%) were unrelated errors. Of the 26 semantic errors, 8 were related to a deficit of the semantic knowledge related to the item and 17 to a deficit in the retrieval of the phonological form of the word. One was associated with a deficit of access to semantic knowledge in the visual modality. The 7 unrelated errors were associated with a loss of semantic knowledge for 4 and deficit of access to the phonological form for 3. In conclusion, this study shows that semantic errors do not systematically reflect a deficit of semantic knowledge in Alzheimer's disease. It also seems that unrelated errors are more frequently related to semantic deficits than semantic errors in this population.     

Retrieval processes in arithmetic production and verification

To investigate whether arithmetic production and verification involve the same retrieval processes, we alternated multiplication production trials (e.g., 9 × 6 = ?) with verification trials (4 × 9 = 36, true or false?) and analyzed positive error priming.Positive error priming is the phenomenon in which errors frequently match correct answers from preceding problems. Production errors were strongly primed by previous production trials (the error-answer matching rate was about 90% greater than expected by chance), but production errors were not strongly primed by previous verification trials (≈13% above chance). Conversely, false-verification errors were primed by previous verification trials (≈25% above chance), but not by production trials. The results indicated that arithmetic production and verification were mediated by different memory processes and suggest a familiarity-based over a retrieval-based model of arithmetic verification.     

Deep dyslexia in the two languages of an Arabic/French bilingual patient.

We present a single case study of an Arabic/French bilingual patient, ZT, who, at the age of 32, suffered a cerebral vascular accident that resulted in a massive infarct in the left peri-sylvian region. ZT's reading displays the characteristics of the deep dyslexia syndrome in both languages, that is, production of semantic, visual, and morphological errors, and concreteness effect in reading aloud and impossibility of reading nonwords. In the first part of this paper, using a three-route model of reading, we account for the patient's performance by positing functional lesions, which affect the non-lexical, the semantic lexical and the non-semantic lexical routes of reading. Phonological priming observed in a cross-language visual lexical decision task indicates that implicit assembled phonological recoding is possible. The above lesions and implicit nonword reading characterize the output form of deep dyslexia. However, error distribution reveals dissociations across languages (e.g. the semantic error rate is higher in French whereas translations are more frequent in the Arabic testing) that cannot be accounted for within a three-route model. In the second part, extensions to Plaut and Shallice's connectionist model (Cognitive Neuropsychology, 10 (5) (1993) 377) are proposed to account for the translinguistic errors observed. ZT's error distribution is compared to that obtained by Plaut and Shallice after lesions had been applied at different locations through the 40-60 network. The overall syndrome of deep dyslexia found in both languages is explained as resulting from lesions along the direct (O-->I) and output (S-->Ip, Ip-->P) pathways of reading. Lesions along the output pathway mostly affecting S-->Ip connections in French and Ip-->P connections in Arabic account for discrepancies in ZT's error pattern across tasks and languages. This case study demonstrates the superiority of a connectionist approach for predicting the error pattern in deep dyslexia.     

When does a deep dyslexic make a semantic error? The roles of age-of-acquisition, concreteness, and frequency

Semantic reading errors are the central and defining feature of deep dyslexia. This study compared the words the deep dyslexic patient LW read correctly with those she omitted and those to which she produced semantic errors in terms of their concreteness, age-of-acquisition, frequency, and length. Semantic errors were made to less concrete, later-acquired, and shorter words than were read correctly; there was no reliable effect of word frequency. More importantly, the actual semantic errors produced were later-acquired than the stimulus words, but they were not more concrete or reliably more frequent. These results implicate age-of-acquisition in the process that produces semantic errors. It is proposed that concreteness determines the specificity of the semantic system to activate a set of candidate responses and that age-of-acquisition biases the ease with which certain words can be selected from this set to be produced as reading responses.     

