Please join us this FRIDAY at 3:00 for our last LF Reading Group of the semester. Friederike Moltmann will give a talk titled “Reifying Terms”. Abstract below. Note the special day, room and time.

SPEAKER: Friederike Moltmann
TITLE: Reifying Terms
TIME: Mon. 5/18, 10:30am
ROOM: 32-D831

In this talk I will discuss the syntax and semantics of NPs of the following sort:

1. a. the person John
   b. the name John
   c. the fictional character Hamlet
   d. the number ten
   e. the numeral ten
   f. the color green

I argue that the semantics of such NPs uniformly consists in introducing an object on the basis of a non-referential occurrence of an expression, hence they are ‘reifying terms’. I argue that numerals, color words, and simple quotations in fact generally are not referential terms and that only with the help of the construction (1) reference to expressions, colors, and numbers as objects is possible.

Prosodic structure refers to the level of linguistic structure above the segmental level, namely phrasal organization and prominence. This talk examines the temporal and structural properties of phrasal organization as reflected in production and perception. Previous research has shown that prosodic phrase boundaries introduce systematic phonetic variation in the temporal properties of segments. Acoustic studies have found that at boundaries segments increase in duration. Articulatory studies have shown that gestures become longer in the vicinity of boundaries and that this articulatory lengthening increases with boundary strength. I will present a series of experimental studies examining a) the effect of prosodic structure on pause duration in utterances, b) the extent of boundary effects as shown in the articulation of gestures near phrase junctures, c) the categoricity and gradiency in the perception of prosodic boundaries, and d) recursion in prosodic structure. The results inform our understanding of the linguistic representation of prosodic structure and its relation to processes involved in producing spoken language. A model incorporating the results of these studies is proposed.

Tseltal, a Mayan language of southern Mexico, has a strict prohibition against vowel hiatus: all vowel hiatus must be resolved either by deleting one of the vowels or epenthesizing a consonant. The choice of the strategy is not predictable from the nature or quality of the vowels involved. In this talk I will argue that the choice of the hiatus resolution strategy can be accounted for by making recourse to syntactic structure; specifically to the presence or absence of strong phase boundaries between the vowels in hiatus. Building on the work of Marvin 2003, Piggott and Newell 2006, and Michaels 2008 I will propose a set of constraints that favor greater faithfulness to material spelled out in an earlier phase to account for the choice of Tseltal vowel hiatus resolution strategy.

http://web.mit.edu/sula5/program.html

And here is what SULA is all about: “The goal of the conference is to bring together researchers working on languages or dialects which do not have an established tradition of work in formal semantics. We especially encourage abstract submissions from those whose work involves primary fieldwork or experimentation as well as analysis.”

There will be very nice talks on the semantics of very exciting languages from North, Central and South America.

Time: Monday 5/4, 5pm
Location: 32-D831

First speaker: Hyesun Cho
Title: The problem of generalization in a statistical learning model of phonotactics

Achieving a proper level of generality in OT-style constraints has not been a problem in most of the phonological grammar learning models up to now (such as GLA (Boersma, 1997); a Maximum Entropy model (Goldwater and Johnson, 2003); a GLA-HG model (Coetzee and Pater, 2008)), because in those models, constraints do not have to be learned. Instead, learning a grammar mainly involves finding rankings or weights of the given set of constraints. In contrast, the phonotactic learning model of Hayes and Wilson (2008) is different from others in that the model derives constraints themselves from the surface forms in the training data. This paper shows that in doing so, it is not trivial to learn constraints with a proper level of generalization.

I ran the Hayes and Wilson model on Korean training data. In the resulting grammar, the model makes overly-broad generalizations, i.e., the constraints that penalize both possible (grammatical) and impossible (ungrammatical) sequences equally, especially when the frequency differences between the possible sequences and the impossible sequences are very small. Because of this, some of the possible sequences are predicted to be worse than impossible sequences. The three problem cases include: post-obstruent tensing, diphthong restrictions, and labials-[ɨ] sequences. I discuss two possible solutions: adjusting feature specifications and employing an additional learning bias.

The grammar learned by the Hayes and Wilson model consists of markedness constraints only. I ran a model that uses both faithfulness constraints and markedness constraints (Goldwater and Johnson, 2003) for the problem cases above. It turns out that a proper level of generalization in the constraints is crucial in a faithfulness model as well.

Second speaker: Patrick Jones
Title: Evidence for the Phonological Stem in Kinande

A number of recent studies on Bantu verbal phonology (Downing 1994, 1998, 1999, 2000; Herman 1996) have argued that the phonological processes which affect the Bantu verb, rather than referring directly to constituents of morpho-syntactic structure (M-Constituents), instead refer exclusively to morpho-prosodic constituents (P-Constituents) which are derived from them. Evidence for this view, which has its origins in the work of Selkirk (1986) and Inkelas (1989), is found in a number of phonological processes whose domains of application approximate constituents defined by morpho-syntactic structure, but do not match them exactly. In this talk, I will argue that the existence of a Phonological Stem (PStem) domain is strongly supported by the verbal phonology of Kinande, and that by recognizing the PStem as a morpho-prosodic domain to which phonological constraints can refer, it is possible to provide straightforward and unified analyses of four distinct phonological processes - Verbal Reduplication, Intonational Tone Assignment, Lexical Tone Assignment, and Purposive Suffix Affixation - that must otherwise be explained in terms of arbitrary and idiosyncratic constraints. However, I will argue against the position that only morpho-prosodic constituents, and not morpho-syntactic ones, may be referred to by phonological processes. I will argue that reference to the MStem as well as reference to the PStem is necessary in order to successfully account for Verbal Reduplication and Lexical Tone Assignment, and will therefore argue for a theory in which phonological processes make necessary, but not exclusive, reference to morpho-prosodic domains.

http://web.media.mit.edu/~dkroy/papers/pdf/roy_aij_2005.pdf

In the unmarked case, the English ing-form expresses a process, that is, a homogenous non-culminated eventuality, simultaneous with that of the main clause:

1. I spent the afternoon sleeping on the couch.

However, exceptions to the simultaneous reading are known to exist. To some extent, gerunds can refer to events following the matrix event (2), or preceding it (3):

2. He entered college at the age of fifteen, graduating four years later at the head of his class. (From Jespersen 1940: 407)
3. Setting sail for the island in the fall of 1740, he reached his destination in the spring of 1741. (From Stump 1985: 97)

This paper is concerned with the English ing-form, the French present participle, the Italian gerund, and the Swedish present participle. It will be shown that the «tense-shifting» property illustrated in (2) and (3) is attested in English, French, and Italian, but not in Swedish. It will be argued that «tense-shifting» as illustrated in (2)-(3) does not follow from the aspectual properties of gerunds but is in fact linked to grammatical Tense. By assumption, then, grammatical Tense is projected in gerundival clauses in English, French, and Italian. In these languages, we observe that (a) clausal negation may be licensed, (b) copular and auxiliary Vs are allowed, and (c) a subject argument is licensed. Swedish differs systematically from the other three languages in disallowing clausal negation, copular and auxiliary Vs, as well as explicit subject arguments. These observations have consequences for a number of Tense-related issues in generative grammar, such as the theoretical status of Finiteness, the relation between Tense and the Aspect-Event system, as well as the acquisition of Tense.

This talk investigates two separate, but related phenomena in the verbal morphology of Moro: tone distribution and the order of object markers. Moro is a Kordofanian language spoken in Sudan; the research is part of the Moro Language Project at UCSD.

