Meredith Brown

Prediction, explanation, and the role of generative models in language processing.

We propose, following Clark, that generative models also play a central role in the perception and interpretation of linguistic signals. The data explanation approach provides a rationale for the role of prediction in language processing and unifies a number of phenomena, including multiple-cue integration, adaptation effects, and cortical responses to violations of linguistic expectations.

Is it or isn’t it: Listeners make rapid use of prosody to infer speaker meanings

A visual world experiment examined the time course for pragmatic inferences derived from visual context and contrastive intonation contours. We used the construction It looks like an X pronounced with either (a) a H(*) pitch accent on the final noun and a low boundary tone, or (b) a contrastive L+H(*) pitch accent and a rising boundary tone, a contour that can support contrastive inference (e.g., It LOOKSL+H*like a zebraL-H%… (but it is not)). When the visual display contained a single related set of contrasting pictures (e.g. a zebra vs. a zebra-like animal), effects of LOOKSL+H* emerged prior to the processing of phonemic information from the target noun. The results indicate that the prosodic processing is incremental and guided by contextually-supported expectations. Additional analyses ruled out explanations based on context-independent heuristics that might substitute for online computation of contrast.

Interpreting prosodic cues in discourse context

Two visual-world experiments investigated whether and how quickly discourse-based expectations about the prosodic realisation of spoken words modulate interpretation of acoustic-prosodic cues. Experiment 1 replicated the effects of segmental lengthening on activation of onset-embedded words (e.g. pumpkin) using resynthetic manipulation of duration and fundamental frequency (F0). In Experiment 2, the same materials were preceded by instructions establishing information-structural differences between competing lexical alternatives (i.e. repeated vs. newly assigned thematic roles) in critical instructions. Eye movements generated upon hearing the critical target word revealed a significant interaction between information structure and target-word realisation: Segmental lengthening and pitch excursion elicited more fixations to the onset-embedded competitor when the target word remained in the same thematic role, but not when its thematic role changed. These results suggest that information structure modulates the interpretation of acoustic-prosodic cues by influencing expectations about fine-grained acoustic-phonetic properties of the unfolding utterance.

Prosody and Intention Recognition

Listeners face multiple challenges in mapping prosody onto intentions: The relevant intentions vary with the general context of an utterance (e.g., the speaker’s goals) and how prosodic contours are realized varies across speakers, accents, and speech conditions. We propose that listeners map acoustic information onto prosodic representations using (rational) probabilistic inference, in the form of generative models, which are updated on the fly based on the match between predictions and the input. We review some ongoing work, motivated by this framework, focusing on the “It looks like an X” construction, which, depending on the pitch contour and context, can be interpreted as “It looks like an X and it is” or “It looks like an X and it isn’t.” We use this construction to investigate the hypothesis that pragmatic processing shows the pattern of adaptation effects that is expected if the mapping of speech onto intentions involves rational inference.

Distinct relative F0 levels elicit categorical effects in F0 maximum and minimum alignment

A standard assumption in intonation research is that the presence and timing of fundamental frequency (F0) maxima and minima relative to segments are crucial phonetic characteristics of intonation patterns. The present experiment tests an alternative hypothesis that the representation of intonation patterns is based on the relative pitch levels of syllables in sequence. Synthetic stimuli were created using the phrase Some lemonade with an overall rising/falling or falling/rising intonation pattern. Cues to the presence and timing of F0 maxima and minima were eliminated by replacing the F0 across lemon‐ with level F0 contours and replacing the sonorant consonants before and after each target vowel nucleus with Gaussian noise. Four continua were created by shifting the F0 levels of one or both syllables in equal 0.5‐ or 0.75‐semitone increments. Thirteen subjects imitated randomized stimuli presented over headphones. Results showed that alignment of maxima and minima in imitations was predictably related to...