This article examines the extent to which lexical bundles (LBs; i.e., frequently recurring strings of words that often span traditional syntactic boundaries) are stored and processed holistically. Three self-paced reading experiments compared sentences containing LBs (e.g., in the middle of the) and matched control sentence fragments (e.g., in the front of the). LBs and sentences containing LBs were read faster than the control sentence fragments in all three experiments. Two follow-up word and sentence recall experiments demonstrated that more sentences containing LBs were correctly We wish to thank R. Harald Baayen for many helpful suggestions and comments as well as to thank several of the participants from the following meetings, where preliminary results of some of this work were discussed: The Symposium on Formulaic Language, held at the University of Tremblay et al.Lexical Bundle Processing remembered. Consistent with construction-type models of language, these results suggest that regular multiword sequences leave memory traces in the brain.

Derivational morphology is one of the most difficult and least studied of all the areas of linguistic description (cf. Lightner, 1968:71). There are two main problems which are largely responsible for this. The first is the question of morpheme recognition or lexical identity: how similar in meaning or in sound do two words have to be in order for the linguist or language learner to identify a common morphemic unit and thus to see a morphological relationship between the words? (This problem is discussed in detail in Derwing, 1973: 122-6.) Many of the morphological rules which are proposed by linguists, whether morphophonemic or phonotactic in presumed character, are posited primarily, if not solely, in order to capture certain kinds of supposed ‘lexical redundancies,’ i.e., systematic variations which appear in the phonological form of the same morpheme when the morpheme occurs in different syntactic constructions. The viability of all such rules is thus directly contingent upon the assumption that the words involved do, in fact, share a common morpheme. Consider, for example, the morphophonemic rule which Chomsky proposes for English which changes a /d/ to an /s/ before the suffix /lv/, and the phonotactic rule which changes a /d/ plus /i/ or /y/ into a /ž/ before a vowel (1964:90); both of these rules are motivated by the presumed fact that the English words decisive and decision, for example, contain in their ‘underlying’ or ‘lexical’ representations the common morpheme decide. But how does one decide whether this claim is justified for ordinary native speakers of the language, particularly in some of the more problematical cases discussed in Derwing (1973)?

Prior research has suggested that Korean might constitute an exception to the proposed notion of the universality of the rhyme. Five experiments were performed to test this hypothesis out with native speakers of that language. Four different experimental tasks were employed: a global sound similarity judgement task, a concept formation task, a unit reduplication task, and a list recall task. In all cases the results indicate that Korean syllables were seen to contain a cohesive CV orbodyunit, in contrast to the VC ofrhymeunit of English. The final experiment, involving list recall, was considered especially important, as it involved the testing of preliterate children, whose results could not have been influenced by knowledge of the orthography. An attempt is made to explain these findings in terms of intrinsic properties of the syllables in the two languages. The theoretical significance of this research is also discussed.

No abstract

A new oral ‘pause-break’ task is described for eliciting judgments about syllable boundaries from speakers of languages of diverse types who have potentially limited educational and/or literacy levels. Results are presented from five different languages, including English. The English results replicate the findings from previous studies using other techniques, while the following generalizations follow from the application of the technique to Arabic, Blackfoot, Korean, and Swiss German: There is (1) a widespread tendency to treat single intervocalic consonants as syllable onsets and (2) an equally broad tendency to break between the two members of intervocalic CC clusters. The treatment of intervocalic geminates or long consonants (as in Arabic and Blackfoot) seems to vary as a function of educational level: In general, the better educated, literate speakers interpret them as disegments and treat them as in (2) above, while the semiliterate subjects often interpret them as single segments and treat them as in (1). The technique itself seems to have proved itself as suitable for use with speakers of a wide range of educational levels and languages.