This workshop will bring together research from the US and abroad for a discussion of the treatment of variation within current formal linguistic theory. Thursday, April 24th events will take place in Campus Center room 904. Friday's events will take place in Campus Center room 804. Saturday's events will be in LGRC 1634 (Math Lounge). More information is available at the Workshop on Locating Variability: Formal Approaches webpage.

The Linguistics Department is sponsoring SALT 18, the Semantics and Linguistics Theory conference. Invited speakers include Lisa Matthewson, Hotze Rullman and Philippe Schlenker. More information is available at the SALT 18 webpage.

End of semester lunch, currently planned for Wednesday May 7 from 12:00 to 1:30, with special graduation-related events between 12:30 and 1:00.

Dan Bodah, BA 1999, and Julia Hanley, BA 2004, both majored in Linguistics, but Bodah is working as a law clerk and Hanley just returned from the Peace Corps. Where will your Linguistics BA degree take you? Ask questions and get ideas from alumni.

Deanna Moore (M.A. 2005) will discuss how she uses Linguistics in her work at National Evaluation Systems.

Please join Linguistics faculty members and fellow Linguistics majors Wednesday, March 5th at 5:30 pm in the Linguistics Department Office [South College 226] and the adjoining Second Floor Lounge. Please come and hang out with us and get to meet your mentor over pizza and refreshments. Please RSVP to Sarah: sarahv@linguist.umass.edu.

Hayes & Londe (2006) argue that Hungarian speakers have implicit knowledge of certain statistical patterns in vowel harmony. For example, stems with a back vowel and a neutral vowel (e.g., farmer 'blue jeans') may take either the back allomorph -nak or the front allomorph -nek of the dative suffix, but higher neutral vowels behave as more transparent, allowing -nak more often. Hayes & Londe found evidence for these patterns in the rates at which real words take each allomorph in a written corpus (the Web) and in subjects' choices in a wug test (Berko 1958). The patterns studied by Hayes & Londe were phonologically natural ones (with one possible exception), well known in the literature. It was therefore unknown whether the patterns seen reflect emergent universal tendencies or rather lexical learning. And, if the patterns reflect lexical learning, are they learnable only because of their phonological naturalness? Would unnatural patterns that happen to be present in the data be equally learnable? To address the question of whether natural and unnatural patterns are equally learnable, we designed a larger-scale wug test. We first identified four unnatural generalizations--albeit stated over phonological categories--in the real-word data. For example, words ending in a bilabial consonant tend to take -nek. Participants, recruited and surveyed over the web, were presented with novel items balanced for the four unnatural generalizations (to the extent that the generalizations are orthogonal) and otherwise reflective of the statistical phonotactics of Hungarian nouns. To avoid item-specific effects, a fresh set of stimuli was generated for every subject. For each item, subjects chose the -nek or -nak form, and also rated each choice. Our results show clear evidence for two of the four unnatural constraints (and five of the six Hayes/Londe constraints; the sixth-- the possibly-unnatural constraint identified above--is applicable to few test items). Thus--assuming we are right to classify these constraints as unnatural--we reject the hypothesis that only natural constraints can be learned and used. More interestingly, our preliminary analysis also argues against the other extreme hypothesis, which is that any constraint stated over phonological categories is equally learnable or usable. When fitting constraint weights not to the experimental data but to the real-word data, we obtain the best match to the experimental data not when all constraints are treated equally, nor when the unnatural constraints are excluded, but rather somewhere in between: unnatural constraints are available, but weighting them strongly is somewhat penalized and their weights are thus lower than they would otherwise be. This echoes the "substantively biased" model proposed in Wilson 2006, where a constraint's phonetic consequences determine learners' a- priori willingness to weight it strongly. In Wilson's case, markedness constraints are biased to remain low-weighted to the extent that they motivate perceptually small changes. In the absence of the detailed phonetic information needed to calculate such detailed bias factors, we merely make a binary distinction between natural and unnatural markedness constraints. Our results thus seem to be a graded version of Becker, Ketrez & Nevins's (2007), in which Turkish speakers extended natural but not unnatural regularities to new items. If at least some unnatural patterns are learnable but penalized (see Pertsova 2004 for another example), studies of differential learnability for real and artificial languages could reach conflicting results depending on whether the strength of the pattern and the difficulty of the task cause the unnatural pattern to fall below an observable-learnability threshold or to hit a ceiling.