The interaction between tone and syllable weight has primarily focused on the distribution of contour tones. In Moro verb roots, high tone is distributed according to syllable size or weight. Heavy syllables preferentially bear high tone, whereas light onsetless syllables cannot bear high tone. I argue in favor of prominence constraints as in onset-sensitive stress systems (Gordon 2005) rather than extraprosodicity, an approach which fails to explain the onsetless syllable?s participation in other prosodic processes. The domain for tonal restrictions in Moro is the macrostem rather than the word. Onsets from within the macrostem (progressive prefix v-) license high tone on the initial root vowel, but those outside the macrostem do not. In addition, macrostem affixes that bear high tone cause deletion of root high tone and prevent the realization of tone-licensing prefixes. The macrostem as a whole only allows a single high tone.

The macrostem constituent is important not only for regulating tonal processes, but also affix order. Object markers in Moro attach as prefixes in imperfective aspect, but the same markers appear as suffixes in perfective aspect, outside the macrostem. Object markers longer than a syllable or without high tone also appear as suffixes. I argue that these data point to a templatic approach to the mobile affixes, in the sense of both a position and prosodic requirements within the macrostem. This is further confirmed by double object markers. In perfective aspect, both object markers appear as suffixes. The linear order of the two objects is determined not based on grammatical role, but by a hierarchy of person/number features (1 > 2 > 3 and pl > sg). In imperfective aspect, the first object marker is realized as a prefix and the second as a suffix; the discontinuous linear order follows the same person/number hierarchy. The Moro data point to a templatic approach to linear ordering (Nordlinger 2008), and also provide support for approaches to morphology that are independent of syntactic operations.

Peter Graff and T. Florian Jaeger of the University of Rochester presented their talk, The OCP is a pressure to keep words distinct: Evidence from Aymara, Dutch and Javanese at the 45th Annual Meeting of the Chicago Linguistic Society.

Jessica Coon was an invited speaker at a “syntax brown bag” at NYU, where she talked about “Predicate Fronting and its Consequences: Ergativity in Chol”. Meanwhile, uptown a bit, Pritty Patel and Patrick Grosz were invited speakers at a CUNY Syntax Supper, where they spoke on “On the Typology of Pronouns: Two Types of Anaphor Resolutions”.

Finally, MIT will soon be invading Manchester, for the 17th Manchester Phonology Meeting (17mfm). Bronwyn Bjorkman will be presenting on “Uniform exponence and reduplication: evidence from Kinande”; Maria Giavazzi will presenting “On the application of velar palatalization in Italian nouns and adjectives”; and Peter Graff will be presenting a joint paper with alum Andrew Nevins, entitled “Vowel harmony provides a figure-ground relation for consonant phonotactics”. Adam Albright will also present, on “Cumulative violations and complexity thresholds: Evidence from Lakhota”

Since the 1980’s, it has been standard to assume that the presupposition of an expression must be entailed by its local context (Heim 1983). But how is a local context derived from the global one? Analyses developed within dynamic semantics offer a lexicalist solution: the meaning of any operator specifies what its ‘Context Change Potential’ is. However the explanatory depth of these solutions has been called into question because they can in effect stipulate in their lexical entries the data to be accounted for. We will offer a reconstruction of local contexts that circumvents this problem, and can be developed within a classical (non-dynamic) semantics. We will also discuss problems that recent experimental results raise for our analysis.

A non-technical summary of our reconstruction of local contexts is available in:
Schlenker, P. 2009. Presuppositions and Local Contexts, Manuscript, Institut Jean-Nicod and NYU

A longer and more technical version is developed in:
Schlenker, P. To appear. Local Contexts. Forthcoming in Semantics and Pragmatics

Both papers are available at https://files.nyu.edu/pds4/public/

In Catalan, conjugation II and III 2sg imperatives consist of a bare root (e.g. [‘tem] ‘fear!’). When pronominal enclitics are added, some extra material (a stem extension) surfaces (e.g. [‘temAla] ‘fear it (fem)!’). The form of the extension can vary from dialect to dialect and from verb to verb ([A], [i], [gA], [igA]), but it is totally predictable. I will argue that the presence of the extension is enforced by a phonological constraint, and that the choice of specific extensions is determined by Lexical Conservatism constraints (Steriade 1999, 2007), epenthesis being blocked by other constraints.

Please join us Wednesday at 3:00 for our talk by Tue Trinh. His talk is titled “Constraining Copy Deletion”

Speaker: Tue Trinh
Title: “Constraining Copy Deletion”
Time: 3:00-4:30 Wed., 4/29
Place: 34-303

More information, incl. the schedule for the rest of the semester:
http://web.mit.edu/linguistics/groups/synsem/index.html

The MIT Linguistics Department is pleased to announce the penultimate linguistics colloquium of the spring semester, which will take place on April 24th, 2009:

Speaker: Daniel Buering, University of California, Los Angeles
Title: At Least and At Most: The Logic of Bounds and Insecurity Time: Friday April 24th, 3:30pm
Location: 32-141

This talk addresses the meaning of the complex determiners at least and at most and their kin in related languages. I explore the idea that the basic meaning of these is ‘exactly n or more/less than n’, and that this meaning triggers an implicature familiar from disjunction: That the speaker is not sure that exactly n, nor that more/less than n. This, I submit, covers the basic meaning of simple sentences with these, which I call speaker insecurity. Adopting a proposal in Klinedinst (2007), I then argue that at least/most trigger embedded implicatures when embedded under modal verbs, resulting in a second reading I call authoritative (making such sentences ambiguous). I then speculate about a third construal in which the determiners are split up, yielding another, stronger authoritative reading. A compositional semantics for the numerical use of these is provided, and the proposal is compared to that in Geurts and Nouwen (2007), which derives the same set of meanings by more semantic means. (This talk is based on my 2007 WCCFL paper (Buering, 2008), but more comprehensive in that it addresses the full range of meanings you get with at most, including the third construal.)

References
Buering, Daniel. 2008. The Least at least Can Do. Proceedings of the 26th West Coast Conference on Formal Linguistics, edited by Charles B. Chang and Hannah J. Haynie, 114–120. Somerville, MA: Cascadilla Proceedings Project.
Geurts, Bart, and Rick Nouwen. 2007. At Least et al : The Semantics of Scalar Modifiers. Language 83:533–559.
Klinedinst, Nathan. 2007. Plurality and Possibility. Ph.D. thesis, UCLA.

The common wisdom about the interpretation of phi-features of pronouns is that they contribute to the meaning the corresponding presuppositions (cf. Heim&Kratzer 1998, Sauerland 2003, etc. etc.). Namely, a pronoun “she" contributes presuppositions about its referent requiring it to be an atom and a female. This particular view of the gender features goes back to Cooper's 1983 book.

However, one important detail is missing: in an intensional environment where some individual have different genders in different sets of worlds under consideration, where must the requirement to be female be fulfilled? While the common wisdom usually does not go that far when talking about gender features; Cooper himself started to investigate the question and came to the conclusion that the features of bound pronouns contribute real normal presuppositions, while the features of free pronouns contribute a special kind of presuppositions - indexical presuppositions, which can only be fulfilled in the actual world.

As a closer look at the relevant data shows, Cooper's was a wrong generalization. After the discussion of relevant examples, I hope you will agree that, first, Cooper was right saying that presuppositions associated with gender features are special - they cannot be accommodated in the way “normal" presuppositions usually can; secondly, that it is not only free pronouns that trigger such special presuppositions, but bound pronouns as well - there is no difference between the two classes (which is probably good news.) The empirical generalizations emerging, however, seem to require a lot of work to accommodate into current semantic frameworks. I will discuss the demands the new data makes of the semantic theory, and will try to sketch a schema of a theory that should be able to accommodate those.

More information, incl. the schedule for the rest of the semester, can be found here.

Based on the patterns of phi-agreement with post-verbal subjects in Hebrew, I argue against the idea that failure to establish a phi-agreement relation between a phi-probe and its putative target (e.g., due to intervention) results in ungrammaticality, or a “crash”; at the same time, I argue that phi-agreement also cannot be optional.

