CLS 46 Schedule

Recently, an alternative semantics originally proposed for questions in English (Hamblin, 1973) has been extended to natural language quantification (Ramchand, 1997; Hagstrom, 1998), including indeterminate phrases (Kratzer and Shimoyama, 2002; Shimoyama, 2006) and disjunction (Alonso-Ovalle, 2006; Hulsey, 2008). On this approach, several facts about the interpretation and distribution of indeterminate phrases and disjunction fall out naturally, including locality conditions. In the paper, we extend the approach to conjunction. We propose that natural language conjunction is a set forming operator and conjoined structures denote the set whose members are Hamblin alternatives created by the conjuncts. We present the observation that modals in gapping with conjunction have both wide and narrow scope readings in Russian. We argue that adopting a Hamblin semantics (Hamblin, 1973; Kratzer and Shimoyama, 2002) for conjunction accounts for the data without complicating the syntax of gapping constructions. The new approach gives a natural explanation for the syntax-semantics of gapping constructions in English and Russian. It provides further evidence for Hamblin alternatives as an analytical tool and sheds light on the nature of existential closure by addressing why a logical possibility people don't normally attend to -- that there might be closure operations with other quantificational force, such as universal -- might actually be realized.

We first introduce and describe two English constructions which have not previously received adequate treatment: the "accommodation" construction ("My car seats four people") and "equative make" construction ("Clowns make good fathers"). We show that, despite their superficial differences, these constructions belong to a previously unrecognized class of middles: namely, one which retains a surface object, unlike canonical middles ("Bureaucrats bribe easily").

Then, we offer an analysis of our new class of middle voice constructions, as well as of canonical middles, that relates them to passives (à la Keyser and Roeper 1984). To do so, we exploit existing syntactic machinery for deriving passives, and show that they can be extended straightforwardly to capture middles. Specifically, we assume (i) that internal arguments can be "smuggled" past intervening DPs before raising into subject position to overcome potential minimality violations (Collins 2005), and (ii) a clause’s external argument is base-generated in [Spec, VoiceP], the head of which bears the clause’s voice morphology (Kratzer 1996).

As a logical extension of Kratzer's proposal, we take the Voice head to be the locus of all featural (and therefore structural) differences among active, passive, and middle clauses. Therefore, we propose that Voice heads come in three flavors: Active, Passive and Middle. Unlike its Active counterpart, Middle does not introduce an external argument; however, it tolerates apparent minimality violations, via smuggling, similar to Passive.

We conclude by exploring the structural and typological predictions made by our proposal, including possible extensions to unaccusative and raising verbs.

A valuable question illuminating the linguistic and extra-linguistic problems of orthography development is why an orthography for a large language like ASL remains controversial in spite of proposals that have enthusiastic partisans (e.g., Sutton SignWriting) or acceptance among groups of users, such as academics (e.g., Stokoe Notation).

We seek to shed light on obstacles to the acceptance and adoption of orthographies proposed for ASL by examining the degree of underrepresentation and the depth of the orthographies, and by probing the socio-political and other attitudes of potential users. Our examination of the attitudes of potential users takes the form of a survey of Deaf attitudes toward Sutton’s and Stokoe’s proposed orthographies, including perceived ease of learning the orthographies and value of the orthographies to the learner. We also elicit responses relating to perceptions of the appropriateness of the orthographies in various genres of written communication. We also evaluate users’ general attitudes toward the need for a writing system for ASL that is distinctive from American English writing.

ASL is an instructive case study in orthography development. The factors contributing to potentials users’ rejection of Stokoe and Sutton emphasize a number of important points relevant to those developing orthographies for unwritten languages, such as the importance of the right balance in emblematicity, the indispensability of technical utility, and finding an appropriate level of divergence from the phonology of the language.

Previous work shows that in short term, laboratory settings, aspects of one's speech can change under exposure to the speech of others (e.g. Goldinger 1998, Nielsen 2007), and that this change is mediated by social variables such as (speaker) gender (e.g. Namy et al. 2002, Pardo 2006). An implicit hypothesis is that these effects can help explain dialect formation and the social stratification of speech. However, it is not known whether such change occurs in natural interaction over a longer term.

This study shows longitudinal change in VOT in a closed linguistic system, mediated by social interaction. Our corpus consists of speech from Big Brother UK (2008), a reality TV show in which 16 contestants live in a house for 93 days, subject to 24 hour audio/video recording. VOT was measured for 4 contestants over the season. We model the effect of time on VOT for different contestants, controlling for known confounds (e.g. speech rate, place of articulation), and allowing for intrinsic differences in the VOTs of different words. Each contestant's modeled VOT time trajectory shows significant longitudinal change, and all trajectories appear to move closer together over time.

We then examine the effect of two measures of social interaction on differences between VOT trajectories for pairs of contestants. The more a pair interacts, as measured via live blog entries by a UK newspaper, the closer their VOTs become. When a pair is on the same side of an artificial divide in the house (present for half the season), their VOTs become closer than if they are on different sides.

Merge - characterized as the one operation that is cost free in narrow syntax - is a symmetric operation that, in the standard case, yields an asymmetric and therefore labelled output (cf. e.g. Boeckx 2008). In this paper two types of structures in which Merge does not yield an asymmetric output are taken into account: nominal root compounding and incorporation in nominal gerunds. In each of these structures Merge combines two syntactic objects in such a way that a point of symmetry arises.

(1a) [_? [_V cut] [_N grass] ] (1b) [_? [_V cut] [_N grass] ]

Roeper (et al.) (2002) postulate an Abstract Clitic Position in root compounds and develop a theory involving leftward movement of the clitic element into a higher speci?er position, however leaving the motivation for movement unclear. Barrie (2006), following Moro (2000) develops a similar account for noun-incorporation in gerund structures and motivates movement in terms of complement-speci?er roll-up. Harley (2009) argues for an abstract Case-feature as the trigger for movement. However, as Boeckx (2008) points out, this type of movement regardless of whether it is motivated by Comp-to- Spec roll-up or by an EPP- or Case-feature is problematic in terms of linearization, because it leads to ambiguous precedence information.

Combining insights from Moro (2000) and phase theory, this paper aims to develop a uniform analysis that can account for each of these structures and at the same time avoids stipulation of any of the above mentioned ad-hoc features and mechanisms.

In this talk I will argue for a theory of meter that conceives it as faithfulness in Optimality-Theoretic terms. Metrical templates, consisting of segmentally unspecified prosodic structure represented as a bracketed grid, act as partial underlying forms of the meter. Standard correspondence theory furnishes the constraints that regulate the similarity of that template to the prosodic structure of the text. The theory is parsimonious, in that it contains no constraints specific to metrics. There are constraints on both prominence and constituency: Max requires strong positions in the meter to be realized with stress; Dep requires the opposite for weak positions; while the Align family calls for metrical and prosodic constituent boundaries to coincide. I will show that this approach leads to a productive analysis of the English iambic meters, including some non-trivial restrictions on the distribution of text-template mismatches. An argument in favor of the meter-as-faithfulness approach is that some previously stipulated properties of meter fall out from the correspondence theory itself. In particular, the monosyllable constraint (which prohibits the main stress of polysyllabic words from occurring in weak positions) results from the properties of Max and Dep. A monosyllable in weak positions violates only Max, while a polysyllable whose main stress is in a weak position violates both Max and Dep -- a categorically worse violation. I will conclude with a brief overview of the typology of English iambic meters from the point of view of the proposed system.