Each human language contains an astronomically large (if not unbounded) number of different sentences. How can something so large and complex possibly be learnt? Over the past decade and a half we've figure out how to define probability distributions over grammars and the linguistic structures they generate, opening up the possibility of Bayesian models of language acquisition. Bayesian approaches are particularly attractive because they can exploit "prior" (e.g., innate) knowledge as well as statistical generalizations from the input. This opens the possibility of an empirical evaluation of the utility of various kinds of innate knowledge. Structured statistical learners have two major advantages over other approaches. First, because the generalizations they learn and the prior knowledge they utilize are both expressed in terms of explicit linguistic representations, it is clear what is learnt and what information is exploited during learning. Second, because of the "curse of dimensionality", learners that identify and exploit structural properties of their input seem to be the only ones that have a chance of "scaling up" to learn real languages. This talk describes Bayesian methods for learning Context-Free Grammars and a generalization of them that we call Adaptor Grammars, and applies them to problems of morphological acquisition and word segmentation. Joint work with Tom Griffiths (Berkeley) and Sharon Goldwater (Stanford)

Prof. Chierchia, (PhD 1984, Linguistics, UMass Amherst) Harvard University, will lecture to a general audience as part of the 100th Anniversary of the UMass Amherst Graduate School. the A Century of Scholarship - 100th Anniversary Colloquia.

This paper explores covert A-movement in several languages (Adyghe, Russian, Greek, and Romanian) proposes the distinction between true and apparent covert A-movement. The existence of covert A-movement has important implications for the analysis of A-movement within formal grammar. First, covert movement cannot be modeled using long-distance Agree alone (as suggested in Chomsky 2000). Backward Raising shows that, in some cases, the moving XP has a genuine syntactic presence in the higher position that cannot be accounted for with just an Agree relation. Agree and covert movement must be kept distinct. Second, Backward Raising shows that Lasnik's (1999) claim that A-movement does not leave copies cannot be correct since an actual copy of A-movement is pronounced in Backward Raising which instantiates covert A-movement. However, covert A-movement is sufficiently rare and an explanation of why this may be the case is still outstanding.

Since Stowell (1982), a common (syntactic) view holds that certain infinitival constructions involve tense. Among the properties used to diagnose infinitival tense (which is typically assumed to be some form of future) are the possibility of eventive predicates, infinitives in nominalizations, and a partial control interpretation of the infinitival subject. Although the core of the predicates singled out by these phenomena overlaps to a large extent, when considered in detail, the different diagnostics show conflicting evidence, yielding the contradictory situation that certain infinitives would have to classified as [+tense] for one property and as [-tense] for another property. In this talk, I will summarize the distribution of tense diagnostics across a range of infinitival constructions and provide some initial suggestions for how the conflicting evidence can be accounted for. I will argue that a simple [+/- tense] view is inadequate and that a three-way distinction is necessary to account for the properties above. I will show that infinitives fall into three classes which are largely predictable from the semantics: syntactically and semantically tenseless infinitives, infinitives involving a syntactically present zero tense, and infinitives lacking a contentful semantic tense but involving a syntactically present modal woll. The different syntactic structures suggested will account directly for the distribution of eventive predicates and will serve as the basis for a unified (though at this point very tentative)account of the different types of control and the distribution of infinitives in nominalizations.

The nature of syntactic representations is a fierce battleground for competing theories of the human language faculty, with some recent high-profile attempts to make them extremely simple going so far as to claim that syntax is wysiwyg (what you see is what you get). On the basis of an examination of ellipsis structures in a number of languages, and a detailed investigation of the distribution of voice in English verb phrase and other ellipses, I show that these approaches cannot be correct, and that any adequate theory of syntax must posit objects (phrases and words) which do not have "surface" (that is, pronounced) manifestations. Briefly, the generalization that emerges is the "large" ellipses (like sluicing and fragment answers) do not tolerate voice mismatches between the elided phrase and its antecedent, while "small" ellipsis (like VP-ellipsis) do; I argue that this is a fact about the organization of the syntax, and not necessarily due to processing factors (pace Frazier 2007). In particular, I argue that Voice is syntactically a separate head in English, and should not be conflated with v (or the head that introduces the external argument in transitives and unergatives, pace Kratzer 1996). Following what I dub "the Johnson strategy" (following Johnson 2001), the triggers of mismatch are outside the ellipsis site: "large" ellipsis will necessarily contain Voice, while VP-ellipsis in English does not (it targets vP, below Voice). Unfortunately, accounting for this set of data seems to require that we posit that the identity relation in ellipsis is at least partially sensitive to syntactic structure (pace Merchant 2001). This apparently leads to the dark side, an exploration of which I venture in the areas of implicit arguments, polarity items, and 'vehicle change' effects; for all these, I show that the Johnson strategy leads us to believe that morphological or "lexical" alternations are controlled or triggered by elements that are remote in the structure (and often unpronounced).

