A PHONOLOGICAL ACCOUNT FOR THE CROSS-LANGUAGE VARIATION IN WORKING MEMORY PROCESSING.
The positive relationship between short-term recall and word articulation rate has been established as a general phenomenon for English words of various lengths (Baddeley, Thomson, & Buchanan, 1975; Cowan, 1992; Cowan, Day, Saults, Keller, Johnson, & Flores, 1992; Hulme, Maughan, & Brown, 1991; Schweickert & Boruff, 1986; Standing, Bond, Smith, & Isely, 1980). Generally speaking, words that could be uttered fast tend to be recalled better. Emphasizing the predictive role of word articulation rate, some previous authors have used it to explain important variations in short-term recall. For instance, Hulme, Thomson, Muir, and Lawrence (1984) demonstrated that the developmental increase in short-term span from 4 years of age to adulthood could be completely explained by a corresponding increase in word articulation rate. A similar conclusion was reached by Nicolson (1981). Kynette, Kemper, Norman, and Cheung (1990) suggested that the reduced short-term span among older adults could as well be explained by a dec line in word articulation rate.
In a similar vein, word articulation rate might also account for cross-language variations in short-term recall. Ellis and Hennelly (1980) found that English-Welsh bilinguals had larger English digit spans than Welsh digit spans, and this was accompanied by a proportionally faster digit pronunciation rate in English than in Welsh. Hoosain and Salili (1987) and Hoosain (1979, 1982) had Cantonese-English bilinguals recall and articulate both Cantonese and English digits. Digit span was significantly greater in Cantonese than in English, and Cantonese digits were articulated more rapidly than English digits. Similar results were obtained by Stigler, Lee, and Stevenson (1986), who demonstrated that Chinese-speaking children had larger digit spans than their age-matched English-speaking counterparts; Chinese digits were also shown to be articulated more rapidly than English digits.
This consistent recall-articulation rate relationship is generally attributed to the articulatory rehearsal component of working memory (Baddeley et al., 1975), which operates like an audio tape running at a constant speed. A fast articulation rate enables a large amount of information being "recorded" onto this tape within a fixed period of time. Recall is thus enhanced. According to this view, the articulation rates of memory items constitute the dominant factor for recall; a constant recall-articulation rate function across different situations is thus expected. In other words, given the same rate of articulation and therefore the same rehearsal efficiency, very similar recall would result, regardless of the type of material, age of subjects, and so forth.
However, a recent study on the short-term recall and articulation of Cantonese and English words in bilingual subjects suggests an important cross-language difference in short-term recall that is not explained by word articulation rate alone. Cheung and Kemper (1993) had Cantonese-English bilinguals recall and articulate visually presented items from the two languages. For each individual subject, recall was regressed on word articulation rate (words per second), both varying over 1-, 2- and 3- syllable items. An analysis of the regression slopes indicated that the function relating recall to word articulation rate was steeper for Cantonese than English words. This result is at variance with the hypothetical constant recall-articulation rate relationship across languages. According to Cheung and Kemper (1993, 1994), this language effect on regression slopes might have to do with differences in the generic syllable structures of verbal items in the two languages. One-syllable English words commonly contain co nsonant clusters (e.g., "flow," "shelf") whereas one-syllable Cantonese words never do (e.g., "choi," "lei"). The effect of consonant clusters on recall has been shown to be negative, in that words with simple (singleton) onsets were recalled better than words comprising cluster onsets, even with a constant number of phonemes (Caplan & Waters, 1994). Because of their lack of clusters, one-syllable Cantonese words are both articulatorily and acoustically simpler than one-syllable English words. As word length increases, however, the canonical shape of the English syllable approaches a simple consonant-vowel (CV) or consonant-vowel-consonant (CVC) structure (e.g., "cham/ber," "re/vi/sion"), which is similar to the syllable shape of multisyllabic Cantonese words (Clements & Keyser, 1983; Hyman, 1975). As a result, the recall advantage of Cantonese over English words caused by uncomplicated syllable structure decreases with increasing word length. This possibly is responsible for the differential recall-articulat ion rate slopes found in the two languages, with both recall and articulation rates varying over numbers of syllables.
