Janet Pierrehumbert - Selected Publications#

1. Needle, J. M, Pierrehumbert, J.B. and Hay, J.B. (2022) Phonological and Morphological Effects in the Acceptability of Pseudowords. In A. Sims and A. Ussishkin, eds. Morphological Typology and Linguistic Cognition. Cambridge University Press, Cambridge, 79-118. (11 citations)

Almost the first observation in any introductory phonology class is that native speakers have sharp intuitions about which forms are possible words in their language, and which are not. This large-scale study obtained judgments for a set of English pseudowords that were systematically varied in length and phonotactic quality. It replicated findings that the acceptability of the forms is a cumulative function of the likelihood of the parts, while also demonstrating that novel forms that appear to have a morphological parse are preferred to those that do not, even if the parse involves a non-existent stem. The results are interpreted in relation to currently competing theories of the mental lexicon.

2. Hofmann, V., Pierrehumbert, J.B. and Schütze, H. (2021) Superbizarre is Not Superb: Improving BERT's Interpretations of Complex Words with Derivational Morphology ACL-ICLNLP 2021. (41 Citations).

One of a series of studies bridging morphological theory with natural language processing, this paper shows how the performance of a highly touted deep learning model of language (BERT) is compromised by poor morphological encoding of complex words, and can be improved by using more linguistically grounded encoding. It was published in the Association for Computational Linguistics conference series, which is the single highest impact venue in the language sciences. (According to Google Scholar, the ACL in fact ranks in the top 50 for impact across all journals in all fields).

3. Röttger, P., Vidgen, B., Nguyen, D, Waseem, Z, Margetts, J. and Pierrehumbert, J. (2021) HateCheck: Functional Tests for Hate Speech Detection Models. ACL-IJNLP 2021. (166 Citations)

This paper, also published by the ACL, uses a linguistically grounded taxonomy of hatespeech (involving concepts such as the use-mention distinction) to define a benchmarking set for hatespeech detection algorithms. The performance of one of the commercial algorithms proved to be so disastrous that it was taken off the market the next day after the paper appeared. The paper provides a signal example of why people in the high-tech sector need to pay attention to linguistics.

4. Racz, P, Beckner, C, Hay, J. B. and Pierrehumbert, J.B. (2020) Morphological convergence as on-line lexical analogy. Language 96(4), 735-770. (10 citations)

Using a novel game-like experimental paradigm, this study was the first to directly demonstrate convergence between interlocutors in their morphological choices. It compares analogy-based and rule-based models of inflectional morphology, and shows that short-term adaptation transpires through rapid on-line analogies. However, it also finds additional evidence for rules in the patterns that participants brought to the baseline condition of the experiment, presumably through longer-term learning.

5. Todd, S, Pierrehumbert, J.B., and Hay, JB. (2019) Word frequency affects in sound change as a consequence of perceptual asymmetries: An exemplar-based model. Cognition 185, 1-- 20. doi: 10.1016/j.cognition.2019.01.004 (88 citations)

Building on recent experiments on speech processing, this paper develops a model that reconciles the conflicting reports about word frequency effects in regular sound changes. It shows how subtle details of speech encoding, word recognition, and storage in memory come together to predict that frequent words will lead in a pull chain, rare words will lead in a push chain, and no frequency effects will occur in changes that do not implicate lexical competition.

6. Pierrehumbert, J.B. (2016) Phonological representation: Beyond abstract versus episodic. Annual Review of Linguistics 2, 33-52. (217 Citations)

This influential paper offered an overall framework for thinking about how statistical patterns in speech and in the lexical inventory allow humans to acquire phonological systems and use them in speech perception and production. It reconciles evidence that phonological cognition is abstract and general with evidence that it reflects experiences of speech in a detailed manner.

7. Hay, J.B., Pierrehumbert, J.B., Walker, A. J. and LaShell, P. (2015) Tracking word frequency effects through 130 years of sound change, Cognition 139, 83-91. (172 Citations)

According to Neogrammarian linguists and their followers, such as William Labov, regular sound changes equally and simultaneously affect all words in the lexicon. This is in contrast to analogical changes such as paradigm levelling that typically affect rare words before common words. Using the ONZE corpus (the longest-running available database of speech recordings), this study demonstrated that less frequent words lead in the New Zealand English vowel push chain. This result supports the general claims of usage-based phonology, according to which individual words can have detailed phonetic representations. However, the finding that rare words -- and not frequent words -- lead in this regular sound change was a surprise to everyone, subsequently explained in Todd et al. (2019).

8. Pierrehumbert, J.B (2006) The Next Toolkit. Journal of Phonetics 34(4) 516-530. (371 Citations)

This invited position paper discusses the big unsolved problems in sociolinguistics. It sketches the methodology that will be needed to create predictive models of how social and cognitive factors interact in shaping linguistic systems in individuals, and in speech communities.

9. Pierrehumbert, J.B. and J. Hirschberg (1990) The Meaning of Intonational contours in the Interpretation of Discourse, in P. Cohen, J. Morgan, and M. Pollack, (eds) Intentions in Communication, MIT Press, Cambridge MA. 271-311. (2611 Citations).

Pierrehumbert and Hirschberg used the linguistic theory of pragmatics, as developed by Horn and others, to provide a compositional theory of intonational meaning in English.

10. Pierrehumbert, J.B. and M. Beckman (1988) Japanese Tone Structure , Linguistic Inquiry Monograph 15, MIT Press, Cambridge. (2411 Citations)

The autosegmental-metrical approach that Pierrehumbert developed in her Ph.D dissertation is here extended and applied to Japanese. The phonetic data overturned previous claims about tone-spreading in Japanese, and led to a new formal model of Japanese lexical accents and phrasal intonation. In addition to its influence with Japanese linguistics, the monograph had a major influence on the development of models for other languages and on the typology of tone, accent, and intonation.