Current psycholinguistic work presents very compelling evidence that interpreting linguistic utterances recruits a highly incremental processing system, where meaning is computed in lock-step with the incoming linguistic signal. However, as has been noted by Grice and many others since, meaning includes not just the conventional meanings associated with linguistic expressions, but also a variety of pragmatic inferences that are derived from conventional meanings. This high degree of incrementality poses a potential challenge for classical Gricean accounts of conversational implicature as applied to the language processing domain, suggesting one of two possibilities. One is that implicatures are not computed incrementally, but rather, lag behind the processing of conventional meaning. The other is that hearers must be able to somehow incrementally compute not only the conventional meaning of the unfolding linguistic signal, but also compute enriched meanings that arise as a result of reasoning about what the speaker might have said but didn’t say, all under the pressures of real-time processing. Several theoretical approaches buy the possibility of incremental processing by entirely by-passing a rationalist inferencing mechanism for at least some implicatures. For example, Chierchia and Levinson argue that scalar implicatures related to Grice’s Quantity maxim are either partially or fully conventionalized, such that the expressions themselves trigger the implicature. In this talk, I will summarize some empirical findings showing that at least under some circumstances, what appears to be fully Gricean Quantity-based inferencing can occur during the course of real time spoken language processing with no apparent slowdown to the system. These inferences also appear to be strikingly sensitive to actual patterns of usage exhibited by speakers, suggesting that hearers may well compute meanings based on an assessment of what a speaker might be expected to say under particular circumstances. Furthermore, they reveal expectations of not just the semantic content of an expression, but also its likely linguistic form. Together, these findings suggest that highly incremental inferencing cannot be used as an argument to motivate accounts of implicature which bypass reasining about speaker rationality and intent. If time permits, I will conclude the talk by making some remarks about two challenges to a Gricean account of implicature that I believe can be illuminated by cognitively-oriented empirical work: 1) The question of whether implicatures can be globally derived on the basis of a propositional unit while being computable incrementally; and 2) The question of the non-uniformity of implicatures in a variety of discourse and linguistic contexts, and how one might arrive at a predictive account of when they occur.

I will examine several instances of asymmetrical concord within the DP that show special properties. In one case, apparent s-deletion in a variety of Central Catalan is subject to three heterogenous conditions: a phonological condition (s must be final in a complex coda and followed by a consonant), a lexical condition (s must be the plural morph), and a syntactic condition (the lexical element ending in s must be prenominal). Thus in bon-s vin-s blanc-s franceso-s ‘good-pl wine-pl white-pl French-pl’, the plural marker in prenominal bon-s doesn’t appear, but plural markers in non-prenominal vin-s and blanc-s show up. Another case is masculine singular allomorphy in Italian and Spanish (It. buon gioc-o – gioc-o buon-o, Sp. buen jueg-o – jueg-o buen-o ‘good play). A third case is the lack of concord of prenominal elements with feminine nouns with initial[á] in Spanish, an extension of el/la allomorphy (aquel (masc.) áre-a (fem.) geogràfic-a (fem.) ‘that geographical area’). Assume N is final within the DP and raising causes agreement with elements appearing to its right, but agreement with the rest takes place at PF. This forces postnominal agreement, but leaves prenominal agreement subject to PF conditions, which now can include allomorphic and phonological conditioning. In the first case, for instance, non-agreeing bon, a bare root, will be preferred to the number-inflected bon-s because it doesn’t violate the marked structure CsC even if it violates (PF) Concord.

Free Choice Any: Two Recalcitrant Problems Veneeta Dayal, Rutgers Two well-known problems in the semantics of Free Choice-any are the partitive (1) and the subtrigging cases (2). We do not have a unified account for them at this time: 1a. Bill may/*must read any of these books. b. Bill may/must read any book he finds. c. Bill may/must read any book that is on his reading list. 2a. Bill read any book *(he found). b. Bill read any book *(that was on his reading list). I explore an account that starts with the intuition that Free Choice Items signal, as a conventional implicature, a declaration of ignorance on the part of the speaker. This intuition can be demonstrated in the context of an interview where precise information is being sought: 3a. Speaker A: Which books did Bill read? b. Speaker B: *#That’s easy, he read any book on his reading list. b’. Speaker B: That’s easy, he read every book on his reading list. We can take the truth conditional import of an any-statement to be the same as that of a regular universal statement but with an additional implicature that the speaker is not in a position to give a strongly exhaustive answer to the corresponding question. The challenge of accounting for the distribution of any then reduces to identifying the contexts in which the truth conditional contribution of an any-statement entails the strongly exhaustive answer. Any-statements will be ruled out in those contexts because the two aspects of meaning will clash. To the extent that this attempt is successful, it could replace notions like widening, quantification over possible individuals, vagueness, strengthening, exclusiveness etc that are empirically and/or conceptually problematic. It may also provide a fresh angle on how to approach the issue of variation within and across languages with respect to the inventory of Free Choice Items.