The role of higher-level prosodic units in sentence processing

This dissertation investigates the role of prosodic units in sentence processing by adults. Speakers use prosody (timing, pitch, amplitude) to create word groups, but we do not know exactly what function these prosodic units have. I suggest that the prosodic units speakers create are those in prosodic phonology (e.g., Nespor & Vogel, 1986). I further suggest that listeners use these same units as domains for processing. The main hypothesis of this research is that processing of prosodic units involves an alternation of close attention to the speaker's signal (tracking) inside units and independent processing by the listener (integration) at ends of units. I focus on one level of unit, the phonological phrase.;My empirical research included three phoneme monitoring studies. All measured monitoring times at consecutive positions. To study tracking, correlation analyses were done for speaker duration and listener monitoring times. Study 1 tested for effects at the intonational phrase (IP) level, using stimuli like (i), where IPs and clauses align. (Targets are underlined.);(i) When federal authorities surrounded the boat, pier crews stopped work ... ;Studies 2 and 3 tested for effects at the phonological phrase (PhP) level, using stimuli like (ii), where PhPs and syntactic phrases align.;(ii) Though officials have tested the prison diet, countless times;The stimuli in Studies 2 and 3 were balanced for modifier position (prison diet versus diet at the prison) and modifier type (adjectival versus nominal). Study 3 spliced contexts (facilitating and neutral) onto the isolated-sentence stimuli from Study 2.;Results of Study 1 were consistent with findings in the literature (faster times for IP-final than nonfinal words, and slow post-IP times). Results of Studies 2 and 3 showed slow post-PhP times, analogous to findings with IPs and reflecting an integration effect. But times for PhP-final words were slower than for nonfinal words, in the opposite direction of the IP-level effect. I argue on the basis of the correlation results, that findings of all three studies are fully consistent with the hypothesized two-part processing.