The primary goal of this study is to develop a theory of the processing of doubly-quantified sentences such as A squirrel picked up every nut, particularly how the scope ambiguity in such sentences is resolved. The research departs from most psycholinguistic work in drawing upon current linguistic theories of LF, the syntax-semantics interface, and formal semantics. First, I investigate the issue of how structural factors affect quantifier scope preferences. I argue that the processor takes an economic stance towards scope assignment. The preferred relative scoping of two quantified phrases is computed from the 'required' LF structure--the LF constructed from required grammatical operations acting on S-structure. Furthermore, I contend that when every has scope over a, the processor does not commit to how many entities the a-phrase represents. Next, I present an analysis of the semantic differences between each and every with respect to event distributivity, in preparation for considering the scope behavior of these quantifiers. I demonstrate that a sentence containing each can only be true of an event which has a totally distributive event structure, where each individual object in the restrictor set of the quantified phrase is associated with its own subevent, and all the subevents are differentiated on some relevant dimension. Every is subject to the weaker requirement that there be at least two different subevents. Finally, I apply the semantic analysis of each and every to the question of how individual quantifiers affect scope preferences. Each has often been said to have a stronger preference for wide scope than every. I argue that this observation arises from cases where each takes wide scope in order to fulfill its condition requiring total event distributivity and differentiation of subevents. Otherwise the scope behavior of each and every is quite similar; they preferentially take wide scope only when that is the scoping computed off the required LF structure. More generally, I hypothesize that a quantifier's scope behavior is driven by the lexical condition(s) which are part of its meaning. Experimental evidence is presented in support of each of these claims.

This work presents an investigation of how prosodic information is used in natural language processing and how prosody should be incorporated into models of sentence comprehension. It is argued that the processing system builds a prosodic representation in the early stages of processing, and is guided by this prosodic representation through multiple stages of analysis. Specifically, the results of four sentence comprehension experiments demonstrate that prosodic phrasing influences syntactic attachment decisions, focus interpretation, and the availability of contextual information in the resolution of lexical ambiguity. Two explicit hypotheses of how prosodic structure is used in processing are proposed to account for these effects: one which accounts for effects of phonological phrasing on syntactic processing decisions and a second which accounts for effects of intonational phrasing on semantic/pragmatic interpretation. Three sources of evidence are provided in support of the central claim that the processor must build and use a prosodic representation from the early stages of processing. First, an experiment on the resolution of prepositional phrase attachment ambiguity demonstrates that syntactic attachment decisions are influenced by the overall pattern of phonological phrasing in utterance, and not simply by prosodic boundaries located at the point of syntactic ambiguity. Thus, the effects of a single kind of prosodic element, at a single level in the prosodic hierarchy, must be accounted for with respect to the larger prosodic structure. A second experiment shows that the interpretation of focus is dependent on both the pattern of pitch accents in the utterance and the pattern of prosodic phrasing, establishing that different kinds of prosodic elements in the prosodic structure are used jointly in processing decisions. Two additional experiments, one on the interpretation of context-sensitive adjectives and a second on the resolution of within-category lexical ambiguity, demonstrate that phonologically distinct levels of prosodic phrasing have separable effects on language processing. Taken together, the four experiments suggest that prosody has a much broader role in sentence comprehension than previously recognized, and that models of sentence processing should be modified to incorporate prosodic structure.