At first glance, these claims—-that phi-agreement is neither optional, nor does its failure result in ungrammaticality—-might seem contradictory. However, I argue that there is a third possibility, which is in fact the only one that can account for the data under consideration: phi-agreement must be attempted by every phi-probe; but if it fails (e.g., due to the presence of an intervener), its failure is systematically tolerated.

Interestingly, this mirrors the behavior of the ruled-based systems of early generative grammar, where rules were composed of a Structural Description (SD) and a Structural Change (SC). In these terms, the effects of phi-agreement, as far as valuing the features on the phi-probe, could be thought of as the SC; the locality conditions associated with phi-agreement (incl. intervention) could be thought of as the SD.

Finally, I note that these result are in conflict with the idea that Case arises as a result of phi-agreement (e.g., as a result of valuing a full phi-set on a probe; Chomsky 2000, et seq.); I show independent evidence—-from empirical domains outside of the ones discussed above—-that a theory claiming that Case is dependent on phi-agreement is untenable.

Time: Monday 4/13, 1:30-3:30pm
Location: 32-D461
Speaker: Bronwyn M. Bjorkman
Title: Uniform Exponence and Reduplication: Evidence from Kinande

In this talk I argue that verbal reduplication in Kinande (a Central Bantu language spoken in parts of the Democratic Republic of the Congo) is subject to constraints enforcing identity between reduplicants of a single root.

Kinande verbal reduplication is typical for a Bantu language: a bisyllabic reduplicant is prefixed to the verbal stem (the root plus any suffixes), and means ‘quickly’ or ‘iteratively’. What is unique about the Kinande system, however, is that reduplication of morphologically complex bases is regulated by a Morpheme Integrity Constraint (MIC, Mutaka and Hyman 1990), which prohibits partial morpheme-copying: individual morphemes must be reduplicated in their entirety or not at all.

What is interesting is the form that reduplicants of morphologically complex verbs take in order to avoid violating the MIC (1b-d): such reduplicants are identical to each other and to the reduplicant of the bare, unsuffixed verb stem (1a):

(1)	a. eri-huk-a	to cook	eri-huka-huk-a
	b. eri-huk-w-a	to be cooked	eri-huka-huk-w-a or eri-hukwa-huk-w-a
	c. eri-huk-ir-a	to cook for	eri-huka-huk-ir-a (*eri-huki-huk-ir-a)
	d. mó-tw-á-huk-ire	we cooked (yes.)	mó-tw-á-huka-huk-ire (*mó-tw-á-huki-huk-ire)

The data in (1) present a challenge for a correspondence-based approach to reduplication (McCarthy and Prince, 1995): in (1b-c) we see that the reduplicant can correspond to a non-contiguous substring of the Base, and in (1d) the reduplicant contains a final [a] that is not present in the base at all.

To account for these data, I propose that Kinande reduplicants are subject to Output-Output (OO) constraints enforcing faithfulness between reduplicative morphemes themselves, not only between morphologically related whole words. Within a set of verbs sharing the same root, reduplicants are thus subject to two separate and sometimes divergent correspondence requirements: they are required by standard Base-Reduplicant (BR) faithfulness to be identical to their linearly adjacent base, but they are also required by OO constraints to be identical to all other reduplicants within the root-defined set (RED-Uniformity). When BR and OO faithfulness requirements compete, the result is optionality, as in (1b). When the MIC rules out the BR faithful candidate, as in (1c-d), the uniform reduplicant is the only grammatical option.

In this talk I will investigate a form of comparison of superiority that one could call `incremental', as in (1) and (2):

(1) Give me (some) more coffee.
(2) Five customers bought a laptop yesterday, and one more customer bought a desktop this morning.

In its incremental reading, the request in (1) is satisfied even if the quantity of coffee that I receive is less than the quantity of coffee that I got before. In the same way, (2) is true even in case only one customer bought a computer this morning. Incremental readings are not attested with all predicates under all conditions, cf. (3) and (4):

(3) Bob was happy right after the talk, and he is going to be happier tonight at the party.
(4) The temperature rose by 4C yesterday afternoon, and it's going to rise some more this afternoon.

(3) entails that Bob will be happier at the party than he was right after the talk — hence, no incremental reading is available. (4) has an incremental reading according to which the temperature might rise by less than 4C this afternoon. And it might even be the case that the temperature fell down during the night, and rose back again before now. However, it has to be the case that the temperature rises from the degree it had reached yesterday afternoon — not from a lower degree. A proper analysis of incremental comparison must capture these restrictions on the availability of incremental readings.

It will be argued that incremental comparison arises from the use of a specific incremental comparison operator. Lexical ambiguity is supported by the absence of incremental comparison in languages that do not lack standard comparison of superiority (eg. German). The incremental comparison operator combines with a property G of eventualities and degrees, and asserts that G is satisfied by an eventuality E to some degree D. It also introduces a presupposition that a specific eventuality E' that is associated with a degree D' precedes E, such that G is satisfied by the sum of E and E', to the degree D plus D'. In other words, the incremental comparison operator asserts that G(E)(D) is true and presupposes that D increments a previous degree D' associated with a previous eventuality E'. It is argued that the reference to a sum of eventualities E+E' in the presupposition suffices to rule out unattested/limited incremental readings with examples such as (3) and (4).

Intriguingly, Guyanese creole Berbice Dutch is a VO language, whereas both its substrate languages (Ijo languages, in particular Kalabari) and its superstrate (16th and 17th century Dutch) are OV (see Kouwenberg (1992)). Ever since the introduction of Bickerton’s bioprogram (Bickerton (1984) et seq), universalist creolists have taken Berbice Dutch to be a perfect illustration of VO as a default setting for basic word order.

We argue that the VO emergence in Berbice Dutch directly results from the grammatical structure of Kalabari and 17th century Dutch and therefore counts as an argument against this universalist claim that Berbice Dutch word order must result from a UG default setting.

Closer inspection on Kalabari and 17th century Dutch reveals (i) that, contrary to what has been assumed in. Kouwenberg (1992) and Lightfoot (2006), Kalabari does not exhibit any Verb Second effects and (ii) that 16th and 17th century Dutch still allowed VO object leakages. Given these facts, VO emergence in Berbice Dutch directly follows:

First Kalabari had no movement causing VO in their native language. Since Kalabari had no way of recognizing the V2 property, Kalabari speakers learning Dutch must have misinterpreted Dutch VO surface strings and subsequently overgeneralized VO to all sentence types. Further input however did not lead Kalabari speakers to reject their initial VO hypothesis and adopt a more complex OV+V2 hypothesis as the VO overgeneralizations were in compliance with the existing Dutch VO leakages. Finally, this explains why Dutch planters adopted counterintuitive VO in depth orderings: those VO constructions were not considered fully ungrammatical in those days. This opened up the way for the next generation to interpret this linguistic input as VO with exceptional leakage to OV. With the loss of syntactic flexibility, finally, word order for Berbice Dutch was set on VO.

Speaker: Shigeto Kawahara (Rutgers University)
Title: Probing knowledge of similarity through puns
Time: Friday April 17, 4-6pm
Location: 32-D831

This talk outlines the aims, results and future prospects of a general research program which investigates knowledge of similarity through the investigation of Japanese imperfect puns, dajare. I argue that speakers attempt to maximize the similarity between corresponding segments in composing puns, just as in phonology where speakers maximize the similarity between, for example, inputs and outputs. In this sense, we find non-trivial parallels between phonology and pun patterns. I further argue that we can take advantage of these parallels, and use puns to investigate our linguistic knowledge of similarity. To develop these arguments, I start with an overview of the results of some recent projects, and follow that with patterns that provide interesting lines of future research.

