Psycholinguistics of Grammar

Psycholinguistics of Grammar

LSA Course 339, Tues/Fri 8:30 – 10:20, Harper 140

Colin Phillips
University of Maryland

Description (from the LSA course listing)

This advanced course will focus on how speakers encode and navigate linguistic representations in memory. Linguists are impressed by the rich grammatical details that natural languages follow. There is now abundant evidence that speakers and comprehenders show fine-grained control over these details during moment-by-moment speaking and understanding, but how do they do this? To make matters more interesting, much recent research provides compelling evidence that language users make use of domain-general memory access mechanisms to retrieve words and phrases and to form linguistic dependencies during comprehension. But these domain-general mechanisms, which access information based primarily on content, are not straightforwardly compatible with pervasive constraints that focus primarily on structural configurations. I will discuss the memory mechanisms, the linguistic constraints, the current evidence on how to reconcile them, and key questions for future research.

Requirements

We’ll have better discussions if you’ve thought about the material ahead of time, and if I have a sense of your thinking.  But your time is limited at the institute. And we are a big group. So here’s how I propose to make the most of this.

Discussion questions will be posted ahead of each session. Read them, think about them, and send an answer to one of the questions in the body of an email before class, by the previous evening if possible. 1-2 paragraphs is fine. More than fine, actually, as I won’t be able to digest 90 long essays. Please send the email to psychogrammar.lsa2015@gmail.com, with the subject line “Discussion Question #X”, rather than to my regular email address. But feel free to use my regular email for anything else relating to the class. Thank you for not sending your discussion comments as a file attachment, as that would make them harder to process. And feel free to attach a picture of yourself to the email message, so that I have a better idea of who you are. Also, the first time that you submit a discussion response, please also tell me a little about yourself – your background, how your interests relate to this course.
First session: there’s not a whole lot of time for this one, of course. If you can send something before class, then great. But I quite understand that this might not work out. So sending afterwards would be fine too.

Registered for a grade: Please send a response to the discussion question for all 4 sessions. And come to class (there’s a sign-in sheet, apparently).
Registered pass/fail: Please send a response to the discussion question for 2 of the sessions (you can choose, but it’s best if you don’t all choose the same sessions!). And come to class.
Auditing: Well, we’re happy to have you along for the ride.  And feel free to send the email address above the background information about yourself that I’m asking of registered participants.

More Discussion – Tour de Coffee

If 8 hours of this is not enough punishment, then let’s talk further. I hear that this is a rather good campus for coffee, but I’m not taking their word for it. So, after each class I’m planning to hang out for a while at one of the local coffee joints. If you’d like to continue the discussion, or would like to talk about anything else related, then come join me. [Warning: check back in case this list is updated, e.g., if I find that a place is closed for the summer.]

Session 1: Plein Air seems like a safe place to start. It’s the new cool coffee shop in town, and unlike the student-run places, we’re fairly confident that it’s open during the summer.

Session 2: Ex Libris. It seems to be one of the few on campus places that’s open during the summer.  And seeing as its in a campus building, it should have a good wifi signal for catching the conclusion of the Queen stage of the Tour de France, which should be happening right after class.

Session 3: Cobb Coffee Shop. “Gas station chic”; most reviews refer to hipsters. That sounds like us, right?

Session 4: Zaleski & Horvath MarketCafé. Z&H is a local staple. Note that they have two places in town. We’ll go to the one at 1323 E 57th St. (And sorry to the person whose suggestion I dismissed — I thought you had in mind the one on 47th.)

Session 1: Goals for a psycholinguistics of grammar

Session 1 is most closely related to Lewis & Phillips 2015, which is a review/position piece. Read at least one of Lewis & Phillips 2015, or Sturt 2003 (on reflexives), or Kazanina et al. 2007 (on Principle C).

Discussion Question: linguists are typically adamant that they are developing theories that are (i) mentalistic, but (ii) not models of real-time language processing. So why should they care about whether real-time processing respects grammatical constraints, if at all? What is the relation between the mechanisms that typical linguistic theories describe and the mechanisms that Sturt or Kazanina et al. are describing? If the goals of linguistics are psychological in nature, then why is it not just a subpart of psycholinguistics?