Such experimental results and interpretations give rise to at least two issues. First, a significant part of the hypothetical effect of generic syllable structure must be nonarticulatory, because the language difference in pattern of recall is not totally explained by word articulation rate. In other words, items having a simple syllable structure (one-syllable Cantonese words) tend to be better remembered not only because they take less time to articulate (this would have been taken care of by regressing recall on articulation rate), but also because they leave simpler acoustic trace, probably in a nonarticulatory phonological store that does not entail any articulatory processes (Baddeley & Lewis, 1981; Baddeley, Lewis, & Vallar, 1984; Besner & Davelaar, 1982). Second, the bilingual subjects tested by Cheung and Kemper (1993) were more proficient in Cantonese than English; differential familiarities with the two languages might have contributed to the language effect on regression slopes. Furthermore, any inherent idiosyncratic features associated with the respective languages other than CV structure could have played a role.
The hypothetical phonological basis of the language effect presumes that articulatory processes and nonarticulatory phonological storage constitute two independent aspects of working memory processing. Experimental evidence for the existence of these two functional components has been obtained from studies employing the concurrent articulation task. Concurrent articulation requires the subject to repeat constantly and softly irrelevant syllables (e.g., the digits "1, 2, 3, 4") during both memory list learning and recall. The vocalization of these syllables preempts the speech apparatus and therefore the subject simply cannot articulate the memory items. Hence, any memory effects that remain under concurrent articulation must have a nonarticulatory origin, as opposed to those that are eliminated by the procedure and are thus considered to be articulatorily based. Using auditory presentation, Baddeley et al. (1984) demonstrated that the word length effect (short words being recalled better than long words) was abolished by concurrent articulation during encoding and recall, whereas the phonological similarity effect (lists of different-sounding words being recalled better than lists of similar-sounding words) remained intact. The authors concluded that the word length effect had originated in articulatory rehearsal, for it disappeared completely when rehearsal was blocked by concurrent articulation. The phonological similarity effect, in contrast, had operated within an independent, nonarticulatory store because it was insensitive to the disruption of rehearsal. These findings were replicated by Longoni, Richardson, and Aiello (1993), who also argued for the functional separation of rehearsal from nonarticulatory storage. The behavior of the irrelevant speech effect (recall being disrupted by unattended speech) resembled that of phonological similarity, in that it tended to remain under concurrent articulation with auditory presentation (Hanley & Broadbent, 1987). Hence, the standard interpretation of the irreleva nt speech effect is that it results from phonological confusion between the unattended speech and the target memory items within the nonarticulatory store (Salame & Baddeley, 1982). In summary, the differential interactions between concurrent articulation and the effects of word length, phonological similarity and irrelevant speech using auditory materials do suggest functionally separate components (articulatory vs. nonarticulatory) within the working memory system.
The contention that processes that are insensitive to concurrent articulation must be nonarticulatory obviously requires another assumption, that concurrent articulation effectively eliminates a very significant portion of the articulatory contribution to working memory processing. Some support for this assumption has come from a study by Gupta and MacWhinney (1995), in which concurrent articulation was systematically compared to a set of control tasks. In one of the control tasks, subjects were required to tap the index finger constantly during list encoding and recall while the irrelevant syllable "the" was repeatedly played. The effect of this task was thus comparable to that of concurrent articulation except that it did not demand the use of the articulatory apparatus on the subject's part. Results showed that a significant part of the disruptive effect of concurrent articulation on serial recall was not replicated with this control task, hence suggesting a prominent articulatory component to serial reca ll and that this component was uniquely susceptible to concurrent articulation. Furthermore, the effect of concurrent articulation is not limited to blocking rehearsal. The task appears to be effective in disrupting other articulatorily based processes as well. One such process is articulatory recoding, or the conversion of printed verbal materials into a phonological code via articulation. Baddeley (1990) showed that the effects of phonological similarity and irrelevant speech were blocked by concurrent articulation under visual presentation. Because the result was not found with auditory materials (Baddeley et al., 1984), it was attributed to the disruptive effect of concurrent articulation on recoding, which was simply not required in auditory presentation. This finding further supports the assumption that concurrent articulation has a unique disruptive effect on the major articulatory mechanisms involved in working memory processing.
