Linguistics Colloquium Series

Past Colloquia

Friday, September 14, 2007 - 4:00 - 311 West Hall

U of M Linguistics Research Roundtable Faculty, Visitors, and Students   Michigan Linguistics Our first colloquium this year will be a "Who Am I?" round table, where faculty, visiting scholars, and graduate students introduce themselves briefly, listing their main research interests and mentioning one or two current projects. Newcomers to the department might also tell us where they've just arrived from. This colloquium will be held in 311 West Hall. You can find a map of central campus on the Campus Information Centers web site: http://umich.edu/~info/central.html An elevator is located at the southern end of the Anthropology department. This can be reached by entering the door marked "Anthropology'' from the West Hall Arch and following the friendly 'elevator' signs. There is a more direct elevator entrance on the side of West Hall facing South University and the School of Social Work building. This glass door can be a bit difficult to find but it is marked "Anthropology/School of Information West'' and opens directly into the elevator lobby. The flyer for the: University of Michigan Linguistics Research Roundtable is available for download.

Friday, September 21, 2007 - 4:00 - 311 West Hall

Some Linguistic Features of Zhaba - A Tibeto-Burman Language in West Sichuan, China Gong Qunhu   Fudan University; Wayne State University Zhaba is the native language of around nine thousand Tibetan people in the high mountains along a river valley in West Sichuan, it is one of the minority languages in China that were not discovered and documented in the large scale survey in the 1960s. This talk is a brief introduction of the phonological, morphological and syntactic features of the language based on fieldwork. Topics such as consonantal clusters, disyllablization, tonogenesis, loanwords from both Tibetan and Chinese, the system of directional affixes for verbs, existential verbs, clause nominalization, etc., will be addressed or touched upon in the talk. Qunhu Gong is professor of linguistics of Fudan University, Shanghai, China. He got his PhD in Shanghai Normal University and worked one year as Postdoctoral Fellow at National University of Singapore. He is currently a visiting professor at Wayne State University. Professor Gong’s research interests include Chinese Sign Language and Sino-Tibetan Linguistics. He is the author of Hantai Guanxici de Shijian Cengci (Chronological strata of Sino-Thai corresponding lexical items, 2002), Zhabayu Yanjiu (A reference grammar of Zhaba, 2007). He is also the Chinese Translator of Roots of Old Chinese (1999) by Laurent Sagart. This colloquium will be held in 311 West Hall. You can find a map of central campus on the Campus Information Centers web site: http://umich.edu/~info/central.html The flyer for this talk: Some Linguistic Features of Zhaba is available for download.

Friday, October 5, 2007 - 4:00 - 311 West Hall

How did Proto-Oceanic *o become Marshallese o? Synchrony, Diachrony, and the Mysterious Intermediate Stage Mark Hale   Concordia University, Montreal A well-known problem in historical linguistics arises when an established sound correspondence shows seriously distinct antecedent and subsequent forms. The most famous example of this type is the development of (antecedent) Proto-Indo-European *dw- to (subsequent) Armenian jerk-. Before such correspondences can be considered established, historical linguists must be able to provide a plausible sequence of diachronic steps which could have lead to the attested reflex of the original form. In this talk we will consider a similarly puzzling case: the question of how a segment which was an *o in Proto-Oceanic times could possibly have developed into something which is pronounced in Marshallese as o. What could the intermediate stage(s) possibly have been? It will turn out that the answer to this seemingly moronic question has wide-ranging implications for our understanding of such fundamental issues as the nature of sound change, phonological 'poverty of the stimulus', the relationship between phonetics and phonology, and, most importantly for my purposes, the relationship between synchronic and diachronic linguistic theory. This colloquium will be held in 311 West Hall. You can find a map of central campus on the Campus Information Centers web site: http://umich.edu/~info/central.html The flyer for this talk is available for download.

