Past linguistics events from 2002
Past linguistics events from 2001
Past linguistics events from 2000
Past linguistics events from 1999
Harvard University
Giuseppe Longobardi, University of Trieste "Semantic Differences and Syntactic Parameters"
Alan Munn, Michigan State University "Three types of coordination asymmetries" 1. First conjunct agreement
Many languages exhibit agreement with the first conjunct of a conjoined set of NPs. One well known example of this comes from Arabic (Aoun, Benmamoun and Sportiche 1994).
(1) a. raaH Kariim w Marwaan
left.3m.sg Kareem and Marwaanb. Kariim w Marwaan raaHo
Kareem and Marwaan left.3plAoun, Benmamoun and Sportiche argue that first conjunct agreement is clausal, and I will provide evidence that such an analysis cannot be upheld, and a phrasal analysis is to be preferred.
2. Unlike category agreement
Another kind of asymmetry found in coordinate structures is unlike category coordination. In particular facts like (2) pose interesting problems for clausal accounts of coordination.
(2) a. John expected Bill to run and that he would win the election
b. *John expected that Bill would run and him to win the election
c. You can count on Bill's character and that he'll be on time
d. *You can count on that Bill will be on time and his great character
3. Gap asymmetries
A third type of asymmetry in coordinate structures can be found in ATB extraction. First and second gaps in an ATB extraction show different reconstruction possibilities with respect to anaphor reconstruction (disallowed in second ATB gaps) and weak crossover (not present in second ATB gaps).
(3) a. Which picture of herself did Mary paint and John buy
b. *Which picture of herself did John buy and Mary paint
c. *Which man did his mother recommend and Bill eventually hire
d. Which man did Mary recommend and his mother hire
These data pattern with parasitic gaps and provide evidence against node sharing accounts while supporting the adjunction analysis of coordination.
For more information, contact karlos@mit.edu.
MIT Speech Group Seminar
Robert E. Hillman
Voice and Speech Laboratory, Mass. Eye and Ear InfirmaryTitle TBA
Communication and Culture Seminar
Christian Kjaer Nelson, University of Massachusetts, Amherst "The conceptualization of turn-taking in conversation analysis:
A critique and reformulation"The Communication and Culture Seminar Series explores the cognitive, sociological, anthropological, linguistic, and philosophical dimensions of communication in social contexts. Meetings feature presentations by invited speakers on topics ranging from theoretical and methodological discussions to studies based on empirical research. The Communication and Culture Seminar is open to the public, and anyone with an interest in these areas of scholarship and research is encouraged to attend.
PARKING: Parking is available at no charge, on a space available basis,at the Broadway Garage, located on Felton St. between Cambridge St. and Broadway. All parkers should identify themselves as participants in the MIDAS Seminar at the CLCS. If there are no spaces available the guard will direct you to another Harvard parking facility.
MIT Phonology Circle
Sharon Manuel "Some Causal Speech Puzzles" In my talk I will discuss this and other cases of assimilation and deletion found in casual speech. These examples suggest that many casual speech phenomena will only be discovered through examination of the acoustic and/or articulatory signal, as powerful perceptual processes may make the results of those phenomena essentially imperceptible to even a very careful listener.
For more information, contact czoll@mit.edu.
Harvard University
Angelika Kratzer, University of Massachusetts, Amherst "Puzzles about Passives"
UMass, Amherst - Colloquium
Michel DeGraff, MIT Title TBA
For more information, contact kaa@oitunix.oit.umass.edu.
MIT Colloquium
Susan Ehrlich, York University
Note: this talk has been postponed until next semester.
MIT Ling-Lunch
Juan Romero "The Person case Constraint revisited" (1) * Ellos le me entregaron
They Cl.3sg.DAT Cl.1sg.ACC handed over
"They handed me over to him"
(2) Ellos me lo entregaron
They Cl.1sg.DAT Cl.3sg.ACC handed over
"They handed it/him over to me"
Bonet states this restriction as follows: If DAT, then ACC=3rd person. That is, if there is a dative clitic/agreement, then the accusative has to be third person.
In this talk (following Ormazabal & Romero, 1998), I am going to argue that this restriction is not morphological but syntactic in nature. I will also try to show that when looked at from this perspective, many apparently unrelated phenomena turn out to be related.
In order to provide an analysis, I am going to propose that languages differ crucially in their agreement encoding abilities. Two different kinds of languages seem to emerge: what we call 3 agreement languages (with dative agreement e.g. Basque, Greek, and Spanish) and 2 agreement languages (without dative agreement e.g. Chichewa, English, KinyaRwanda). I will show that this distinction correlates with a number of other properties such as differences in passivization or unaccusative formation.
References:
Bonet, Eulalia (1991): Morphology After Syntax: Pronominal Clitics in Romance. MIT PhD. dissertation, distributed by MITWPL.
Ormazabal, Javier and Juan Romero (1998): On the syntactic nature of the `me-lui' and the Person Case Constraint. ms, UPV-MIT
Perlmutter, David (1980): Deep and surface structure constraints in syntax. New York: Rinehart & Wilson Inc.
For more information, contact karlos@mit.edu.
Harvard University: The Edwin O. Reischauer Institute of Japanese Studies Japan Forum
Akihiko Uechi, Postdoctoral Fellow
The Reischauer Institute"What Is Wa-marking, and How To Use It"
MIT Phonology Circle
Morris Halle,
Massachusetts Institute of Technology"Stress Assignment in Indo-European" In the body of the paper I will review the evolution of this system into those found in the different daughter languages. These are basically of two kinds:
a) initial stress throughout (Celtic, Italic, Germanic, Latvian, West Slavic)b) stress is assigned to the last accented morpheme, and, in the absence of accented morphemes, to the initial syllable of the world.
The latter is not found intact in any of the surviving daughter languages. Clear traces of this system, however, are found in Armenian and classical Greek. In Armenian, stress is assigned to the last full vowel and, in the absence of full vowels, to the initial syllable (Vaux 1998). In Greek stress is assigned to the last accented morpheme and, in the absence of accented morphemes, to the penult or antepenult. It will be shown that these stress distributions are obtained by systems that differ from that of the proto-language only in relatively minor details.
Stress assignment in Latin, Polish, and other modern IE languages is a further modification of the Greek system.
The paper will conclude with a review of vowel shortening in Slavic. The major difference between the accentual system of Slavic and that of the IE proto-language is the result of a rule shifting the stress from a short syllable to the one immediately following. (Illich-Svitych 1963, Dybo 1981). The mechanism of the shift will be discussed, and it will shown that the distribution of vowel shortening in the West Slavic languages Slovak and Czech reflects the accentual structure of words at a point in time after the introduction of the rule of stress shift and before its lexicalization. This result provides a striking parallel to Verner's Law in Germanic (Verner 1876), where the (irregular) distribution of voicing in continuants was shown to reflect the position of stress in the proto-language although, like in Slovak and Czech, stress in Germanic had been initial for many centuries.