Increased multilingualism frequently leads to a loss of linguistic diversity as speakers of less-widely-used languages shift to more powerful languages of wider communication, leading to extinction. In some places, however, stable multilingualism has developed. In this paper I document cases of stable multilingualism in three language communities in Tajikistan. In the process, I present ways to measure factors which lead to language maintenance.

Yaghnobi, the most widely-used of the three languages, has only 13,000 speakers; the other two, Parya and Ishkashimi, have 5000 and 1000 speakers, respectively. All three communities interact regularly with neighboring Tajik or Uzbek-speaking communities. None of the three languages have official governmental status; none are used in the schools. Furthermore, our research indicates that speakers in all three communities have high levels of proficiency in Tajik or Uzbek. In spite of these negative factors, each of them is actively used by all age groups. The traditional vernaculars and the dominant regional languages have developed into stable diglossic relationships: The traditional vernacular is used in the home and community, while the dominant language is used in official settings.

Initial observations indicated that the strength of the vernaculars was due to two factors claimed to support language maintenance elsewhere: dense, complex social networks and the role of language in ethnic identity. I present tools we developed to measure these factors, and show that both are strong in the communities under consideration. I close with suggestions concerning how to support stable multilingualism.

We investigate the performance of the Hayes and Wilson (2008) Maximum Entropy (MaxEnt) Phonotactic Learner on Amharic, a language with gradient phonotactic restrictions. We evaluate both the log-likelihood of the training data and the correlation between model predictions and speaker judgments. Amharic verb roots show OCP violations for place of articulation (Bender & Fulass 1978, Rose & King 2007). OCP-Place in Amharic poses two challenges: (1) Constraint length. OCP-Place spans up to three consonants. (2) Gradiency. OCP-Place is stronger in some word positions and for certain places of articulation. We evaluated automatically learned grammars and a (MaxEnt weighted) hand-written grammar encoding OCP-Place restrictions, assessing model quality by cross-validated log-likelihood on training data and on accuracy in predicting speaker judgments on nonce forms. The hand-written grammar achieved better correlation with speaker judgments (r=0.45) than any automatically-learned grammar (r=0.39). Furthermore, the relationship between cross-validated log likelihood and correlation with speaker judgments was complex: in some cases, adding a constraint to the model improved log-likelihood but reducedcorrelation with speaker judgments. These results call into question the linking hypothesis between probability and well-formedness judgments in gradient phonotactic models, suggesting that under-representation of sound sequences may not always correspond to active speaker dispreferences.

The topic of this talk are certain interpretations of sequences of imperative and declarative clauses, sometimes with a coordinator 'and' between them. (1) is an example.

(1) Give him a thousand bucks, if you like. He (still) won't budge

= (Give him $1000, if you like.) Even if you do that, he won't budge

There are three characteristics of these constructions that we will account for: (i) A scalar particle is mandatory for an unconditional interpretation, esp. if there is a coordinator between the sentences and the first sentence is not 'pragmatically extreme' (2).

(2) Talk to him #(even) all night and he #(still) won't feel any better

(ii) The imperative in (1) is clearly not interpreted as expressing a command/obligation. It can only be interpreted as a permission. (iii) A sequence of an existential modal clause and a declarative clause can also convey an unconditional meaning. This also holds if there is an overt coordinator between the sentences (3).

(3) You can give him $1000 and he (still) won't change his mind

= Even if you give him $1000, he won't change his mind

Analysis: Permissions and existential modals feature in above sequences because of their augmenting nature - they expand the respective information state; universal modals shrink it. This conditions what propositional referents these sentences make available for the subsequent discourse. Adopting an anaphoric approach to modal subordination, the modal base of 'will' in the second sentence has to be appropriately resolved. Different resolutions are then possible with different first sentences in conjunction and juxtaposition. Finally, the scalar particles have the standard presuppositions, which license the unconditional inference.

It is well-known that negative polarity items (NPIs) can occur in the antecedent of hypothetical conditionals such as (1) if they are law-like (Heim 1984).

(1) If Jesse lifts a finger to help Alex, I will be disappointed.

A constellation hitherto not discussed is the distribution of NPIs in biscuit conditionals (BCs), which is even more restricted.

(2) ?? If you want to lift a finger to help, there are ten boxes in that corner that you can unpack.

I account for this difference in distribution by applying a pragmatic theory of NPI licensing along the lines of Krifka (1995) to a Kratzer-style semantics of conditionals, showing that NPIs like lift a finger are only licensed if the ordering source reflects a certain correlation between the "strength" of the antecedent and the informativity of the overall sentence. This is almost never the case in biscuit conditionals, rendering them bad most of the time.

The notion of linguistic drift, first proposed by Sapir (1921), focuses on inherited or natural tendencies of distinct linguistic varieties. Trudgill, Gordon, Lewis, & MacLagan (2000) furthers this discussion, distinguishing two types of linguistic drift: Drift 1, in which inherited linguistic changes in progress from the parent variety are independently continued in disconnected child varieties, and Drift 2, in which inherited propensities toward linguistic changes from a parent variety are independently realized in disconnected child varieties.

We explore linguistic drift by analyzing two Insular Catalan varieties that can be considered child language varieties of a Peninsular Catalan variety. Though we report Insular Catalan evidence for Drift 1 and Drift 2, our discussion focuses on the existence of vitalic Insular phonological and morphological variants that are vestigial and/or archaic (cf. Trudgill 1999) in Peninsular Catalan and therefore are not able to be accounted for by either type of linguistic drift.

In order to account for the presence of relic variants in diaspora varieties, linguistic drift must take into account certain social factors. We propose that first-order social networks contribute to relic variant maintenance. Additionally, we emphasize the role of language contact in the unique historical development of Algherese, discussing its implications for linguistic drift theory and referencing previous treatments of language contact as a motivator behind strictly apparent reversals of linguistic drift and as an inhibitor of linguistic innovation.

In various languages (in)definiteness is or can be morphologically expressed on adjectives, either as an affix or as a clitic. I argue that the clitic/affix distinction has semantic ramifications arising from the different ways in which affixes and clitics combine with their hosts. I observe that definite affixes, but not definite clitics, systematically restrict the types of readings an adjective may have. Specifically, adjectives that are ambiguous between an intersective and subsective reading (e.g., Kamp & Partee 1995, Siegel 1976) in indefinite contexts lose the subsective reading in their definite form, but only in languages where definite markers are morphologically affixal. I propose that this difference follows from the compositional semantics, but only on the assumption that affixes combine with their adjectival host prior to combining with the noun (1a), while clitics pattern syntactically with free determiners, taking the NP consisting of adjective and noun as their complements (1b).

(1) a. [[ADJ-DEF_AFFIX] NOUN]. b. [DEF_CLITIC [ADJ NOUN]].