Speaker: Diana Apoussidou (UMass Amhert/University of Amsterdam)
Title: Modeling the acquisition of French liaison using allomorphy
Time: 4/6 5pm
Location: 32-D831

As language acquisition research shows (e.g. Chevrot et al. 2008), children learning French are creative when segmenting nouns starting with a vowel. Words like arbre ‘tree’ are in adult speech rarely produced in isolation and undergo a liaison with the final consonant of the preceding word, e.g. un arbre is pronounced as oe.narbr, or des arbre as de.zarbr. Children until the age of 4;6 therefore produce errors such as narbr or zarbr. Chevrot et al. (2008) analyze these errors in terms of templates that the children use in the course of development. The templates are made up of un+/Nword2/ or deux+/Zword2/ etc., where the extra consonant in front of a word depends on the preceding word. I propose instead that the errors produced by the children can be analyzed in terms of allomorphy: children hypothesize different underlying representations for words (e.g. literally /arbr/, /narbr/ and /zarbr/ for ‘tree’) depending on what they can observe. This can be modeled with an optimization-based grammar where different underlying forms of a word are represented by lexical constraints. The results show that even with a resulting ‘correct’ lexicon (e.g. vowel-initial /arbr/ as underlying representation of ‘tree’), interference with the grammar can lead to the use of allomorphs in production (e.g. /narbr/ in combination with un, yielding /oe#narbr/ instead of /oen#arbre).

Last week, Claire Halpert returned from Tervuren, Belgium where she presented a paper on “Superiority Effects in Zulu and Kinande Inversion” at a special workshop on Bantu inversion constrations at the 3rd International Conference on Bantu Languages.

Conference news of the present…

Meanwhile, this weekend was an active one for talks by the MIT linguistics community!

Third-year grad student Bronwyn Bjorkman and first-year grad students Igor Yanovich and Rafael Nonato all presented papers at ECO-5, the “Maryland-MIT-Harvard-UMass-UConn Workshop in Formal Linguistics”, held this year at Maryland. Bronwyn’s talk was entitled ‘Go Get, Come See: the Syntax of a Double Verb Construction in North American English’; Igor’s talk was called “How likely to be viable is a PF theory for A-reconstruction?’; and Rafael’s talk asked the question “What is quantification again?”

At the same time, a few states away, second-year student Jeremy Hartman presented his paper “The semantic effects of non-A-bar traces: evidence from ellipsis parallelism” at Semantics and Linguistic Theory (a.k.a. SALT) at Ohio State.

More or less simultaneously with all these talks, one state to the west, fourth-year student Jessica Coon and second-year student Guillaume Thomas presented papers at the 14th annual Workshop on Structure and Constituency in the Languages of the Americas (WSCLA 14) held at Purdue. Jessica’s paper was entitled “A biclausal analysis of aspect based split ergativity”, and Guillaume’s was “Incremental comparatives and inherently evaluative ‘many’ in Mbya”. Conor Quinn, who was a post-doc at MIT from 2006 through last Spring, also presented a paper at WSCLA, entitled “Incorporated verbal classifiers in a predictive typology of noun incorporation”.

And finally, one more state to the west, Adam Albright was at the University of Chicago, giving a linguistics colloquium talk about “Phonetic faithfulness and affix-by-affix differences in derived words”, and a talk in the Workshop on Language, Cognition, and Computation series entitled “Why are cumulative markedness effects so rare?”

Conference news of the future…

Peter Graff’s joint paper with Florian Jaeger entitled The OCP is a pressure to keep words distinct: Evidence from Aymara, Dutch and Javanese” has been accepted for presentation at the upcoming meeting of the Chicago Linguistic Society

Papers…

Omer Preminger’s paper “Breaking Agreements: Distinguishing Agreement and Clitic-Doubling by Their Failures” has been accepted for publication by Linguistic Inquiry and should appear next Fall.

And please remember…

Please remember to send us your news items about talks and papers so we can announce them in Whamit!

The program for the 6th Workshop on Altaic Formal Linguistics (May 22-24, Nagoya, Japan) has been announced.

Speaker: Peter Graff (with Florian Jaeger)
Title: The OCP is a pressure to keep words perceptually distinct: Evidence from Javanese
Time: Friday 4/10 3:30pm, 32-D831

In this study we advance two claims about co-occurrence restrictions on consonants (OCP; Leben 1973) based on a case study of Javanese: i) belonging to the same perceptually salient natural class significantly decreases the likelihood of two consonants co-occurring, ii) that this probabilistic penalty increases linearly with the number of similar segments within a root evidencing cumulativity of OCP effects. Generalizing from perceptual experiments, we hypothesize that the OCP functions as a lexical optimization constraint to keep the words of a language perceptually distinct.

In the first part of this study we investigate whether perceptually salient natural classes have stronger OCP effects associated with them than other sets. In order to not over-parameterize the model we chose a subset of possible natural classes, some with perceptual correlates (e.g. rhotic, lateral, strident) and some with articulatory correlates (e.g. alveolar, glide, palatal). Of 9,261 theoretically possible C1VC2VC3-templates, 1,913 are attested (Uhlenbeck, 1978). We use logistic regression to test whether C1VC2VC3-templates where any two of C1, C2, C3 belong to a natural class are less likely to occur. We simultaneously control for the frequency of C1, C2, and C3 in their respective positions as well as identity (C1=C2), which is known to be favored in Javanese. We find highly significant OCP effects of both articulatory and perceptually motivated classes. By far the strongest similarity avoidance effects, however, are observed for features that are independently known to be highly perceptually salient (rhotic-/r/ and lateral-/l/, Heid and Hawkins 2000; β/r/=-3.86,p<0.0001; β/l/=-2.47,p<0.0001; cf. mean β’s for other OCP effects=-1.47).

Gallagher (2008) shows that, for some features, listeners are better at discriminating words with 0 instances of a feature from words with 1 or 2, than at distinguishing words with 1 instance from words with 2. . Given this result, we generalize that if the OCP is a pressure to optimize perceptual distinctness of words, then additional similar segmentsmake roots even less likely. Indeed, model comparison shows that a cumulative model explains the data significantly better than a non-cumulative model (Bayesian Information Criterion difference=82.5).

Our aim is to place this study in a larger context of logistic regression models of five more languages on which we are currently conducting similar studies. We hope to see i) whether perceptually salient classes of segments co-occur less and ii) whether OCP effects are cumulative as expected under our hypothesis. We will compare our models to other models of similarity avoidance (Frisch et al. 2004, Coetzee and Pater 2008) to see whether our generalizations hold up independent of modeling approach and whether any of these models has an inherent advantage in predicting possible roots.

In GB, A-movement was characterized in two, parallel ways. First, A-movement targets a potential theta position (thus A(rgument) movement) while A’-movement is to a non-theta position. Second, A-movement is Case-driven. The first distinction became obsolete with the advent of the predicate-internal subject hypothesis, which deprives Spec,TP of ever being a theta-position. This leaves only the second characterization for defining A-movement. S. Takahashi (2006) and S. Takahashi and Hulsey (in press, LI) propose an intriguing Case-based analysis for A-movement within MP. In this talk, I will suggest an alternative to Case by exploring instances of A-movement across a number of languages that do not involve Case (e.g., Finnish, Japanese). Based on these cases, I will introduce an entirely different approach to distinguishing A- and A’-movements that takes advantage of the phase architecture of grammar ― what I term the “Phase-Based Characterization of Chains” (PBCC) (Miyagawa, in press). This proposal notes that movements that do not cross a Transfer Domain have A-movement properties while those that cross a Transfer Domain have A’-properties. The analysis provides a straightforward account of not only the familiar A- and A’-movements including scrambling, but it also successfully accounts for more exotic and mysterious types of movements that rely on the notion of“mixed A/A’ position” found in languages such as Finnish (Holmberg and Nikanne 2002).

Miyagawa, Shigeru. In press. Why Agree? Why Move? Unifying Agreement-based and Discourse Configurational Languages. Linguistic Inquiry Monograph 54, MIT Press.

