A growing body of evidence suggests that a hierarchy of word‐based constituents influences the phonetic shape of utterances. These constituents include word‐sized and larger constituents, e.g. prosodic words, clitic groups, phonological phrases, and intonational phrases. Previous experiments have shown that word‐rhyme durations in phrasally‐stressed words depend on the number of syllables in the word (e.g. –un in tuna shorter than –un in tune), and possibly larger units (Huggins 1975, Beckman & Edwards 1990, Turk & Shattuck‐Hufnagel 2000). Polysyllabic shortening is one of the mechanisms proposed to account for shorter durations in constituents with more syllables.
In this paper, we ask whether polysyllabic shortening evidence supports another type of proposed prosodic constituent: the cross‐word, or Abercrombian foot (Abercrombie 1965). Abercrombian feet consist of a phrasally prominent syllable followed by non‐phrasally stressed syllables up to, but not including the following phrasally prominent syllable. Table I shows that for some word sequences containing phrasal prominences on both content words, word‐based constituents and Abercrombian feet are isomorphic. However, for e.g. bake elixirs and bake avocadoes, Abercrombian feet, but not word‐based constituents, include word fragments.
| Word‐based constituents | Abercrombian Feet |
| [Bake] [apples | [Bake] [apples |
| [Baking] [apples | [Baking] [apples |
| [Bake us] [apples | [Bake us] [apples |
| [Bake] [avocadoes | [Bake avo‐] [‐cadoes |
| [Bake] [elixirs | [Bake e‐] [‐lixirs |
Our materials consisted of phrases like those listed in Table 1, created out of 10 monosyllabic verb stems (–bake, pick, cook, tab, bag, stop, track, grab, crib, catch). Our materials were recorded by six speakers of a variety of American English. We embedded each phrase in the 4th line of a limerick, in order to ensure reliable placement of phrasal prominences on each content word in the target sequence, and to encourage the production of near‐isochronous inter‐stress intervals.
We predicted that if polysyllabic shortening occurs within Abercrombian feet, e.g. –ake in bake avocadoes and bake elixirs should be shorter than –ake in bake apples. On the other hand, if polysyllabic shortening occurs within word‐based constituents, but not Abercrombian feet, we expected e.g. shorter –ake in baking apples than in bake apples, but no difference between e.g. –ake in bake apples and bake avocadoes.
Preliminary results for three speakers suggest that word‐based constituents are more influential than Abercrombian feet, even in limerick speech, which is highly encouraging of rhythmicity. In particular, we found that 1) inter‐stress intervals are not isochronous: interval durations increase with increasing number of syllables; 2) wordbased constituents influenced duration patterns for all three speakers; 3) speakers inserted other types of boundary markers (e.g. silence, glottalization) at word boundaries within Abercrombian feet, and 4) only one of the three speakers showed evidence of polysyllabic shortening within Abercrombian feet. The magnitude of shortening within this constituent was less than what we observed within word‐based constituents, but suggests that speakers may have the option of using both word‐based and Abercrombian feet simultaneously.