Three rating experiments were conducted: two baseline experiments (where responses were expected to be similar across the two languages) and a critical Wh-Experiment. The Wh-Experiment manipulated extraction (Subject-/Object-), Subject-type (Bare/Which), and Object-type (Bare/Which). The manipulations of the Subject-/Object-type were included to test the discourse-linking hypothesis (Pesetsky, 1987), according to which changing the bare forms of wh-words to which-forms improves superiority violations. The data in the Wh-Experiment revealed cross-linguistic differences between English and Russian. Specifically, there were reliable superiority effects in English, but not in Russian. Furthermore, the English data provide support for the discourse-linking hypothesis. In conclusion, contrary to the predictions of the processing account, there was no suggestion of superiority effects in Russian. The results therefore demonstrate that whatever the underlying cause of the superiority violation constraint in English is, it does not apply in Russian. Implications for the accounts of superiority violations will be discussed.

In Korean, N can often appear without any suffixes whereas V cannot. Owing
to this, the coda distribution and its historical change in Middle Korean
monosyllabic morphemes were quite different between N and V. For example:

1) Intervocalic lenition appeared only in the inflection form of V as a
rule.
2) Allomorphs which were different in codas were mainly seen in N.
3) In Contemporary Korean (CK), V is more stable and retains the MK codas.
N tends to change based on the isolated form.

In this way, N codas have undergone more changes than V codas. In
particular, coronal codas greatly changed between CK and MK. In order to
explain this, I examined the data of not only monosyllabic words but also
of disyllabic/trisyllabic words. As a result, it turned out that in MK the
coronal codas already had a biased distribution--that is, few examples of
coda -t (unaspirated) and -th (aspirated) whereas many examples of -s etc.

Finally, I try to examine some unique coda changes:

a) The reanalysis of codas in compound words.
b) The difference between compound words and semi-compound words.
c) The preservation of codas in derived stems.

            Previous studies (e.g., Eady et al., 1986) have found differences in duration and F0 between contrastive focus elements in question-answer pairs and pitch-accented elements in all-new, “broad focus” sentences. Such studies have generally examined contrastive focus in isolation, i.e., in contexts where everything else in the sentence is given in the discourse. This approach does not allow one to determine how contrastive focus interacts with presentational/informational focus with regard to sentence intonation.

            Lisa Selkirk and I conducted an elicitation experiment at Umass Amherst last year in which the two types of focus were both realized within the same sentence. The stimuli were sentences with double complements, each one with some kind of focus. With contrastive focus represented as FOC and presentational focus represented as foc, the conditions were (a) FOC followed by foc; (b) foc FOC; and (c) foc foc. Rather than question-answer pairs, subjects were given varying discourse situations for each condition and asked to utter the sentences in a way that fit the discourse. Contrastive readings were forced in part by association with only or some other focus-sensitive particle:

            a. She only goes to [Wal-mart]FOC on [Monday]foc

            ‘It is only Wal-Mart that she goes to on Monday.’

            b. She only goes to [Wal-mart]foc on [Monday]FOC

            ‘It is only on Monday that she goes to Wal-Mart.’

            c. She goes to [Walmart]foc on [Monday]foc.

            The study found that FOC complements were realized with greater duration and higher F0 than foc complements in both sentence-medial and sentence-final position. In addition, FOC complements resulted in greater pitch compression on succeeding elements, although some pitch compression was found after foc elements as well. Patterns are considerably more complicated when the “all-new” C condition is taken into account. On Tuesday, I’ll present the results of the experiment. I hope to discuss how the data can be accounted for in terms of stress prominence, phonological phrasing, and the syntax-phonology interface. The interpretation of the results is very much an open question, and I hope those in attendance will share their opinions.

Producing voicing is difficult in a velar stop, while producing an intense burst is difficultin a bilabial stop. These difficulties follow from differences between the places of articulation inhow fast intraoral air pressure rises and how much the speaker can manipulate the speed of thatrise. They also lead us to expect that languages like Dutch, which lacks /g/ but has /b,d/, andArabic, which lacks /p/ but has /t, k/, should be more common than languages like Mixe, whichlacks /b/ but has /d, g/, and Usan, which lacks /k/ but has /p, t/. However, if we examine co-occurrence frequencies statistically, we find that a stop with a particular laryngeal articulation atone major place of articulation occurs significantly more often than expected when stops with thesame laryngeal articulation occur at all the other major places of articulation. Moreover, stopswith some marked laryngeal articulations, [+voice] and [+constricted glottis], do not occursignificantly more often than expected when a minimally different, less marked stop does. Only[+spread glottis] stops occur significantly more than expected when their less markedcounterparts do. With this exception, stops with one laryngeal articulation depend on other stopswith the same laryngeal articulation at other places of articulation, but don’t depend on otherstops with less marked laryngeal articulations and the same place of articulation. These statistical analyses suggest that although languages like Dutch and Arabic are expected onphonetic grounds, they are just as pathological as Mixe and Usan, because all four (kinds of)languages fail to fill out their inventories of voiced or voiceless stops.