In this talk I examine instances of sluicing whereby the ellipsis site is not structurally isomorphic to its antecedent. The data are presented in three incremental steps: (1) copular clauses can be used in sluicing to circumvent preposition stranding violations in non- preposition stranding languages; (2) such copular rescue is blocked in languages with morphological case marking; (3) this blocking is overruled when (a) the underlying copular clause is case-sensitive, or (b) the sluiced wh-phrase is syncretic between the case assigned by the preposition and the case found in a copular clause. As none of the existing constraints on accommodation of ellipsis antecedents (Fox 1999, 2000; Sauerland 2004; Hardt 2004, 2005) can account for this data pattern, I propose a new constraint, which states that an ellipsis remnant has to be in the licensing potential (cf. Chung, Ladusaw and McCloskey 1995) of both the actual and the accommodated antecedent. This proposal will be shown to receive support from pragmatically controlled sluicing and the interaction between spading and morphological case in dialect Dutch.

4/6 Diana Apoussidou
4/13 Bronwyn Bjorkman
4/20 Patriots Day
4/27 Eulàlia Bonet
5/4 Peter Graff
5/11 Jelena Krivokapić

Speaker: Patrick Grosz
Title: “Movement and Agreement in Right-Node Raising Constructions”
Time: Thurs 12:30-1:45
Place: 32-D461

Quantification over times and worlds in natural language is traditionally considered to be syntactically implicit. More recently tenses and modals have been treated as syntactically explicit quantifiers. I propose that new kind of evidence for the explicitly quantificational character of raising verbs of the “begin” and “threaten” type can be obtained from their scope interaction with the subject.

Consider the following two scenarios:

“HI scenario”
Who is getting good roles before April? Mary: no; Susan: no; Eva: yes
after April? Mary: yes; Susan: no; Eva: yes

“LO scenario”
Who is getting good roles before April? Mary: yes/no; Susan: no; Eva: yes
after April? Mary: yes; Susan: no; Eva: no

The following English sentence may describe either situation (the ambiguity is enhanced by the presence of the temporal adjunct). Notice that HI and the LO readings are logically independent.

In April only Mary began to get good roles.

HI: ‘only Mary went from not getting good roles to getting them’
LO: ‘it began to be the case that only Mary is getting good roles’

Other languages, Hungarian and Shupamem among them, have verb-initial orders that unambiguously carry the LO reading. They do that in two different ways. In Hungarian, “only Mary” is a nominative subject inside the infinitival complement (see Szabolcsi 2009). This talk will focus on the Shupamem type, where the fronting of “begin” appears to assign “begin” wide scope over the operator subject that has properly raised to the tensed clause. I will argue that “begin” does not only have quantificational content but explicitly quantifies over a time argument. Possibly the argument carries over to “threaten” type verbs binding a world argument.

Kusumoto 2005, On the quantification over times in natural language. Natural language Semantics 13: 317-357.
Lechner 2007, Interpretive effects of head movement. http://ling.auf.net/lingBuzz/000178
Schlenker 2006, Ontological symmetry in language: A brief manifesto. Mind and Language 21: 504-539
Szabolcsi 2009, Overt nominative subjects in infinitival complements in Hungarian. To appear in den Dikken & Vago, eds. http://semanticsarchive.net/Archive/jBkM2QxZ/

Detailed information will be posted on the website: http://people.umass.edu/harris/snews/snews.html.

In recent minimalist literature, transformations (Move) have been re-conceptualized as iterative applications of the basic syntactic operation Merge. We will begin by adapting Heim and Kratzer’s (1998) semantic treatment of movement in light of this new state of affairs. We will then propose a generalization of the movement interpretation rule that can handle cases of multidominance other than the canonical movement configuration. Finally we will propose a modified and generalized Predicate Composition and Modification rule which will also depend on syntactic sharing. These new rules will provide an original insight into a number of well known syntactic and semantic puzzles such as extraposition, ECM, object control, small clauses and complex causative constructions.

(Work in progress with Roni Katzir)

Second-year student Jeremy Hartman’s paper “The semantic effects of non-A-bar traces: evidence from ellipsis parallelism” has been accepted for presentation at the conference Semantics and Linguistic Theory (a.k.a. SALT) at Ohio State.

A paper by second-year graduate student Pritty Patel, third-year student Patrick Grosz, Ted Gibson and Evelina Fedorenko entitled “Experimental evidence against a strict version of the Formal Link Condition on E-Type Pronouns” has been accepted as a poster presentation at the 2009 CUNY Conference on Sentence Processing

The Malagasy cleft: what and why

There are two goals of this talk. One is to discuss the particular characteristics of a certain construction in Malagasy that is used both for focus and for wh-questions. The second goal is to investigate the different ways one can go about creating an analysis for a construction that, on the surface, can look similar to an English construction but in a language that is otherwise quite different from English. Malagasy has a construction, sometimes called a no-[nu]-construction, named for the no particle that it contains. It has the following format where the pre-no material encodes new information.

Rasoa no mividy ny vary
Rasoa no pres.at.buy det rice
‘It is Rasoa who buys the rice.’

Many papers have been written on the Malagasy no construction since Keenan’s (1976) seminal paper (e.g. Law 2005, Paul 2001, Pearson 2006, Potsdam 2004), but the exact nature of the construction is still being debated. Much of the controversy has centred around three issues.

(i) the nature of the [no XP] (clause or DP?),
(ii) the nature of no (Det, Focus head, or Comp?), and
(iii) the relation between the pre-no constituent and the following material (movement, predication, or something else?).

In this talk I revisit these issues and bring new data into the discussion arguing in the end that (i) the [no XP] is nominal, (ii) no is in Det, and (iii) the pre-no constituent has not moved from the [no XP].

While the details of the analysis are partly driven by the data, they are also partly driven by the inherent nature of Malagasy within a language typology delineated by a movement typology. I have argued elsewhere (Travis 2005, 2006) that languages differ as to whether a feature triggers XP or X0 movement. In a language like English (or Italian), a V feature triggers X0 movement while a D feature triggers XP movement, and in a language like Malagasy, the reverse is true (there is VP and D0 movement). Given that Malagasy differs fundamentally from English, we might expect that a surface similar construction would have a fundamentally different analysis.

References
Keenan, Edward L. 1976. Remarkable Subjects in Malagasy. In Subject and Topic, ed. Charles Li, 249-301. New York: Academic Press.
Law, Paul. 2005. Questions and clefts in Malagasy. In Proceedings of Austronesian Formal Linguistics Association, eds. Jeffrey Heinz and Dimitris Ntelitheos, 195-209. UCLA Working Papers in Linguistics.
Paul, Ileana. 2001. Concealed pseudo-clefts. Lingua 111:707-727.
Pearson, Matt. 2006. What’s No? Clause linking in Malagasy. San Diego: Workshop on Comparative Austronesian Syntax.
Potsdam, Eric. 2004. Wh-questions in Malagasy. In Proceedings of the 11th Meeting of the Austronesian Formal Linguistics Association ed. Paul Law, 244-258. ZAS Working Papers in Linguistics.
Travis, Lisa deMena. 2005. VP, D0 movement languages. In Negation, Tense and Clausal Architecture: Cross-linguistic Investigations, eds. Raffaella Zanuttini, Héctor Campos, Elena Herburger and Paul Portner. Washington, DC: Georgetown University Press.
Travis, Lisa deMena. 2006. Voice Morphology in Malagasy as Clitic Left Dislocation or Malagasy in Wonderland: through the looking glass. In Clause structure and adjuncts in Austronesian languages, eds. Hans-Martin Gärtner, Paul Law and Joachim Sabel, 281-318. Berlin: Mouton de Gruyter.

