Edward E. Smith

Semantic Distance and the Verification of Semantic Relations.

Four experiments dealt with the verification of semantic relations. In Experiment I, subjects decided whether an instance was a member of a specified category. For some categories (for example, birds) verification was faster when the target category was a direct superordinate (bird) than a higher level superordinate (animal), while for another category (mammal) this finding reversed. Experiment II obtained ratings of semantic distance that accounted for the previously obtained verification results. Multidimensional scaling of the ratings suggested that semantic distance could be represented as Euclidean distance in a semantic space. Experiments III and IV indicated that semantic distance could predict RTs in another categorization task and choices in an analogies task. These results place constraints on a theory of semantic memory.

The role of syllables in perceptual processing

Abstract Three experiments examined tachistoscopic reports for numbers and words as a function of the number of syllables in the item. Experiment I showed that the number of syllables in the vocalization of a two-digit number had no effect on report accuracy, indicating that tachistoscopic accuracy is unaffected by the duration of any implicit speech process. Experiment II showed that tachistoscopic accuracy was greater for one- than for two-syllable words, and it was hypothesized that this difference was due to the syllable functioning as a single perceptual unit. Subsequent probability analyses indicated that one-syllable words were indeed processed in a more unitary manner than two-syllable words, but suggested that the critical perceptual unit was a vocalic center group (Hansen & Rodgers, 1965) rather than a syllable. Experiment III further indicated that the difference between one- and two-syllable words was a perceptual one. The results were discussed in terms of processing stage models of word perception.

Fact retrieval and the paradox of interference

Many theories of memory incorporate the notion of retrieval interference, which says that facts learned about the same concept can interfere with one another during retrieval. This notion implies that the more facts learned about a concept the greater the retrieval interference. But this makes it paradoxical how one ever becomes knowledgeable about a topic. To resolve this paradox, we argue that one can overcome interference by using relevant world knowledge to integrate various facts learned about a concept. Three experiments demonstrate that integration can overcome interference. All used a recognition paradigm developed by Anderson and Bower 1973. Subjects first memorized a set of facts about particular people (e.g., Marly, the banker), and then decided whether or not test sentences were from the memorized set. Experiments 1 and 2 manipulated (a) the number of facts learned about a particular person and (b) whether or not they were integrated. With unintegrated facts, the time to recognize a test sentence increased with the number of facts learned about the person mentioned in that sentence; with integrated facts, no such increase was found. Experiment 3 further showed that subjects had difficulty rejecting distractors that were thematically related to integrated facts. To account for these effects, two hypotheses were considered, one an alteration of Anderson and Bowers HAM model, and the other an elaboration of Schank and Abelsons 1977 script analysis.

The role of orthographic and phonotactic rules in perceiving letter patterns.

Three experiments examined the role of orthographic and phonotactic rules in the tachistoscopic recognition of letter strings. Experiment 1 showed that the presence of a vowel or multiletter spelling patterns facilitates perceptual accuracy. To account for these results a model was proposed in which an input string is first parsed into syllablelike units, which are then recorded into speech. It was demonstrated that the perceptual accuracy for a string is correlated with the number of recoding steps needed to convert that string into speech. Experiment 2 further demonstrated that this recoding process can predict perceptibility differences among strings with varying numbers of phonotactic violations, and Experiment 3 assessed some of the specific assumptions of the recoding process.

The role of familiarity in determining typicality

Ashcraft (1978b) found that people tend to know more properties of instances they rate as typical of a category than of instances they rate as atypical. This suggests that variations in typicality result from variations in familiarity. Three experiments are presented that challenge or qualify this suggestion. Experiment 1 showed that subjects sometimes produce more properties for items they rate as low in typicality. Experiment 2 showed that in a large, random sample of items, there was a tendency to produce fewer properties for atypical items, but Experiment 3 indicated that part of the reason for this result was a response bias to assign low typicality ratings to unfamiliar words, rather than a reflection of low perceived typicality of the referents themselves. These results suggest that variations in typicality can exist independent of variations in familiarity, although familiarity may also play a role.

Semantic memory and psychological semantics

Publisher Summary This chapter presents a theoretical approach to semantic memory, which is applicable to a wide range of semantic phenomena. The chapter discusses the topic of semantic memory in psycholinguistic perspective and then demonstrates how semantic propositions are verified. A semantic feature representation is first assumed, which distinguishes between defining and characteristic features. This representation is then coupled with a two-stage processing model, and then the resulting Feature Comparison model is applied to the results of studies requiring the verification of simple subset statements. This model offers an explicit explanation of semantic relatedness and category size effects in this paradigm. The Feature Comparison model is then extended to accommodate findings from recent Same-Different experiments. The extended model proves capable of encompassing a range of semantic relatedness findings, including some newly reported effects which seem problematic for other models. The basic Feature Comparison model is applied for the verification of simple property statements. While the representation of property information necessitates several new structural and processing considerations, the basic model provides an explanation of various semantic effects on verification.

Units of word recognition: Evidence for the use of multiple units

The present experiments were concerned with subsyllabic units in word recognition. The studies employed a same-different RT task. A one-syllable word and a probe were presented simultaneously, and subjects responded “same” if the probe was an integral subset of the word (BLAST:BL), and “different” otherwise (BLAST:BC). Probes included single letters (BLAST:L), doublets (BLAST:LA), triplets (BLAST:LAS), and the whole word (BLAST: BLAST). The first experiment suggested that a variety of spelling patterns, as well as the whole word, served as units of processing. In particular, initial consonant clusters were processed as quickly as single letters, and only slightly faster than the whole words. Consonant strings were used in the second experiment, which demonstrated spelling pattern effects with permissible strings (BLGST:BL) as compared to nonpermissable strings (BXJPK:BX). A third experiment demonstrated that the particular effective units depend on task demands.

Imaginal versus verbal coding and the primary-secondary memory distinction

This experiment tested the hypothesis that the beneficial effect of imagery over repetition instructions on paired-associate learning is due to a facilitation of secondary memory (SM), but not of primary memory (PM). On each trial S was presented with a list of 13 paired-associates and then probed, with the first member of one of the pairs, for the second member of that pair. Half the Ss performed under imagery instructions, and half under repetition instructions. Imagery instructions clearly benefited recall from SM, but not from PM. When the data were corrected for the contribution of SM by the Waugh and Norman (1965) model, imagery instructions led to poorer recall from PM than repetition instructions did. These results place certain constraints on the nature of the PM and SM representations.

Cross-category structure in semantic memory

In Experiment I, subjects made similarity judgments about all 56 category terms listed in the Battig and Montague (1969) norms. These judgments were then subjected to a hierarchical clustering analysis. Experiment II demonstrated that the relations among the category labels are very similar to the relations among the high dominance exemplars of these categories. Experiment III showed that the distances between the category terms in the hierarchical clustering analysis could predict RTs in a same-different paradigm.

When are semantic targets detected faster than visual or acoustic ones

Three experiments explored G. Cohen’s (1970) finding that semantic targets can be detected faster than visual or acoustic ones when searching through meaningful prose. In all three experiments, the subject searched a sentence for a target defined visually, acoustically, or semantically. In Experiment I, the subject detected semantic targets faster than visual- or acoustic-syllable targets, regardless of whether he was instructed to read the sentence for meaning or simply scan it. Experiments II and III showed that this advantage of semantic targets over visual or acoustic ones was greatly reduced when all three types of targets were words. The results were interpreted as indicating that when words are presented in context, either: (a) the unit of word perception is the entire word, or (b) the unit of identification or awareness is at least as large as the entire word.