به سوی یک مدل قطع لوب فرونتال نارساخوانی عمیق: نقش بازخورد معنایی در حافظه کاذب آواشناسی

We use a false memory paradigm to investigate the extent to which phonologically relations of a target word may be implicated in semantically driven access of that target word. We are specifically interested in understanding how access to orthography may limit the scope of this implication. Our interest in these questions is directly motivated by the claim that the probability of deep dyslexics making a semantic error in reading a word is affected by the number of close phonological relations of the word (Buchanan, L., Hildebrandt, N., & MacKinnon, G. E. (1994). Phonological processing of non-words by a deep dyslexic patient: A rowse is implicitly a rose. Journal of Neurolinguistics, 8, 163–181; Buchanan, L., Hildebrandt, N., & MacKinnon, G. E. (1996). Phonological processing of non-words in deep dyslexia: Typical and independent? Journal of Neurolinguistics, 9, 113–133). We use the study list in a false memory paradigm to semantically prime words. We look for false recognition effects among phonological relations of the semantically primed words in both modalities, with and without simultaneous orthographic overlap. The critical lures of interest in the experiments are words that are related by orthographic or phonological overlap to the unseen semantic target. We found increased false recognition rates for phonological associates of the unseen semantically-primed target in the auditory modality, only among words with no orthographic overlap with the unseen semantically-primed target. The locus of overlap and the phonological neighborhood size of the phonological relation also plays a role in the false memory rate. In a related experiment, we showed that the phonological neighborhood and concreteness of critical lures also mediates the probability of a semantic false memory. These results are discussed with respect to a frontal lobe disconnection theory of deep dyslexia, which posits that semantic errors in deep dyslexia are result from impoverished constraints to prefrontal regions that are implicated in semantic and phonological access of written words.

A fluent reader has the impression, in reading a single word, of simply working on that one word. Psycholinguistic research has demonstrated that a great deal of complex computational effort underlies that impression of focused simplicity. We know that a skilled reader is sensitive to the frequency, regularity, and number of orthographic neighbors (among other variables) of each word, because each of these factors measurably impacts how easily the word can be accessed. We therefore know that a reader of a single word has not only accessed the word, but set into motion a great sweep of spreading activation that has implicated a great many other words as well. This paper focuses on a different aspect of language that may appear equally simple and unitary from a phenomenological perspective, namely, accessing a word from its semantics. If asked to name a four-legged animal that barks and is often kept as a pet, most of us can quickly produce the word dog. In the easiest cases, the impression we have is much like that impression we have in reading a word: we feel that we have plucked only the single word we need directly from a stored lexicon. The work in this paper explores the possibility that the act of selecting a word using semantic input is much like selecting a word using visual input: underlain by a great sweep of spreading activation that implicates many other words en route to the target word. In particular, our interest lies in investigating the extent to which phonologically related words of a target word may be implicated in semantically driven access of that word. We are also specifically interested in understanding how access to orthography may limit the scope of this implication. Our interest in these questions is directly motivated by the hypothesis that the probability of making a semantic error in reading a word is affected by the number of close phonological relations of the word. This hypothesis derives from work with deep dyslexia. Deep dyslexia occurs in previously literate adults following brain damage. Patients with the syndrome show numerous difficulties in single word reading, exhibiting visual/phonological errors, derivational errors, and a profound difficulty in sounding out non-words. However, the defining feature of deep dyslexia is the production of semantic errors during word reading (e.g. reading leg as foot). Such errors are produced by deep dyslexics without awareness. Buchanan et al., 1994 and Buchanan et al., 1996 proposed that the reading deficits characteristic of deep dyslexia result from a selection impairment in the phonological output lexicon, where phonological information is assembled during lexical access. In order to test this hypothesis, they manipulated the number of phonological neighbors in a word reading task. A phonological neighbor is a word that differs by a single phoneme from a target word. For example, words pot and kit are both phonological neighbors of cot. The phonological neighborhood of a word (PN) is the number of words that differ from a target word by exactly one phoneme. PN is used as a measure of phonological representativeness. Words with a large PN are, by definition, very similar phonologically to many other words. Words