Special session: ‘The History of Phonological Theory’ featuring John Goldsmith, D. Robert Ladd, and Tobias Scheer, and with a contribution from Morris Halle. The session will be introduced by Jacques Durand.

Held in Manchester, UK. Organised through a collaboration of phonologists at the University of Edinburgh, the University of Manchester, the Universite Toulouse-Le Mirail and elsewhere.

Conference website: http://www.englang.ed.ac.uk/mfm/17mfm.html

3/2 [open]
3/9 [open]
3/16 Maria Giavazzi
3/23 Spring Break
3/30 [open]
4/6 Diana Apoussidou
4/13 Bronwyn Bjorkman
4/20 Patriots Day
4/27 Eulàlia Bonet
5/4 Peter Graff
5/11 Jelena Krivokapić

Speaker: T. Florian Jaeger (University of Rochester)
Title: Efficient Language Production?
Time: Friday, February 27th, 2009, 3:30pm
Place: 32-141

Abstract:

In this talk, I return to a question that has fascinated language researchers from various disciplines for a long time (Zipf, 1929; Mandelbrot, 1965; Givon, 1987; Hawkins, 1994; Hale, 2001; Bybee, 2002; Genzel & Charniak, 2002; Manin, 2006, among many others), although it has arguably never been at the heart of language research: To what extent is human language processing efficient?

More specifically, I ask to what extent speakers structure their utterances so as to be communicatively efficient. I present a series of studies that test the Uniform Information Density hypothesis (Jaeger, 2006; Levy & Jaeger, 2007; based on Genzel & Charniak, 2002): Within the bounds defined by grammar, speakers prefer to structure their utterances so that information is distributed uniformly across the signal (information density; where the information content of a linguistic unit is defined information theoretically, Shannon, 1948, as -log p(unit)). Where speakers can choose between several variants to encode their message, they prefer the variant with more uniform information density. Uniform Information Density is theoretically optimal in that it maximizes the amount of successfully transferred information and minimizes average processing load.

I discuss evidence from phonetic, phonological, morphosyntactic, and syntactic reduction (word durations; weak vs. full vowels; t/d deletion; contractions such as he’s vs. he is; whiz-deletion in passive subject-extracted relative clauses), as well as studies on discourse planning beyond the level of the clause. I also present new experimental evidence from the distribution of disfluencies and gestures in information dense stretches of speech. The results of all these studies lend to support to the hypothesis that language production is organized to be efficient. When encoding their intended message into linguistic utterances, speakers are sensitive to the information density of the variants they can choose from.

(In collaboration with: Susan Wagner Cook, Austin Frank, Carlos Gomez Gallo, Ting Qian, and Matt Post)

Speaker: Tara McAllister
Title: Articulatory and Perceptual Factors in a Child-Specific Error Process
Time: Tues 2/17, 5pm
Location: 32-D831

Several commonly observed processes in child speech lack counterparts in adult phonologies. Particularly problematic are child processes of neutralization in prosodically strong positions, which contravene our understanding of positional neutralization as a phonetically motivated process governed by the relative strength of perceptual cues. Previous analyses have implicated both child-specific patterns of perception (Dinnsen & Farris-Trimble, 2008) and limitations of the immature articulatory apparatus (Inkelas & Rose, 2008) as the source of this reversal of the adult pattern. I will evaluate the evidence for child-specific perceptual and articulatory factors in an experimental investigation of one child’s pattern of velar fronting in strong position. It will be shown that this child exhibited an adult-like perceptual advantage for contrasts in word-initial position. This suggests that articulatory rather than perceptual factors are responsible for his pattern of neutralization in strong position. Acoustic data from neutralized /d/ and /g/ tokens will also be presented to extend our understanding of articulatory factors that contribute to the process of velar fronting.

In addition, the most recent issue of Natural Language and Linguistic Theory contains an article by graduate student Marie-Christine Meyer with alum Uli Sauerland, entitled “A pragmatic constraint on ambiguity detection: A rejoinder to Büring and Hartmann and to Reis”.

(Pay no attention to the fact that clicking on the website for the conference currently produces an error screen with the message “site désactivé”. We have it on reliable authority that the conference will happen and will be great!)

This year they are looking for three presenters from each of the 5 schools. Titles and abstracts (for the program) will be due March 21st and we hope to have a preliminary list of presenters by March 7th. Abstracts are just for the program. Unless there is a huge amount of interest, you will be accepted to the conference just by sending an email to MIT’s ECO5 liaison, Jessica Coon.

Please let Jessica know if you have any questions and if you are interested in presenting or attending. It’s possible that we will be able to get a carpool together, so if you need or are able to offer a ride, please let her know as well. Also, flights are quite cheap right now (about $120), so that is another option.

When: Registration Day, February 2, 2009
Where: 32-D850 (Lounge)
Time: 2:00—4:00

Sabine Iatridou and Agustin Rayo will be hosting games starting at 3:00 PM.

Welcome Back!

Class descriptions will be added to this week’s newsletter throughout the week.

Please submit items for inclusion in Whamit!

If you would like to give a presentation this semester but can’t attend the organizational meeting, please send a message to Adam to let him know.

January 30th, 4:00pm
Boylston Hall 104
Elisabeth Selkirk, University of Massachusetts, Amherst
“Spelling out syntactic constituents as prosodic domains: match constraints and the syntax-prosodic structure interface”

Domain-sensitive phenomena of sentence phonology provide evidence for a parsing of phonological representation into three basic levels of constituency above the foot: intonational phrase (ι), phonological phrase (ϕ) and prosodic word (ω), as shown, for example, in the Kisseberth 1994 study of the tonal phonology of the Bantu language Tsonga. On the basis of data from Tsonga and other languages it will be argued that the relation between syntactic constituency and phonological domain structure should be captured in terms of syntactic structure faithfulness constraints calling for syntactic clauses, phrases and words to match up with corresponding constituents of an independent prosodic structure representation (namely ι, ϕ and ω, respectively). Fully satisfying these interface Match constraints would produce a prosodic constituent structure that is isomorphic to the syntactic constituency and at variance with the so-called Strict Layer Hypothesis, creating systematic violations of the alleged prosodic markedness constraints Exhaustivity and Nonrecursivity (contra Selkirk 1986, 1995, Truckenbrodt 1999). This paper argues that the evidence supports a theory of the interface which takes an isomorphism between syntactic and prosodic constituency as an ideal. But the evidence also shows that this ideal may fail to be met, due to the role for prosodic structure markedness constraints like Exhaustivity, Nonrecursivity, prosodic minimality and so on, which may force the domain structure to conform to the (potentially conflicting) ideal of a phonological organization that is appropriate for pronunciation. Indeed, it is the role prosodic markedness constraints play in the characterization of phonological domain structure that makes the case that domain-sensitive phenomena are sensitive to an independent prosodic structure and not directly to syntactic structure. The working hypothesis is that the prosodic structure realization component of a grammar consists of an optimality theoretic ranking of interface Match constraints and prosodic structure markedness constraints. The claim is that such a grammar allows for a descriptively accurate account of the range of attested phonological domain structures in individual languages, and for a characterization of typological differences in domain organization found cross-linguistically.

Any theory of constraints on the relation between syntactic constituency and prosodic constituency must specify which types of syntactic constituency are relevant to phonology. It will be argued that Match Phrase and Match Clause each stand for a family of constraints. Match Phrase includes Match constraints distinguishing between phrases that are specifiers and those that are complement of the phasal heads ν (Chomky 2001) and Top⁰ (Kratzer and Selkirk 2007); the Match Clause constraints distinguish between comma phrases (Potts 2005) and those that are the complement of the phasal head C (Chomsky 2001, Pak 2008). Motivation for distinguishing subtypes of Match constraints comes from differences in their interaction with prosodic structure markedness constraints in different languages, expressible in terms of distinct constraint rankings.