Languages that this presentation mainly concentrates on are (i) Icelandic and Serbo-Croatian (the affix-type) and (ii) Bulgarian and Romanian (the clitic-type). In addition to shedding further light on the semantics of definiteness, this paper argues that the lowering of the clitic to the adjective in Bulgarian must be at a post-syntactic level that does not feed semantic interpretation, as proposed in Embick and Noyer (2001), and argues against the lexicalist account on which the [Noun/Adj - def. article] sequence in this language is derived in the lexicon.

Special syntactic patterns often go hand-in-hand with special meanings. A well-documented case is found in Hungarian, where the immediate preverbal position is argued to be semantically exhaustive (Szalbosci 1981). This claim is extended to the English it-clefts (Kiss 1998). In this paper, I explore the issue in French, investigating how non-canonical argument realization differs in interpretation from SVO forms. I argue the French c'est-cleft is indeed special regarding exhaustivity, yet this meaning is not truth-conditional. In line with Horn (1981), I claim the exhaustive inference is a conversational implicature and can be either reinforced or cancelled. Thus, I advocate a pragmatic account of the inference (Wedgwood 2005 and Onéa&Beaver 2009), and explain the cleft construction is a tool that selects the best variable answering the question under discussion and restricts the domain of alternatives determined by the context; the upper bound being the universal quantifier and the lower bound being the empty set. I present two types of new evidence to support my account: (1) Experimental data showing the level of exhaustivity in the cleft is higher than in SVO forms and lower than in only-sentences, and (2) corpus data showing a variety of particles including quantifiers felicitously occur in the cleft.

Korean verbs show numerous alternations. Children acquiring Korean are reported to go through a stage in which they produce some alternations but not others (Do, to appear). The current simulation uses the MaxEnt Grammar Tool (Hayes, 2006) to explore why some alternations are acquired late. It shows that the frequency of different alternations in Korean can give rise to the attested intermediate stage, without assuming intrinsic bias.

The learner is assumed to have a set of constraints, and weights them according to the frequency of violations in the data. Following Hayes (2004), learning is simulated in two stages; phonotactic and morphological learning. The phonotactic learning stage models the fact that even prior to learning morphology, learners master phonotactic distributions. Once morphological relations are discovered, learners may wish to avoid alternations. I model this with output-output faithfulness constraints (OO-F), demanding faithful to a base form.

After two-stage learning, the learner was able to demote all OO-F to master all alternations. The interest of the current study, though, is to simulate a stage in which learning is incomplete. I simulated this by controlling how freely the model can change the weights of OO-F. By decreasing sigma, the learner is biased to leave the weights closer to their high initial weights.

The grammar trained purely by the frequency of different alternations perfectly predicts the child forms. Therefore, this study demonstrates that even without intrinsic bias, the statistics of Korean gives rise to the attested intermediate learning stage.

The question of whether and to what extent information from different linguistic levels is integrated during language processing remains unresolved. Here, we explore the maximal range of such interactive effects by testing the effect of pragmatics, arguably the top-most level, on phonetic perception, the bottom-most. We report on a phonetic category perception experiment in which (i) pragmatic inference guides expectations about pronominal co-reference and (ii) the pronouns are acoustically ambiguous. The goal is to establish whether listeners' top-down discourse expectations about which individual will be mentioned next can influence their bottom-up perception of speech. We also explore the time-course of the effect by using gated portions of the pronoun at multiple time-points to assess whether the bias obtains only with exposure to the entire word, or earlier in processing.

The results show that female-biasing contexts do yield higher 'she' ratings compared to male-biasing contexts (p<0.01). These results extend existing work on phonetic category perception showing that perception is sensitive not only to previously identified cues such as lexical status and prosodic boundaries, but also to discourse-level pragmatic cues. Furthermore the time-course results suggest this effect is present at the earliest stage of processing, reflecting expectation, and again at the final resolution of ambiguity when processing the whole word. Thus, we show that the range of interacting cues is larger than previously thought and that phonetic perception interacts with higher-level causal inferencing about events and co-reference across clauses in a discourse.

Gass and Ard (1984) proposed that only the “core” meanings of a given form transfer across languages. More recently, several studies targeting aspectual coercion have also shown evidence of difficulty when the L1 and L2 exhibit similar properties (Gabriele et al. 2009; Slabakova and Montrul, 2007), raising the question of whether coerced meanings do not transfer. We examine the interpretation of the simple present and the present progressive in English by native speakers of Spanish and Japanese. Both forms allow ‘shifts’ in meaning depending on the particular context or the presence of certain temporal adverbials (Binnick, 1991; Copley, 2007; Krifka et al., 1995; De Swart, 1998). For example, although the progressive is usually used to refer to an activity in progress (Sophia is playing basketball now), in certain contexts the progressive can refer to a habitual event (John is drinking wine this week), a futurate event (John is leaving for Italy next week) or a temporary state (Sophia is wanting a new job) A subset of the coerced meanings are possible in all languages (progressive habitual) while others are possible only in English (stative/futurate progressive). Results of a grammaticality judgment task suggested, similar to previous studies, that the coerced meanings were more difficult regardless of similarity in the L1 and L2. However, results of an interpretation task showed target-like performance, suggesting that coerced meanings can transfer and that L2 learners can successfully integrate knowledge of discourse constraints in the interpretation of a given form (cf. Sorace, 2005).

We will present an S(hortest)M(ove)C(ondition)-compatible implementation of multiple-wh-interrogatives based on the wh-clustering approach by Grewendorf (2001) and Sabel (2001) within the framework of Stablerian Minimalist Grammars. We will further argue that application of this approach to all languages (that possess multiple-wh-interrogatives) is superior to limiting its scope to specific individual languages. The proper analysis of wh-indefinites and "additional wh-effects" belongs among the major design challenges.

Ethnic adjectives (EAs) such as French have been attributed two distinct uses: a "thematic" use, typical with nominalizations ((1a)), and a "classificatory" use ((1b)), most obviously found with nonderived nominals though also possible with nominalizations (Kayne 1984, Bosque & Picallo 1996, Fábregas 2007, Alexiadou&Stavrou, to appear).

(1) a. French agreement (to join the negotiations) b. French wine

On the thematic use, the adjective describes a participant in the situation described by the verb underlying the nominalization. This has led some authors to analyze them as proper arguments of the nominalization. (1a) is then equivalent to (2) in a strict sense, and ethnic adjectives are nouns at some level of representation.

(2) agreement by France (to join the negotiations)

We challenge this view, offering an analysis which maintains that ethnic adjectives are adjectives and gives a unified account of the semantics of thematic and classificatory uses of EAs, while at the same time accounting for the apparent synonymy between (1a) and (2). In this analysis, ethnic adjectives are not syntactic arguments of the nominalizations, and the role-like interpretation arises through the interaction between the semantics of the EA and that of the head noun.

