HLP Lab | Invited_talks

UID in Early Modern English

2018-12-06T00:00:00-05:00

Previous research analyzing the usage of that-complementizer has provided evidence that its use in Present Day English is guided by the principle of uniform information density. It has been shown that speakers tend to leave the complementizer out in those cases, where its production would lead to a decrease in the information transmission rate. Similarly, the complementizer is often inserted in those locations where a peak in the rate at which information is transmitted would occur in the case of zero linking.

If the usage/omission of that is indeed a result of the sensitivity of language producers to information density, we may assume the same principle to govern syntactic choices at earlier stages of the development of the English language. The language production should, too, be optimized in order to avoid spikes in processing difficulty.

In my project, I test in how far this assumption is reflected in the data of Early Modern period on the basis of reconstructed speech from the Corpus of Early English Dialogues 1560–1760. I explore whether the syntactic variation can be explained solely by the processing constraints, or whether other factors play a role. The examined factors include the length and position of the complement clause, ambiguity at the onset of the complement clause, subject type, as well as the year in which the speech was first published (in order to explore the small-scale diachronic development of the phenomenon).

Phonetic convergence of Hong Kong English towards native English accents and the effect of language attitudes

2018-11-29T00:00:00-05:00

The interactive-alignment model (Pickering & Garrod 2004) suggests that convergence is an automatic process in a conversation and it occurs not only at phonetic level but also at syntactic and pragmatic levels. Evidence of convergence at phonetic level was found in a few studies (Babel 2010, 2012; Pardo 2006; Pardo, Gibbons, Suppes & Krauss 2012), however, these studies mainly focused on native speakers, convergence between non-native speakers and native speakers received less attention. Present study aims to examine Hong Kong English (HKE) speakers’ speech accommodation towards Received Pronunciation (RP) and General American English (GenAmE) during and after 1 hour’s conversation with an RP/GenAmE interlocutor.

The presentation will focus on the HKE speakers’ convergence on two vowels (THOUGHT and PATH vowel) and three consonants (rhoticity, fricative /z/ and fricative /θ/) in the two exposure conditions. The effect of language attitudes in convergence will be reported too. Exemplar-based theories are adopted to explain convergence; however, the accounts do not seem to explain the divergence found in the present study.

References:

Babel, M. (2010). Dialect divergence and convergence in New Zealand. Language in Society, 39(4), 437–456.
Babel, M. (2012). Evidence for phonetic and social selectivity in spontaneous phonetic imitation. Journal of Phonetics, 40(1), 177–189.
Pardo, J. S. (2006). On phonetic convergence during conversational interaction. The Journal of the Acoustical Society of America, 119(4), 2382.
Pardo, J. S., Gibbons, R., Suppes, A., Krauss, R. M. (2012). Phonetic convergence in college roommates. Journal of Phonetics, 40(1), 190–197.
Pickering, M. J., & Garrod, S. (2004). Toward a mechanistic psychology of dialogue. The Behavioral and Brain Sciences, 27(2), 169-190.

Wanna replicate embodiment effects? Do you have the power?

2018-11-15T00:00:00-05:00

Prediction and adaptation in prosodic processing

2018-10-30T00:00:00-04:00

Prosody plays an integral role in comprehending spoken utterances. The use of prosody to signal communicative functions also exhibits a substantial degree of variability, raising the question as to how listeners reliably use prosody to infer meaning. Despite this variability, listeners can rapidly integrate prosodic information to anticipate communicative intentions.

However, what parts of the signal listeners use to anticipate speaker intentions and how they adapt their anticipatory behavior in light of experiences is not well understood yet. This talk presents mouse tracking data that explore these open questions. I will present evidence that (i) listeners can use both the presence of salient pitch events and the absence of otherwise expected pitch events to predict upcoming referential intentions; (ii) some parts of the prosodic signal are systematically ignored when making predictions; and (iii) listeners dynamically adapt their predictive behavior in light of accumulating evidence.

It is argued that listeners rapidly integrate bottom-up acoustic information and weigh it against their top-down expectations about likely prosodic patterns. Moreover, listeners are rational when they evaluate the reliability, that is the usefulness, of parts of the signal to predict what the speaker intends to communicate.