The foregoing discussion justifies the use of concurrent articulation as a technique to discern the relative roles of articulation and Nonarticulatory phonological storage in cross-language working memory processing. Using bilingual subjects, Experiment 1 of the present study obtained serial recall and word articulation rate data with both Cantonese and English words. Subjects were told to perform the recall task either with or without concurrent articulation. If a significant part of the langauge effect on regression slopes was nonarticulatory, the effect would be observed even under concurrent articulation, which blocked any articulatory processes. Moreover, the articulatory component underlying the language effect would be represented by its interaction with concurrent articulation. Experiment 2 had monolingual English speakers recall and articulate pseudowords preserving the CV structures of the words used in Experiment 1. If the language effect on regression slopes was caused by syllable-CV variations r ather than differential proficiencies in the two languages, the effect would be replicated with the monolingual subjects recalling and articulating the pseudowords. This method also eliminated other alternative explanations for the language effect such as variations in prosodic features and response criteria, because the subjects' monolingual background and the nonlexicality of the pseudowords should equate the two sets of items on every aspect except syllable-CV structure, which was the dimension under manipulation.
Design. Experiment 1 administered an immediate serial recall task and a word articulation task with both English and Cantonese words. For the recall task, a 2 (language: Cantonese and English words) x 3 (word length: 1-, 2-, and 3-syllable words) x 2 (concurrent articulation: with and without) within-subject design was used. Trials were blocked by language and word length; presentation of blocks was counterbalanced. The word articulation task used a 2 x 3 within-subject design with language (English and Cantonese words) and word length (1-, 2-, and 3-syllable words) as the independent factors. Trial presentation was counterbalanced.
Subjects. Subjects were 33 female and 3 male students at the Hong Kong Polytechnic, who were Cantonese-English bilinguals with native Cantonese fluency and English as their second language. At the time of testing, they had been studying English for at least 14 years. None of these subjects had taken the Test of English as a Foreign Language (TOEFL). They, however, were asked to complete a brief personal history which detailed their exposure to and use of English, and to rate their own oral fluency, and listening, reading, and writing abilities in English on a 5-point scale with 1 = very bad and 5 = excellent. A composite score totaling the ratings for the four abilities (maximum = 20) was obtained for each subject. The personal histories showed that this was a homogeneous group in terms of their use of and exposure to English as a second language. The subjects all considered themselves to be fluent users of English. The composite self-rating score ranged from 8 to 18 with a mean of 14.7 (standard deviation = 2.0).
Materials. For immediate serial recall, ten 1-syllable, ten 2-syllable, and ten 3-syllable words in each language were randomly selected within certain frequency ranges. Mean occurrence frequencies of the English words were 35.4, 37.3, and 29.8 per one million words for the 1-, 2-, and 3-syllable items, respectively; mean occurrence frequencies of the Cantonese words were 36.7, 37.0, and 31.5 per one million words for the 1-, 2-, and 3-syllable items, respectively. T tests did not reveal any significant differences in occurrence frequency between any two sets of words (all ps [greater than] .10). Out of the ten 1-syllable English words, seven contained syllables with consonant clusters (e.g., "straw," "flow," "cream"); the figures were down to two and one for 2- and 3-syllable English words, respectively (e.g., "si/lence", "at/trac/tion"). None of the Cantonese items comprised any syllables that were laden with consonant clusters (e.g., "choi," "man/hok," "lap/fong/tie"). The Cantonese words were presented a nd recalled in the traditional Chinese script, which is based on nonalphabetic "characters" with no spellings. The English words were selected from Francis and Kucera (1982), and the Cantonese words were selected from a frequency count developed by the Hong Kong Department of Education (1986). Three word pairs at each word length in each language were selected from the same sources for word articulation. In both languages, the occurrence frequencies of the 1-, 2-, and 3-syllable word pairs were 17, 19, and 21 per one million words, respectively. These words were different from those used in the recall task. The Cantonese items were also printed in Chinese characters.
Procedure. The recall task was always administered before the word articulation task. In the recall task, three 7-word lists were used in each of the 12 experimental conditions. They were three random subsets of the 10 words selected at each word length from each language. Word lists were visually presented and written recall was required. Individual words were printed on file cards; during presentation, the seven cards which made up a list were sequentially displayed at a rate of one word per second. Each card was placed on top of the preceding card in a stack before the subject. A final blank card after the seventh card signaled the end of a list and that the subject had to begin written recall of the list. Subjects were required to recall all the words in the list in the correct serial order. Response forms were provided so that subjects could record their recall. In the concurrent articulation condition, subjects were required to repeat softly the digits "1, 2, 3, 4" during both word presentation and wri tten recall. The digits were always spoken in the same language as the recall list. Subjects learned and recalled the word lists silently in the no concurrent articulation condition. In each of the 12 experimental conditions, three recall scores based on three lists were obtained. Median recall out of the three was taken as the recall score in that particular test condition.
