The acquisition of prosody in L2 is relatively understudied, however, the acquisition of stress – especially in English as L2 – has recently received some attention (e.g. Archibald 1993, 1994, 1995, 1997, 1998, Pater 1993, 1997).  The results of such research, however, are somewhat inconclusive for several reasons.  The numbers of subjects and stimuli are typically too small to serve as the basis for robust generalizations, and to the extent that real words are examined, it is unclear whether correct production reflects a subject’s knowledge of generalizations (possibly in terms of parameters) about English stress, or merely that s/he has learned the position of stress for each lexical item.

The present study examines the ability of native English speakers and speakers of different L1s to perceive stress in English (2-, 3- and 4-yllable nonce words).  The position of stress and the structure of the syllables (i.e. CV vs. CVG or CVC, with either a schwa, lax or tense vowel) are systematically varied. We will show that while native speakers of English have little difficulty with the task, significantly different results arise for the L2 speakers due to their L1.

2/15/2002, 3:00 pm. Gore 304

2/22/2002, 3:00 pm.

This talk deals with the question of how the class of raising verbs (e.g. _seem_, _appear_, _happen (to)_, _tend (to)_) is learned by children. A body of work inspired by Landau and Gleitman (1985) has demonstrated the importance of knowing a verb's argument structure (which and how many NP theta-roles a verb assigns) in learning the verb's meaning (Naigles and Gleitman 1989, Kako 1998, i.a.). While this information does not point to the lexical meaning of the verb, it EXCLUDES a range of meanings, thereby narrowing the hypothesis space (_gorp_ in `The dax gorped the rom' might mean "hit" or "eat", but it cannot mean "sleep").  What about verbs that do not assign any NP theta-roles? Raising verbs, such as _seem_, do not assign a theta-role to their subject in a sentence like (1). Moreover, nothing about the meaning of the subject or the meaning of the predicate points to (or excludes) anything in the lexical meaning of _seem_.

(1)    Mary seemed to be happy.

We know that _Mary_ is not an argument of _seem_ because _seem_ can take an expletive subject as in (2),

(2)    It seemed that Mary was happy

and from the semantics: (1) is "about" Mary being happy, not about Mary seeming. But how does a child know? In particular, how does a child know that the subject in (1') IS an argument of the verb, and that the _it_ subject in (2') is not an expletive? The strings in (1) and (2) are both compatible with non-raising verbs, and consequently, with very different structures:

(1')    Mary wanted to be happy. (*It wanted that Mary was happy)
(2')    It said that Mary was happy. (_it_ is not an expletive)

Now, if we answer question (a)

(a) How do children figure out the lexical meaning of _seem_ and its ilk?

by saying that there is something about the structure of _seem_ sentences that cues the learner (e.g. the lack of NP theta-roles), then we first have to find out (b)

(b) How do children figure out the structure of sentences with _seem_?

Although this second question appears at first glance to have an easy answer (the child only needs to hear the verb _seem_ with an expletive subject), I will argue that this answer is insufficient, and that the problem is not trivial.

I will present results from a series of experiments with adults, showing that although adults are surprisingly resistant to positing raising verbs even in highly plausible environments, such as (3),

(3)      It _____ that Robert loved Zoe.

there are nevertheless strong cues, such as subject (in)animacy and eventivity/stativity of the lower predicate, which children are sensitive to.

I will also present some corpus-based data on children's early production of raising verbs and propose a series of experiments to test children's comprehension of the syntax and semantics of raising verbs.

03/08/02, 3:00 pm; Gore 304

Verbs in Navajo are made up of a stem and two classes of prefixes: disjunct prefixes and conjunct prefixes.  The disjunct prefixes form a class with stems because they are lexical categories and they have the full range of phonological structures.  These properties distinguish disjunct prefixes and stems from conjunct prefixes, which are not lexical items and have a highly restricted phonological inventory.  One chief aim of this work is to give a natural account of the differences between these
two morphological classes, i.e., one that characterizes the lexical/non-lexical distinction together with the phonological
differences.

A set of hypotheses is developed in Optimality Theory that explain this correlation in terms of the faithfulness properties of the morphological classes.  In particular, lexical items are argued to have a privileged faithfulness status in the sense that they require a stronger form of identity to their underlying representation (Beckman 1998, Casali 1997). Disjunct prefixes and stems are lexical items, and so they may have the full range of phonological structures because of their privileged faithfulness status.  Conjunct prefixes, on the other hand, do not have the same status, and consequently, have a more restricted phonological
inventory.  Furthermore, the faithfulness-based analysis is argued to give a principled explanation for the different phonological processes that affect disjunct and conjunct prefixes, a result that distinguishes this approach from other plausible analyses of the distinct prefix classes in terms of positional markedness (Steriade 1997, Zoll 1998).

04/12/02, 3:00 pm; Gore 304

 A distinction often associated with Wasow (1977) holds that there are two different types of passive, and thus two types of passive participle. The  first type, the verbal passive, was held to be the product of a syntactic derivation. The second type, often called Adjectival Passive, was held to be the result of Lexical rules.  In the present talk, I will address the distinction between participles in terms of a non-Lexicalist theory, Distributed Morphology. There is no Lexicon in this theory, so that all derivations have to be syntactic; the distinction between participle types cannot be a result of a distinction in module between syntactic and lexical derivation.  My emphasis will be on capturing differences between types of participle in terms of different syntactic structures in which they appear. In the domain of what were formerly called Adjectival passives, I will identify a further distinction, between (1) participles that have an eventive subcomponent, and (2) those that are interpreted purely as states, more like familiar `adjectives' (cf. also Kratzer (1995,2001) for further distinctions along these lines).  The differences between these types will be explored in terms of a height of attachment approach, in which Aspectual nodes attach at different points in a verbalized (containing small v) or non-verbalized (bare root) structure.  The approach is similar to a line of research found in the literature on nominalizations. In developing this approach to the participles, I will address several subissues, including the presence of verbalizing morphology and resultative secondary predicates in stative participles, and allomorphy patterns in the different types of participle.

05/03/02, 3:00 pm  Gore 304

Much research in linguistic theory seeks to uncover how semantic information and pragmatic inference jointly contribute to the way humans interpret utterances in context. Until recently, however, little was known about how the ability to integrate lexically encoded and contextually inferred aspects of meaning develops during language learning. This talk begins to address this question focusing on one of the best-known cases of pragmatic inference, scalar inferences (SIs; cf. Some professors are famous -> Not all professors are famous). I report the results from two experiments which investigated preschoolers' understanding of SIs in environments involving quantifiers ('some'), numerals ('two') and aspectual verbs ('start'). It was found that, given appropriate contextual conditions, 5-year-old children had some success in deriving SIs. Furthermore, children's success depended crucially on the nature of the scalar term: number expressions were more successful SI-triggers than other scalar expressions. These results suggest that there are interesting asymmetries within the semantic class of scalar expressions; they also show that these asymmetries are reflected in the architecture of early conversational inferences. I explore one possible source of scalar asymmetries, the 'discrete' nature of numerical modifiers, and show that it correctly predicts the behavior of other discrete modifiers (e.g. 'half') and the early acquisitional patterns of number words vs. vague quantifiers. I conclude by discussing some implications of this work for the development of the semantics/pragmatics interface.