Friday, October 19, 2007 - 4:00 - 311 West Hall

A Laboratory Phonologist Studies Sexual Orientation (or, Out of the Streets and into the Sound Booths) Benjamin Munson   University of Minnesota, Twin Cities Stereotypes about sexual orientation and speech abound in popular culture, enshrined in such terms as 'gaydar' and the 'gay lisp', and in numerous popular-media portrayals of lesbians and gay men. This talk will review research that our collective has done examining phonetic variation and sexual orientation. The first part of this talk summarizes our work on this topic, giving equal time to both the questions we set out to answer, and the yet-to-be-answered questions that our findings raise, including: Can people really perceive sexual orientation through speech? Studies by our collective show that self-identified GLB people are rated, as a group, to sound more GLB than heterosexual-identified individuals. We also found considerable overlap between the groups in the extent to which individual talkers were rated to sound GLB or heterosexual. We also found overlap the acoustic characteristics of GLB and heterosexual people's speech. Perceived sexual orientation ratings were found to be strongly predicted by selected acoustic characteristics of vowels and consonants in a way that suggest that listeners are not merely assessing the perceptual distance between individual talkers and prototypically masculine or feminine voices. What do people perceive when they perceive sexual orientation in speech? Is people's perception of sexual orientation through speech mediated by their perception of some other, arguably more-primary, information? Here, we found some limited evidence that listeners perceive sexual orientation through their perception of other variables, such as perceived height, age, and speech clarity. We also found evidence that individuals' ratings of sexual orientation are just as strongly affected by phonetic variation that is associated with sexual orientation (high peak-frequency tokens of /s/), as by unattested variation that is nonetheless characteristic of social stereotypes about GLB people ('lisped' /s/). Are social stereotypes about GLB speech invoked during speech perception? Does the perceived sexual orientation of a person's voice influence listeners' perception of aspects of the signal that are not directly related to sexual orientation itself? The answer to this is a qualified 'yes'. Categorization boundaries for synthetic speech continua are affected by the perceived sexual orientation of the voice of the talker to which the continua are appended, and response times in a voice-recognition task are affected by how GLB-sounding male talkers' voices are. The second part of this talk summarizes our ongoing work on this topic. Here, I will discuss what our collective believes to be the two most important issues that must be resolved as this work goes forward. The most critical of these challenges, we believe, is to find new and innovative ways to assess the structure of indexical meanings more generally. This is important be cause we believe that a basic understanding of the set of contrastive elements is needed before we can understand how phonetic variation might convey speech styles. The second concerns the emergence of these speech styles in development. In particular, we believe that it is crucial to determine whether indexical categories arise in development as the inevitable consequence of individuals' parsing of variation in the speech stream.

Friday, November 2, 2007 - 4:00 - 311 West Hall

Language Across the Lifespan Susan Kemper   University of Kansas This talk will address how aging affects language production and comprehension. Dr. Kemper will include research on how to enhance older adults’ comprehension through "elderspeak," a set of special speech modifications intended for older adults. She will report on a series of studies documenting age-related changes to language production, focusing on the production of complex grammatical structures and the effects of dual task demands on production. Susan Kemper is the Roy A. Roberts Distinguished Professor of Psychology and Senior Scientist in Gerontology at the University of Kansas. This colloquium will be held in 311 West Hall. You can find a map of central campus on the Campus Information Centers web site: http://umich.edu/~info/central.html

Friday, November 16, 2007 - 4:00 - 311 West Hall

The Path to Optimality John McCarthy   University of Massachusetts This talk examines a version of Optimality Theory that incorporates serial derivations. Called Harmonic Serialism (HS), this approach was introduced, briefly discussed, and ultimately set aside by Prince and Smolensky (1993/2004). The candidate-generating component GEN is the main locus of difference between "classic" OT and HS. In classic OT, GEN can produce candidates that differ from the input in multiple ways: e.g., GEN(/pat/) = {pat, pati, pad, padi, peti, ...}. HS's candidates are allowed only one difference from the input at a time: GEN(/pat/) = {pat, pati, pad, ...}. In HS, the EVAL component selects the optimal member of this limited candidate set, which then becomes the input to another pass through GEN and EVAL. This process continues until convergence: the input to GEN and the output of EVAL are identical. In HS, the path to ultimate optimality is gradual rather than immediate. When the input and ultimate output differ in two or more ways, HS requires two or more passes through GEN and EVAL, whereas classic OT always requires just one. To get from /pat/ to [padi], classic OT requires a constraint ranking in which [padi] is more harmonic than [pat], [pad], [pati], etc. HS has another requirement: some form intermediate between /pat/ and [padi], either [pati] or [pad], has to be more harmonic than [pat] and less harmonic than [padi]. If plausible constraints and possible rankings don’t provide such a form, then the mapping from /pat/ to [padi] will be impossible in HS but possible in classic OT, ceteris paribus. Is HS's greater restrictiveness supported by the facts? To answer that question, I'll discuss some examples of the too-many-solutions (AKA too-many-repairs) problem. This problem arises whenever ranking permutation predicts unattested ways of satisfying a markedness constraint. For instance, consonant cluster simplification always targets the first consonant, though deleting either one would satisfy NO-CODA: /patka/ [paka], *[pata]. I'll argue that the unattested outcome requires an intermediate step, [patha], that is not more harmonic than [patka]. I'll also discuss syncope of unstressed vowels from a HS perspective. I'll conclude by suggesting some extraphonological connections of this work. It fits better than classic OT with perceptually-based theories of phonology like the P-Map (Steriade 2001), since the HS derivation provides the needed point of comparison for perceptual similarity. And it accords with ideas that have been commonplace in syntax for quite some time -- e.g., seemingly global operations like long-distance wh movement are best understood as the result of a succession of local operations.