In the word articulation task, three word pairs were used in each language/word-length condition. Each word pair was printed on a file card and visually presented to the subject. The subject was asked to repeat aloud the word pair as quickly as possible for 20 times. Subjects were recorded and reading time for the middle 10 repetitions was measured using a stopwatch. Three measures based on the three word pairs were obtained in each of the six test conditions. Median reading time over the three trials in each condition was converted into a word articulation rate in words per second.
Mean recall (max = 7) and word articulation rates are presented in Table 1.
Recall. The recall data were submitted to a 2 x 3 x 2 within-subject analysis of variance (ANOVA). All main effects were significant: language, F(1, 35) = 52.17, MSE = 1.92, p [less than] .001; word length, F(2, 70) = 29.13, MSE = 1 .59, p [less than] .001; concurrent articulation, F(1, 35) = 41.72, MSE = 1.84, p [less than] .001. Cantonese words were better recalled than English words; short words were better recalled than long words; concurrent articulation reduced recall significantly.
These main effects were qualified by some 2-way interactions: Language x Word Length, F(2, 70) = 13.40, MSE = 1.33, p [less than] .001; Word Length x Concurrent Articulation, F(2, 70) = 4.89, MSE = 1.37, p = .01; Language x Concurrent Articulation, F(1, 35) = 25.55, MSE = 0.98, p [less than] .001. These 2-way interactions could be understood in terms of the significant 3-way interaction, F(2, 70) = 4.45, MSE= 1.19, p [less than] .02. The 3-way interaction indicated that the detrimental effect of increasing word length on recall was more pronounced for Cantonese than English, and this interaction was stronger without concurrent articulation. However, planned analyses revealed that the Language x Word Length interaction was significant both with, F(2, 70) = 15.83, MSE = 1.06, p [less than] .001, and without concurrent articulation, F(2, 70) = 41.19, MSE = 1.25, p [less than] .001. Concurrent articulation reduced, but did not eliminate, the Language x Word Length interaction.
The effect of word length was decomposed into 1- vs. 2- and 2- vs. 3-syllable contrasts; their independent contributions to the Language x Word Length interaction were examined. For the 1- vs. 2-syllable contrast, the Language x Word Length interaction was significant, F(1, 35) = 8.92, MSE = 1.59, p = .005; for the 2- vs. 3-syllable words contrast, the Language x Word Length interaction was not significant, F(1, 35) = 3.56, MSE 1.26, p [greater than] .05. Therefore, increasing word length from 1 to 2 syllables reduced Cantonese recall more than English recall; further lengthening the words from 2 to 3 syllables had no differential effects on recall in the two languages.
Word articulation rate. Word articulation rates were analyzed by a 2 x 3 within-subject ANOVA. The language main effect was significant, F(1, 35) = 78.16, MSE = 0.21, p [less than] .001, as well as the word-length main effect, F(2, 70) = 247.97, MSE = 0.32, p [less than] .001. The 2-way interaction was also significant, F(2, 70) = 12.67, MSE = 0.16, p [less than] .001. Cantonese words were articulated faster than English words; short words were articulated faster than long ones. The decrease in word articulation rate caused by increasing word length was greater for Cantonese than English words. The word length simple effect was significant in both language conditions: English, F(2, 70) = 110.32, MSE = 0.26, p [less than] .001; Cantonese, F(2, 70) = 244.20, MSE = 0.21, p [less than] .001.
Relationship between recall and word articulation rate. For each individual subject in each language/concurrent articulation condition, recall was regressed on word articulation rate, both varying over word lengths (i.e., number of syllables). The resulting regression slopes ([beta]s) were analyzed by a 2 (Language) x 2 (Concurrent Articulation) within-subject ANOVA. With concurrent articulation, mean [beta]s were -0.006 and 0.426 for English and Cantonese words, respectively. Without concurrent articulation, the corresponding figures were 0.251 and 1.223. The language main effect was significant, F(1, 35) = 17.97, MSE = 3.13, p [less than] .001, as well as the concurrent articulation main effect, F(1, 35) = 6.97, MSE = 1.96, p [less than] .02, and the 2-way interaction, F(1, 35) = 4.62, MSE = 1.45, p [less than] .04. The effect of language on [beta]s was significantly reduced, but not eliminated, by concurrent articulation. The language simple effect was significant under concurrent articulation, F(1, 35) = 10.97, MSE = 1.10, p [less than] .003, and without concurrent articulation, F(1, 35) = 14.63, MSE = 3.48, p [less than] .002.