2/27/09 - T. Florian Jaeger (University of Rochester)
3/6/09 - Lisa Travis (McGill)
3/20/09 - Anna Szabolcsi (NYU)
4/3/09 - Jeroen van Craenenbroeck (KU Brussels)
4/17/09 - Sharon Rose (UCSD)
4/24/09 - Daniel Buring (UCLA)
5/1/09 - Philippe Schlenker (Institut Jean-Nicod)

Unless otherwise announced, the talks will take place at 3:30pm, in
room 32-141.

for the 3rd annual Cornell Undergraduate Linguistics Colloquium
Cornell University, Ithaca, New York
April 4-5, 2009

UnderLings, the Cornell University undergraduate linguistics association, requests abstract submissions for the third annual Cornell Undergraduate Linguistics Colloquium. Student submissions at all levels are encouraged in a variety of subfields of linguistics, including but not limited to phonetics, phonology, syntax, historical linguistics, sociolinguistics, psycholinguistics, and language acquisition. Applicants pursuing a B.A., B.S., or equivalent degree are invited to submit a one-page abstract for a talk of no more than twenty minutes in length or for a poster presentation at our poster session. Abstracts should be submitted to culc2009@gmail.com by Friday March 6th, 2009. Please indicate whether you would like to be considered for a talk or for the poster session or both. There is a high probability that the conference proceedings will be published afterward, most likely in an online, widely-accessible format.

More information about the colloquium and online pre-registration will be available soon at http://conf.ling.cornell.edu/culc2009.

Please direct any questions to culc2009@gmail.com.

Mon Jan 26 thru Fri Jan 30, 1:00 - 3:00PM
Room 46-3189 M-R 1/26 – 1/29 and 46-3310 for F 1/30
http://stellar.mit.edu/S/course/9/ia09/9.savfdaai/index.html

Description:

A whirl-wind tour of the statistics used in behavioral science research, covering topics including: data visualization, building your own null-hypothesis distribution through permutation, useful parametric distributions, the generalized linear model, and model- based analyses more generally. Familiarity with Matlab, Octave, or R will be useful, prior experience with statistics will be helpful but is not essential. This course is intended to be a ground-up sketch of a coherent, alternative perspective to the “null-hypothesis significance testing” method for behavioral research (but don’t worry if you don’t know what this means).

Course Outline:

Visualization. Creating a visualization to understand experimental results. Simple univariate displays. Conventional multivariate displays. The repertoire of visual variables. Introduction of examples to be used throughout the course: simple behavioral experiments, complex behavioral experiments, and eye-tracking.

Permutation. Understanding what would have happened “by chance” through non-parametric tests, confidence bounds, and measures of effect size. Discussion of null-hypothesis significance testing and its limitations.

Distribution. Understanding the spread of data. Inferring parametric forms (Binomial, Gaussian, Poisson, etc.) as a convenient way of describing the structure of data. Effect size and Bayesian derivation of tests for parametric distributions, inc. binomial test, t-test, Cohen’s d, etc.

Models of Data 1: The Linear Model. What is a “model of data.” Basic assumptions of the linear model. The standard and generalized linear model and relationship to ANOVA. Bayesian derivation of the LM. Link functions and logistic regression. Effect size in a linear model. Introduction to multilevel models.

Models of Data 2: Bayesian Models. Constructing and testing more complex models of data. Bayesian models as a tool for creating models with complex task assumptions. Brief introduction to basic techniques for Bayesian inference.

“Symmetric and Interacting Alternatives for Implicature and Accommodation.”

The talk is geared to two interdisciplinary audiences (the Institute of Cognitive Science at Carleton University, and the departments of linguistics and philosophy at Yale), so feedback from psychologists and philosophers would be especially welcome.

Composed of talks by department faculty and graduate students of past and present.

Thursday, January 29th, and Friday, January 30th
9:30 am-5:45 pm
The Stata Center, 32-D461

All are welcome.

-> more info

Enrollment limits and prerequisites: Permission of instructor (email graff@mit.edu).
No background in Statistics assumed.
Enrollment limited to 10 students.
Priority will be given to Linguistics, Psychology and Brain and Cognitive Science students and faculty.

Description:

This mini-course will introduce basic concepts of probability theory and statistical models applicable to the quantitative study of linguistic phenomena. Participants will learn how to implement statistical models and graphically depict data in the statistical programming language R. Concepts that will be covered include: scales, distributions, hypothesis testing, experimental design, random variables, Central Limit Theorem, Bayes’ Rule, sampling, t-test, ANOVA, regression, logistic regression, mixed models.

Day 1: R language, Graphics, Descriptive Statistics, Scales, Variables
Day 2: Probability Theory - Random Variables, Distributions, Hypothesis Testing, Bayes’ Rule, Sampling, Central Limit Theorem
Day 3: Statistical Models - T-Test, Regression, ANOVA, Mixed Models, Logistic Regression
Day 4: Experimental Design

Participants are expected to complete all assigned readings and problem sets.
URL: http://stellar.mit.edu/S/project/mathmod/index.html

Attendance requirements: Participants requested to attend all sessions

Peter Graff is organizing a workshop entitled ‘The Culture-Phonology Interface: Implications of Laboratory Sociophonetics for Phonological Theory’. Edward Flemming will be a discussant. Peter is involved in two of the workshop talks:

Peter Graff (Massachusetts Institute of Technology), Benjamin Munson (University of Minnesota): Studying the culture-phonology interface.

Peter Graff (Massachusetts Institute of Technology), Kie Zuraw (University of California, Los Angeles), Kuniko Y. Nielsen (University of California, Los Angeles): Investigating preferential imitation.

Giorgio Magri will give a poster on ‘Modeling the order of acquisition of Dutch syllable structures’.

Several recent MIT alumni will be presenting as well:

Andrew Nevins (Harvard University): Contributions of Sign Language morphology to the agreement/cliticization distinction

Ivona Kucerova (University College London): Nulls subjects and the extension requirement

Martin Hackl (Pomona College): Decomposing complex quantifiers: Evidence from verification

Seth Cable (University of Massachusetts): Use of subordinate clauses as matrix utterances in the Pacific Northwest

Title: The Ranking Problem in Optimality Theory
Time: Wed 12/10, 5pm, 32-D831

Every learning problem can be formulated as an optimization problem: within a given typology, pick the grammar that is best given the data. Very little work has been done within Generative Grammar on how to actually solve the optimization problem itself. Recently, a small but growing body of literature within the framework of Optimality Theory (OT) has started to study the optimization problem itself. I will introduce the main ideas of this enterprise and I will illustrate it by focusing on a problem currently open in computational OT, namely that of devising incremental ranking algorithms that perform both promotion and demotion. I will explain why the problem is interesting. I will review one such algorithm, namely Boersma’s (1994) GLA. I will discuss why the algorithm does not work, by explaining in detail what goes wrong in the case of a counterexample discovered by Pater (2007). I will then note that the ranking problem within OT can be described as a linear feasibility problem. This very simple observation has far reaching consquences. In particular, it offers a straightforward way do device the desired incremental algorithm.

The conference will include an introduction by Kai von Fintel, talks by Philippe Schlenker and Jason Stanley, and a roundtable discussion with Bob Stalnaker, Alex Byrne, and Gennaro Chierchia. Dinner (on Friday) and breakfast and lunch (on Saturday) will be provided for everyone.

There is a website with more detailed information (including location and schedule).

Title: Why Agree? Why Move? Unifying Agreement-based and Discourse-configurational Languages
Time: December 2nd, Tuesday
Location: Boylston Hall Room 303; 6 p.m.