For more information, contact czoll@mit.edu.
MIT RTG Talk
Mark Steedman, University of Edinburgh "Information Structure and the Syntax-Phonology Interface" The theory is framed in Combinatory Categorial Grammar (CCG), which directly pairs phonological and logical forms without intermediary representational levels.
For more information, contact dgrodner@psyche.mit.edu.
Harvard University
C.-T.James Huang
University of California, Irvine"Passives"
UMass, Amherst - Colloquium
Arto Anttila, Boston University "Morphologically conditioned phonological rules" The argument is based on a detailed analysis of the Finnish nominal declension. Finnish is particularly instructive as the system is undergoing diachronic change and new morphological and lexical conditions have emerged over the past 100 years. First, I show that out of the large number of logically possible declensions only very few--the phonologically best ones--are found. Second, stems with particular phonological shapes are predicted to prefer certain declensional types numerically. This is confirmed by means of an on-line dictionary (about 6,000 relevant nominal stems). Third, in the absence of morphological and lexical conditioning, free variation is predicted in stems of a particular phonological shape. About 1,000 such variable stems are found in the dictionary. Fourth, morphological and lexical conditions turn out to emerge in environments where phonology does not numerically favor any particular option.
I will conclude by considering what these results mean for alternative optimality-theoretic approaches to phonology/morphology interaction that either have been or could be proposed.
For more information, contact kaa@oitunix.oit.umass.edu.
MIT Ling-Lunch
Cédric Boeckx, University of Connecticut "Raising Questions about wh-in situ in French"
For more information, contact karlos@mit.edu.
MIT Speech Group Seminar
W. Tecumseh Fitch
Harvard/MIT Speech and Hearing Sciences, Harvard UniversityTitle not available at this time.
MIT Phonology Circle
Shigeko Shinohara "Emergence of UG in foreign word adaptations" Some of our results are as follows. The accent of English adaptations reflects the source while French adaptations assign their accent by default. This is analyzed in terms of a trochaic footing with nonfinality--a metrical structure that previous research has shown to play a role in hypocoristics, truncations and speech disguises (Tateishi 1991, Ito and Mester 1992, etc.) Reflections of UG constraints that appear in the accentuation include the avoidance of prominence on epenthetic vowels. In the adaptation of dental plosives before high vowels, a UG preference for voiceless affricates over voiced ones emerges. The input sequence [tu] is adapted with an affricate [ts] (Tours --> [tsuuru]), while [du] adaptation avoids the affrication by lowering the vowel to [o] (Pompidou --> [poNpidoo]). This aspect appears also in the context of epenthesis. The vowel [u] is used after most of the consonants. But [o] appears only after dental plosives. When the vowel is epenthetic, the constraint barring a *[tu] sequence and faithful rendering of the plosive are both respected by lowering the vowel to [o] (estrade --> [esutoraddo]). It is noteworthy that the unnatural change, [u] --> [o]/t, d __ when analyzed in rule-based derivational approach, makes sense in constraint-based approaches. We also study the gemination for input word final consonants (archeveque --> [arusjubekku], madeleine --> [madoreenu]) as a result of the UG stem-syllable edge alignment. The type of gemination observed (vowel or consonant) is constrained by the length of the preceding vowel and also by the UG preference for voiceless over voiced geminates.
Most of the grammars which emerged in the process of adaptation are common in other languages. The preference of voiceless over voiced affricates is reflected in Russian and German. The avoidance of epenthetic (or more generally non prominent) vowels in the accentuation is observed in Armenian, Hebrew, Palestinian Arabic, Mohawk, etc. Stem-syllable alignment is also found in Axininca Campa, Malay, Biblical Hebrew, etc.
For more information, contact czoll@mit.edu.
Harvard University
Richard Kayne, New York University "Prepositions, Complementizers and Constituent Structure"
MIT Colloquium
Roger Schwarzschild, Rutgers University "Interval Analysis of Clausal Comparatives"
UMass, Amherst - Colloquium
Danny Fox, Harvard University Title TBA
For more information, contact kaa@oitunix.oit.umass.edu.
Boston University
Marco Haverkort, Boston University
and Rijksuniversiteit Groningen, the Netherlands"Positron Emission Tomography and Language Processing"
For more information, contact program@louis-xiv.bu.edu.
MIT Phonology Circle
Bert Vaux "The Forty Vowels of Musa Dagh: Long-Distance Compensatory Lengthening and the Prosodic Weight of Clitics in Kistin9g"The addition of enclitics to nouns is especially interesting, as it triggers a variety of changes in the final vowel of the root: low vowels and diphthongs become long, and non-low vowels diphthongize (some lengthen as well). These changes initially appear to result from long-distance compensatory lengthening (of the sort found in Friulian and Ponapean), triggered by deletion of the clitic when preceded by a consonant: cf. /tsar-n/ 'tree-def.' --> [tsa:r] vs. /gaedeu-n/ 'cat-def.' --> [gaedeun]. However, some clitics trigger lengthening even when they do not delete, e.g. /tsar-i/ 'tree-is' --> [tsa:ri]. The clitics in Kistin9g thus fall into two classes: those that trigger lengthening only when deleted (class 1), and those that invariably trigger lengthening whether they are deleted or not (class 2).
I suggest that the different behavior of these two classes follows from their phonological makeup: class 1 clitics consist of a single consonant, whereas class 2 clitics all contain at least one vowel. Class 2 clitics will therefore inevitably be assigned at least one mora, whereas class 1 clitics will not when they follow vowel-final words. If we assume that at some stage in the derivation clitics lose their ability to bear moras, any moras they have already been assigned will be available to attach to the nominal root, producing the attested lengthening effects. However, a class 1 clitic following a vowel-final root, as in /gaedeu-n/, will *not* be assigned a mora, under the assumption that only syllable nuclei can bear moras in this language. Consequently, the clitic -n has no mora to contribute to the noun in words of this type, and lengthening does not occur.
For more information, contact czoll@mit.edu.
Boston University Conference on Language Development
Harvard University
James Higginbotham, University of Oxford "Why is sequence of tense obligatory?"
MIT Colloquium
Bill Idsardi, University of Delaware "Segholate Opacities"
MIT Ling-Lunch
Miriam Engelhardt, Hebrew University of Jerusalem "Nominalization and Control Theory"
For more information, contact karlos@mit.edu.
MIT RLE Speech Communication Group Seminar
Stefanie Shattuck-Hufnagel
Research Laboratory of Electronics, MIT"The role of prosody in speech production planning"
Harvard University
Gennaro Chierchia, University of Milan "Ontology and grammar: mass nouns, plurals, and kinds across languages"
MIT Phonology Circle
Eric Bakovic "Dominance and Markedness in Maasai " Maasai (Nilotic; Tucker & Mpaayei 1955, Hall et al. 1974, Levergood 1984, Cole 1987, Archangeli & Pulleyblank 1994) has a vowel harmony system of the dominant-recessive type, where the harmonic feature is [ATR] (relative tongue root advancement/retraction). In the simplest case, either all the vowels in a word must be [+ATR], in agreement with a [+ATR] dominant vowel, or they must all be [-ATR] (the recessive value of the feature, in the absence of dominant vowels).
