A message from your QP Fest committee:
Remember to save the date for QP Fest: April 25th, 2-5 p.m., followed by a Friday Social. Check out the QP Fest 2014 web page here for the schedule of speakers, talk titles, and abstracts. And see you next Friday, April 25th, for grad student talks, snacks, and socializing!
Anna Greenwood (UCSC) will be giving a talk today at noon in the Greenberg Room for the phonology workshop. The abstract is given below.
Unpacking naturalness and complexity biases in stress pattern learning
Abstract: Complexity and naturalness biases are said to constrain the space of typological variation due to their effects on learnability (see Moreton and Pater 2012). Cross-experimentally, the effect of complexity is quite robust in that patterns that reference a single feature are learned more successfully than patterns that reference multiple features. The role that naturalness plays on pattern acquisition, however, is still highly contested in the literature.
The experiment adds to the complexity vs. naturalness debate by comparing subjects’ ability to acquire unattested stress patterns: one which is complex (but phonetically natural) and another which is unnatural (but simple). Participants learned one of three weight-based stress systems, where peninitial heavy syllables disrupted default initial stress. The two simple conditions deﬁned heavy syllables on one factor: presence (I) vs. absence (II) of a coda (C). In the two natural conditions, heavy syllables were closed (I), or closed with a nonhigh (a) vowel (III).
The experiment found an effect for both complexity and naturalness biases, the latter of which is anomalous due to the lack of a naturalness effect in other experiments (e.g. Pycha et al. 2003, Kuo 2009). The paper posits that the naturalness bias may arise due to a performance error: subjects are more likely to misperceive unnatural stress patterns than natural patterns. The robust effect of complexity likely arises from competence error.
Chris Donlay (UC Santa Barbara) will visit the Fieldwork Workshop on Monday (4/21) at 1 PM in the Chair’s Office. He’ll talk about Katso (Sino-Tibetan, China) as well as his fieldwork design and experiences. Light snacks and drinks will be available, and there will be an informal discussion about fieldwork afterward as well.
Prema Nadathur will be presenting at SMircle this Monday at 4PM in the Greenberg Room about Weak Crossover. Her title and abstract are given below.
Weak Crossover and the Direct Association Hypothesis
Abstract: Weak crossover has figured prominently in the debate over the existence of traces, as it has been claimed to provide evidence for their necessity in long-distance dependencies. This paper argues against this claim by providing a treatment of weak crossover that does away with the need for empty categories. This is achieved by use of the Direct Association Hypothesis (Pickering and Barry 1991), which proposes that filler-gap dependencies are characterized by a direct link between an extracted element and its subcategorizer. I show that the new treatment not only accounts for uncontroversial crossover data, but also fares better than the preceding LFG treatments on some key examples. Finally, I outline some consequences of the new proposal, as well as directions for further inquiry, and argue that direct association may provide a robust starting point for reexamining a number of phenomena involving filler-gap dependencies.
Geoff Pullum (University of Edinburgh) will be presenting at SPLaT! April 24 at 4PM in the Greenberg Room. Come 15 minutes early for tea and snacks. The title and abstract are given below.
The stimulus poverty story: misconduct or incompetence?
Abstract: A reconsideration of the present state of the dispute about the so-called argument from poverty of the stimulus suggests that a similar situation in a discipline maintaining scientific standards would lead to retractions of papers, and charges of scientific misconduct or at least rank incompetence. Instead what we find among generative linguists and sympathetic cognitive sciences is a complacent consensus around placidly received wisdom repeated in looser and looser forms. This talk offers a reminder that linguists should be embarrassed about the situation.
I will assume that the audience is broadly familiar with the thesis of linguistic nativism and the claim that infant learners cannot learn a first language from experience because their environments provide insufficient information. As minimal preparatory reading I would suggest comparing the appallingly incompetent Wikipedia article on the topic with the early sections of Pullum and Scholz 2002. Payne et al. 2013 is also highly relevant, and detailed exemplification will be drawn from it, but since it is already published and widely accessible I will attempt to open up discussion of it rather than present its content in full.
A message from your QP Fest committee:
Remember to save the date for QP Fest: April 25th, 2-5 p.m., followed by a Friday Social. We’ve got a great line-up of speakers: Sam Bowman, Phil Crone, Lelia Glass, Masoud Jasbi, and Bonnie Krejci. And just a reminder if you’re still considering a talk: Today’s the deadline to let us know, thanks! (please direct email to Robin Melnick)
Please join us for Sociolunch this quarter, every Wednesday at 11:30 AM – 1:00 PM in the Greenberg Room. (Note the new time!).
To start the quarter off, this Wednesday, April 9, Mark Myslín (UCSD, his page here) will be sharing about his work on codeswitching in Czech-English bilinguals. All are welcome! His abstract is given below.
Codeswitching and predictability of meaning in discourse
Abstract: Codeswitching is often not a simple matter of lexical gap-filling, instead depending on sociolinguistic and discourse functions. We hypothesize that such variation is affected by meaning predictability: codeswitches correlate with meanings of low predictability, allowing language choice to be a formal marker of information content along with familiar means such as prosody and syntax. More generally, we argue for the importance of communicative and social context in understanding linguistic variation, and for rigorous approaches in the analysis of natural discourse data.
In grammatical choice, speakers often convey less predictable meanings with distinct, extensive encodings. In bilingual discourse, a similar division of labor is possible: less predictable meanings can be more saliently encoded through a switch to the less frequently used language. In our corpus of three hours of Czech-English conversation, most switches were to English at points of less predictable meanings. To quantify this, we developed a guessing game (Shannon, 1951) in which new participants listened to full conversations and guessed missing words. With 3,458 guesses across 725 items, concepts that had been expressed through English switches were indeed more difficult to guess. Meaning predictability was a key factor in a logistic regression with 12 control factors from sociolinguistic, psycholinguistic, and discourse-functional traditions.
Codeswitching allows salient encoding of high-information meanings, but unlike phonetic or syntactic reduction of low-information content, the strategy is not more temporally efficient for the speaker. This, especially given our comprehender-based metric of information, supports an account in which language choice functions to draw comprehender attention to novel information.