Broadly similar conclusions follow from statistical analyses of the co-occurrence ofvoiced and voiceless non-strident fricatives and the corresponding stops. It’s been suggested thatthe voiced, non-strident fricatives /, ð, / occur in many languages as lenited pronunciations ofearlier /b, d, g/. Lenition would ease the aerodynamic conflict between obstruency and voicingby letting some of the air in the oral cavity escape and maintaining the tranglottal pressure dropon which voicing depends. In accord with this hypothesis, // is indeed significantly more likelyto occur if /b/ is absent, and // is more likely to occur when /g/ is absent. However, thedifference between observed and expected frequencies isn’t significant for the velar stops, eventhough on aerodynamic grounds we’d expect /g/ to lenite to // far more often than /b/ lenites to// and no difference is observed for /ð/ and /d/. Even worse, // occurs significantly more oftenthan expected when /p/ is absent, and /x/ is more likely to occur when /k/ is absent. Again, thedifference between observed and expected frequencies isn’t significant for the velars, andobserved and expected frequencies don’t differ for // and /t/. No aerodynamic advantage isobtained by leniting a voiceless stop, at any place of articulation. That lenition depends on placeof articulation in the same way for voiceless stops as voiced ones despite its meeting nofunctional demand in the former suggests that pattern congruity is at least as potent a force inshaping phonological inventories as the aerodynamics of speaking or other phonetic forces.

Oroqen is a Tungusic language spoken in northeast China. It is giving way to Mandarin at a very fast rate, and will likely loose its last fluent speakers within the next 10 or 15 years. Data for this presentation were collected during fieldwork in July, 2005, with Profs. Lindsay Whaley (Dartmouth College) and Fengxiang Li (Cal. State U., Chico). Fieldwork expenses associated with the phonetic data collection were partially funded by the MIT Department of Linguistics and Philosophy, the Speech Communication Group at RLE, and the Harvard-MIT Division of Health Sciences and Technology.

   Not available

   A traditional generative view of phonology assumes a single abstract underlying representation for each lexical item. Under this view, observed differences in the realization of a given form (in terms of reduction, coarticulation, duration, etc.) in different utterances must follow from factors conditioning the production of those utterances. It has been claimed, however, that such models are overly simplistic and cannot account for the range of subphonemic detail argued to play a role in the production, perception, and representation of lexical entries. Such detail can be encoded within an exemplar model, in which individual instances of a particular form are stored and can, consequently, affect the long-term representation of lexical items. In this paper, we investigate the predictions made by these models by considering related claims made by Bybee (2001) and Jurafsky et al. (2001), which together predict that the effects of (token) frequency on phonetic duration are such that more
frequent lexical items should have shorter durations. The best evidence in support of this claim comes from cases where frequency differences correlate with the difference between function and content words. It is well known that function words show more reduced and variable realization than content words (as discussed, for example, by Jurafksky et al. 2001
and Lavoie 2002). The question is whether this correlation holds more generally: If we control for the distinction between content and function words, will we still find an effect?

Our study investigates the phonetic durations of heterographic pairs of homophonous English nouns that differ in token frequency. Fourteen such pairs were grouped into three categories based on the magnitude of the frequency difference between the members of each pair, as determined by relative frequencies in five large corpora. This included Large Difference pairs (e.g., time ~ thyme, way ~whey), Medium Difference pairs (e.g., pain ~ pane, gate, gait), and Little or No Difference pairs (e.g., son ~ sun, peace~ piece). Results reported here are for four native speakers of American English who participated in two experiments. In the first experiment, the speakers were recorded reading four repetitions of a randomized list of the target words in a frame sentence. In the second experiment, a subset of these words was read in composed sentences with controlled prosodic structures. The phonetic duration of each target word was then measured in Praat, and the ratio more frequent/less frequent was calculated for each repetition of each pair. Statistical analysis was done with a mixed model of repeated measures.