NPI licensing in clausal comparatives (Hoeksema 1983, von Stechow 1984, Giannakidou 1998, among others) is generally thought to be free. Yet why NPIs would be licensed in comparatives does not follow from any of the current theories of NPIs, and it remains controversial whether clausal comparatives are negative, downward entailing, or nonveridical (Larson 1988; Schwarzschild and Wilkinson 2002; Heim 2006; Giannakidou 1998). In this paper, we want to make a primarily empirical point. We show that the comparative is not, in fact, a licensing environment for NPIs, and that the appearance of NPIs in comparative is quite restricted. We show that (a) it is easy to confuse NPIs with free choice items (FCIs) in the comparative; (b) strong NPIs (and we illustrate with English, Greek and Korean) are excluded; and (c) only the weakest NPI type appears in the comparative, i.e. those that need not be sanctioned via licensing (i.e. by being in the scope of an appropriate expression at LF), but via rescuing (in the sense of Giannakidou 2006, i.e. sanctioned via a global pragmatic inference, without being in the scope of a appropriate operator at LF). Any and English minimizers are typical cases of such NPIs. Our observation that stronger NPI types (that can only be licensed in the scope of an antiveridical operator, often negation) do not appear in the comparative is consistent with analyses that do not posit negation in the comparative clause; and we show that if an antiveridical expression is introduced, such as the negative metalinguistic negation charari (Giannakidou and Yoon (GY) 2009), strong NPIs become fine. In conclusion, if NPI sanctioning in comparatives is actually limited to global pragmatic sanctioning, as we argue, and there is no NPI-licensing in comparatives, the difficulty in proving downward entailment or nonveridicality for the comparative is no longer surprising.

This talk investigates the German particle doch, contrasting it with the particle ja (Weydt 1969). I propose that in declaratives, ja and doch are weak and strong counterparts of each other, in the following sense. They share a core meaning component (uncontroversiality/>y, cf. Kratzer 1999), but doch has an additional meaning component (contrast/correction, cf. Thurmair 1991). It follows that ja and doch on their own are in competition. The particle ja is used when the presuppositions for doch are not met; in contrast, doch is used when its presuppositions are met, due to Maximize Presupposition (Heim 1991).

In my analysis of doch, I argue that the correction component operates on propositional alternatives (doch reinforces the modified proposition p in contrast to a contextually salient alternative q that contradicts p) and is presuppositional in nature. I argue that doch makes use of an alternative semantics, associating with focus. This predicts correctly that doch triggers intervention effects (Beck 2006): It cannot associate with the same focus as another focus-sensitive element, such as nur ‘only’. My analysis accounts for ordering restrictions, which permit ja doch, but rule out doch ja. Kratzer (1999) argues that ja operates on complete propositions and cannot occur between a quantifier and a variable that it binds. I show that ja also cannot intervene between a focus-sensitive particle, like nur ‘only’ and the focus. Given that doch is focus-sensitive, we correctly rule out doch ja, but not ja doch.

Our experiment investigated whether any of the two possible ways of characterizing the ordinary vs. contrastive topic distinction in Hungarian, based on information-structural vs. syntactic criteria, correlate with systematic prosodic differences.

From an information-structural perspective, constituents that denote what the sentence is about are contrasted to those that in addition “indicate alternative aboutness topics” (Krifka 2007). From the perspective of syntax, a different distinction has been emphasized: that between arguments of the verb having the features [+referential] and [+specific], which can appear in the [Spec,TopP] position of the sentence without strong prosodic prominence of pronounced changes of pitch, and those that can only occupy the same position if they are pronounced with a “contrastive intonation comprised of a brief fall and a long rise” (É.Kiss 2002), which correlates with an “interpretational surplus” (Szabolcsi 1981).

In our experiment, pairs of speakers were asked to read out mini-dialogues forming prototypical contexts for ordinary vs. contrastive topic interpretation. Accent tones of the topic and the focus do not differ with respect to topic type: initial accents were rising and focus accents were falling. For [+referential] and [+specific] constituents (proper names), pitch patterns between the topic and the focus accent were different, however: HL was more frequent for ordinary and LH for contrastive topics, while H was common for both. According to a Chi-square test, constituents outside the above group showed the same pattern as the contrastive topics within it. The theoretical and cross-linguistic implications of these findings are also discussed in the talk.

We investigate embedded identity question with post-copular phrases, observing that - unexpectedly - not all definite descriptions are felicitous in this position (Who do you think is the violinist? vs. #Who do you think is the man on the left?). Since extraction out of copular questions is possible only for predicational sentences (Higgins 1973), the question raised by this contrast is why certain definite descriptions can function as predicates while others cannot.

We compare the infelicitous cases to demonstrative descriptions like #Who do you think is that man?. For demonstratives, we follow Kaplan (1989) in analyzing the demonstrative phrase as directly referential and argue that directly referential phrases cannot function as predicates. Diverging from Kaplan, we argue that direct reference is not a semantic (or lexical) property of demonstratives, and propose that direct reference is also found for definite descriptions, in contexts where a property in a description is perceived by the interlocutors as holding of the referent in the physical context (we call these perceptually grounded descriptions). As predicted, the felicity of embedded identity questions can be manipulated by changing properties of the physical context of utterance, which affects whether the definite description is perceptually grounded. Interestingly, the notion of perceptually grounded definite descriptions also plays a role in the interpretation of counterfactuals, where we find that perceptually grounded definites act like (demonstrated) demonstratives in taking obligatory wide scope over the modal. Perceptual grounded description shed new light on the interconnections between the semantic composition and the pragmatics of referring expressions.

According to Li (1999), nominal expressions in Mandarin Chinese, a classifier language, can be analyzed in terms of the structure [_DP D [_NumP numeral [Num [_ClP classifier [_NP N]]]]], where the classifier takes an NP complement to head Classifier Phrase (ClP), which in turn complements Num, whose Spec is occupied by a numeral expression. If this is the only UG option available for classifier languages, Japanese should also observe it, and behave the same as Mandarin in the structurally relevant respects. Two facts contradict this prediction, however. First, classifiers fail to occur without a numeral in Japanese, while this is possible in Mandarin. Second, Japanese, unlike Mandarin, allows the membership of a group noun to be enumerated in apparently the same way as other, non-group nouns (e.g., sanninno kazoku ‘family of three’). It seems therefore that Japanese classifiers have a more distant relationship to the head noun than the Cl-NP complementation suggests, but rather have a closer relationship to the numeral than the structural configuration held between SpecNumP and Cl suggests.

This paper proposes that in Japanese the classifier obligatorily takes a numeral complement, not an NP complement, to form a constituent (i.e., [_ClP (Spec) [[numeral] classifier]]), severing itself and the numeral from the rest of DP (i.e., [_DP [_NumP [_NP N] Num] D]). Numerical specification is then implemented by a ClP’s adjunction to one of the nominal extended projections, namely NP, NumP, and DP.

The present hypothesis, we argue, explains a number of facts, including those two mentioned above.

An ongoing debate focuses on whether ergative case is inherent (Woolford 1997, Massam 2002, Legate 2005, 2008 among others) or structural (Marantz 1991, Bobaljik 1993, Bittner and Hale 1996, Preminger 2009, among others). In this paper we present data from the ergative language Kapampangan (Austronesian; Philippines) and argue that ergative case cannot be inherent in this language. Kapampangan pronouns are second position clitics that may double the argument NPs with which they agree in person, number and case. Ergative clitics appear in two patterns where they cannot be associated with a thematic subject. The first is weather verbs and the second are unaccusatives with perfective aspect. In these patterns the clitic pronoun cluster exhibits the pattern [Erg.3sg+Abs.a]. Since the absolutive clitic crossreferences the sole thematic argument, the fixed 3rd person singular ergative clitic must crossreference something else. It is difficult to explain the presence of fixed 3rd person singular agreement on the ergative clitic in just these environments without relating these features to the feature specification of T, since they cannot be related to any thematic position. We argue that the 3rd person singular ergative clitic doubles a null expletive in [Spec, TP], which stands in an Agree relation with T.

