Marilyn C. Smith

Tracing the time course of picture--word processing.

A number of independent lines of research have suggested that semantic and articulatory information become available differentially from pictures and words. The first of the experiments reported here sought to clarify the time course by which information about pictures and words becomes available by considering the pattern of interference generated when incongruent pictures and words are presented simultaneously in a Stroop-like situation. Previous investigators report that picture naming is easily disrupted by the presence of a distracting word but that word naming is relatively immune to interference from an incongruent picture. Under the assumption that information available from a completed process may disrupt an ongoing process, these results suggest that words access articulatory information more rapidly than do pictures. Experiment 1 extended this paradigm by requiring subjects to verify the category of the target stimulus. In accordance with the hypothesis that picture access the semantic code more rapidly than words, there was a reversal in the interference pattern: Word categorization suffered considerable disruption, whereas picture categorization was minimally affected by the presence of an incongruent word. Experiment 2 sought to further test the hypothesis that access to semantic and articulatory codes is different for pictures and words by examining memory for those items following naming or categorization. Categorized words were better recognized than named words, whereas the reverse was true for pictures, a result which suggests that picture naming involves more extensive processing than picture categorization. Experiment 3 replicated this result under conditions in which viewing time was held constant. The last experiment extended the investigation of memory differences to a situation in which subjects were required to generate the superordinate category name. Here, memory for categorized pictures was as good as memory for named pictures. Category generation also influenced memory for words, memory performance being superior to that following a yes--no verification of category membership. These experiments suggest a model of information access whereby pictures access semantic information were readily than name information, with the reverse being true for words. Memory for both pictures and words was a function of the amount of processing required to access a particular type of information as well as the extent of response differentiation necessitated by the task.

The bilingual lexicon: Language-specific units in an integrated network

Five experiments were conducted to evaluate hypotheses concerning lexical organization in bilinguals. Previous work has shown that the repetition effect in lexical decision is restricted to intralingual conditions for English and Hindi. Experiments 1–3 each involved 12 English—French bilinguals and demonstrated that a previous finding generalizes to more similar languages and orthographies (English and French) and that it is unaffected by mixed-language test conditions. The results also demonstrated that although interlingual transfer occurs in response to translative activity during encoding, transfer does not occur as a result of tasks which emphasize meaning rather than translation. Experiment 4 involved 12 English—Hindi bilinguals in a two-word lexical decision task with semantically related and unrelated words under pure and mixed-language conditions. The results confirmed Meyer and Ruddys (1974, Bilingual word recognition: Organization and retrieval of a alternative lexical codes) unpublished report that semantic priming is present under mixed as well as pure language conditions, and further showed that when language is defined by orthography—as is the case for English and Hindi—there is no overall deficit for mixed-language conditions. Experiment 5 demonstrated that for interlingual semantic priming to occur under successive presentation conditions, the related concepts must follow one another immediately. The overall pattern of results suggests that although the unit of lexical representation in bilinguals is language specific, the units function in an integrated network.

The relationship between contextual facilitation and depth of processing

Two experiments examined how depth of processing on a contextual priming item affects (a) the amount of priming obtained in the processing of a target item in a lexical decision task (LDT) and (b) subsequent episodic memory for the prime and target items. Experiment 1, in which prime and target items were presented sequentially, yielded three main results: (a) The magnitude of the priming effect, measured by the difference between lexical decision times to word targets preceded by related and unrelated primes, increased as the depth of prime processing increased. (b) In an unexpected postsession recognition test, episodic memory for a prime was dependent on the depth to which it had been processed, whereas memory for a target was unaffected by the depth of processing that had occurred on its prime. (c) Episodic memory for both primes and targets was greater when they had appeared in related pairs rather than unrelated pairs in the LDT. However, unlike immediate contextual priming, the magnitude of the semantic relatedness effect in episodic recognition was not affected by level of processing. In Experiment 2 a two-word LDT was used in which a yes response was made only if two simultaneously presented letter strings were both words. Depth of processing was varied by using different types of nonword distractors: pronounceable nonwords, random letter strings, or strings of Xs. As in Experiment 1, priming was greater and episodic recognition was better, the deeper the level of processing that occurred in the LDT. Similarly, episodic recognition memory was greater for items that had appeared in related pairs in the LDT. Although these data suggest that similar processes modulate both contextual priming effects and episodic recognition, the dissociation in some conditions between the occurrence of contextual priming in the LDT and later relatedness effects in episodic recognition indicate that the underlying mechanisms are not identical.

Modality effects in word identification.

An experiment was designed to investigate the locus of persistence of information about presentation modality for verbal stimuli. Twenty-four Ss were presented with a continuous series of 672 letter sequences for word/nonword categorization. The sequences were divided equally between words and nonwords, and each item was presented twice in the series, either in the same or in a different modality. Repetition facilitation, the advantage resulting from a second presentation, was greatest in the intramodality conditions for both words (+re responses) and nonwords (−ve responses). Facilitation in these conditions declined from 170 msec at Lag 0 (4 sec) to approximately 40 msec at Lag 63. Facilitation was reduced in the cross-modality condition for words and was absent from the cross-modality condition for nonwords. The modality-specific component of the repetition effect found in the word/nonword categorization paradigm may be attributed to persistence in the nonlexical, as distinct from lexical, component of the word categorization process.