An interesting complication arises when there is a low vowel in the word, which must be [-ATR] due to the general incompatibility of lowness and [+ATR] (Archangeli & Pulleyblank 1994). When there is a dominant vowel in a morpheme to the left of a low vowel (that is, in a less peripheral morpheme), the low vowel shifts to a non-low [+ATR] vowel. When there is a dominant vowel to the right of a low vowel (in a more peripheral morpheme), the low vowel remains low and [-ATR] and blocks the propogation of [+ATR] to the less peripheral morphemes to its left.
I propose an optimality-theoretic account of dominant-recessive harmony in general and of the pattern in Maasai in particular, explaining the asymmetry in directionality just described in terms of the general morphological structure of the language. To make a long story short, the ambiguous behavior of the low vowel is analyzed as a residual effect of stem control, since whether or not the shift from a low [-ATR] to a non-low [+ATR] vowel takes place is dependent on the value of [ATR] in a less peripheral morpheme.
Selected References:
Archangeli, D. and D. Pulleyblank. 1994. Grounded Phonology. MIT Press, Cambridge.
Cole, J. 1987. Planar Phonology and Morphology. Garland, New York.
Hall, B. L., R. M. R. Hall, M. D. Pam, A. Myers, S. A. Antell, and G. K. Cherono. 1974. African Vowel Harmony Systems from the Vantage Point of Kalenjin. Africa und Uebersee 57, 241-267.
Levergood, B. 1984. Rule Governed Vowel Harmony and the Strict Cycle. Texas Linguistic Forum 24, 33-55.
Tucker, A. N. and T. ole Mpaayei. 1955. A Maasai Grammar, with Vocabulary. Longmans, Green and Co., London.
For more information, contact czoll@mit.edu.
MIT Ling-Lunch
Anders Holmberg, University of Tromsø "Deriving OV order in Finnish"
For more information, contact karlos@mit.edu.
MIT Speech Group Seminar
Mike Philips
Applied Language Technologies, Inc.Title TBA
MIT Phonology Circle
Brad Waller "OT Grammar Spaces: What Computational Phonology Can Tell Us"
For more information, contact czoll@mit.edu.
Speech and Hearing Foundation of Massachusetts Lecture
Ursula Bellugi Director of the Laboratory for Cognitive Neuroscience, the Salk Institute for Biological Studies, La Jolla, CA "Sign Language and Brain Organization"
For more information, contact Jennifer Donovan at 824-8536.
Transformations: Technology, Foreign Languages, and Undergraduate Education, MIT
MIT Colloquium
Jay Jasanoff, Harvard University "Lachmann's Law Revisited" Three major classes of solutions have been proposed:
1.the classic Neogrammarian approach, due to Saussure, according to which PIE surface forms of the type *akto- /ag-to-/ were analogically reshaped to surface *agto- under the influence of synchronically related forms where the voiced stop stood before a vowel (cf. Lat. ago: 'I do', agis 'you do', etc.). The new *agto- then underwent LL (> *a:gto-), and a second application of voicing assimilation produced the attested a:ctus.2.another analogical approach, anticipated by earlier scholars but chiefly associated with Kurylowicz, according to which the standard formulation of LL is an incorrect generalization. The long vowel of a:ctus (and le:ctus /leg-to-/ 'read', fu:sus /fud-to-/ 'melted', etc.) is explained as an analogical echo of the long vowel of the corresponding perfect stem (cf. e:g-, le:g-, fu:d-, etc.).
3.the early generative solution, due to Kiparsky, according to which LL represents a case of non-chronological rule addition ("rule insertion"). A synchronic rule lengthening vowels before voiced + voiceless clusters was supposedly added to the grammar of Latin at a point in the ordered sequence of rules where it applied to underlying /ag-to-/ before the operation of the -gt- > -kt- rule.
Of these, 1) seems hopelessly ad hoc, while 2), which incorrectly predicts forms like fa:ctus 'done' beside perf. fe:c- and ve:ctus 'conveyed' beside perf. perf. ve:x-, is observationally inadequate. Solution 3), which challenged the foundations of the Neogrammarian model of linguistic change, attracted a great deal of attention in the "classic" period of generative phonology, giving rise, e.g., to four squibs in LI in 1978-79 alone. Ultimately, however, this interpretation of LL was largely abandoned - partly because it was unable to account for a small number of apparent counterexamples, and partly because comparable instances of rule insertion turned out to be difficult or impossible to find elsewhere.
The claim made in this paper is that the much-despised Neogrammarian solution (PIE *akto- > pre-Lat. surface *agto- > Lat. a:ctus) is actually correct. The essential new insight is provided by the superlative ma:ximus 'greatest', which was shown by Cowgill in 1966 to go back to earlier *magisamos. This preform underwent two changes on the way to classical Latin: 1) syncope of the *-i- in the second syllable, yielding the previously disallowed voiced + voiceless cluster *-gs-; and 2) reassimilation of the new -gs- to -ks-, with attendant lengthening of the preceding vowel. This development does not prove, of course, that PIE *akto- actually was remade to *agto- in pre-Latin. But it does show that voiced + voiceless clusters of the required type were phonotactically possible at an earlier stage in the prehistory of Latin, and that such clusters regularly induced LL-type vowel lengthening before becoming completely voiceless. The Neogrammarian account of LL is thus not quite so arbitrary and stipulative as it has been thought to be.
Apart from matters of detail, the only real question for the Neogrammarian analysis is whether the presumed analogical remodeling of *akto- to *agto- would have been a plausible morphological development. In this connection it is instructive to examine a typological parallel in Ukrainian and dialectal Russian. Proto-Slavic had only voiceless + voiceless and voiced + voiced clusters; morphemes like the infinitive suffix -ti thus induced devoicing of a preceding voiced obstruent (cf. OCS vesti < vez- 'carry by vehicle'). In the individual Slavic languages, however, "mixed" (voiced + voiceless) clusters became phonetically possible following the syncope of the reduced vowels (yers) in the early Middle Ages. Such clusters were quickly eliminated over most of the Slavic area, but Ukrainian maintained and extended them, even replacing infinitives of the type vesti by analogical vezty. It willbe seen, however, that vezty is a form of precisely the same type as pre-Lat. *agto-; the parallel with the first step in the Neogrammarian account of LL is complete.
UMass, Amherst - Colloquium
Ken Hale, MIT "Eccentric Agreement"
For more information, contact kaa@oitunix.oit.umass.edu.
MIT Phonology Circle / Ling-Lunch
Ellen Kaisse, University of Washington "Are laterals [-continuant]?"
For more information, contact czoll@mit.edu.
MIT Speech Communication Group Seminar
Victor Zue
Laboratory for Computer Science, MIT"What's up at the SLS Group?"