with a small PN are phonologically more distinct. Buchanan et al. showed that deep dyslexics tended to produce more semantic errors in reading words with large rather than small PNs. They also showed there was no similar effect for the number of orthographic neighbors. The experiments reported here constitute an attempt to find a normal model of this sensitivity of semantic access to PN, and to study in more detail whether it may be modulated by modality (visual versus auditory), position of overlap (early versus late) between the semantic word and its phonological neighbors, and nature of overlap (purely phonological versus phonological and orthographic). We propose a neurologically-motivated hypothesis—frontal lobe disconnection theory of deep dyslexia—for the locus of the semantics/phonological N interaction effect. As an initial test of the hypothesis, we conducted an experiment to see if there is a concreteness effect in the sensitivity of semantic access. Although no direct demonstration of this specific sensitivity of semantics to PN has previously been shown in normals, there has been a great deal of experimental work examining the closely related phenomenon of an interaction between semantics and phonology. Some of this work has examined the interaction by manipulating semantic variables. Strain, Patterson, and Seidenberg (1995) showed that reading times were slower for exception words with abstract meanings than for either exception words with concrete meanings or abstract words with regular spellings—a semantic variable impacting on the orthography-to-phonology decoding process. Hino and Lupker (1996); see also (Pexman & Lupker, 1999) reported a polysemy effect (easier lexical access in lexical decision and naming for words with more than one meaning), which they attributed to enhanced feedback from semantics to orthography for polysemous words. Related work has looked at ambiguous words, which produce more semantic activation than nonambiguous words by activating multiple meanings. Hino, Lupker, and Pexman (2002) manipulated synonymy and task to demonstrate semantic interaction effects in lexical decision, naming, and semantic categorization. They showed that ambiguity was a disadvantage when the task was not semantic, allowing feedback to play a role. In contrast, ambiguity did not play a role when the task was semantic. In another related set of experiments by the same authors, Pexman, Lupker, and Hino (2002) manipulated the number of semantic features of non-polysemous concrete words. They demonstrated that words with many semantic features were accessed more easily, using lexical decision and naming tasks, than words with fewer semantic features, and also showed (using pseudohomophone stimuli) that sensitivity to the number of semantic features was greater when a task required greater phonological processing. Other work is more directly relevant to the present work, because it has demonstrated activation spreading to words that were never presented, instead of demonstrating interactivity between phonological and semantic variables associated with presented words. These experiments have used semantically mediated priming of a word's phonology. O'Seaghda and Marin (1997) found a weak facilitatory effect in word naming (described as ‘slender but real’) following semantically mediated auditory phonological priming. For example, a word like night has a small priming effect on the word dare, due to the overlap of the onset between the target dare and the word day, a close semantic associate of the prime night. O'Seaghda and Marin noted that the effect was usually not significant by items, a finding that weakens the conclusion that this effect is capturing something essential about the organization of the mental lexicon. In a series of follow-up experiments, Farrar, Van Orden and Hamouz (2001) extended these findings. They demonstrated that O'Seaghda and Marin's semantically-mediated phonological priming effect in word naming might be explained by inconsistency between the orthographic body and its pronunciation (i.e. by the way the orthographic representation maps onto the phonological rime). This explanation is consistent with the predictions of interactive activation models. According to such models, words with inconsistent phonology cohere more slowly. Farrar et al. hypothesized and demonstrated that this makes phonologically inconsistent words better items for use in semantically-mediated phonological priming experiments. This finding is also consistent with Strain, Patterson, and Seidenberg's (1995) demonstration that words with an inconsistent spelling-to-sound mapping were most sensitive to variations in semantics in word naming. This inconsistency results in slow phonological coherence (and resultant slow time to access; Chateau and Jared, 2003, Jared, 1997, Norris and Brown, 1985 and Seidenberg, 1985), thereby allowing more time for semantic constraints to play a role. In English, body-rime inconsistency is of particular importance because the rime is an important feature in phonological assembly (Patterson & Morton, 1985) and because in English many spelling bodies are ambiguous (Stone, Vanhoy, & van Orden, 1997). Treiman, Mullennix, Bijeljac-Babic, and Richmond-Wiley (1995) showed that naming RTs for low frequency words are significantly correlated with pronunciation consistency measures of the rime.