Skype talk

2017-04-12T00:00:00-04:00

Expectations and adaptation in prosodic processing

2017-02-28T00:00:00-05:00

Prosody - the rhythm, intensity, and intonation of speech - is a crucial aspect of human communication. However, there are still many questions as to how listeners process and represent prosodic information. Prosodic processing has usually been thought of as a mainly bottom-up process, where listeners hear cues and build up their prosodic representations based on this acoustic information. However, because prosody signals information about many other levels of language processing (e.g., syntax, pragmatics, meta-linguistic), it is likely that these other levels of language processing might affect how listeners interpret prosodic information. We decided to study this possibility by investigating the relationship between prosodic and syntactic processing. In a series of studies, we explored how listeners’ representation of prosodic boundaries were affected by their expectations as to where these boundaries should occur given the syntactic structure of the sentence they heard. We find that listeners are most likely to report boundaries at locations where they were more syntactically plausible, despite the acoustic information they received in the message. Additionally, we explore whether listeners can adapt to new prosody-syntax mappings after being given feedback based on their interpretations of ambiguous sentences. Our results suggest that listeners are able to adapt to novel (even contradictory) prosody-syntax mappings after enough exposure. Together, these series of studies shed light on how listeners process prosodic information, suggesting that prosodic representations are somewhat flexible, and are influenced by listeners’ expectations and experience with language.

Uniformity in talker-specific phonetic realization: Evidence from sibilant fricatives in American English and Czech

2017-02-23T00:00:00-05:00

The acoustic-phonetic properties of speech sounds vary substantially across languages, and even across talkers within a single language. There is, however, considerable evidence that phonetic properties that the realizations of different speech sounds do not vary independently but are highly structured (mutually predictable). Evidence for structured variation of this type can be found in the covariation of vowels in the F1xF2 plane, with talkers forming relatively congruent but shifted formant patterns, but shifted vowel spaces (e.g., Joos, 1948; Nearey, 1978), and covariation of voice onset time (VOT), as talkers primarily differ in overall mean VOT, especially among the aspirated stops [pʰ tʰ kʰ] (e.g., Chodroff & Wilson, 2017). Cases of structured variation reveal that there are constraints on phonetic variation in language- and talker-specific sound systems (e.g., Liljencrants & Lindblom, 1972).

This talk examines the predictions of two uniformity constraints that require similar (or uniform) phonetic realization of distinctive features across members of a natural class. To evaluate the predictions, I present data on the realization of sibilant fricatives across talkers of American English and Czech. The findings support the hypothesis that talker variation is highly constrained, and also have important implications for rapid perceptual adaptation to novel talkers.

Skype talk

2017-02-14T00:00:00-05:00

Categorical vs. Episodic Memory for Pitch Accents in American English

2017-02-02T00:00:00-05:00

Phonological accounts of speech perception postulate that listeners map variable instances of speech to categorical features and remember only those categories. Other research maintains that listeners perceive and remember subcategorical phonetic detail. Our study probes memory to investigate the reality of categorical encoding for prosody—when listeners hear a pitch accent, what do they remember? Two types of prosodic variation are tested: phonological variation (presence vs. absence of a pitch accent), and variation in phonetic cues to pitch accent (F0 peak, word duration). We report results from six experiments that test memory for phonological pitch accent vs. phonetic cues. Our results suggest that listeners encode both categorical distinctions and phonetic detail in memory, but categorical distinctions are more reliably retrieved than cues in later tests of episodic memory. They also show that listeners may vary in the degree to which they remember prosodic detail.

Models of retrieval in sentence comprehension: A computational evaluation using Bayesian hierarchical modeling

2017-01-23T00:00:00-05:00

Research on similarity-based interference has provided extensive evidence that the formation of dependencies between non-adjacent words relies on a cue-based retrieval mechanism. There are two different models that can account for one of the main predictions of interference, i.e., a slowdown at a retrieval site, when several items share a feature associated with a retrieval cue: Lewis and Vasishth’s (2005) activation-based model and McElree’s (2000) direct access model. Even though these two models have been used almost interchangeably, they are based on different assumptions and predict differences in the relationship between reading times and response accuracy. The activation-based model follows the assumptions of the ACT-R framework, and its retrieval process behaves as a lognormal race between accumulators of evidence with a single variance. Under this model, accuracy of the retrieval is determined by the winner of the race and retrieval time by its rate of accumulation. In contrast, the direct access model assumes a model of memory where only the probability of retrieval can be affected, while the retrieval time is constant; in this model, differences in latencies are a by-product of the possibility of backtracking and repairing incorrect retrievals. We implemented both models in a Bayesian hierarchical framework in order to evaluate them and compare them. We show that some aspects of the data are better fit under the direct access model than under the activation-based model. We suggest that this finding does not rule out the possibility that retrieval may be behaving as a race model with assumptions that follow less closely the ones from the ACT-R framework. We show that by introducing a modification of the activation model, i.e, by assuming that the accumulation of evidence for retrieval of incorrect items is not only slower but noisier (i.e., different variances for the correct and incorrect items), the model can provide a fit as good as the one of the direct access model.