Research demonstrating that both infants and adults track statistical distributions of acoustic-phonetic cues and use this information when making phonetic category judgements (Maye et al., 2002; Clayards et al., 2008) has led to interest in computational models of phonetic category acquisition, which can shed light on the requirements and limitations of statistical learning. While initial forays into this area have yielded encouraging results (e.g., de Boer and Kuhl, 2003; Vallabha et al., 2007), these studies have considered relatively well-separated, low-dimensional contrasts, such as subsets of the American English vowel space. One might expect contrasts which are not as robustly cued -- instances of so-called near mergers or incomplete neutralizations -- to present greater difficulty for purely statistical learning approaches.

This paper considers the transmission of near mergers in light of a general unsupervised learning approach to the acquisition of phonetic category structure, where the number of categories to be learned is not specified in advance. If a contrast between two categories X and Y is sufficiently robust, the learner is predicted to converge on the distributions which generated the input; otherwise, the result should be (complete) neutralization, with the learner positing just a single category. The results of training and testing on data from English, Korean, Dutch, and Turkish indicate that a purely statistical learner is capable of correctly inducing extremely subtle phonetic contrasts, suggesting that it is possible for an incompletely neutralized contrast to persist in the transfer from one generation to the next.

The study compares the vowel structures of Mexican Heritage English (MHE) and the regional Anglo dialect in the Albany Park neighborhood of Chicago. Analysis of temporal properties of MHE vowels (i.e., durations and formant trajectories) provides evidence of a Spanish language influence not captured in traditional vowel plots. The detailed characterization of the resultant MHE vowel structure provides insight into the interaction of the disparate Spanish and English vowel systems.

Neighborhood participants comprised four groups: Anglo speakers; MHE speakers; Mexican late learners of English (L2E); and native Mexican Spanish speakers. English wordlist productions included at least 10 tokens of each of eleven English vowels from a range of consonantal contexts. A Spanish wordlist was used to produce a corresponding analysis.

Unlike L2E, MHE displays a well-differentiated inventory of eleven vowels whose first and second formant centers are very similar to those of the Anglo speakers. However, consistent differences between MHE and Anglo patterns occur in the dynamic features. Vowel lengthening before voiced consonants and long/short duration ratios vary significantly between the two groups. Formant trajectory for [æ] also differs significantly. Overall, MHE shares the category structure of the Anglo variety while differing in the secondary vocalic characteristics observed in Spanish.

The study corroborates recent research suggesting [æ] as a key component of MHE distinctiveness (Roeder, 2006), but shows that the trajectory of [æ] is the distinguishing feature. The results contrast with California MHE patterns (Godinez and Maddieson, 1985) in that vowel duration is a defining feature of Chicago MHE.

According to Steriade (1997), phonetic environments can be ranked from more to less favorable for the presence of voicing contrasts word initially. The most highly ranked and optimal environment for preservation of voicing in obstruents is pre-vocalically (O_[vd]V); next, and less optimal is the environment before sonorants (O_[vd]R); even less optimal is the environment before voiced obstruents O_[vd]O_[vd]; and the most lowly ranked and least optimal phonetic environment for the preservation of voicing contrasts is before voiceless obstruents (O_[vd]O_[vcls]). Steriade claims that this hierarchy, which accounts for cross-linguistic typological distribution of voicing contrasts, is perceptually based. Her claim: voicing contrasts are more likely to occur in more highly ranked environments because the more highly an environment is ranked, the more perceptual cues it provides. In this study I investigate the production end, the phonetic realization of voicing in different phonetic environments. Results of a phonetic experiment show: while voicing contrasts are realized in all target environments, what varies is the rate at which voicing is realized in each environment. When voiced obstruents precede a vowel they are voiced in 100% of the tokens; preceding sonorants in 79%; preceding voiced obstruents in 57%, and when preceding voiceless obstruents in only 31%. Surprisingly, and unexpectedly, voiced segments surface as voiceless in the environments #O_[vd]R and #O_[vd]O_[vd], where voicing is expected to be realized because there is no trigger for voicelessness in the immediate phonetic environment. This is attributed to failure to complete, or properly coordinate the gestures necessary for the initiation of voicing.

The meaning of aspectual adverbs has been much discussed in the literature (König, 1977; Hoepelman and Rohrer, 1981; Löbner, 1989, 1999; Mittwoch, 1993; van der Auwera, 1993, 1998; Michaelis, 1993, 1996; Smessaert and ter Meulen, 2004). However, few systematic analyses of their polarity sensitivity have been proposed although they display a nice set of NPI and PPI pairs in the eventuality domain. Israel (1995) is the only work that deals with this issue, which provides an account based on pragmatic scales. In this paper, I argue that their polarity sensitivity derives from their logical monotonicity properties rather than defeasible pragmatic implicatures. Israel’s pragmatic analysis cannot explain why polarity failure results in ungrammaticality, rather than pragmatic awkwardness. Contrary to his claim, the use of aspectual adverbs does not require some pragmatic norm or the speaker expectation of earliness or lateness. I show that aspectual NPIs yet and anymore are licensed by negative downward-entailing expressions, while aspectual PPIs already and STILL (emphatic still) repel them (Ladusaw, 1977; Kadmon and Landman, 1993). Aspectual polarity items in light of a (non)veridicality analysis of PIs (Giannakidou, 1998, 2006) are also discussed. I argue that the polarity sensitivity of aspectual adverbs derives from their aspectual meaning, as they make an explicit temporal contrast between the described situation and an alternative (presupposed) situation. A semantic analysis of aspectual PIs is provided, which also explains why only still, among aspectual adverbs, requires focus stress to be an NPI. This grammatically driven approach to aspectual polarity is in line with recent developments in scalar inferences and polarity items (Krifka, 2003; Chierchia, 2006), which treats scalar inferences as part of recursive semantics, just like truth-conditional meanings.

This paper investigates second language (L2) learners’ knowledge of the constraints on Korean numeral quantifier float and aims to shed light on the role of Universal Grammar (UG) in second language acquisition. In essence, the idea of the accessibility of UG is motivated by the fact that (some) L2 learners are able to overcome learnability issues and successfully acquire complex grammars that constitute poverty-of-the-stimulus problems. In Korean, floated numeral quantifiers (NQ) can be associated with the direct object of transitive verbs, but not with the subject of transitive verbs. Another asymmetry is also found with the two types of intransitive verbs; NQ-float is allowed on the sole argument of unaccusative verbs, but not on the sole argument of unergative verbs. These constrains on Korean floating NQs constitute poverty-of-the-stimulus problems for L2 learners of Korean whose L1 is English, in that English does not exhibit NQs on the one hand and English floating quantifiers (mainly universal quantifiers) are not bound by the same type of constraints found with Korean floating NQs on the other. Thirty-seven advanced English-speaking L2 learners of Korean participated in a written acceptability judgment task, and the results show that most of them (n=30) were able to distinguish between the subject and the direct object of transitive verbs, despite the differing properties of English and Korean floating quantifiers. Furthermore, some of them (n=5) were able to detect both types of asymmetry (transitive and intransitive), which add support to the claim that L2 acquisition may be mediated by UG.