Deep dyslexia and semantic errors: a test of the failure of inhibition hypothesis using a semantic blocking paradigm

Deep dyslexia is an acquired reading disorder that involves the production of semantic errors and the inability to read aloud nonwords successfully. Several explanations for this reading impairment posit multiple loci of damage to account for the various error types produced in deep dyslexia. In contrast, the failure of inhibition hypothesis suggests that damage in the phonological output lexicon alone can explain these errors. Specifically, this hypothesis proposes normal processing via orthographic and phonological reading routes, as well as an intact semantic system. However, slowed or reduced inhibitory connections result in the failure to suppress spuriously activated neighbours in the phonological output lexicon, where neighbourhood can be defined in terms of phonology, orthography, or semantics. Given a failure to inhibit semantically related candidates, semantic reading errors occur. Important to the test of this hypothesis is that it evolves several predictions that are contrary to performance observed in the normal population. In particular, semantic errors are predicted to be greater in conditions where words are blocked according to semantic category than in random presentations. In addition, a semantic interference effect is expected. The results of semantic blocking were consistent with these predictions and lend support to the failure of inhibition hypothesis.     

Lesioning an attractor network: investigations of acquired dyslexia

A recurrent connectionist network was trained to output semantic feature vectors when presented with letter strings. When damaged, the network exhibited characteristics that resembled several of the phenomena found in deep dyslexia and semantic-access dyslexia. Damaged networks sometimes settled to the semantic vectors for semantically similar but visually dissimilar words. With severe damage, a forced-choice decision between categories was possible even when the choice of the particular semantic vector within the category was not possible. The damaged networks typically exhibited many mixed visual and semantic errors in which the output corresponded to a word that was both visually and semantically similar. Surprisingly, damage near the output sometimes caused pure visual errors. Indeed, the characteristic error pattern of deep dyslexia occurred with damage to virtually any part of the network.     

Visual factors in auditory localization

Previous experiments showing the importance of visual factors in auditory localization are shown to have been insufficiently quantitative.

In the first Experiment, bells were rung and lights shone on the same or different vectors, eleven subjects indicating which bell had rung. In the second Experiment, a puff of steam was seen to issue from a kettle whistle with no whistling sound, while similar whistles were sounded by compressed air on that or another vector. Twenty-one subjects cooperated.

The addition of a visual stimulus at 0° deviation increased the percentage of correct responses significantly in the second, and insignificantly in the first experiment. At 20°-30° deviation the proportion of naive responses to the visual cue was 43 per cent. in the first and 97 per cent, in the second experiment. At greater angular deviations, the proportion of naive responses fell to chance level in the first, but remained significant in the second experiment, even at 90°. The “visuo-auditory threshold” was found to be 20°-30°, but might be much larger if there were more grounds for supposing the two stimuli to be from the same source in space.     

Normative Data for the Boston Naming Test in Native Dutch-Speaking Belgian Elderly

A normative study of the 60-item version of the Boston Naming Test (BNT) was performed in a group of 200 native Dutch-speaking Flemish elderly. Analysis of test results revealed that BNT performance in Dutch is significantly affected by age, years of education, and gender. Error analysis disclosed verbal semantic paraphasias to occur as the most frequent error type (1/3 errors). “Don't know responses,” verbal semantic paraphasias, and adequate circumlocutions were found on at least 30 different BNT items and constituted the most diffusely distributed error types. Following a careful review of other normative BNT studies, group characteristics rather than cultural differences were found to account for the difference in the overall mean scores. Our study surprisingly revealed that, as far as American–English, Australian–English and Dutch-speaking elderly are concerned, linguistics do not have an impact on the overall mean BNT score. A linguistic impact, however, clearly holds on the qualitative levels of performance, reflected by fundamental differences in the error distribution in different languages. Language-related BNT characteristics therefore stress the need for specific adaptations of norms.     