Experiment 1 showed that increasing word length had a greater detrimental effect on Cantonese than English recall in bilinguals. This tendency was significantly reduced, but not abolished, by having subjects repeat irrelevant digits during list learning and recall. The concurrent articulation of digits was supposed to block any articulatory mechanisms; the fact that it reduced the language effect does point to an articulatory component of the effect. However, the fact that it remained under concurrent articulation suggests that a significant part of it must be nonarticulatory. This is further supported by the [beta] (slope) analysis. Regressing recall on word articulation rate varying over word lengths resulted in steeper slopes for Cantonese than English, and this effect held under concurrent articulation. Word articulation rate did not totally explain the cross-language recall difference, as indicated by the discrepancy in slope. Concurrent articulation significantly reduced but did not eliminate the langu age effect, thus suggesting both an articulatory and a nonarticulatory origin.
One likely source of the nonarticulatory effect is a short-term phonological store that is independent of articulatory rehearsal, yet sensitive to the phonological structures of memory items (Baddeley & Lewis, 1981; Besner & Davelaar, 1982; Longoni et al., 1993). The between-language difference in syllable structure was obvious with the 1-syllable items, but it largely disappeared when word length increased to 2 and 3 syllables. This analysis is consistent with the current result that the Language x Word-Length interaction effect on recall was based on the 1-vs. 2-syllable contrast; the 2- vs. 3-syllable contrast contributed little to the interaction. Therefore, increasing word length from 1 to 2 syllables (1) created different additional articulatory demands for items in the two languages, and at the same time (2) produced differentially complex acoustic trace in the nonarticulatory store for items in the two languages. The 1-syllable English item left more complex trace than the 1-syllable Cantonese item b ecause of the former's high probability of containing consonant clusters. This language difference in the likelihood of within-syllable clusters decreased sharply with 2-syllable items and, consequently, the corresponding language difference in memory trace complexity was markedly reduced. Concurrent articulation blocked mechanism (1) but not (2), therefore the language effect remained even when the system was taxed by concurrent articulation.
Nevertheless, it is still unclear whether syllable-CV variation was the critical factor underlying the language effect. The current subjects were more fluent in Cantonese than English, and such imbalance might have contributed to differential processing mechanisms for items in the two languages. Moreover, the English and Cantonese items were presented in different scripts (the Roman alphabet vs. the Chinese character), and such scriptal variation might have induced very different processing strategies for the two sets of items. Possible prosodic and response criterion variations were also not controlled for. Experiment 2 examined these possibilities by using monolingual English speakers and pseudowords preserving the CV structures of the test words used in Experiment 1. All pseudowords were romanized and spelled in English letters. If syllable-CV structure was a critical factor, the same pattern of recall in relation to word articulation rates would be replicated with these pseudowords.
Design. The design was identical to that of Experiment 1, except that the concurrent articulation factor was omitted, resulting in a 2 (language) x 3 (word length) within-subject factorial for both the recall and the articulation task.
Participants. A group of 36 native American English speakers, 29 females and 7 males, participated in the present experiment. They were undergraduates at the University of Kansas, USA, who obtained course credit for their participation.
Materials. Ten pseudowords in each language/word-length condition were constructed for immediate serial recall. These pseudowords preserved the CV structures of the test words used in the preceding experiment. Numbers of items containing "cluster-syllables" were therefore identical to the corresponding numbers in the previous experiment. Ten pseudoEnglish and ten pseudoCantonese items were produced at each word length. Following the same procedure, three pairs of pseudowords were created for each language/word length condition in the articulation task. The pseudoCantonese items were romanized and presented in the form of "pinyin," which is a widely used Cantonese-to-English phonetic transcription system based on the 26 English letters. Spellings of these pseudowords were smoothed so that they resembled real English words. Nevertheless, they preserved strictly the CV structures of their real word counterparts used in the previous experiment. T tests were performed to examine any difference in number of letter s between the pseudoEnglish and the pseudoCantonese items at each word length. No significant differences were found (all ps [greater than] .10). Examples of pseudowords are given in Table 2.
