Architectures and Mechanisms for Language Processing (review)

Architectures and Mechanisms for Language Processing (AMLaP) is an international conference that has established itself as the European counterpart of the annual CUNY conference on sentence processing. The current volume contains revised versions of presentations originally made at the first AMLaP conference in 1995. The papers are anything but obsolete, thanks to the editors’ decision to invite position papers rather than reports on specific results that were current in 1995.

In ‘Architectures and mechanisms in sentence comprehension’, the three editors provide an overview of some of the most important concepts underlying the research reported in this volume.

In ‘Evaluating models of human sentence processing’, Charles Clifton, Jr. presents experimental data consistent with viewing human parsing as the result of multiple, distinct processes. C also expresses continued support for models such as the garden-path model in which a single analysis is processed ‘in depth’, in contrast to models in which the parser may consider multiple analyses at once, as is the case in current constraint-satisfaction models.

In ‘Specifying architectures for language processing: Process, control, and memory in parsing and interpretation’, Richard L. Lewis argues that theories of language processing need to include specifications of the memories used in parsing and interpretation as well as a statement of how automatic and controlled processes interact. The chapter goes on to describe the specific theory adopted by Lewis that meets these requirements, NL-Soar.

In ‘Modeling thematic and discourse context effects with a multiple constraints approach: Implications for the architecture of the language comprehension system’, Michael K. Tanenhaus, Michael J. Spivey-Knowlton, and Joy E. Hanna, demonstrate that differentially weighted constraints allow one-stage models of parsing to model results previously cited as evidence for two-stage models of processing. Based on these findings, Tanenhaus et al. argue that the burden of proof is on proponents of multiple-stage models to motivate the additional assumption of separate stages. Taken together, these opening chapters, and indeed the entire volume, represent a debate among some of the most influential researchers in the field of sentence processing.

‘Late closure in context: Some consequences for parsimony’, by Gerry T. M. Altmann, discusses some problems with Altmann’s earlier claims about the ways in which discourse information, such as information about referent uniqueness, is utilized in comprehension. ‘The modular statistical hypothesis: Exploring lexical ambiguity’, by Steffan Corley and Matthew W. Crocker, contains a not very tightly argued critique of interactive models and a demonstration of how the authors’ model handles syntactic-category disambiguation. Of particular interest within the context of this collection is the acknowledgment that statistical and modular approaches are compatible in principle since processes within a module may rely on statistical information.

‘Lexical syntax and parsing architecture’, by Paola Merlo and Suzanne Stevenson, provides an account of the processing complexity of reduced relative clauses based on lexical structure of the ambiguous verb, making interesting predictions about the classes of verbs that should give rise to processing difficulties. ‘Constituency, context, and connectionism in syntactic parsing’, by James Henderson, describes an approach towards representing constituency in connectionist networks and applies this approach to unbounded dependency constructions.

The remaining papers cover a range of areas and specific topics, from event-related potential research (‘On the electrophysiology of language comprehension: Implications for the human language system’ by Colin Brown and Peter Hagoort) to reading (‘Parsing and incremental understanding during reading’ by Martin J. Pickering and Matthew J. Traxler) and from the interpretation of relative clauses (‘Syntactic attachment and anaphor resolution: The two sides of relative clause attachment’ by Barbara Hemforth, Lars Konieczny and Christoph Scheepers) to ‘Cross-linguistic psycholinguistics’ by Maria De Vincenzi. The last two papers touch...