Emirati Arabic has two major types of wh-constructions, i.e. wh-fronting ([ʃu ʔɛʃtr-ɛt ʔmɛs?] ‘what bought.2sg.M yesterday’) and wh-cleft ([ʃu ɛlli ʔɛʃtrɛt-ah ʔmɛs?] ‘what that bought-3sg.M yesterday’). Wh-fronting is defined by a sentence-initial wh followed by a wh-gap at the base position, whereas wh-cleft is characterized by a relative marker [?lli] and a resumptive prounoun. The two wh-constructions differ in various aspects and therefore involve distinctive configurations. The first is about island constraints: wh-fronting observes all island constraints, whereas wh-cleft does not. Second, wh-fronting allows the use of all types of wh-expressions, whereas wh-cleft bans wh-adjuncts (e.g. ‘how’ and ‘why’). Third, wh-fronting (but not wh-cleft) can form headless free relatives. We argue that wh-fronting is a movement construction, whereas wh-cleft is a non-movement structure formed by a base-generated wh-word and a cleft construction. However the two constructions converge on reconstruction. Reconstruction on both wh-constructions is allowed only if the wh-phrase and the base position are not separated by islands. This resonates with Aoun and Li’s 2003 observation in Lebanese Arabic, in which island constraints condition wh-reconstruction while having no effect on movement per se. Emirati Arabic shows dialectal difference in that only strong islands preclude wh-reconstruction. This strengthens the hypothesis that reconstruction is not a diagnostic of movement. Moreover, that wh-movement observes all island constraints whereas wh-reconstruction observes some of them reveals an important design of grammar, i.e. island constraints can be stated as a derivational constraint within the narrow syntax, or as a representational constraint defined the interface level.

The insertion of [ɪz] (e.g. house -> hizouse) is a phenomenon that is prevalent in the American English hip-hop community in music and, increasingly, spontaneous speech. This investigation draws on data from a corpus (Lindsay 2006, 2009) comprising lyrics of 34,444 songs from the Original Hip-Hop Lyrics Archive (ohhla.com). Infixation into monosyllabic words has a clear pattern of insertion between onset and rime, while the infixation pattern for bisyllabic base words illustrates that the iz-infixation form is crucially linked to the uninfixed form's stress (also attested in Viau 2006, Lindsay 2006). Further, all infixed forms have penultimate stress, regardless of the position of the infix. Thus, ['mɛ.ɾəl] has the iz-infixed counterpart, [mɪ.'zɛ.ɾəl], while [bə.'haɪnd] has the infixed form [bə.'hɪ.zaɪnd]. However, some multisyllabic infixed forms do not follow the predicted pattern, such as ['sʌ.bɚb], which appears in the corpus as [sʌ.'bɪ.zɚb] (rather than [sɪ.'zʌ.bɚb]). In this paper, I show that the usage of iz-infixation appears to be encouraged by factors that are exterior to the word itself -- namely, constraints on textsetting/meter (following Hayes 2005) and poetic rhyme (following Zwicky 1976 and Holtman 1996). Infixation can occur in a syllable that is unstressed in the base form if it satisfies these prosodic constraints. I also demonstrate that these factors can be used to explain the multisyllabic forms that do not fit the purely stress-governed pattern. Lastly, I discuss segmental anomalies related to iz-infixation, such as the apparent gap for words whose stressed syllables begin with [Cj] and its relationship to Pig Latin analyses.

This paper examines the occurrence of evaluative adverbs, such as “unfortunately”, in questions in French and Catalan. We extend Bonami and Godard (2008)'s proposal for declaratives in French to provide a unitary account that includes questions, and explain the more restrictive behavior found in Catalan.

French allows evaluative adverbs both in assertions and (polar and wh) questions. In wh-interrogatives, the adverb is not interpreted inside each proposition constituting the denotation of the question. In order to explain this behavior, we propose that, in sentences containing a parenthetical in French, meaning is conveyed at two different tiers: the Truth-Conditional tier contains a proposition p or a set thereof p (depending on the clause type), while the Use-Conditional tier contains a force operator and the contribution of “unfortunately” applied to p or p. When the argument is p, the meaning conveyed is that no matter which p in p is true, it is unfortunate that p holds. This does not obtain when the argument is p.

As for Catalan, evaluative adverbs are usually claimed to be unacceptable in questions. Nevertheless, we show that they are acceptable in confirmatory questions, biased questions and wh-questions in which the speaker knows (part of) the answer. We propose that this is due to the fact that Catalan evaluative adverbs can only take a proposition as argument. Thus, they can only occur in a question if the speaker believes that one particular proposition in the question denotation is more likely to be true than the rest.

This talk investigates so-called “genus-species splits” (GSS) in German, a kind of topicalization construction where an overt lexical noun surfaces in the base position of the left-peripheral topic DP:

(1) Raubvögel kennt Peter nur wenige Bussarde

birds-of-prey knows Peter only few buzzards

The dependency between the topicalized DP and the “filled gap” in its base position is shown to be island-sensitive, hence the result of movement. We follow Fanselow and Cavar (2002), who appeal to Distributed Deletion in their account of regular split-topicalization. We argue that GSS arise due to conflicting interface requirements: assignment of topic/focus features of the head noun requires double pronunciation, which, however, is ruled out on general grounds. A repair mechanism replaces the focal noun with a noun that denotes an instance of the topic noun, constrained by the condition in (3):

(2) [wenige_FOC Raubvögel_TOP/FOC] kennt Peter nur <[_DP wenige_FOC Raubvögel_TOP/FOC]> replacement

[wenige_FOC Raubvögel_TOP/FOC] kennt Peter nur <[_DP wenige_FOC Bussarde_TOP/FOC]>

(3) Existential Closure (EC) of the lower/focal DP must entail the EC of the topic DP.

This will allow for the replacements leading to (1) and (4) while correctly ruling out those in (5):

(4) Gefährliche Raubvögel hat Peter nur drei Bussarde gesehen? dangerous birds-of-prey has Peter only three buzzards seen

(5) *Raubvögel hat Peter {keine/alle} Bussarde gesehen? *birds of prey has Peter no/all buzzards seen

The talk will present arguments against alternative analyses of the phenomenon and end with a discussion of the status of the proposed repair mechanism in grammar.

This paper focuses on the object-verb agreement pattern present in the perfective aspect in Kutchi Gujarati (Indo-Aryan). The language has a 3rd person object reflexive NP, whose presence triggers agreement mismatch: The verb appears to reflect the features of the subject instead of the object. Such agreement displacement also occurs in clauses where the subject consists of two conjoined DPs and a plural reflexive object, pot-pothane. Canonically, in clauses with conjuncts in argument positions, the verb typically shows plural agreement. However the presence of this 3rd person reflexive results in 1st conjunct agreement (Benmamoun 1992).