A model-driven analysis of severity, response characteristics, and partial recovery in aphasics' picture naming

Dell, Schwartz, Martin, Saffran, and Gagnon (DSMSG; 1997) presented a computational analysis of aphasic naming that, among other things, purports to explain why some error types correlate with naming severity while others do not. It does so in terms of chance response opportunities, which differ among error types and which come into play particularly when activation levels are small. The present study looks at error frequencies in relation to severity at two points in time: at study entry and after a period of partial recovery. Results support the model's distinction between severity-sensitive errors (nonwords. formal paraphasias, and unrelated errors) and those that are severity insensitive (semantic; mixed). Additionally, we show that the degree of target overlap in nonwords is sensitive to severity but various measures of monitoring and error correction are not. While these results generally support DSMSG, effects at the level of individual patients underscore the difficulties that their model encounters in explaining some pure error dissociations.     

The use of source memory to identify one's own episodic confusion errors

In 4 category cued recall experiments, participants falsely recalled nonlist common members, a semantic confusion error. Errors were more likely if critical nonlist words were presented on an incidental task, causing source memory failures called episodic confusion errors. Participants could better identify the source of falsely recalled words if they had deeply processed the words on the incidental task. For deep but not shallow processing, participants could reliably include or exclude incidentally shown category members in recall. The illusion that critical items actually appeared on categorized lists was diminished but not eradicated when participants identified episodic confusion errors post hoc among their own recalled responses; participants often believed that critical items had been on both the incidental task and the study list. Improved source monitoring can potentially mitigate episodic (but not semantic) confusion errors.     

Simultaneous activation of reading mechanisms: evidence from a case of deep dyslexia

We report the performance of LC, a deep dyslexic. We investigated extensively her errors according to serial cognitive neuropsychological models of oral reading. Initial evaluation of her reading suggested impaired access to the phonological output lexicon (POL). Impaired grapheme-to-phoneme conversion (GPC) and semantic errors in reading suggested that LC read via an impoverished semantic route. However, a serial model of oral reading could not explain error differences in reading, picture naming, spontaneous speech, and repetition. Neologisms occurred in oral reading but not in spontaneous speech and repetition. Semantic errors in naming exceeded those in oral reading. To account for these different error patterns we propose that the semantic route, the direct route from the orthographic input lexicon to the POL, and GPC activate simultaneously during reading, converging at the POL to constrain phonological selection. These routes are modular but not functionally encapsulated. For LC, the POL receives ambiguous information due to degradation of all routes, causing reading errors.     

Variability in the semantic errors produced by brain-injured patients

One of the most common types of errors produced by aphasic patients during oral word production is semantic errors. However, although aphasia semantic errors are often treated as a single homogenous group, there are, in fact, several subtypes defined by the nature of the error-target relationship: paradigmatic, if the two words are category coordinates; syntagmatic, if they are associatively related but from different semantic categories; and superordinate, if the meaning of the error is broader than the meaning of the target. The goal of this study was to investigate whether or not these various subtypes of semantic errors have a similar processing origin. With this objective, we compared the patterns of semantic errors made by a group of Alzheimer patients in a picture-naming task with those made by a group of aphasic patients. We examined the percentages of the different error types, the degree of association between target and error, and the frequency values both of errors and targets. The results suggest that the three subtypes of semantic errors have different origins: the superordinate appear to arise at the semantic level, the syntagmatic at the lexical level, and the paradigmatic at both levels of processing.     

Field articulation and working memory

This experiment assessed the proposal that performance differences associated with field independence/field articulation reflect differential efficiency in the use of a limited capacity information-processing system termed working memory. Extreme scores on a group-administered version of the embedded figures test were used to identify subjects “high” and “low” in degree of field articulation. The task required that a subject retain a series of digits for subsequent report while performing a semantic modification of a target phrase. Concurrent information load on working memory was varied in two ways: by increasing memory load from 3 to 6 digits, and by increasing the complexity of the semantic processing required. In low information load conditions no differences were found between high FA and low FA groups. In high information load conditions low FA subjects made more errors in digit recall and took longer to perform the difficult semantic modification.     