MIT Phonology Circle
Sam Rosenthall, MIT "Weak Roots in Arabic" (1) active perfect a. /ktb/ 'to write' b. /rmy/ 'to throw' c. /gwl/ 'to say' kataba rama: ga:la katabat ramat ga:lat katabta ramayta gulta passive perfect d. kutiba e. rumiya f. gi:la kutibat rumiyat gi:lat kutibta rumi:ta giltaThe different surface forms of the weak roots are shown to be a consequence of constraints on syllable structure and uniformity inverb paradigms. The satisfaction of these constraints compels violations of the faithfulness constraints.
Uniformity in the verb paradigm plays a crucial in this analysis. The stem for verb suffixes (equivalent to the measures, see McCarthy 1979) are proposed to be the same for strong and weak verbs (cf. Brame 1970). It will be shown that only by making reference to output uniformity (as opposed to input/output faithfulness) that the surface forms of weak verbs have a natural explanation.
For more information, contact czoll@mit.edu.
MIT Ling-Lunch
Charles Yang, Massachusetts Institute of Technology "A Variational Theory of Language Acquisition"
For more information, contact karlos@mit.edu.
MIT Speech Group Seminar
Michael Cohen
Cognitive and Neural Systems, Boston UniversityKrishna Govindarajan
Applied Language Technologies, Inc."A sequential model of context effects in perception of stop consonant clusters" In the talk, we will present a simple sequential model that is capable of modeling listeners' percepts in the perception of stop consonant clusters. The four-parameter, autoregressive, adaptation model computes the exponentially-weighted average of the prior closure durations. This average, which is corrupted by noise, is used in determing the number of stops that are heard. Using maximum likelihood estimation, the parameters were fit to data collected from several psychophysical experiments. The model is shown not only to capture 99.5% of the variance of pooled psychometric functions, but also 85% of the variance of individuals' trial-by-trial performance. Finally, we will discuss the results of this sequential model in relation to stationary models and related signal detection theory approaches, the reliability of the constructed fit, and the goodness-of-fit of the parameters.
UMass, Amherst - Colloquium
Chris Kennedy, Northwestern University "New and old perspectives on comparative (sub)deletion"
For more information, contact kaa@oitunix.oit.umass.edu.
Harvard University Linguistics Department Colloquium
Zeljko Boskovic, University of Connecticut "Wh-Movement, Focus Movement, and Economy"
MIT Ling-Lunch
Calixto Agüero-Bautista, Massachusetts Institute of Technology
"Identity effects in Spanish diminutives"
Marie-Hélène Côté, Massachusetts Institute of Technology
"Syllable structure and domain-final strengthening: Evidence from Basque"
MIT Speech Group Seminar
John Ingram
Linguistics, University of Queensland"Experiments in inter-language Phonetics/Phonology: and what they say about speech perception" I contrast alternative analytical models of loanword formation: a phonological constraint-based of three constraints: syllable preservation (NOSPLIT), a length constraint (LENGTH), and a bimoraic stem (BISTEM) constraint. These three constraints are of interest because they differ in their domain of application and language specificity, which has implications for notions of constraint re-ranking in (second) language learning.
Two approaches are taken to analyzing the subjects blend preferences: a non-standard application of OT, based on constraint ranking, and a linear modeling (LM) approach to optimization, familiar to behavioral scientists. The story told by each is similar, though LM enjoys certain methodological advantages, and further suggests that the strong requirement of strict ranking among constraints may not hold when OT is applied to preference data.
References:
Bailey, D.C., 1962. A Glossary of Japanese Neologisms, University of Arizona Press.
Kubozono, H. (1995). Perceptual evidence for the mora in Japanese. In B. Connell & A. Arvaniti (Eds.), Phonology and Phonetic evidence: Papers in Laboratory Phonology IV.(pp.141 -156), Cambridge: Cambridge University Press.
Kubozono, H. (1990). Phonological constraints on blending in English as a case for phonology-morphology interface. Yearbook of Morphology 3, 1-20.
Paradis C. and Lacharit D. (1997) Preservation and minimality in loanword adaptation. Journal of Linguistics, 33, 379-430.
Silverman, D. (1992) Multiple scansions in loanword phonology: evidence from Cantonese. Phonology, 9. 289-328.
MIT Phonology Circle
Yoonjung Kang, MIT "Reversal of Markedness in Consonantal Place of Articulation" P&S's hierarchy predicts that coronals will always have a wider distribution than other places of articulation (POA), but many counterexamples exist: (1) In Korean stem-final consonant clusters, one of the consonants deletes when the stem is word-final or is followed by a consonant-initial suffix. The deleted segment is always the coronal: ps>p, ks>k, lm>m, lp>p, lk>k, etc. (Kenstowicz 1993, Iverson and Lee 1994). (2) In English, French and Attic Greek, a coronal cannot be the first consonant of morpheme-internal stop clusters: pC, kC but *tC (Clements 1990, Yip 1991, Rice 1992). (3) Cross-linguistically, coronals are more likely to be the target of place assimilation than other POAs (Mohanan 1993, Jun 1995). (4) In Tagalog, Cebuano Bisayan, Leti and Moa, when the first consonant of a cluster is a coronal, metathesis applies: atip~ apt-an 'roof' (Tagalog), inum~ imn-a 'drink' (CB), *tanem> tomna (Leti), tamna (Moa) 'to plant' (Blust 1979). In all four cases above, coronals are prohibited in unreleased and mostly preconsonantal positions. This restriction is fulfilled in a static morpheme structure condition^Was in (2) above^Wand also through various repair strategies: deletion^W(1), assimilation^W(2) and metathesis^W(4). Kiparsky's and Jun's faithfulness hierarchy cannot account for the metathesis strategy or the generalization that all these processes conspire to avoid a coronal in unreleased/preconsonantal positions.
We propose that in addition to context-free markedness constraints, there is a set of perceptually grounded context-specific markedness constraints which universally ranks over the context-free ones: *[CORONAL, -RELEASE]>> *[LABIAL, -RELEASE]>> *[DORSAL, -RELEASE]>> *DORSAL, *LABIAL>> *CORONAL. Since the tongue tip is the fastest of the three articulators, the coronal gesture has the shortest formant transition and is easily masked by a following gesture under overlap resulting in loss of audible release (Browman and Goldstein 1990, Byrd 1994). Thus coronals are perceptually less favored than other POAs, leading to the markedness reversal. This makes the interesting prediction that the markedness-reversal pattern of place restriction occurs only in the unreleased/preconsonantal contexts, while the default markedness pattern occurs only when the context of perceptual markedness constraints is not met, vacuously satisfying them. In Korean and English, cluster-initial consonants are not audibly released and consequently coronals are disfavored; whereas in Finnish, consonants are consistently released and codas only allow coronals (Kim-Renaud 1974, Henderson and Repp 1982, Yip 1991). Furthermore, assimilation always show the markedness-reversal pattern since gestural distance across morpheme boundaries is not reliably maintained (Byrd 1994, Cho 1997). This confirms our prediction that the choice of markedness hierarchies correlates with the specific gestural organization of the different languages.