Procedure. The experimental procedure was identical to that in Experiment 1, except that no concurrent articulation was required in the recall task.
Mean recall (max = 7) and pseudoword articulation rates are presented in Table 3.
Recall. Recall data were submitted to a 2 x 3 within-subject ANOVA. The significant effects were: word length, F(2, 70) = 58.02, MSE = 0.65, p [less than] .001; and 2-way, F(2, 70) = 27.56, MSE = 0.36, p [less than] .001. The 2-way interaction suggested that increasing word length had a greater detrimental effect on Cantonese than English pseudoword recall. Additional analyses showed that the Language x Word-Length interaction was based on the 1- vs. 2-syllable contrast. The contrast between 2- and 3-syllable pseudowords contributed little to the interaction: 1- vs. 2-syllable, F(1, 35) = 29.08, MSE = 0.56, p [less than] .001; and 2-vs. 3-syllable, F(1, 35) = 1.18, MSE= 0.36, p [greater than] .20.
Pseudoword articulation rate. Pseudoword articulation rates were analyzed by a 2 x 3 within-subject ANOVA. The word length main effect was significant, F(2, 70) = 441.06, MSE = 0.09, p [less than] .001: Short pseudowords were articulated faster than long pseudowords. The language main effect was not significant, although the 2-way interaction was, F(2, 70) = 42.88, MSE = 0.05, p [less than] .001. Increasing word length had a larger effect on the articulation speed of pseudoCantonese than pseudoEnglish items.
Relationship between recall and articulation rate. For each individual subject, pseudoword recall was regressed on articulation rate in each language condition, both varying over word lengths. A slope ([beta]) was thus derived in each language condition. Mean [beta] for the pseudoCantonese items was 1.260 and that for the pseudoEnglish items was 0.594. [beta]s were submitted to a one-way within-subject ANOVA. The language main effect was significant, F(1, 35) = 28.98, MSE = 0.40, p [less than] .001.
The interaction between language and word length reported in Experiment 1, which points to language-dependent differential reduction in recall caused by increasing word length, was replicated in the present experiment with pseudowords. Furthermore, the effect was primarily caused by comparing 1-syllable to 2-syllable pseudowords. These findings are in harmony with the phonological explanation that the canonical (C)V(C) structure of 1-syllable Cantonese words gives rise to a recall advantage over single syllable English words, which often involve consonant clusters. This recall advantage disappears with multisyllabic words, as the canonical syllabic structures of the two languages become similar. This explains why increasing word length from 1 to 2 syllables reduced recall more for Cantonese than English words, and why increasing word length from 2 to 3 syllables had similar effects for the two languages. Regression slopes relating recall to articulation rates were steeper for Cantonese than English pseudowor ds, thus replicating the finding of Experiment 1. The slope difference indicates that articulation rate is not the only factor for cross-language variations in short-term recall. A nonarticulatory component must be postulated.
One point that is worth noting has to do with the lack of significance of the language main effect on overall recall and word articulation rates. The fact that pseudowords and monolingual English-speaking subjects were used in the present experiment might have contributed to this, in that language proficiency did affect overall recall and articulation speed. Using pseudowords therefore eliminated the language main effects on recall and articulation rates. Nevertheless, a language proficiency explanation for the differential recall-articulation rate relationships in the two languages was rejected because the language main effect on regression slopes remained. In other words, although the bilingual's imbalanced proficiencies could explain why words in the more familiar language are generally recalled better and uttered faster, they cannot possibly account for the fact that recall and articulation speed are differentially related in the respective languages. It is this latter fact upon which the main argument o f the present experiment lies, and it does require a phonological account.
The present study provides evidence for two arguments concerning cross-language variations in working memory processing. First, the language effect on regression slopes has a significant articulatory as well as a nonarticulatory component, because concurrent articulation, which blocked any articulatorily based memory processes, interacted with but did not totally abolish the language effect (Experiment 1). While the rehearsal process should be responsible for the articulatory effect, one likely locus of the nonarticulatory effect is the phonological store, which contains acoustic memory trace left by verbal input (Baddeley & Lewis, 1981). Previous evidence has pointed to the functional separation of articulatory memory processes from the nonarticulatory phonological store (Baddeley, 1990; Baddeley et al., 1984; Longoni et al., 1993). Second, the critical factor for the language effect is phonological in nature, rather than having to do with the bilingual's imbalanced proficiencies in the respective languages . It is because the effect was replicated with pseudowords following the CV structures of previous test words, using monolingual subjects. The possibility that other inherent linguistic features associated with the languages have played a part was also eliminated. Taken together, these arguments and findings suggest the involvement of both articulatorily based mechanisms and nonarticulatory phonological storage in cross-language working memory processing. Both components are highly sensitive to the phonological structures of memory items, as revealed by the fact that differential patterns of recall emerge when items from two languages differing in canonical syllable shape are compared. This interpretation is entirely consistent with Baddeley's (1986) formulation of working memory, the phonological loop component of which contains an articulatory control process and a separate, nonarticulatory phonological store.