If the Bybee/Jurafsky hypothesis is correct and greater frequency leads to shorter duration, then these ratios should systematically fall below 1 for the Large Difference and Medium Difference pairs, while those for the Little or No Difference group should be approximately 1. No systematic differences were found for individual speakers or across speakers in either the frame sentences or the composed sentences. The lack of positive correlation between duration and token frequency calls into question the hypothesis being tested, namely that greater frequency leads to shorter duration, as well as the implication that differences between function and content words follow from differences in frequency. In discussing these results, we consider other studies on homophones and duration (e.g., Whalen ms and Guion 1995) and the implication of these results for issues of the lexicon and speech processing. Our study supports the conclusion that at least one aspect of lexical representation is coarse-grained (e.g.
Pierrehumbert 2003, Beckman 2003), corresponding roughly to what is thought of as a phonemic representation. Our results, taken in light of previous studies, underline the need for a better understanding of the locus of frequency effects in both the lexicon and speech production.

I. Purpose: This paper proposes and motivates the notion of Secondary Agree with
respect to multiple Case constructions in Korean/Japanese.

II. Movements in Parallel: So-called Possessor-Raising is a genitive-nominative
conversion process inducing a multiple Case construction as shown in the Korean
example (1):

(1) John-uy/-i son-i khu-ta.
J.-GEN/-NOM hand-NOM big-DEC
?John?s hand is big. / John is such that his hand is big.?

The newly-created multiple Case (nominative in this case) phrase like John-i in
(1) shows both A- and A?-movement properties simultaneously: it induces the
A-movement effects with respect to Binding Conditions A and C,
subject-oriented honorification, adverb placement, etc., as well as
A?-movement effects like focus/topic effects and intervention effects (in
terms of the cartographic Relativised Minimality (Rizzi 2004a)). Note that the
A- and A?-movement properties of the newly-created multiple Case phrase are
simultaneous in that John-i in (1) functions as subject and focus at the same
time. One might argue that it is nothing special since we may assume that
John-i in (1) functions as focus in A?-position (Spec-C) and its trace as
subject in A-position (Spec-T) under the assumption that it undergoes Focus
Movement from Spec-T to Spec-C. But recently the legitimacy of movement from
Spec-T to Spec-C was called into question (Chomsky 2005). Furthermore, I will
show evidence that the movement of the newly-created multiple Case phrase to
Spec-T (as well as to Spec-C) should be subject to criterial freezing (cf.
Rizzi?s (2004b) Subject Criterion; Chomsky?s (2005) freezing effect of
Agree/Move). So, I claim that John-i in (1) should undergo movements in
parallel to Spec-T and Spec-C at the same time (Chomsky 2005).

III. Criterial Attraction: For the evidence that even the A-movement of the
newly-created multiple Case phrase undergoes criterial attraction (Rizzi
2004b), consider another multiple Case construction called Major Subject
Construction as shown in (2):
(2) enehak-i chwicik-i elyep-ta.
Linguistics-NOM finding-jobs-NOM difficult-DEC
?Linguistics is such that finding a job is difficult.?
The newly-created multiple Case phrase enehak-i, so-called Major Subject, in (2)
shows the same A- and A?-movement properties as John-i in (1) simultaneously;
hence, I claim we may also call John-i in (1) or any newly-created multiple
nominative phrase a Major Subject; in fact they all carry the ?aboutness?
relation to the following clause, the characteristic property of Major Subject
(Heycock 1993). I propose to capture the property of ?aboutness? relation
for the newly-created multiple Case phrases by the notion of Subject Criterion
(Rizzi 2004b); hence, the A-movement of newly-created multiple Case phrases
should be subject to criterial freezing. One might claim that the
?aboutness? relation is a property of the A?-movement of the
newly-created multiple Case phrase, i.e., due to the Topic Criterion. I have
three arguments against such a claim. First, the notion of topic and the notion
of the ?aboutness? relation are related but clearly distinct. Second, there
is little motivation to exceptionally allow the newly-created multiple Case
phrases to move to two criterial (freezing) positions in Spec-C, i.e.,
Spec-Focus and Spec-Topic, at the same time. Third, there is a bona fide
topical sentence counterpart of (2) with the nominative particle ?i simply
replaced by the topic particle ?un, so that the topic phrase enehak-un
legitimately undergoes A?-movement to Spec-Topic to satisfy the Topic
Criterion. To conclude, the newly-created multiple Case phrase undergoes two
criterial attractions to satisfy Subject Criterion and Focus Criterion at the
same time. Furthermore, I claim that the newly-created multiple Case phrase
undergoes only criterial attractions. Why is it so?