For more information, contact czoll@mit.edu.
MIT Colloquium
Edward Keenan, University of California, Los Angeles "The Historical Creation of Anaphors in English " In this talk I document these changes and account for them by "systemic" forces (rather than parameter changes). I draw on one language specific change: Function Word Proclisis: alive < on+live, alone < all+one, because <bi+cause, myself < me+self and three more general forces: Inertia (= Modulo decay, things stay as they are unless acted upon by an outside force); Sufficient Learning (language learners learn just enough to speak correctly) and one positive feedback process: I Do Therefore I Should. A principle of Minimal Expressivity also plays a minor role.
I push (to see where it will go) the idea that properties of the anaphora profile of Present Day English are explained historically (in the same sense in which the postnominal position of alive, asleep, afire, .. is: they go where PPs go because they used to be PPs and nothing has happened to change their position (Inertia)). Some cases are easy: -self forms agree in number and gender with their antecedents. More surpirisng cross linguistically: -self forms don't occur as possessors. And flat out contentious: -self forms are locally bound.
I conclude with a brief assessment of the asdxioms for Binding Theory, both classical and a la Reinhart & Reuland, applied to Old English and Middle English.
For more information, contact Sonny Vu.
MIT Speech Group Seminar
Alec Marantz
Linguistics and Philosophy, MITTitle not available at this time.
Boston University: Walter Rodney Seminar Series
Chege Githiora, Boston University "Problems of Definition and Equivalence Between Non-Kindred Languages: Insights from Spanish, Swahili, and Gikuyu"
MIT Phonology Circle
Elsa Gomez-Imbert "Voicing and nasality in some Eastern Tukanoan languages" Surprisingly, Walker's recent typological proposal (1998), whose aim is to capture cross-linguistic variation in nasal harmony systems, doesn't refer explicitly to voicing. Mainly, her phonetically-grounded constraint hierarchy that ranks segments according to their (in)compatibility with nasalization doesn't make any distinction between voiced and voiceless stops with respect to nasal harmony.
Although the idea that transparennt segments pattern with targets and should be regarded as belonging to the same set of segments is an improvement in thecomprehension of transparency versus blocking effects, first hand data from Eastern Tukanoan languages, mainly Barasana and Tatuyo, strongly suggest the need for a finer grained scale which includes voicing, in order to account for nasal harmony patterns observed across the world.
For more information, contact czoll@mit.edu.
MIT Colloquium
Kyle Johnson, University of Massachusetts, Amherst "Another Way to Hide QR"
MIT Speech Group Seminar
Cheryl Zoll
Linguistics and Philosophy, MITTitle not available at this time.
MIT Speech Group Seminar
Mario Svirsky
Otolaryngology, Head & Neck, Indiana University School of MedicineSpeech Perception and Perceptual Learning in Cochlear Implant Users: Insights from the multidimensional Phoneme Identifical (MPI) Model 1) vowel perception of normal and "conflicting-cue" vowels by cochlear implant users who receive analog stimulation;
2) vowel and consonant perception by users of pulsatile stimulation strategies;
3) differences in vowel perception with different frequency-to-electrode maps by users of pulsatile stimulation strategies;
4) adaptation to a modified peripheral frequency map.
Point number 4) addresses the adaptation shown by CI users to the percepts elicited by electrical stimulation, which are of higher pitch than the percepts elicited by acoustic stimulation, due to the more-basal-than-normal location of the stimulating electrodes . Study of this topic is important not only because it will provide clinically and scientifically useful insights about speech perception with a CI, but also because the CI population presents a unique opportunity to investigate human adaptation to a modified frequency map, and may allow us to obtain new knowledge about plasticity of the central auditory system in adults. The MPI model provides a new way to measure and quantify auditory adaptation in response to a modified frequency map.
MIT Colloquium
Yoad Winter, Utrecht University "Characterizing Semantic Number" (1) a. The/those/five students (I know and respect) are a good team.
(1) b. *All the/no/exactly five students (I know and respect) are a good team.
(2) a. All the/no/exactly five students gathered in the hall.
(2) b. *All the/no/exactly five students are a good team. (=(1b))
(3) a. All the/no/at least two students met yesterday at school.
(3) b. *Every/no/more than one student met yesterday at school.(Surprising contrasts as in (2) were first pointed out in Dowty (1987)).
How do collective readings of sentences come about in the "a" sentences above? Why are they ruled out in the "b" sentences?The talk will propose an answer to these questions that is based on the foundational modeltheoretic semanticdistinction between *atomic* individuals and *plural* individuals. Atomic individuals are *primitives*: they are elements of the basic definition of a model. Plural individuals are *sets* of such atoms.
Three main factors govern collective interpretations of sentences in the proposed theory:
1.distinction between "quantificational" NPs as in (1a) and "potentially predicative" (so called "referential") NPs as in (1b): only the latter allow mappings from sets to atoms, which gives collectivity in (1a).2.A distinction between predicates like "gathered in the hall" in (2a) and predicates like "be a good team" in (2b): the former predicates range over plural individuals whereas the latter range only over atoms.
3.A definition of the relation between the morphological number of natural language predicates and their denotation. Plural nouns and verbs as in (3a) can range over plural individuals whereas singular NL predicates as in (3b) cannot.
The interaction between these principles leads to a general system that predicts the collectivity potential of sentences, hence their *semantic number*, in various circumstances.
A select reference: Dowty, D. (1987). "Collective predicates, distributive predicates and ALL". Proceedings of ESCOL3.
MIT Speech Group Seminar
Samuel J. Keyser
Linguistics and Philosophy, MITKenneth N. Stevens
Electrical Engineering and Computer Science, MITThe Vocal Tract, Tree Geometry and Enhancement: A Model of Speech Production We hypothesize that during the planning of an utterance, a speaker selects words from the stored lexicon and organizes these words in a working memory, each word being in the form of a sequence of a vocal-tract-like structures. Once these sequences are in place, a final planning step attaches enhancing instructions to appropriate articulators. These instructions are then converted to motor commands to the articulators.
MIT Phonology Circle
Cheryl Zoll, MIT "Tone Melodies and Richness of the Base"
MIT Speech Group Seminar
Dirk Janssen
Max Planck Institute for PsycholinguisticsTitle TBA
MIT Phonology Circle
Friday, May 8, 1998 at 2:30 at MIT, Bldg. E39, Rm 335
Prof. Sang-Cheol Ahn
"Partial Reduplication in Korean"
Abstract: This paper proposes an optimality theoretic account of the so-called "internal" partial reduplication in Korean. I will first show an outline of Korean reduplication in general and then introduce the typical nature of the ideophonic partial reduplication. I will propose that there are only two types of affixation in Korean, prefixal and suffixal, and argue that the "internal" reduplication is a subset of the suffixal type. Here I will show how the current Optimality Theory account brings out the uniformity in the description of the overall Korean partial reduplication. The analysis extends to cases of emphatic suffixation as well, where the phonetic characteristics of the suffixal segments will be discussed. Finally, I will show that the same account employed for the suffixal type can be extended to the interpretation of the prefixal reduplication.