Effects of near and distant semantic neighbors on word production

One way to examine the dynamics of word processing is to investigate how processing is affected by the co-activation of similar words (“neighbors”). A unique prediction of attractor dynamical models is that near neighbors should exert inhibitory effects and distant neighbors should exert facilitative effects. In study 1, data from 62 unselected chronic aphasia patients revealed a higher rate of semantic errors for words with many near semantic neighbors and fewer semantic errors for words with many distant semantic neighbors. In study 2, this basic result was replicated in controls using a speeded picture-naming paradigm. Together, these two studies provide strong new evidence consistent with the attractor dynamics view of neighborhood effects. In addition, analyses of correlations between effect sizes and lesion locations, and comparisons with the existing literature on semantic deficits in aphasia and the speeded picture-naming paradigm, all provide converging evidence that the semantic error patterns found in the present studies were due to disruptions of cognitive control mechanisms.     

Semantic Factors in the Production of Number Agreement

This paper examines the role of semantic factors in the production of subject–verb number agreement. As an ostensibly grammatical process, number agreement provides an interesting case for examining the flow and interaction of semantic and syntactic information through the language-production system. Using a sentence-completion task, agreement errors can be elicited from subjects by presenting them with sentence fragments containing a complex noun-phrase, in which the nonhead noun is plural (e.g., The key to the cabinets...WERE missing.). Previous research has demonstrated that the probability of making an error can be affected by varying the properties of the nouns in the complex noun phrase. By investigating which variables do and do not affect error rates, constraints on the flow of information through the production system can be inferred. In three experiments, we investigated the possible effects of three different semantic manipulations of the nouns in the complex NP: animacy, semantic overlap, and plausibility of modification by the sentence predicate. We found that both animacy and semantic relatedness had reliable effects on error rates, indicating that the mechanism involved in implementing agreement cannot be blind to semantic information. However, the plausibility with which each noun could serve as the subject of the sentence predicate had no effect on error rates. Taken together, these results suggest that while semantic information is visible to the agreement mechanism, there are still constraints on when this information can affect the process. Specifically, it may be the case that only information contained within the complex NP is considered for the purposes of implementing agreement.     

Developmental Deep Dyslexia in Japanese: A Case Study

This report demonstrates the existence of developmental deep dyslexia involving Japanese orthography. When asked to read (or name) isolated kanji, T.S., a sixth grader with a normal IQ and no speech impairment, produced a number of visual, selection, and semantic errors. It is suggested that these errors arise from an interaction between phonological coding impairment and the relative sparing of direct visual processing, which are taken to be characteristic of developmental deep dyslexia. A selection hypothesis is presented to explain selection errors, semantic errors, and some peculiar phenomena observed in acquired dyslexics.     

The learning of surface structure clues to deep structure by a puppet show technique

Fifty-six children were classified in terms of their developmental level on the linguistic structure “John is easy/eager to please”, and an experiment was carried out teaching them “new” (nonsense) words in related transformations by a puppet show technique. These transformations serve as frames for differentiating those new words which indicate a surface structure derived from a transformed deep structure, from those which do not. Results showed that children used different strategies at different developmental levels. Of children who did not use fixed strategies, only children who had correctly performed on all normal adult words (“Passers”) were able to learn which deep structures were called for by some of the new words; all other groups performed at chance levels on all types of new words. However, even these Passers performed correctly only on words given in structures indicating that the tested structure was to be treated as a transformation. The possibility of a language-learning principle related to the marked/unmarked distinction in natural languages is discussed.     

Serial position effects in semantic memory: reconstructing the order of verses of hymns

Serial position effects (primacy and recency) have been consistently demonstrated in both short- and long-term episodic memory tasks. The search for corresponding effects in semantic memory tasks (e.g., reconstructing the order of U.S. presidents) has been confounded by factors such as differential exposure to stimuli. In the present study, the stimuli were six-verse hymns that would have been sung from the first to the last verse by churchgoers on numerous occasions. Participants were presented with the verses of each hymn in random order and were required to reconstruct the correct order. Primacy and recency effects were significantly more evident for churchgoers than for nonchurchgoers. Moreover, error gradients were steeper than chance for churchgoers but not for nonchurchgoers; in other words, churchgoers' errors were more likely to be close to the correct position than further away. These findings provide the first unequivocal demonstration of serial position effects in semantic memory.