Friday, November 30, 2007 - 4:00 - 311 West Hall

Towards a New Empiricism in Linguistics John A. Goldsmith   University of Chicago Empiricist views of knowledge have traditionally been guided by four ideas: that the prototype of knowledge is sensorial; that what is innately known is not rich in information; that occurrences of events can be counted and measured profitably; and that knowledge must always be labeled by a degree of (un)certainty. Rationalism, by contrast, has been guided by four counter-principles: that the prototype of knowledge is mathematics; that innate knowledge is like any other kind of knowledge; that what is epistemologically important does not occur at a particular moment, but is rather a timeless generalization; and that knowledge, by definition, is certain. Empiricism offers two prospects today that are very attractive to many linguists (though by no means all!). It offers a coherent scientific characterization of linguistics that is not based on psychological reality; it is not necessary, on such a view, to make the claim that grammars are models of the brain's operation in order to have a fully explicit and functioning science---one that looks remarkably like what real linguists do on a daily basis, in fact. In addition, it offers a coherent account for why the careful organization of linguistic data is not an activity of value only once the data have been shown to be relevant to evaluating one theoretical proposal with respect to another. An empiricist account of knowledge explicitly associates equal importance to theoretical elegance and to empirical description. The development of empiricism is closely linked historically to the rise of the theory of probability and its application to scientific questions, and this is the aspect of the question that I will discuss in the most detail, describing some aspects of bayesian analysis in linguistics. In a nutshell, probabilistic models are quantitative knowledge of evidence, and they are excellent models to explore and exploit in the case of evidence that is as rich, varied, and structured as the data that linguists collect and study. The flyer for this talk: Towards a New Empiricism in Linguistics is available.

Friday, December 7, 2007 - 4:00 - 311 West Hall

Student Colloquium Graduate Students   University of Michigan

Friday, January 11, 2008 - 4:00 - 4448 East Hall

Russian case morphology and the syntactic categories David Pesetsky   MIT Sometimes it is the oddest facts that provide the best clues to significant properties of language, because their very oddity limits the space in which we are likely to search for possible explanations. In this talk, I argue that the strange behavior of Russian nominal phrases with paucal numerals ('two', 'three' and 'four') provide clues of just this type concerning the syntactic side of morphological case. When a nominal phrase like the Russian counterpart of 'these last two beautiful tables' occupies a nominative environment, the pre-numeral demonstrative and adjective ('these last') bear nominative plural morphology, and the numeral itself is nominative. The post-numeral adjective ('beautfiul'), however, is genitive plural; and the noun ('table') is genitive singular -- a situation that the illustrious Russian grammarian Peshkovsky (1956) characterized as "a typical example of the degree to which grammatical and logical thinking may diverge". I suggest that the behavior of these phrases is actually entirely logical -- once one adopts a particular structural analysis of the Russian DP. and a particular view of the nature of case morphology. Following in part Richards (2007), I propose that Russian is a covert case-stacking language -- with stacking restricted, however, by the spelling out of phases (Chomsky 1995; 2001). Furthermore, case affixes, traditionally classified using case-specific terminology (nominative, genitive, etc.), are actually instantiations of the various syntactic categories. In particular, genitive is N, nominative is D, accusative is V, and the obliques are P. The proposal concerning genitive (in essence, that nouns are born genitive) is the most surprising, and will be a focus of particular attention. More generally, this work raises at least the possibility that there is no special theory of morphological case.