For more information, contact Cheryl Zoll, czoll@mit.edu.
SALT 8: Semantics and Linguistic Theory
Eighth annual meeting, to be held at MIT May 8-10, 1998
Harvard University, Undergraduate Thesis Colloquium
Friday, May 8, 1998, 3:00, Sever 110
Anne Charity: "The Idiolect of Bessie Smith: A Descriptive Analysis"
Alex Sepulveda: "On Pronominal Clitic Doubling in Spanish"
For more information, contact Linguistics Dept. Administrator, Mary Violette, violette@fas.harvard.edu.
French/American Colloquium on Syntax and Semantics
Hosted by the MIT Department of Linguistics and Philosophy, to be held at MIT May 7, 1998
Harvard University
Tuesday, May 5, 1998, 10:00 to noon, Lamont Forum Room (Fifth floor of the Lamont Library, Harvard Yard)
Jane Grimshaw, Rutgers University
"Constraints on Constraints in Optimality Theoretic Syntax"
Reception to follow immediately after talk in Lamont Forum Room. For more information, contact Linguistics Dept. Administrator, Mary Violette, violette@fas.harvard.edu.
MIT
Monday, May 4, 1998 from 3:00 to 5:00 at MIT, Bldg. 56, Rm 114
Mamoru Saito
"Wh-Licensing in Japanese"
Harvard University
Monday, May 4, 1998, 4:00-6:00, Lamont Forum Room (Fifth floor of the Lamont Library, Harvard Yard)
Jane Grimshaw, Rutgers University
"The Best Place to Put a Clitic"
For more information, contact Linguistics Dept. Administrator, Mary Violette, violette@fas.harvard.edu.
MIT Phonology Circle
Friday, May 1, 1998 at 2:30 at MIT, Bldg. E39, Rm 335
Yoon Jung Kang
"Constraint Underspecification and Markedness in Place Features"
Abstract: In OT, the wellformedness constraints regulate only the fully specified output structure and with the principle of Richness of the Base, one cannot impose any restriction on the input structure. Unless a serialism is adopted, one can no longer resort to Underspecification as an analytical device. (But, this is not necessarily a defect of the theory, since there have been plenty of arguments against underspecification anayway.) Then one wonders how the new theory deals with the phenomena that were previously dealt with through underspecification. Of those, I will discuss the treatment of segmental markedness in OT, especially the unmarked status of coronal place.
The asymmetry between Coronal and other place features shows up in various ways.
(1) Implicational Inventory Universal
(2) Epenthesis
(3) Neutralization or Coda Condition
(4) Cluster Condition
(5) Assimilation
(6) Transparency to V spreading
(7) More contrasts allowed
(8) Consonant Harmony
(9) Order of Acquisition.
Smolensky(1993) attempted to explain the generalizations with recourse to the markedness hierarcy : *PL/Labial, *PL/Dorsal >> *PL/Coronal but only with limitted success. The markedness hierarchy always predicts that if there is any change at all from input to output, it is always in the direction to increase unmarked structures and decrease marked structures but never the other way around. But, place assimilation in clusters shows exactly the changes in the opposite direction. After a couple of revisions I reach the temporary conclusion that the constraints (Markedness and Faithfulness) can specifically refer only to the marked elements. That is, there are constraints, Faith-Labial and *Pl/Labial but no constraints Faith-Coronal or *PL/Coronal. This sysmtem predicts that markedness opposition is always binary. We address the problem raised by the markednes hierarhcy *Labial, *Dorsal >> *Coronal >>*Pharyngeal (proposed in Lombardi(1997)).
For more information, contact Cheryl Zoll, czoll@mit.edu.
MIT Colloquium
Friday, May 1, 1998 at 4:00, MIT Bldg. E51, Rm 395
Dianne Jonas, Yale University
"Formal Features, Licensing and Clause Structure"
For more information, contact Sonny X. Vu, son@mit.edu.
Boston University Applied Linguistics Colloquium
Friday, May 1, 1998 at 4:00, in the African Studies Center, 270 Bay State Road, 4th floor
Jon Aske, Salem State College
"A new look at word order typology in terms of the information relations topic and focus"
For more information, contact John Hutchison, hutch@bu.edu.
MIT Phonology Circle
Friday, April 24, 1998 at 2:30 at MIT, Bldg. E39, Rm 335
Degif Petron
" Continuancy Dissimilation in Chaha:
Its Interaction with Reduplicative Identity and Unmarked Features"
For more information, contact Cheryl Zoll, czoll@mit.edu.
MIT Phonology Circle
Friday, April 17, 1998 at 2:30 in E39, Rm 335
Jim Harris, MIT
"The Distribution of Nasal Consonants in Spanish
Abstract: Harris (1983, "H83") proposed a rule of Nasal Depalatalization (ND) in Spanish that changes palatal ñ in a syllable rhyme to coronal n. H83 claimed that this rule, together with independently motivated aspects of morphological structure, supports an argument for cyclic application of phonological rules. Later, Harris (1984a,b) eliminated the phonological motivation for ND and Harris (1991a,b) invalidated its morphological underpinnings. But some stories die hard: references to the supposed implications of ND for phonological theory continue to appear. Last year alone, three published works dealt at length with ND as though it were still a live issue: Bakovic (1977), Pensado (1977), and Peperkamp (1977). I attempt to give ND a decent burial within a comprehensive account of nasal consonants in Spanish that, inter alia, (a) shows how H83 misanalyzed ND and (b) points out consequences for Bakovic's, Pensado's, and Peperkamp's investigations.
Suggested reading: Calabrese (1995)
Bakovic, Eric. 1997. Spanish codas and overapplication. In Schwegler, Tranel & Uribe-Etxebarria (eds.), Romance Linguistics: Theoretical Perspectives. Amsterdam: John Benjamins.
Calabrese, Andrea. 1995. A constraint-based theory of phonological markedness and simplification procedures. LI 26:373-463.
Harris, James. 1983. Syllable Structure and Stress in Spanish. MIT Press.
Harris, James. 1984a. Autosegmental phonology, lexical phonology, and Spanish nasals. In Aronoff & Oehrle (eds.), Language Sound Structure: Studies in Phonology Presented to Morris Halle by his Teacher and Students. MIT Press.
Harris, James. 1984b. Theories of phonological representation and nasal consonants in Spanish. In Baldi (ed.), Papers from the XIIth Linguistic Symposium on Romance Languages. Amsterdam: John Benjamins.
Harris, James. 1991a. The exponence of gender in Spanish. LI 22:27-62.
Harris, James. 1991b. The form classes of Spanish Substantives. Morphology Yearbook 1:65-88.
