Alan Munn

Processing Coordinate Structures

Coordination often involves syntactically like categories. Based on the results of four reading time studies, it is argued here that the syntactic like-category restriction is not grammatical. Coordination of unlike categories can be just as acceptable as coordination of like categories. However, syntactically like category coordination is processed faster than coordination of unlike categories even when the two sentence types are judged to be fully acceptable. Further, parallelism of conjuncts facilitates processing regardless of whether it is parallelism in the category of the conjuncts (a property which the grammar might regulate) or parallelism in the internal structure of the conjuncts (a property which the grammar does not regulate, on anyones view). Parallelism did not facilitate processing when the structure of a subject and object were manipulated, implying that parallelism effects are largely limited to the conjuncts of a coordinate structure and not due simply to the repetition of a phrase with a particular shape.

First Conjunct Agreement: Against a Clausal Analysis

Aoun, Benmamoun, and Sportiche (1994) propose that first conjunct agreement in Arabic is derived from clausal conjunction. This article shows that the clausal account of first conjunct agreement is empirically inadequate because it fails to distinguish clearly between syntactic and semantic agreement. It argues that an analysis of coordination as adjunction proposed in Munn 1992, 1993 accounts both for the Arabic facts and for the fact that first conjunct agreement is dependent on head government. It also shows that there are asymmetries between governed agreement and specifier-head agreement that are independent of coordination, lending support to a unified analysis of first conjunct agreement in terms of government or its equivalent.

The effects of scrambling on Spanish and Korean agrammatic interpretation: Why linear models fail and structural models survive

Several models of comprehension deficits in agrammatic aphasia rely heavily on linear considerations in the assignment of thematic roles to structural positions (e.g., the Trace-Deletion Hypothesis, the Mapping Hypothesis, and the Argument-Linking Hypothesis). These accounts predict that constructions in languages with rules that affect syntactic structure but preserve relative linear order should be unimpaired. Other models [e.g., the Double-Dependency Hypothesis, (DDH)] do not resort to linearity but are purely structural in conception and therefore should be immune to word-order effects. We tested linear and nonlinear accounts with scrambling structures in Korean and topicalization structures in Spanish. The results are very clear. The (nonlinear) DDH is entirely compatible with the evidence, but the linear accounts are not.

Double-Agents and Trace-Deletion in Agrammatism

The Trace-Deletion Hypothesis (henceforth TDH; Grodzinsky 1986, 1995) states that syntactic traces are deleted in agrammatism and that whenever a trace is deleted, a default strategy is activated. The default strategy assigns the role of Agent to the first NP. In structures where a second NP receives the Agent role syntactically, the consequence is that the agrammatic representation contains two conflicting Agents for the same action. This is the mechanism that induces guessing and the random performance on comprehension tests that has often been observed for passives and certain other structures. In this paper, we isolate the default strategy of the TDH, using a sentence-picture matching task in which one of the pictures matches the meaning arrived at by the default strategy. Our results show that an agrammatic representation does not involve double-Agents, and thus the default strategy (and therefore the TDH) is refuted.

Harnessing technology to improve formative assessment of student conceptions in STEM: Forging a national network

Concept inventories, consisting of multiple-choice questions designed around common student misconceptions, are designed to reveal student thinking. However, students often have complex, heterogeneous ideas about scientific concepts. Constructed-response assessments, in which students must create their own answer, may better reveal students’ thinking, but are time- and resource-intensive to evaluate. This report describes the initial meeting of a National Science Foundation–funded cross-institutional collaboration of interdisciplinary science, technology, engineering, and mathematics (STEM) education researchers interested in exploring the use of automated text analysis to evaluate constructed-response assessments. Participants at the meeting shared existing work on lexical analysis and concept inventories, participated in technology demonstrations and workshops, and discussed research goals. We are seeking interested collaborators to join our research community.

On the identity requirement of ATB extraction

Across-the-Board (ATB) movement from coordinate structures is usually subject to an identity requirement: the same element must be extracted from each conjunct. In this paper I show that there are cases in which the identity requirement appears not to be met, in that different answers can be felicitously given to a single ATB question. I show that the non-ATB readings are instances of sloppy identity of the argument of a functional wh, and are subject to grammatical restrictions.