Friday, January 18, 2008 - 4:00 - 4448 East Hall

MLK Colloquium Benjamin Bailey   University of Massachusetts In this talk I examine ways in which language is used to reflect and reconstitute social identities. Through talk we position ourselves and others relative to co-present interlocutors, the communicative activities in which we are engaged, and various dimensions of the wider world, including social identity categories and their relative value. The indexical dimension of language is a key to these negotiations, as linguistic forms and ways of speaking invoke particular historical and social associations. Indexical meanings are not fixed: while popular indexical associations may pre-exist any given interaction, the meanings of any form are also negotiated and contested in individual interactions. Because social identity negotiations involve both received and negotiated meanings, they provide a means of linking meaning making processes in interactions at the local level with larger social and historical processes, e.g., racial formation. These ideas will be illustrated with examples that highlight the agency of individuals and the indexical power of language in the constitution of ethnic and racial identities. Benjamin Bailey is an Associate Professor in the Department of Communication at UMass-Amherst. His research addresses questions of language, identity, and social categories, especially focused on issues of race. He has published work on multilingualism, interethnic communication and conflict, and immigrant communities in the United States, particularly Dominican Americans. Note: This talk will be held in 4448 East Hall.

Friday, February 1, 2008 - 4:00 - 311 West Hall

Universal and language specific ways of Infant Directed Speech Reiko Mazuka   Duke University & Riken Brain Science Institute, Japan Linguistic, psycholinguistic, and brain imaging study of Japanese infant directed speech. It has long been known that when adults speak to infants and young children, they modify their speech, using higher pitch, exaggerated intonation, and shorter, slower phasing. Infants prefer to listen to such modified speech, sometimes called Infant Directed Speech (IDS).　 Although many IDS characteristics are assumed to be universal, not much is known about how the language-specific properties of a given language manifest itself in IDS. In this talk, we present results from an on-going project at RIKEN Brain Science Institute in Japan to examine the role of infant-directed speech, combining three types of methods: 1) detailed analysis of Japanese IDS speech including segmental, lexical, and intonational phrasing; 2) infants’ responsiveness to IDS in their own languages and those of foreign languages; and 3) an fMRI study of IDS to examine the role of IDS to mothers. Results from these studies show that although the overall properties of IDS may be universal -- e.g., “exaggerated prosody” -- the specific ways to achieve it may differ from language to language, and infants and mothers are tuned in to the specific cues of IDS in their own language.

Friday, February 15, 2008 - 4:00 - 311 West Hall

The Surprising Nature of Short-Term Memory in Language Processing Rick Lewis   University of Michigan Many linguistic relations appear to be non-local. Consider the question in (1): (1) What kind of short-term memory system do you think natural language, with its many long-distance relations, requires? It may not be elegant prose, but example (1) shows that most of us can quickly compute relations (such as "what kind...." is the object of "requires") even though the heads of such relations may be separated by many words (about 15 in (1) above) or annoying interrupting remarks (see examples earlier in this sentence) of a parenthetical nature. Yet, classic measures of "verbal short term memory" in cognitive psychology suggest that most of us can't even reliably remember a 7-digit phone number. And modern theories of short-term memory (STM) are even more constrained than the traditional views, suggesting a distinguished capacity of only about 2 items---not even close to 7! How can the human system process language under such constraints? I'll present new theoretical and empirical work that attempts to provide some answers. The theory is a straightforward application of recent STM theory to sentence processing, but it makes some unexpected predictions, including: serial order information is not really needed in comprehension, certain patently ungrammatical structures yield the illusion of grammaticality, and sometimes longer-distance relations are processed more quickly. A wide range of cross-linguistic evidence now supports this theory, which has been newly extended to language production. The key is parallel, cue-based retrieval working over rich linguistic representations.