IV. Secondary Agree: I propose Secondary Agree Principle (3):
(3) a. Secondary Agree Activation
A default Case/?-feature may be derivationally assigned to the head
of a DP/PP so that the DP/PP may undergo Secondary Agree, which is an optional
free-ride ?defective? Agree in addition to the obligatory ?full? Agree
(Primary Agree) for a probe head. (The default Case/?-feature is determined
parametrically for each language.)
b. Secondary Agree Condition
Secondary Agree Activation applies only for optional overt criterial attraction.
According to Secondary Agree Principle (3), the newly-created multiple Case
phrases John-i and enehak-i in (1) and (2) are derived as follows. According to
(3a), the genitive phrase John-uy and the adjunct PP (with abstract P)
enehak-ABOUT are assigned Secondary Agree features, forming John-uy+AGREE and
enehak-ABOUT+AGREE respectively, which in turn become John-i and enehak-i by
the Agree Feature Valuation (Pesetsky and Torrego 2004) realizing AGREE as
nominative and the Case Stacking Condition deleting the original particles like
-uy. Then, according to (3b), John-i and enehak-i in (1) and (2) undergo two
criterial attractions each to satisfy Subject Criterion and Focus Criterion,
which are apparently licensed by the Secondary Agree features assigned to them
by (3a) and the focus features assigned to them by (4), respectively:
(4) The phrase to which a non-canonical Case particle is attached receives a
focus feature.
The mere assignments of the Agree feature by (3a) and the focus feature by (4),
however, do not guarantee the overt A- and A?-movements of the phrase, since
those features can be checked covertly in Korean/Japanese. Note that the
intervention effects of the newly-created multiple Case phrases mentioned
earlier crucially depend on their overt A?-movements. What guarantees their
overt A- and A?-movement is (3b): Secondary Agree induces overt criterial
attraction. To conclude, Secondary Agree is only for some output effects of
criterial attraction.

V. Further/Cross-linguistic Application of Secondary Agree: Secondary Agree
accounts for the overt A- and A?-movement properties of all the other
multiple Case constructions, including Case-marked adjunct/adverbial
constructions, Case-stacking constructions, etc., and partially for ECM and
Scrambling constructions (Miyagawa 2004) in Korean/Japanese, Locative Inversion
constructions in English, and quirky Case constructions in Icelandic (Boeckx
2003), as far as relevant conditions are met.

For further information, contact Victor Manfredi.

(1) I came to Boston last year.
(2) I've been in Boston for the last year.

Based on a corpus analysis of the tokens of (the) last/next from the Brown Corpus and the 1996 LA Times Corpus, Chesterman's (1991) discussion of extensivity and the null article, and Allen and Hill's (1979) notion of predicated time, this paper presents a comprehensive account of when the determiners last and next combine with null or the; why last/next followed by singular temporal nouns occur with null, as in (1), or with the, as in (2); and why only singular temporal nouns, but not plural temporal nouns or nontemporal nouns, can combine with null + last/next.

Sponsored by the James Geddes, Jr., Lecture Series.
For further information, contact C. Neidle.

 (1) p and (q or r) is equivalent to...
      a.   ...( p and q) or (p and  r)    [true]
      b.   ...((p and q) or  p) and r     [false]

 Along with assigning a meaning, the grammar also assigns a prosodic structure to each expression. Prosody disambiguates (1a) and (1b), although they are string identical when spoken. The phrasing reflects the internal structure, as is evidenced for example by final lengthening effects that depend on boundary strength.

 The idea is now to identify principles that recursively negotiate the relative  prosody between sister constituents, analogous to the principles on the  semantics side that negotiate the composition of meaning. The global prosody in the proposed system is assigned based on decisions made locally, as in the transformational cycle of early generative  approaches to phonology (Chomsky and Halle 1968), but incorporating insights from more recent research on stress and prosody (cf. Selkirk 1984, Gussenhoven 1984, Inkelas and Zec 1990, Jacobs 1991, Cinque 1993, Truckenbrodt 1995, Arregi 2002).

 We will look at the fragment of English that only consists of names and the functors 'and' and 'or', which turns out to have a fairly complex combinatorial power and an intricate prosodic pattern.  A single mapping principle, 'sister matching', and its interaction with the syntactic cycle derives the correct phrasing (e.g. for structures such as (1)).