Pensado, Carmen. 1997. On the Spanish depalatalization of /ñ/ and /l/ in rhymes. In Martínez-Gil & Morales-Front (eds.), Issues in the Phonology and Morphology of the Major Iberian Languages. Washington: Georgetown University Press.
Peperkamp, Sharon. 1997. Prosodic words. Ph.D. dissertation, Universiteit van Amsterdam. The Hague: Holland Academic Graphics.
For more information, contact Cheryl Zoll, czoll@mit.edu
MIT Colloquium
Friday, April 17, 1998 at 4:00, MIT Bldg. E51, Rm 395
Rolf Noyer, University of Pennsylvania
"Morphological Merger and Locality"
Abstract: Morphological Merger (Marantz 1984), is a general constraint on mappings between levels of morphosyntactic representation. Construed operationally, Merger may move an element X to adjoin to the head of X's complement domain. In the canonical example, which we call Local Dislocation, an X-0 trades its relation of adjacency to a following constituent with a relation of affixation to the 'linear' head (peripheral zero-level category) of that constituent. Typical examples include certain second-position clitics (Marantz 1988). The reordered elements are necessarily string-adjacent. In Lowering Merger, such as the movement of T to V in English, however, an X-0 trades its 'head-to-complement' relation with its complement for a relation of affixation to the 'structural' head of its complement. Because headedness is defined structurally, an intervening adjunct, such as an adverb, is not a barrier to Lowering Merger (Bobaljik 1994).
In Distributed Morphology, it is proposed that syntax manipulates abstract categories, whose phonological content is provided only after syntax by a process called Vocabulary Insertion (Halle & Marantz 1993). By hypothesis, Vocabulary Insertion is accompanied by the imposition of linear order on a string; prior to Vocabulary Insertion, the property of linear adjacency cannot be defined on the constituents of a syntactic structure. Because the string adjacency condition on Local Dislocation results from a definition of headedness which refers to linear order, it follows that Local Dislocation must follow Vocabulary Insertion. Hence, we formulate the following hypothesis:
The Local Dislocation Hypothesis (LDH)
(1) If a Merger operation refers to the identity of Vocabulary Items - whether diacritic or phonological properties of these - it must occur under string adjacency.
(2) If a Merger operation is sensitive only to syntactic category, it cannot be constrained to apply under string adjacency.
The LDH follows as an automatic consequence of features of Distributed Morphology, and so provides an important test for the correctness of the theory. We review known examples of Merger, including some previously undiscussed data from Finno-Ugric languages, in support of the LDH, concluding that the LDH shows considerable promise.
For more information, contact Sonny X. Vu, son@mit.edu.
Boston University
Tuesday, April 14, 1998, 9:00-10:30 AM, SED 250
Rachel Mayberry
Director, School of Communication Sciences and Disorders
McGill University
"The Critical Period for Language Acquisition:
A Psycholinguistic Perspective from Studies of American Sign Language"
Abstract: The hypothesis that language acquisition can only take place during early childhood is known as the "critical period for language acquisition." Currently there is no consensus as to whether a critical period constrains language acquisition and, if so, what the precise nature of these effects might be on language comprehension and production. We have found that the critical period exerts robust effects on first language acquisition and reduced effects on second language acquisition in a series of studies of signed and spoken language processing in deaf and hearing individuals. This paper summarizes these six studies and proposes a model for how the critical period differentially affects first and second language comprehension.
Dr. Mayberry is a candidate for a position in Deaf Studies. An ASL interpreter will be present.For more information, contact Prof. R. Hoffmeister, rhoff@bu.edu.
Boston University
Tuesday, April 14, 1998, 5:30-7:00 PM, CAS 432
François Grosjean
Professor of psycholinguistics and language processing
and Director of the Language and Speech Processing Laboratory,
Université de Neuchâtel, Switzerland
"Studying Bilinguals: Methodological and Conceptual Issues"For more information, please contact Maria Brisk, brisk@acs.bu.edu.
MIT Phonology Circle
Friday, April 10, 1998 at 2:30 at MIT, Bldg. E39, Rm 335
Hyang-Sook Sohn
"More Focus and Phrasing in Northern Kyungsang Korean"
MIT Colloquium
Friday, April 10, 1998 at 4:00, MIT Bldg. E51, Rm 395
Manfred Krifka, University of Texas, Austin
"Non-novel Indefinites in Adverbial Quantification"Abstract: It has been observed that accent differences lead to differences in the interpretation of sentences with adverbial quantifiers:
1. A FRESHman usually wears a baseball cap.
'Most wearers of baseball caps are freshmen.'
2. A freshman usually wears a BASEball cap.
'Most freshmen wear a baseball cap.'
This phenomenon has been analyzed as an instance of association with focus by Rooth (1985), who assumes that focus is on "a freshman" in (1) and on "wears a baseball cap" in (2). But the familiar ways of interpreting the influence of focus leads to a technical problem, called the problem of "requantification" by von Fintel (1994). I will argue that we should rather assume that "a baseball cap" in (1) and "a freshman" in (2) are a special kind of indefinites that presuppose that their index is already introduced, and which are marked by destressing and/or topic accent. I will show how, in quantificational structures, this presupposition is accommodated, and why we do not find such indefinites in non-quantificational structures. I will also discuss similar accentual effects in the protasis of conditionals.
MIT Phonology Circle
Friday, April 3, 1998 at 2:30, MIT Bldg. E39, Rm. 335
Morris Halle, MIT
"English Stress"
MIT Colloquium
Friday, April 3, 1998 at 4:00, MIT Bldg. E51, Rm 395
Sigrid Beck, University of Connecticut
"'Different' and Plurals"Abstract: In both (1) and (2) below, the interpretation of 'different' seems to depend on the preceding NP:
(1) Jill and Debbie read different books.
(2) Every student read a different book.
(1') and (2') are paraphrases of the readings in question.
(1') The book that Jill read was different from the book that Debbie read.
(2') Every student read a book that was different from the book that every other student read.
Carlson (1987) and Moltmann (1992) provide semantic analyses for these data that rely on a semantic relation between the NP and 'different'. That relation is the same for the plural NP case and the universal NP case. I will argue that the two cases should not receive a parallel analysis. A condition on the availability of the universal NP dependent reading is that the universal can take scope over 'different'. There is no such condition in the plural NP case. I propose an analysis for the plural NP case that assumes a pragmatic relation between the plural NP and 'different'. I rely on Schwarzschild's (1996) theory of reciprocity to get the NP dependent reading.