Friday, March 7, 2008 - 4:00 - 311 West Hall

Student Colloquium: Chris Odato & Li Yang Christopher V. Odato -- Relevance in the Eye of the Beholder: How Does Age Matter in the Evaluation of Speech? Research on older adults' speech identifies social or cognitive mechanisms that lead individuals to produce speech that is regarded as rambling and/or off topic. Largely unconsidered in this work is the possibility that when listeners evaluate older adults' speech, their expectations about speakers and speech situations contribute to their judgments. In this study we disentangle speakers from their speech by using a constant set of texts and ascribing them to different speakers and speech situations. Forty older adults (age 70+) and forty college students evaluated transcribed narratives on the extent to which the speaker was off-target and identified the portions that they considered off-target. The results show a complex interaction of age and speech situation in the evaluation of speech. This paper argues that relevance must be treated as subjective and points to the importance of identifying the analytic categories raters/listeners bring to the task of evaluating others' speech. Li Yang -- Named entity recoginition with syntactic and semantic features Named entity recoginition (NER) is an important task in the field of information extraction. The NER task is to identify the named entities in the input document and classify them into predefined entity types, such as people names, organizations, locations, product names etc. Traditionally, two approaches are exmployed to tackling the entity recoginition problem. Rule-based systems utilize manually built syntactic patterns, while machine learning-based systems depend much on sophsiticated statistical algorithms and less on language-dependent syntactic and semantic features. The two approaches have left three questions unanswered. First, what syntactic and semantic structures contribute to the recognition of named entities? Second, how would such syntactic and semantic structures contribute to NER? Third, would such stuctures help NER across different domains? We present an NER system that is built to answer the above three research questions. The system relies on the meaning of the syntactic structures that define named entities. The syntactic heads of the defining structures are learned from the training data and are later used to interpret the meaning of the syntactic structures in the test data. Semantic features are extracted from the structures that define named entities and are used to build a decision tree classifier. The mapping phase iteratively matches a previously labeled name with other mentions of the same name and with conjunctions in the same document. Evaluation on two test sets shows this system can generalize to data from different domains and perform moderately well.

Friday, March 28, 2008 - 4:00 - 311 West Hall

Competence and Performance in mismatching VP ellipsis John T. Hale   Michigan State University joint work with: Christina S. Kim University of Rochester Jeffrey T. Runner University of Rochester Gregory M. Kobele Humboldt University, Berlin It is by now well-established that some VP ellipsis constructions are acceptable even if the match between elided element and antecedent is not surface-exact (Tanenhaus & Carlson 90, Arregui et al 06, Kertz 08). For instance in 1a the ellipsis site Δ matches the antecedent VP "funded by the university." By contrast, in 1b the elided phrase "fund Jane's project" clashes in Voice with the passive antecedent "Jane's project was funded" and yet this example is still quite acceptable. 1a Jane's project was funded by the university, and Greg's project was Δ too b Jane's project was funded by the university, and the pharmaceutical company did, Δ too This talk presents our evolving work on a processing model that handles the kind of mismatching ellipsis constructions exemplified in 1b. The model includes a kind of Minimalist Grammar (Stabler 97) that has been extended to support hypothetical reasoning. The particular account formalizes the traditional syntactic deletion-under-identity idea with a view towards avoiding the overgeneration that plagues semantic approaches to ellipsis. The model's predictions are compared to acceptability data collected using magnitude estimation. More generally, we advocate for a methodology that applies competence grammars in cognitive models of the human sentence processor. Ana Arregui, Jr Charles Clifton, Lyn Frazier, and Keir Moulton. Processing elided verb phrases with flawed antecedents: the recycling hypothesis. Journal of Memory and Language, 55:232-246, 2006. Laura Kertz. Verb phrase ellipsis in context. Talk presented at the Annual Meeting of the Linguistic Society of America, 2008. Gregory M. Kobele. A formal foundation for A and A-bar movement. To Apppear in UCLA Working Papers in Linguistics. Edward P. Stabler, Jr. Derivational Minimalism. In Christian Retoré (ed) Logical Aspects of Computational Linguistics volume 1328 of Lecture Notes in Computer Science. Springer-Verlag, Berlin pages 68-95, 1997. Michael Tanenhaus and Greg Carlson. Comprehension of deep and surface verb phrase anaphors. Language and Cognitive Processess, 5:257-280, 1990