Supplemental:
The standard statement of the mentalistic goals of grammatical theory can be found in Chapter 1 of Chomsky 1965.
An interesting more recent perspective can be found in Jackendoff 2002.
Townsend & Bever 2001, chapter 2, has a good summary of the “received wisdom” about 1960s attempts to connect linguistic theories with psycholinguistic theories. [Note: this is a locked PDF, so you can read it online, but cannot print it out. That’s what I was able to find, sorry.]
Marr 1982 is about vision. It’s a key source for discussions of levels of analysis in cognitive theories.

Day 2: Linguistic dependencies and memory architecture

Session 2 is most closely related to the overall goals of  Phillips, Wagers, & Lau, 2011. But the picture in that article is superseded by a number of more recent findings, including Dillon et al. 2013 (on reflexives), Parker & Phillips 2015 (on NPIs; draft of July 18th 2015), Kush et al. 2015 (on quantifier-variable relations). Read at least one of these. I recommend that you not always choose the overview/review article, as reading the primary reports is valuable.

Key background: two key pieces of background on language processing and content-addressable memory are McElree et al. 2003 (speed-accuracy tradeoff and unbounded dependencies), and Lewis, Vasishth, & Van Dyke, 2006 (review and summary of ACT-R computational model).

Discussion questions: answer one of these, based on which article you read.
If you read Dillon et al. 2013: What does it mean to say that agreement and reflexive licensing show contrasting intrusion profiles? Why does it matter to the authors that the intrusion/interference effects for agreement lead to facilitation rather than inhibition? What are some reasons why agreement and reflexives might show different profiles?
If you read Kush et al. 2015: Why does bound variable licensing present a challenge for models based on a content-addressable memory architecture? Speculate on ways that a language processor could successfully target c-commanding positions for retrieval.
If you read Parker & Phillips 2015: Why is the effect of timing here surprising? What can we conclude from the fact that NPI licensing and agreement licensing are differentially affected by the timing manipulation?
If you read Phillips et al. 2011: Most of the evidence in the paper is drawn from English. Should we expect to find the same profiles for the same linguistic phenomena in other languages, or should we be unsurprised to find different profiles in other languages, e.g., due to differences in morphology or word order?

Day 3: Linguistic predictions and long-term memory

[Links and additional detail to be added shortly.]

The primary readings for this session are Chow et al. 2015 (on English comprehension), Chow et al. 2015 (on Mandarin comprehension), or Momma et al. 2015 (on Japanese production).

Background: an important study that launched much of the recent research on (apparent) misinterpretation is Kim & Osterhout 2005.

Discussion Question:
If you read one of the Chow et al. articles: Why does it matter to the authors specifically that they see non-effects on the N400 ERP response, while they pay less attention to the P600 ERP response? Identifying subjects, objects, etc. should be easy. So why should this information have a delayed effect on N400 responses?
If you read the Momma et al. article: How is look-ahead in production planning different than prediction in comprehension?

Day 4: Arbitration and reductionism: what timing evidence is(n’t) good for

This session will aim to discuss two main ways in which researchers have tried to use evidence from the timing or difficulty of linguistic processes to decide among competing high-level theoretical accounts.

Readings to be chosen from:
Phillips 2013 (on reductionist accounts in general; review), Phillips 2013 (on reductionist accounts of islands; review), Yoshida et al. 2014 (on reductionist accounts of islands; experiment).
Phillips & Wagers 2007 (on theory arbitration for filler-gap dependencies); Phillips & Parker 2014 (on theory arbitration for ellipsis).

Discussion Question:
If you read the papers about theory arbitration (P&W07 or P&P14): These papers make the apparently negative argument that psycholinguistic methods aren’t very good at deciding disputes among competing theories of wh-movement and ellipsis. What’s the basis for their skepticism? Do you think it should be possible to do better in the future — with finer-grained experimental measures, or with new linguistic theories? Explain.
If you read one of the papers on reductionist accounts of linguistic constraints: In your view, what is the most persuasive type of evidence in favor of a reductionist account of a linguistic constraint, and what is the most persuasive type of evidence that a linguistic constraint cannot be reduced to extralinguistic mechanisms? In answering this, explain what is your criterion for viewing an argument as more/less ‘persuasive’.