Boston University Applied Linguistics Colloquium
Friday, March 20, 1998 at 4:00, in room 416 of the African Studies Center (on Bay State Road)
Rose-Marie Déchaine, University of British Columbia, Vancouver
"Transitive States"
MIT Phonology Circle
Friday, March 13, 1998, at 2:30, at MIT in E39-335
Rob Pensalfini, University of Chicago
"Morpheme-structure Constraints in Central Australian Languages"
Abstract: Breen and Pensalfini (in press, hereafter B&P) make a strong argument for considering Arrernte, an Arandic language of Central Australia, to be a language which lacks syllable onsets at the word level entirely, and which syllabifies all consonants as codas. If their argument can be upheld, it will no longer be possible to claim, as has been accepted for many years (since at least Jakobson 1962), that all languages have CV syllables (syllables consisting of a consonant followed by a vowel) among their inventory. This is a point of significant theoretical importance, as in several phonological theories (including Government Phonology (Kaye, Lowenstamm and Vergnaud 1985, Harris 1990), and Optimality Theory (Prince and Smolensky 1993, McCarthy and Prince 1993a, b), henceforth OT), this universal follows directly from certain primitive assumptions of the theory.
I will argue that there are some reasons to consider that Arrernte allows syllables with onsets, and indeed prefers them in the unmarked case, with the apparent VC(C) syllabification resulting from morpheme-structure constraints which outrank (in an Optimality Theoretic sense) the syllable structure constraints. OT is the theory best suited to handling the Arrernte facts because it allows exactly this kind of competition between the phonological demands of syllable-structure and the morphological demands of morpheme structure.
These same morpheme structure constraints are seen to be responsible for patterns of reduplication found in unrelated languages much further to the North in Australia which have long proved recalcitrant to analysis. I argue that a simple re-ranking of these constraints with respect to input-output faithfulness constraints could have resulted in the loss of morpheme-initial consonants and morpheme-final vowels that gives modern Arrernte its appearance of onsetless syllabification.
The morpheme structure constraints in question are alignment constraints of the general type proposed by McCarthy and Prince (1993). Though active in several unrelated languages in the Australian continent, it seems unlikely that they could be considered universal constraints, as they do not reflect a universal cross-linguistic tendency (Sherrard 1997). Rather they are language-specific instantiations of the universal Generalized Alignment meta-constraint (Green 1993, Russell 1995). However, the analysis proposed here for the shift to an Arrernte-style language assumes the existence of both language-specific morpheme structure constraints and underlying representations, contra Russell 1995. The interplay between morpheme-structure constraints and purely prosodic constraints also allows us to account for the interaction of stress with affixation in central Australian languages, which has proven problematic under cyclic and OT approaches alike.
MIT Colloquium
Friday, March 13, 1998 at 4:00, MIT Bldg.E51, Rm 395
Jose Hualde, University of Illinois
"There Are No Underlying Representations in Basque
(or, Should We Abandon the Derivational Metaphor?)"
MIT RLE Speech Communication Group Seminar
Wednesday, March 11, 1998, at 3:00 at MIT in 34-401B (Building 34, Grier Room B)
Katherine Demuth, Brown University
"Phonological Constraints on Children's Early Words "
Abstract: It has long been observed that children tend to omit syllables from some of their early words and utterances, with words like banana. Various perceptual, articulatory and rhythmic explanations have been proposed to account for this phenomenon, but none of these provide an adequate explanation of either the crosslinguistic data or an account of how this changes over time. This talk presents a phonological account of this problem by appealing to the Prosodic Hierarchy (Selkirk 1984, McCarthy & Prince 1993) and the notion of phonological constraints (Prince & Smolensky 1993) which are gradually reranked over time. It concludes by exploring the implications of phonological constraints for the acquisition of morphology and for a developmental model of language production.
Refreshments will be served at 2:45.
MIT Phonology Circle
Friday, March 6, 1998, at 2:30, at MIT in E39-229:
Arto Anttila, Boston University
"Morphological Paradigms"
Abstract: I present morphological evidence for the hypothesis that grammars are partial orderings of ranked constraints and discuss its implications for the phonology/morphology interface. To date, the best evidence for partial ranking comes from patterns of optionality and statistical variation (Kiparsky 1993, Anttila 1997, Anttila and Cho in press, Nagy and Reynolds 1997, Ringen and Heinamaki 1997, among others). I show that the same theory extends naturally to a phenomenon which at first seems completely unrelated to variation: the structure of inflectional paradigms. According to an authoritative dictionary, Finnish has 82 phonologically irreducible nominal declensions. A closer look shows that most of the declensions are variations on a few phonological themes. The following three paradigms are typical:
Declension 1
Declension 2
Declension 3
nom. sg.
karahka
asema
matala
gen. sg.
karahka-n
asema-n
matala-n
par. pl.
karahko-j-a
asem-i-a
matalo-i-ta/matal-i-a
gen. pl.
karahko-j-en
asem-i-en
matalo-i-den/matal-i-en
s
'branch'
'station'
'shallow'
The declensions differ in the plural. In Declension 1, the stem-final /a/ is mutated to [o], in Declension 2, /a/ is deleted, and in Declension 3 either option is possible. Intriguingly, declensions are only found in stems with an odd number of syllables. Even-numbered stems are phonologically predictable: /a/-deletion is categorical after syllables with a rounded vowel (muna--mun-i-a/*muno-j-a), /a/ --> [o] applies elsewhere (kana--kano-j-a/*kan-i-a). Deletion is a consequence of the Obligatory Contour Principle which prohibits rounded vowels in adjacent syllables; mutation is the default process.
I assume the constraints OCP_round and OCP_height, both enforced more stringently within feet than across feet, and Faithfulness constraints against deletion and mutation. A partial ranking of these constraints yields a phonologically strictly limited range of mutation/deletion patterns which subsume the observed declensions. In even-numbered stems, deletion is correctly predicted to be categorical after rounded vowels. In odd-numbered stems, both mutation and deletion are permitted everywhere, hence the possibility of diverging declensions. The system also predicts the existence of mixed declensions (Declension 3) where mutation and deletion occur in free variation. Finally, the grammar makes the soft prediction that mutation is strongly preferred after /i/ and strongly dispreferred after /o/. This prediction is confirmed: the declension of the variable suffix /-jA/ shows mutation only after /i/, and deletion elsewhere, and deletion after /o/ is virtually categorical.
I conclude that morphological declensions are a species of phonological variation due to partial ranking. However, here phonological variation is not free, but exploited by morphology which assigns different stem classes to different rankings within the same partial ranking. This also explains why morphological paradigms reflect the general phonological patterns of the language: paradigms are subgrammars (rankings) embedded in the invariant phonology of the language.
MIT Colloquium
Friday, March 6, 1998 at 4:00, MIT Bldg.E51, Rm 395
Guglielmo Cinque, University of Venice
"Clitic Climbing and other Transparency Effects"
Abstract: In the talk, I will present a new way of analysing transparency effects (Clitic Climbing, long DP-movement, long auxiliary selection...). All the analyses so far proposed have been unable to offer a principled account of why such effects are found with just modal, aspectual and motion verbs (across languages). Building on a recent analysis of the functional structure of the clause, which takes it to be made up of many Mood, Modal, Tense and Aspect heads, I will propose that transparency effects are only possible with matrix verbs which happen to correspond to functional heads, and can thus be directly inserted in the extended projection of a lexical verb, in a strictly monoclausal structure.
