Embedded Words and Lexical Segmentation in Spoken Language

Introduction to Embedded Word Research

A number of studies completed by researchers concern embedded words and their practicalities and implications in common language usage throughout society. The articles written about those studies present conclusions that are both thought-provoking and enlightening.

In Jeffrey Bowers et al.'s study published in 2004, experiments were categorized by congruent and incongruent classifications, with results showing that responses were slower in the incongruent category. As the authors report, "Responses were slower and less accurate in the incongruent conditions suggesting that subsets and supersets were processed to the level of meaning" (Bowers et al. 2004, p. 131). This finding — that lexical access extends to the meaning of embedded sub-words — set a productive foundation for subsequent research in the field.

Cross-Linguistic Approaches to Speech Segmentation

An earlier study by Anne Cutler showed how language can be compared cross-linguistically, stating that "even universal characteristics of language structure can only be accurately observed cross-linguistically" (Cutler 1997, p. 3). Cutler concluded that there are universal characteristics of language and that they should all be studied, further asserting that "Spoken-language processing by human listeners cannot, in effect, be understood unless it is studied comparatively" (Cutler, p. 15).

If, according to Cutler, spoken languages cannot be understood unless studied comparatively, then experiments studying how embedded words in the English language are processed would be of clear interest to the researcher. Anne Cutler also helped set the stage for word segmentation by showing that both English and other European languages contain a large number of closed-class items recognized as separate words. She notes that most of these words are short but are not orthographically distinct from open-class words, and that "the two groups of words may play different roles in conveying linguistic messages" (Cutler 1992, p. 110). Cutler used visual stimulation to verify her evidence.

Jean Vrooman's study offered a partially contrasting perspective by using the Finnish language rather than English. Vrooman's study concluded that "the present study showed that Finnish word boundaries are signaled by vowel disharmony and word stress" (Vrooman et al. 1998, p. 147), suggesting that the acoustic cues listeners rely upon are at least partly language-specific.

Lexical Competition and Ambiguity in Embedded Words

In his 1997 study, Matt Davis used auditory responses to demonstrate that "one case in which lexical competition has been argued to be especially valuable is in the identification of words that are embedded in the onset of longer words, e.g., cat embedded in catalogue" (Davis et al. 1997, p. 1). This work highlighted how spoken word recognition involves active competition among candidate words during online processing.

In a subsequent 2002 study, Davis et al. continued in the same vein as the earlier 1997 research, finding that "words in fluent speech are not separated by silence in the same way that printed words are...This perceptual experience clearly reflects acquired language-specific knowledge, because listeners do not have the same experience when hearing an unfamiliar language" (Davis et al. 2002, p. 218).

Another 1997 study by Davis appeared somewhat ambiguous regarding the assertion that embedded words were of central importance, and concluded that "Empirical evidence on this issue is inconclusive at present...indeed it may be that providing psychologically realistic accounts of subjects' sensitivity to duration differences in the speech stream presents a substantial challenge to accounts both with and without lexical competition" (Davis et al. 1997, p. 12). This study seemed to follow up on a 1995 study by a different researcher that made a similar observation, though not as completely in context.

In that 1995 study, David W. Gow Jr. observed that in spoken language, as compared to the written word, "words usually abut one another without intervening silences and are often woven together through co-articulation to form a seamless stream of acoustic-phonetic information" (Gow and Gordon 1995, p. 344). Gow used auditory methods to explore how acoustics affect the phonetic flow of information.