Why do we find agreement in human language? And why is there movement? I will propose a unified answer to these questions based on a specific design for human language that we have assumed in generative grammar since the early 1980s — in GB, LFG, MP and others. What I will show is that agreement must be a universal phenomenon, occurring in every language, and with it, movement. This is obviously a more abstract notion of agreement than phi-feature agreement since many languages lack such agreement. I will show that informational structural features such as topic and focus play a role in “agreementless” languages that is computationally equivalent to phi-features. That is, topic/focus and phi-feature agreement are two sides of the same coin, both there to implement “agreement” and “movement” within exactly the same mechanism.

Title: Learning to Model Text Structure
Speaker: Regina Barzilay, CSAIL
When: Wednesday Dec 3, 3-4:30pm
Where: 26-310

Discourse models capture relations across different sentences in a document. These models are crucial in applications where it is important to generate coherent text. Traditionally, rule-based approaches have been predominant in discourse research. However, these models are hard to incorporate as-is in modern systems: they rely on handcrafted rules, valid only for limited domains, with no guarantee of scalability or portability.

In this talk, I will present discourse models that can be effectively learned from a collection of unannotated texts. The key premise of our work is that the distribution of entities in coherent texts exhibits certain regularities. The models I will be presenting operate over an automatically-computed representation that reflects distributional, syntactic, and referential information about discourse entities. This representation allows us to induce the properties of coherent texts from a given corpus, without recourse to manual annotation or a predefined knowledge base. To conclude my talk, I will show how these models can be effectively integrated in statistical generation and summarization systems.

This is joint work with Mirella Lapata and Lillian Lee.

Title: Summing Up Constraint Interaction: Chain Shifts in a Split Additive Model
Time: Wed 12/3, 5pm, 32-D831

Peter Graff
“The Culture-Phonology Interface: Implications of Laboratory Sociophonetics for Phonological Theory”
Thursday, Dec. 4
12:30-1:45
Room 32-D461

In this talk we will address the impact of social factors on phonological generalizations. The notion of a social variable in speech will be defined. An overview of laboratory and corpus-based studies of the interaction of phonetic variation and different social variables will be provided. Common factors of categories affected by sociophonetic variation and possible causes for sociophonetic variation will be identified. Cross-linguistic evidence for high-level phonological alternations interacting with social factors will be presented and possible models will be discussed.

In the second part of the talk we report on research probing selective imitation in the laboratory, comparing male and female subjects’ imitation of male and female talkers. We also compare male and female subjects’ imitation of the same stimuli when they are phonetically ambiguous but labeled as “Michael” or “Jessica” to shed light on the question of whether socially motivated phonetic imitation is conditioned by acoustic or cultural factors.

Time: Friday, December 5th, 2008, 3:30pm
Place: Room 32-141

There are two postverbal strategies for asking ‘why’ in Zulu. The first is the purpose applicative question, which is akin to English what for questions. This is in a sense a bipartite strategy, exhibiting a wh enclitic (-ni ‘what’) and a applicative verbal affix that licenses it. The second strategy uses the word ngani ‘why’ and is used only to question the reason of a negative clause. While in both cases the wh element is postverbal, it is argued to be in very different syntactic positions in the two cases. The enclitic -ni ‘what’ of the purpose applicative is shown to be below the inflectional domain, while ngani of a reason question is in the complementiser domain.

In these questions, several types of evidence show that purpose WHY is below the inflectional domain in Zulu, including the distribution of conjoint and disjoint verb forms and the point of attachment of the applicative morpheme. Furthermore, purpose questions are shown to exhibit transparency effects, in the sense that within a “restructuring domain”, both parts of this question (the particle -ni ‘what’ and the applicative verbal affix that licenses it) attach to the lower verb but are interpreted on the upper verb. Beginning with Rizzi’s (1999) analysis of Italian, it has been claimed for a growing number of languages that, unlike other wh phrases, reason WHY is introduced in the complementiser field rather than moving there from a position below the inflectional domain. In all of the languages for which such an analysis has been proposed, the ‘why’ word appears in some left-peripheral or otherwise preverbal position. Zulu ngani ‘why’ is argued to need a similar analysis, even though it appears in postverbal position. Arguments for the analysis are made on the basis of the distribution of conjoint and disjoint verb forms, interactions between WHY and negation, the absence of transparency effects, and previous analyses of Zulu’s elocutionary force particles yini and na. Specifically, ngani is argued to be an Int0 head (a head in the complementiser domain), around which the IP must move. It is suggested that reason WHY (as opposed to purpose WHY) is universally introduced above negation.

Title: Clefts, Inverse Copulas, and Passives: Understanding Their Delayed Acquisition as Phasal Difficulties
Time: Tuesday, November 18th, 6 p.m. Boylston Hall Room 303

Title: Phonetic similarity in an English hip-hop corpus Time: Wed 11/12, 5pm, 32-D831

In this talk, I present preliminary results from a corpus study of hip-hop. Previous studies on half-rhyme in Romanian poetry (Steriade 2003) and Japanese hip-hop and imperfect puns (Kawahara 2007, 2008) have established that the frequency of specific imperfect rhymes varies with the phonetic distance between the correspondents involved in the rhyme. The current study extends that finding to English hip-hop. The complex nature of the data poses special challenges for data extraction and analysis. I’ll discuss in some detail how the corpus was constructed and what the proper statistical methods are for testing generalizations about half-rhymes.

Vanja de Lint
“Argument Structure in Classifier Constructions in ASL”

Thursday, Nov. 20
12:30-1:45
Room 32-D461

“Special Cases”

Friday, November 21st, 2008, 3:30pm

Room 32-141

There will be a party in honor of Ora, beginning at 6:30pm. Directions to the party will be provided at the talk venue.

Abstract in pdf format

Title: Formerly stressed vowels in Western Armenian
Time: Wed 11/12, 5pm, 32-D831

In this talk, I examine the stressed and unstressed vowels of related forms of Western Armenian. Stressed high vowels [i] and [u] either change to [ə], as seen in (1) or delete, seen in (2) and a stressed diphthong [uj] changes to [u], as seen in (3), when stress shifts off of them. The rest of the vowels and diphthongs (mostly) are unaffected.

1. • “letter” [kír] → [kər-él] “to write”
   • “lie” [súd] → [səd-él] “to make false”
2. • “monkey” [gabíg] → [gabg-él] “to mime”
   • “clean”[makúr] → [makr-él] “clean”
3. • “color” [kújn] → [kun-avór] “colorful”
   • “culture” [məʃagújt] → [məʃagut-ajín] “cultural”

Many languages exhibit phonological differences between stressed and unstressed vowels. There are languages, like Catalan, Bulgarian and Russian, where all unstressed vowels reduce. In another set of languages, like Palauan, Romanian and Armenian, only a certain set of vowels that once bore stress reduce when stress shifts.

Unlike Romanian and Palauan, Armenian has a phonological process of ə-epenthesis. I argue that the surface schwas that seem to be corresponding to the once stressed high vowels are also part of this epenthesis process. Without positing a new phonological mechanism to account for the high vowel disappearance, I incorporate the derived environment effect into the already present phonology of Western Armenian.

Title: Spoken Conversational Systems
Speaker: Stephanie Seneff, Spoken Language Systems group, CSAIL
When: Wednesday Nov 12, 3-4:30pm
Where: 26-310

The Spoken Language Systems group in CSAIL has been developing multimodal dialogue systems for over two decades. These systems typically provide information on a specific topic such as flight scheduling, weather, geographical information, calendar management, etc. Our goal has been to build systems that engage in natural spoken conversation, using a so-called “mixed-initiative” dialogue strategy. In this talk, I will first give a high level overview of system architecture and components. The main content of the talk will emphasize discourse and dialogue modelling, in the context of these spoken dialogue systems. I will also describe techniques used to stress test the system and guide system development, such as simulated dialogue interaction. I will mainly use the multimodal restaurant domain and the telephone-access flight scheduling domain as illustrative examples. Audio and video clips will be played to demonstrate system capabilities.