Articulatory loop explanations of memory span

In recent years, a number of memory span findings have been attributed to the operation of an articulatory loop (Baddeley & Hitch, 1974). These attributions have been made on the basis of finding a correspondence between span differences and pronunciation rate differences. This experiment explored articulatory loop explanations for two material effects in memory span: the word-frequency effect (span for high-frequency words is larger than span for low-frequency words) and the word-class effect (span for function words is smaller than span for either nouns or adjectives). The results indicate that it is possible to obtain span differences without finding corresponding pronunciation rate differences. Moreover, span differences were as pronounced under articulatory suppression conditions as they were under rehearsal conditions. Both of these results limit the generality of articulatory loop explanations of memory span. This is the author post-print of: Tehan, Gerald and Humphreys, Michael S. (1988) Articulatory loop explanations of memory span and pronunciation rate correspondences: a cautionary note. Bulletin of the Psychonomic Society, 26 . pp. 293-296. ISSN 0090-5054 Explanations for performance on the memory span task have varied greatly over the years; however, recent evidence indicates that one of the best predictors of memory span is the time taken to pronounce the to-be-remembered items. Data from a number of studies suggest that span is equivalent to the number of items that can be pronounced in about 2 sec (Baddeley, Thomson, & Buchanan, 1975; Schweickert & Boruff, 1986; Standing, Bond, Smith, & Isley, 1980). The relationship between span and pronunciation rate was first noted by Baddeley et al. (1975). In a series of experiments, Baddeley et al. found that short words were better remembered than long words in the memory span task. Word length was measured in a couple of different ways: (1) the syllables in the word were counted, and (2) in the case where words had been matched for frequency, number of syllables, and number of phonemes, the length of pronunciation was estimated by measuring the waveform produced by the pronunciation of each to-be-remembered word. Baddeley et al. argued that this word-length effect was consistent with the operation of an articulatory loop. The articulatory loop had previously been proposed as the mechanism in working memory that underlies performance on the span task (Baddeley & Hitch, 1974). It was seen as a phonological store that was driven by an articulatory control process (Baddeley, 1986). In the span task, for visual presentation, items are registered on the loop by way of subvocalization. The items in the loop decay rapidly, but can be refreshed by rehearsal. Thus span recall is a function of rehearsal and decay rates, such that the more that can be rehearsed in a given period of time, the more will be remembered. The word-length effect is due, then, to the fact that more short words than long words can be rehearsed in any given unit of time. Additional confirming evidence for the operation of a rehearsal-driven articulatory loop is obtained when rehearsal is prevented by means of articulatory suppression. Under suppression conditions, at least for visual presentation, the word-length effect disappears. Furthermore, the concept of a rehearsal-driven articulatory loop accounted for individual differences in span, in that measures of rehearsal speed correlated quite highly with span scores. In the present experiment two measures of rehearsal speed were used: (1) an articulation measure, in which 3 words were rehearsed 10 times as quickly as possible, and (2) a reading measure, in which 50 words were read as quickly as possible (Baddeley et al., 1975). In addition to the word-length effect, there is evidence that differences in span for a variety of materials covaries with rehearsal rates (pronunciation rates) for those materials. Watkins (1977) demonstrated a difference in span for highand lowfrequency words, whereas Wright (1979) showed a corresponding difference in the time taken to read 25-item lists of highand low-frequency words. Schweickert and Boruff (1986) and Standing et al. (1980) showed that memory span for a number of different materials (e.g., digits vs. letters vs. words) covaries with pronunciation rates for these materials. Differences in pronunciation rates also have accounted for crosscultural differences in memory span. Ellis and Hennelly (1980) explained the difference in digit span between bilingual Welsh and English speakers in terms of differential pronunciation rates for the two types of material. Stigler, Lee, and Stevenson (1986) found the same result in comparing bilingual Chinese and English speakers. Between-subject differences in memory span for digits in English, Spanish, Hebrew, and Arabic have been shown to be correlated with the average number of syllables in the number words of the respective languages (Naveh-Benjamin & Ayers, 1986). These correspondences between memory span and pronunciation rate have generally been interpreted as evidence for the involvement of the articulatory loop (for an exception, see Schweikert & Boruff, 1986). However, one piece of evidence was briefly reported that may not be consistent with articulatory loop explanations of memory span. Humphreys, Lynch, Revelle, and Hall (1983) reported a study in which it was found that children's span for function words (articles, prepositions, and conjunctions) was poor relative to span for other grammatical classes. This is odd, since these function words tend to be used very frequently in the language and thus should be particularly well recalled (Watkins, 1977). In addition, there seems to be little reason to suppose that function words should be rehearsed or read more slowly than the other word classes. The present experiment had two aims. First, we sought to confirm that the word-class effect could be found with an adult population. Second, we sought to establish how consistent both the word-class effect and the word-frequency effect were with articulatory loop explanations of memory span. The subjects in the experiment studied lists of high-frequency and low-frequency nouns, adjectives, and function words for immediate serial recall. These lists were studied under conditions in which the subject was free to rehearse and under conditions in which rehearsal was suppressed. Span estimates and rehearsal time measures were taken for the different materials. In this experiment, the six different types of words were matched for spoken duration, using procedures similar to those used by Baddeley et al. (1975). If memory span is purely a matter of word length, matching the various materials for spoken duration should result in equivalent span estimates for the different materials, since the same number of items in each case should fit on the articulatory loop. However, Baddeley and Lewis (1984) suggested that the best determinant of span is the functional rate of rehearsal. Thus, even if words are matched for spoken duration, span differences could still be present if one set of material is rehearsed faster than another. Geffen and Luszcz (1983) showed that this is the case for word frequency. Even when highfrequency and low-frequency words have the same spoken duration, high-frequency words can be read more quickly than low-frequency words. One final prediction can be derived from the concept of the articulatory loop. If corresponding rehearsal and span differences emerge under conditions in which the subject is free to rehearse, the articulatory loop hypothesis would predict that these differences in span should disappear once rehearsal has been suppressed, provided that the stimuli were presented visually (Baddeley et al., 1975). Failure to find rehearsal differences and corresponding span differences or the failure of span differences to disappear under suppression would severely limit the generality of the concept of the articulatory loop. It would also raise doubt as to the source of the other span differences mentioned.