 Data from a production experiment provides evidence for a recursive prosodic system, where boundaries are scaled in strength relative to earlier produced boundaries. The results from coordination structures can be used to investigate other syntactic structures, two cases of interest are right-node raising and possessor constructions.

This seminar will be held on Wednesday, April 6, 2005, from 12:00 noon to 1:30 PM in the Center conference room, Room 16-151. The seminar is open to all members of the MIT community, free of charge.

Registration is required. You can register on the web at http://web.mit.edu/hr/worklife/se
minar.html, or call the Center at 617.253.1592, or email the Center at worklife@mit.edu with the name of the workshop and your name, telephone number, and email address.

We look forward to seeing you there.

Sponsored by the Boston University Undergraduate Lingusitics Association (BULA)

(1) Lo vimos a Guille
him-ACC 1P-saw to-Guille
'We saw Guille' (Jaeggli 1982)

Constructions like (1) have inspired the emergence of two competing views of the syntax of cliticization: the argument view and the agreement view. Under the argument view, object clitics are internal arguments of the verb and the post-verbal NP is to be analyzed as some kind of adjunct. By contrast, the agreement view of clitics takes the post verbal NP to be the object of the verb and the clitic is treated as a marker of agreement between the given NP and the verb (or some other functional head). These differences imply different claims about the structure of (1). Thus, the argument view implies that there is a pronoun in the structure of (1), whereas the agreement view implies that there is no pronoun in that structure. This difference in turns leads to the following predictions: a) doubling constructions should show pronominal effects (the argument view), and b) doubling constructions should not show pronominal effects (the agreement view).

I will show that constructions like (1) do show several pronominal effects that cannot be captured within the agreement view of the phenomenon. One such effect, for instance, is the fact that doubling constructions with a definite description in the position of the post-verbal NP behave like pronouns and not like descriptions (in terms of presupposition projection) when occurring within the scope of an intensional verb like want.

I will therefore adopt the argument view of clitics, and will develop a semantics of clitic-doubling in which object clitics are versions of the definite determiner (cf. Uriagereka 1995) that take a null complement NP with an adjoined empty pronominal element. I will assume that the clitic moves as a DP pi-piping its null complement and tucking-in in the sense of Richards (1997) below the subject in IP. In moving qua a DP, a clitic will create a derived predicate in the sense of Heim and Kratzer (1998). I will argue that the double enters the interpretation of the sentence by adjoining at LF to the derived predicate created by the moved clitic. The consequence of this adjunction is that the double must be type-shifted by an operator along the lines of Partee's (1987) BE operator to compose semantically with the type of the derived predicate by predicate conjunction. It will be shown that the semantic effect of interpreting the double in this way amounts to adding the information of the double as a restriction to the clitic (hence the title of the talk).

The semantics of the BE operator in the context of clitic-doubling is such that it applies to NPs in their generalized quantifier type (i.e. the type ááe,tñ, tñ ) to yield predicates (cf. Montague (1973), Partee (1987)) that apply to the elements in the singleton witnesses of the quantifier. This means that generalized quantifiers that do not contain singleton witnesses (in the normal cases) (i.e. universals, negative quantifiers, proportional quantifiers, numeral quantifiers, etc.) will not yield sensible predicates and wont be able to occur in the doubling construction. On the other hand, as Partee (1987) shows, indefinite and definite NPs are the two classes of NPs that yield sensible predicates by combining with the BE operator. The present analysis then predicts that only such NPs will make good doubles in languages like Spanish and Romanian.

The analysis, thus, makes the cross-linguistic prediction that only those NPs that can be turned into predicates by the BE or some other type-shifting operator (e.g., Partee's (1987) ident operator) available in the given language will make good doubles in that language. It will be shown that the various semantic restrictions on clitic-doubling across languages can be accounted for in this way (for example the fact that only definite NPs make good doubles in Modern Greek (Anagnostopoulou 1999)).

Background reading:
Partee, Barbara (1987). Noun Phrase interpretation and type-shifting principles. In Jeroen Groenendijk, Dick de Jongh, and Martin Stokhof (eds.), Studies in Discourse Representation Theory and the Theory of Generalized Quantifiers. (pp. 115-143) Dordrecht: Foris Publications.
Sportiche, Dominique (1999). Pronominal clitic dependencies. In Henk van Riemsdijk (ed.), Clitics in the Languages of Europe.(pp. 679-708). Berlin: Mouton de Gruyter.

Boston University Department of Modern Foreign Languages and Literatures. Dr. Ippolito is a candidate for a faculty position. For further information, contact C. Neidle.