I propose reflexives cannot trigger verbal agreement (an anaphor agreement effect, Rizzi 1990), causing agreement displacement: The verb actually agrees with the subject rather than the reflexive object. I argue plural agreement is established via Agree under c-command (Chomsky 2001), a relationship between a probe and a goal, established when an agreeing head is merged. Contrastively, I argue that first conjunct agreement is a second cycle effect (Bejar and Rezac 2009), where a probe on v can search as high as its specifier if it does not find a suitable goal in its c-command domain. I argue subject &P’s are phases, and spell out at the next phase level; the difference between plural agreement and first conjunct agreement is whether the &P has spelled out (which results in first conjunct) or not (which results in plural agreement). It follows that agreement displacement happens late in the derivation, explaining 1st conjunct agreement.

Converbs are non-finite verb forms which mainly indicate adverbial subordination. Such forms pose various challenges for morphology, syntax, and pragmatics. In this work we show that the distribution of nine different kinds of converbal derivation patterns in Sakha (a head-final, suffixing language with SOV word order) can be captured by a unique lexical rule. This rule is general enough to allow the derived converbs to project phrases that occur in very different constructions (i.e. with an implicit subject co-referential with the subject of the modified finite VP, or with an overt subject which is not co-referential with the modified verbal structure). The converbal structures are assumed to function as modifier and therefore can also occur in chaining constructions, where they are stacked one after the other with a purpose of advancing the narration. Interestingly, the illocutionary scope in Sakha converbal constructions varies depending on the placement of focus, expressed by a pitch accent. In this paper we show that the principles that govern focus assignment in Sakha chaining constructions and other languages like English are essentially the same.

A number of studies have investigated the extent to which formant transitions, in addition to fricative noise, are used as a cue to fricative consonants, but with varying results (e.g. Harris (1958), Heinz & Stevens (1961), Jongman (1989)). Wagner et al (2006) aimed to resolve some of this ambiguity, and showed that the presence of spectrally similar fricatives in a listener's native phoneme inventory increased attention to formant transition cues in a phoneme monitoring task. Therefore, depending on the native language background, some listeners are predicted to weight particular cues more heavily than others when identifying fricatives. The present study provides a new line of evidence for cue weightings that are dependent on a listener's phoneme inventory. This study compares English and Mexican Spanish listeners in a fricative monitoring task, both in clear speech and noise conditions. Following Cooke's (2006) 'glimpsing' model of speech perception in noise, English listeners were predicted to be less affected by noise than Mexican Spanish listeners, since they can potentially extract more information from the signal. Sixteen listeners, eight native English and eight native Mexican Spanish, carried out the task. Stimuli consisted of pseudowords, produced by a native French speaker, and participants were instructed to respond upon hearing /f/ or /s/. Hit rate results revealed a statistically significant interaction between language, condition and target fricative ([F(1,14) = 5, p < 0.05]); Mexican Spanish speakers were more affected by noise than English listeners when listening for /f/. These results support language-specific accounts of perception, particularly highlighting the role of phoneme inventories.

Following the results of a diverse typological study, I argue that “dependent” tag questions (“DTQs”: Mary left, didn’t she?) are not unique to English (as has been claimed), but can in fact be found in several languages, and they consistently pattern with cases of VP ellipsis (VPE) found elsewhere in those languages. Thus, I claim that DTQs are derived by VPE. I support this claim first in English, by showing that DTQs exhibit hallmark properties of VPE with respect to auxiliary stranding and VPE licensing. Then, I show that DTQs behave like VPE in Taiwanese, Danish, Brazilian Portuguese, Scottish Gaelic, Persian, and likely Samoan. Thus, this study expands the empirical domain of ellipsis to include DTQs, and it also expands the empirical domain of DTQs to include languages other than English.

The purpose of this paper is to investigate the semantics and pragmatics of the Japanese minimizer PPIs and to consider the (non) parallelism between the meaning of the ‘amount minimizer’ as in (1) and that of the ‘expressive minimizer’ as in (2) in terms of the semantics-pragmatics interface:

(1) Kono sao-wa {chotto/sukoshi} magat-teiru.

This rod-TOP a little bend-PERF

‘This rod is a little bent.’

(2) {Chotto/*sukoshi} jikan-ga nai-desu.

A little time-NOM NEG-PRED.POL

At-issue: ‘I don’t have time.’

CI: The degree of commitment of the assertion is low.

The minimizers in (1) modify a lower-closed scale adjective (e.g. Kennedy 2007). Thus, (1) is interpreted as, ‘The degree of bentness of this rod exceeds the minimum standard by a small amount.’ By contrast, chotto in (2) is used to weaken the degree of illocutionary force (Matsumoto 1985). (Note that sukoshi does not have an expressive use.) In terms of the semantics/pragmatics interface, we can say that the amount minimizer contributes to ‘what is said’, while the expressive minimizer contributes to a CI (e.g. Grice 1975; Potts 2005).

I argue that although the meanings of the amount and expressive minimizers are logically and dimensionally different, they share the same scalar meaning. It will be shown that the semantic difference comes from the kind of argument a minimizer takes (i.e. a speech act, or an individual). This paper also shows that the multidimensional approach (Potts 2005) to polarity sensitivity can naturally explain the seemingly puzzling distribution patterns of the Japanese minimizer PPIs.

Speakers have reliable truth-judgments when comparing pluralities. The semantics of these constructions, however, cannot straightforwardly follow from the semantics generally assumed for comparatives (e.g., von Stechow 1984, Heim 1985, Kennedy 1997) or plurals (e.g., Link 1983, Landman 1989, Schwarzschild 1996). Past work on plural comparison (Matushanksy and Ruys, 2006) attempts to capture speakers’ intuitions in a semantics that reduces plural comparison to a multitude of comparisons between the individual members of compared pluralities. We present experimental evidence that plural comparison does not reduce to the comparison of degrees true of individual members, but rather to the comparison of collective degrees inferred from the pluralities involved.

Our results support the hypothesis that a plurality can have a single degree associated with it that differs from the maximal degrees true of each of its parts, and that this degree is calculated by averaging the maximal degrees of the individuals belonging to the plurality. Thus, collective properties of pluralities are compared. Plural comparison then proceeds just as singular comparison, where the property relevant for comparison is inferred by averaging the degrees associated with the individual members of each plurality. Translating differences between pluralities into a probabilistic truth value significantly improves the model’s fit to human data. Ongoing work investigates how the gradience in human judgments arises.

Bimodal bilinguals represent a unique class of dual language users because unlike unimodal bilinguals, they are fluent in a spoken and a signed language. While previous research suggests that unimodal bilinguals activate both of their languages in parallel (Marian & Spivey, 2003; Blumenfeld & Marian, 2007), this phenomenon can often be attributed to structural overlap between the languages. It is unclear whether languages that do not share form (or modality) are also activated in parallel during language comprehension. To investigate this question, twenty-six participants (13 hearing ASL-English bilinguals and 13 monolingual English speakers) were shown four images on a computer screen and were instructed, in English, to select a target item (e.g., “click on the cheese”). In competitor trials, the display contained a picture of an item whose ASL translation overlapped with that of the target. Bimodal bilinguals activated their signed language when receiving input exclusively in their spoken language. These results indicate that language information is readily transmitted across modalities, such that two highly unrelated languages can be activated simultaneously, even when phonological information from one of the two languages is absent. Similarities between users of spoken and signed languages in cognitive processing patterns (Emmorey, McCullough & Brentari, 2003; Pettito and Marentette, 1991) suggest that the mechanisms required to process language are capable of utilizing various forms of linguistic input. In conclusion, we propose that the language system should be considered modality-independent and capable of processing linguistic information equally, regardless of whether it is auditory or sensorimotor in nature.

