Journal of Experimental Psychology: Learning, Memory, and Cognition 1988, Vol. 14, No. 1, 145-152 Copyright 1988 by the American Psychological Association, Inc. 0278-7 393/88/500.75 Short-Term Memory Errors for Spoken Syllables Are Affected by the Linguistic Structure of the Syllables Rebecca Treiman and Catalina Danis Wayne State University Three experiments on short-term serial memory for spoken syllables are reported. The stimuli were CVC (consonant-vowel-consonant) syllables in Experiment 1, CCVs in Experiment 2, and VCCs in Experiment 3. Analyses of subjects' errors showed that the phonemes within a syllable were not equally free to break apart and recombine. Certain groups of phonemes—the vowel- final consonant group of a CVC, the initial cluster of a CC V, and a vowel-liquid group within a VCC—tended to behave as units. These results are consistent with the view that syllables are coded in terms of an onset (initial consonant or cluster) and a rime (remainder). Errors in short- term memory for spoken syllables are affected by the linguistic structure of the syllables. Many studies (e.g., Conrad, 1964; Drewnowski & Murdock, 1980) have shown that errors in short-term memory for linguistic materials tend to be phonologically related to the to-be-remembered stimuli. To explore the nature of the pho- nological relations, we asked whether short-term memory errors for spoken syllables are more likely to preserve as units some groups of phonemes than others. For example, suppose the stimuli are CVC (consonant-vowel-consonant) syllables like /ger/. (See Figure 1 for a key to the notation.) If an error retains two phonemes of /ger/, are these phonemes equally likely to be C, and V, C, and C 2 , or V and C 2 ? That is, are some groups of phonemes more cohesive than others in short- term memory? Previous linguistic and psycholinguistic work suggests an answer to the foregoing questions. This work shows that syllables are not simply linear strings of phonemes, but that they have a hierarchical internal structure. The major con- stituents of the syllables are the onset and the rime. The onset is the initial consonant or cluster. The rime is the vowel and any following consonants. Thus, /ger/ has the onset /g/ and the rime /er/. Evidence that syllables contain onset and rime units comes from linguistic constraints on the distributions of phonemes within syllables, errors in the production of speech, and experiments in which subjects divide syllables at different points. (See Treiman, in press, for a review.) If the same constituents are involved in short-term memory, retentions of VC 2 from a CVC should outnumber retentions of C|V and CiC 2 . Such a result would suggest that phonemes do not Portions of the data presented in this article werefirstpresented at the meeling of the Psychonomic Society, San Antonio, Texas, No- vember 1984. This research was supported by National Institute of Child Health and Human Development Grants HD18387 and 202276. Thanks go to Peggy Ericson, Brett Kessler, Ray Kiefer, Jane Ann Riedford, and Tom Rizzo for their assistance. We thank Douglas Medin, Robert Crowder, and two anonymous reviewers for their comments. Correspondence concerning this article should be addressed to Rebecca Treiman, Department of Psychology, Wayne State Univer- sity, 71 West Warren Avenue, Detroit, Michigan 48202. behave as independent units in short-term memory. Rather, syllables are coded in terms of onset and rime units. Preliminary evidence for onset/rime coding was provided by Brady, Shankweiler, and Mann (1983). They reported that children's errors more often preserved the V and final C of a stimulus than the initial C (or CC) and V. We sought to provide further evidence for linguistic units in short-term memory by systematically comparing errors on syllables of different structures. As well as asking whether syllables are coded in terms of onset and rime units, we also asked whether rimes are coded in terms of subunits. Some evidence for subunits of the rime was found in a study by Treiman (1984), in which subjects could divide VCCs after the V or after the first C. VC/C divisions predominated when the postvocalic C was a liquid (/I/ or /r/). When the postvocalic C was a nasal (/m/, /n/, or /rj/), VC/C divisions and V/CC divisions did not differ in frequency. Finally, when an obstruent (like / p / or /$/) fol- lowed the vowel, V/CC divisions tended to outnumber VC/ C divisions. These results were interpreted in terms of the linguistic notion of a sonority hierarchy (Hooper, 1976; Ki- parsky, 1979). Liquids, which are high on the sonority hier- archy or vowel-like, seem to be grouped with the preceding vowel. Nasals, which are intermediate in sonority, are some- what less likely to form a unit with the vowel. Obstruents are low in sonority and have least cohesion with the preceding vowel. To ask whether short-term memory errors differ as a function of the postvocalic consonant, the present experi- ments compared syllables with postvocalic liquids, nasals, and obstruents. In addition to the question of syllable structure, our study examined two subsidiary questions. First, do retained pho- nemes and groups of phonemes tend to maintain their posi- tion within an incorrectly remembered syllable? An affirma- tive answer would suggest that units are somehow "tagged" for their original position within the syllable. Brady et al. (1983) and Ellis (1980) reported maintenance of syllable position; we examined this issue for a wider range of syllable types. A second subsidiary question concerns the degree to which retained phonemes and groups of phonemes preserve 145