We to express our gratitude to Dr. Charles Rice and two anonymous reviewers for their constructive comments on an earlier version of the paper.
BADDELEY, A. D. (1986). Working memory. Oxford: Oxford University Press.
BADDELEY, A. D. (1990). The development of the concept of working memory: Implications and contributions of neuropsychology. In G. Vallar & T. Shallice (Eds.), Neuropsychological impairments of short-term memory. Cambridge, England: Cambridge University Press.
BADDELEY, A. D., & LEWIS, V. J. (1981). Inner active processes in reading: the inner voice, the inner ear and the inner eye. In A. M. Lesgold & C. A. Perfetti (Eds.), Interactive processes in reading. Hillsdale, NJ: Erlbaum.
BADDELEY, A. D., LEWIS, V. J., & VALLAR, G. (1984). Exploring the articulatory loop. Quarterly Journal of Experimental Psychology, 36, 233-252.
BADDELEY, A. D., THOMSON, N., & BUCHANAN, M. (1975). Word length and the structure of short-term memory. Journal of Verbal Learning and Verbal Behavior, 14, 575-589
BESNER, D., & DAVELAAR, E. (1982). Basic processes in reading: Two phonological codes. Canadian Journal of Psychology, 36 (4), 701-711.
CAPLAN, D., & WATERS, G. S. (1994). Articulatory length and phonological similarity in span tasks: A reply to Baddeley and Andrade. The Quarterly Journal of Experimental Psychology, 47A(4), 1055-1062.
CHEUNG, H., & KEMPER, S. (1993). Recall and articulation of English and Chinese words by Chinese-English bilinguals. Memory and Cognition, 21(5), 666-670.
CHEUNG, H., & KEMPER, S. (1994). Recall and articulation of English and Chinese words under memory preload conditions. Language and Speech, 37(2), 147-161.
CLEMENTS, G. N., & KEYSER, S. J. (1983). CV phonology: A generative theory of the syllable. Cambridge, MA: MIT Press.
COWAN, N. (1992). Verbal memory span and the timing of spoken recall. Journal of Memory and Language, 31, 668-684.
COWAN, N., DAY, L., SAULTS, J. S., KELLER, T. A., JOHNSON, T., & FLORES, L. (1992). The role of verbal output time in the effects of word length on immediate memory. Journal of Memory and Language, 31, 1-17.
ELLIS, N. C., & HENNELLY, R. A. (1980). A bilingual word-length effect: Implications for intelligence testing and the relative ease of mental calculation in Welsh and English. British Journal of Psychology, 71, 43-51.
FRANCIS, W. N., & KUCERA, H. (1982). Frequency analysis of English usage. Boston: Houghton Mifflin Company.
GUPTA, P., & MACWHINNEY, B. (1995). Is the articulatory loop articulatory or auditory? Reexamining the effects of concurrent articulation on immediate serial recall. Journal of Memory and Language, 34, 63-88.
HANLEY, R. J., & BROADBENT, C. (1987). The effect of unattended speech on serial recall following auditory presentation. British Journal of Psychology, 78, 287-297.
HONG KONG DEPARTMENT OF EDUCATION. (1986). A study of Chinese words used by Hong Kong junior high school students. HK Department of Education.
HOOSAIN, R. (1979). Forward and backward digit span in the languages of the bilingual. The Journal of Genetic Psychology, 135, 263-268.
HOOSAIN, R. (1982). Correlation between pronunciation speed and digit span size. Perceptual and Motor Skill, 55, 1128.
HOOSAIN, R., & SALILI, F. (1987). Language differences in pronunciation speed for numbers, digit span, and mathematical ability. Psychologia, 30, 34-38.