In this paper, I show that a unified model of modal/temporal syntax-semantics for modals as Condoravdi (2002) proposes is not generalizable to a language like Greek, where the temporal and modal information are teased apart clearly by separate temporal and modal morphology. Moreover, I argue that Greek modals are compatible with both nonpast and past complements illustrating the existence of inner and outer tense. Specifically, I contend that the perspective of Greek modals is always set in the present (or, nonpast), and conclude that languages vary on how their modals contribute temporal information. Greek modals convey more transparent morphological options than English modals, and show that the modal and the embedded clause can be tensed, tense and modality are interpreted separately and can both have their compositional contribution. Hence, Condoravdi’s account cannot be proposed as a general theory of interaction of tense and modality.

Schegloff and his colleagues (1977) proposed that the operation of repair is a communication phenomenon that facilitates the collaborative construction of conversations. The vast majority of successive studies are not only unconcerned with relationship difference among the interlocutors but their results are not presented within a statistical analysis. The present study demonstrates operational and functional aspects of self-repair in two sets of conversations, which differ in terms of the relationship between the participants. The data of this study consist of 12 tape-recorded and transcribed Japanese conversations, six of which are between friends and the other six between strangers.

The examination of statistical differences on the repair use between the two sets of data reveals the following: In conversations between strangers, most of the insertion-type repair (i.e. inserting modifiers as in ‘school-* Japanese schoolmates’) and replacement-type repair (i.e. ‘I was studying ... a language.. English at that time’) were used to elaborate and slightly adjust semantic coding. Moreover, the interlocutor attempts to repair by rephrasing her/his utterance when detecting any problems, rather than abandoning. Conversely, in the interaction with friends, they utilized the insertion-type repair to add emphatic effect (i.e. by inserting emphatic adverbs such as zettai ‘definitely’) and the replacement-type repair to correct or cancel overt errors. These differences are statistically significant (Chi-square analysis confirms that there is only 0.01% of chance that the differences occurred by chance alone) and reveal how the repair use manifests the difference in the degree of shared information resulted from the interlocutor’s relationship.

Coronalization (velar softening) is a widely attested historical sound change where velar stops become coronal sibilant affricates preceding high front vowels: k → {tʃ ts} / _i. This phonological development is conventionally treated as a case of articulatory assimilation (e.g. Pope 1934, Chomsky & Halle 1968, Clements & Hume 1995, Calabrese 2005, Recasens & Espinosa 2009), however others have argued that it is motivated by misperception, where frication in the stop release burst causes the velar to be interpreted as a coronal affricate (e.g. Ohala 1992, Guion 1998, Flemming 2002). Another common phonological process, assibilation, occurs in similar environments; typically a coronal stop becomes a strident affricate when preceded by a high front vowel: t → {ts tʃ} / _i. Unlike coronalization, assibilation is usually considered to be a result of perceptual factors (Hall & Hamann 2006, Kim 2001, Kirchner 2001). Based on typological and phonological evidence, I will argue that coronalization and assibilation have the same underlying perceptual motivation and that coronalization is best characterized as a type of assibilation. The phonological evidence includes cases of coronalization where k seems to have become ts directly, with an intermediate tʃ stage, as can be demonstrated for many members of the Athabascan language family (Young 1983, Leer 2008). This proposal will be demonstrated using perceptually-motivated constraints within the framework of Optimality Theory to account for Romanian morphological palatalization, a process that results in both assibilation and coronalization (Chitoran 2002).

This talk investigates successive cyclic movement by looking at agreement in the left periphery in Wolof, an Atlantic language. Examination of two types of agreeing elements, floated quantifiers and left peripheral heads, shows that Wolof provides overt morphological evidence for successive cyclic wh-movement. In addition, the Wolof patterns also yield insight into the pathway of movement. Specifically, Wolof data indicate that successive cyclic movement targets at least two positions per clause, not only the specifier of CP.

Social meaning has received increasing attention from sociolinguists in recent years (e.g., Eckert 2008, Podesva 2007). However, researchers have focused primarily on phonetic variables. This study furthers their efforts by using a perception-focused methodology to investigate how orthography – specifically, the use of unconventional spelling in brand names (e.g., Equipt) (Davies 1987) – conveys meaning. It argues that the more unusual the spelling, the more likely it is to convey negative meanings (e.g., affectedness) and evoke polarized perceptions.

This study focuses on three unconventional spellings of the English word slick: SLIQ, SLIC, and SLIK. To gather perception data, interviews were conducted with 160 native English speakers (80 men and 80 women) with similar social backgrounds. Interviewees were shown one of the four spellings (SLIQ, SLIC, SLIK, SLICK) and were told that it is a car name. Based on the name alone, they were asked to describe the car and the car owner in detail. Those who were exposed to the one of the three unconventional spellings were subsequently asked if their impression of the car and/or the owner would change if the conventional spelling was used instead (and if so, how).

This study contributes to the “sociolinguistics of spelling” (Sebba 2007) by uncovering the polarizing nature of unconventional spelling and by proposing a perception-focused methodology for investigating the social meaning of orthography. It also lays the foundations for quantitative research on the effects of interviewees’ demographic categories on the perception of unconventional spelling.

Previous studies have suggested two interesting asymmetries in compensatory lengthening (CL) in which the deletion of coda or sonorant consonants is likely to trigger CL whereas the deletion of onset or obstruent consonants never does (Hayes 1989; Kavitskaya 2002). Results of my cross-linguistic survey, however, show a considerable number of counterexamples to these asymmetries: CL through onset loss in 28 languages (e.g., Samothraki Greek, /ruxa/[u:xa] ‘clothes’), and CL through obstruent loss in 28 languages (e.g., Lhasa Tibetan, /tsik/[tsik]~[tsi:] ‘one’). Crucially, I claim that CL typology is not a “hard” universal as suggested but more adequately described as implicational relationships. In addition, it will be shown that these typological patterns of CL are not randomly attested but originate from relevant phonetic and perceptual groundings. Several different sources suggest that humans perceive the duration change of codas and sonorants more sensitively than that of onsets and obstruents. Based on this relative perceptibility, I explain the typological patterns of CL by proposing two universal rankings of perceptually-driven faithfulness constraints, adopting the P-map hypothesis (Steriade 2009): (i) Max-[X]_VC >> Max-[X]_CV, and (ii) Max-[X]_V//[+son] >> Max-[X]_V//[-son]. Also, language-specific patterns of CL can be analyzed in terms of variable interactions between the proposed faithfulness constraints and Ident[V-length] prohibiting the change of input vowel length. To conclude, this phonetically-based OT account can explain both universal and language-specific patterns of CL in a unified way.