Models of visual word recognition : When obscuring the stimulus yields a clearer view

It is well known that visual word recognition is influenced by context, word frequency, and stimulus quality. A processing account is outlined in which stimulus quality affects the orthographic input lexicon, whereas context influences both the orthographic input lexicon and the semantic system. Word frequency exerts its primary effects on the pathways that link lexical systems with each other and with the semantic system. Previous findings that are problematic for alternative models along with the results of two new experiments are consistent with this account

Auditory and visual word-stem completion: Separating data-driven and conceptually driven processes

Two experiments investigated the contributions of data-driven and conceptually driven processing on an implicit word-stem completion task. In Experiment 1, individual words were studied either visually or auditorily and were tested using either visual or auditory word-stems. Keeping modality the same from study to test led to more priming than did changing modality, but there was reliable cross-modal priming. In Experiment 2, subjects read sentences like The boat travelled underwater and inferred the subject noun (i.e. “submarine”) or sentences like The submarine travelled underwater and categorized the subject noun (i.e. “boat”). At test, there was reliable priming for both actually read nouns and inferred nouns. In addition, a modality effect was evident for the actually read nouns but not for the inferred nouns. Taken together, these results imply that there is a basic conceptually driven contribution to priming plus an additional contribution of data-driven processing when surface form is the same at study and test.

Contextual Facilitation in a Letter Search Task Depends on How the Prime Is Processed.

In a variety of situations it has been observed that the processing of a verbal stimulus is facilitated when it is preceded by an associated word. This article is concerned with determining whether such facilitation occurs automatically upon prime presentation or whether facilitation depends on the manner of processing the prime. Experiment 1 demonstrated that a letter search in a target word was facilitated when the target was preceded by either an identical or semantically associated word. If, however, a letter search was required in the prime as well as in the target (Experiment 2), the relative advantage enjoyed by targets preceded by identical-word primes disappeared. Experiment 3 replicated this loss of facilitation using semantically associated word pairs. Contextual facilitation thus appears to depend upon the mode of analysis of the prime. If the prime is analyzed as a meaningful unit, facilitation occurs. If, however, it is subjected to a more discrete, letter-by-letter analysis, the priming effect vanishes.

The effect of varying information on the psychological refractory period

Abstract The two main theories of the psychological refractory period, single-channel theory and expectancy theory, make different predictions about the expected delay in RT2 as a function of the time required to select the first response. According to a single-channel theory, delays in RT2 should be proportional to the time required for RT1, while an expectancy theory predicts that RT2 is influenced only by range of ISIs employed, and should not change with changes in RT1. By varying the number of alternatives associated with the first response, it was found that delays in a two-choice verbal RT2 increased with increasing number of alternatives of RT1, supporting the prediction of the single-channel model. In a second experiment the influence of varying information of RT2 on the latency of RT1 was examined and it was found that latency of RT1 increased as the number of alternatives of the second response was increased. A third experiment was therefore done to determine whether the delay in RT1 was due merely to a change in set, or to an increase in amount of attention required to keep the second signal in store. It was found that corresponding increases were not found on trials where S was expecting the second signal but it was not presented, thus suggesting that some channel capacity is required for the maintenance of the second signal.

Basic processes in reading: Is the orthographic depth hypothesis sinking?

Publisher Summary This chapter discusses oral reading in alphabetic English, a deep orthography; and the orthographic depth hypothesis and the evidence that purports to support it. Orthographies may be defined as either “shallow” or “deep,” depending on the ease of predicting the pronunciation of a word from its spelling. In shallow orthographies, the spelling-sound correspondence is direct: given the rules, anyone can immediately “name” the words correctly. In contrast, in deep orthographies the relationship is less direct, and readers must learn the arbitrary or unusual pronunciations of irregular words such as “yacht.” A consequence of this linguistic difference between deep and shallow orthographies is that it is often assumed that the oral reading of shallow orthographies is qualitatively different from the oral reading of deep orthographies. Old and new evidence from studies of Persian, Spanish, Dutch, Italian, and Croatian, which appears to undermine the essential tenents of the orthographic depth hypothesis is reviewed and assessed in the chapter. The orthographic depth hypothesis in its strong form makes a very simple claim: there is no orthographic input lexicon in the minds of readers processing orthographies, which consist entirely of words with consistent spelling-sound correspondences. The argument is that orthographic access to semantics and the direct mapping from orthographic input lexicon to phonological output lexicon only exists in scripts with inconsistent spelling-sound correspondences, and does so precisely because of this inconsistency.

Semantic priming in visual word recognition: Activation blocking and domains of processing

The fact that letter search on a prime eliminates the typically robust semantic priming effect in lexical decision is often attributed to the “shallowness” of the prime-processing task. In three experiments we investigated this claim by using two different “shallow” prime-processing tasks: letter search and color identification. Consistent with previous reports, lexical decisions to semantically related targets were not facilitated when subjects searched the prime for a probe letter. In contrast, semantic priming was observed following a color discrimination task on the prime. We suggest that a levels-of-processing interpretation is not an adequate framework for understanding these data. Instead, a domain-specific processing account is offered in which explicit processing at the letter level (as in letter search) makes demands on resources (e.g., activation) that drives processing at the semantic level. This competition is resolved by establishing a temporary activation block at the lexical-semantic interface, which results in the elimination or attenuation of semantic priming. In contrast, global judgment of color is viewed as a domain that does not make demands on the resources that drive the visual word recognition machinery. There is therefore no need for an activation block, and semantic priming is not prevented.