HULME, C., MAUGHAN, S., & BROWN, G. D. A. (1991). Memory for familiar and unfamiliar words: evidence for a long-term memory contribution to short-term memory span. Journal of Memory and Language, 30, 685-701.
HULME, C., THOMSON, N., MUIR, C., & LAWRENCE, A. (1984). Speech rate and the development of short-term memory span. Journal of Experimental Child Psychology, 38, 241-253.
HYMAN, L. (1975). Phonology: Theory and analysis. New York: Holt, Rinehart, and Winston.
KYNETTE, D., KEMPER, S., NORMAN, S., & CHEUNG, H. (1990). Adults' word recall and word repetition. Experimental Aging Research, 16(3), 117-121.
LONGONI, A. M., RICHARDSON, J. T. E., & AIELLO, A. (1993). Articulatory rehearsal and phonological storage in working memory. Memory and Cognition, 21(1), 11-22.
NICOLSON, R. (1981). The relationship between memory span and processing speed. In M. Friedman, J. P. Das, & N. O'Connor (Eds.), Intelligence and learning (pp. 179-1 84). New York: Plenum Press.
SALAME, P., & BADDELEY, A. (1982). Disruption of short-term memory by unattended speech: Implications for the structure of working memory. Journal of Verbal Learning and Verbal Behavior, 21, 150-164.
SCHWEICKERT, R., & BORUFF, B. (1986). Short-term memory capacity: Magic number or magic spell? Journal of Experimental Psychology: Learning, Memory and Cognition, 12, 419-425.
STANDING, L., BOND, B., SMITH, R, & ISELY, C. (1980). Is the immediate memory span determined by subvocalization rate? British Journal of Psychology, 71, 525-539.
STIGLER, J.W., LEE, S.Y., & STEVENSON, H.W. (1986). Digit memory in Chinese and English: Evidence for a temporally limited store. Cognition, 23, 1-20.
Mean Recall and Articulation Rates in Experiment 1 Number of Syllables 1 2 3 English Recall Concurrent Articulation 314 2.72 2.86 (1.18) (1.21) (1.42) No Concurrent Articulation 3.56 3.19 3.06 (1.48) (1.49) (1.22) Articulation rates 3.57 2.32 1.83 (words per second) (0.87) (0.52) (0.29) Cantonese Recall Concurrent Articulation 3.94 3.36 2.86 (1.22) (1.29) (1.22) No Concurrent Articulation 6.19 4.22 3.72 (1.33) (1.79) (1.32) Articulation rates 4.40 2.94 2.01 (words per second) (0.80) (0.59) (0.35) Note. Standard deviations are given in parentheses; maximum recall = 7. Examples of Pseudowords Used in Experiment 2 and Their Word Counterparts in Emperiment 1 English Cantonese words pseudowords words [*] pseudowords 1-syllable p/ug p/og tong kong coast coyst yee yiu 2-syllable so/lid si/lid man/hok mun/hok sad/dle sot/tle da/lo ba/lo 3-syllable em/pha/sis am/pa/kis fei/ka/yik fa/ka/yik cer/tain/ty kor/ten/ty dan/paak/chud dan/kar/mud Notes. "/" denotes syllable boundary; within-syllable consonant clusters are underlined. (*.)These worlds were presented in Chinese characters to the bilingual subjects in Experient 1. Mean Recall and Articulation Rates in Experiment 2 Number of Syllables 1 2 3 PseudoEnglish Recall 2.47 2.33 1.50 (0.85) (1.07) (0.61) Articulation rates 2.96 1.92 1.53 (pseudowords per second) (0.53) (0.37) (0.26) PseudoCantonese Recall 3.61 1.78 0.97 (1.10) (0.68) (0.51) Articulation rates 3.37 1.97 1.17 (pseudowords per second) (0.53) (0.38) (0.21) Note. Standard deviations are given in parantheses; maximum recall = 7.
|Printer friendly Cite/link Email Feedback|
|Author:||CHEUNG, HIM; KEMPER, SUSAN; LEUNG, E.|
|Publication:||The Psychological Record|
|Date:||Mar 22, 2000|
|Previous Article:||EFFECT OF UNSIGNALED DELAY TO REINFORCEMENT ON WITHIN-SESSION RESPONDING.|
|Next Article:||FIXED-INTERVAL RESPONDING DURING HUMAN COMPUTER-INTERACTIVE PROBLEM SOLVING.|