Interface strategies: Optimal and costly computations (review)

Interface Strategies proposes a highly ingenious framework for the interface between the computational system (CS) and the other cognitive systems (concept, inference and context systems) within the hypothesis of the Minimalist Program (Chomsky 2000) that language is an optimal solution to interface conditions. Reinhart identifies four areas where the outputs of the CS are insufficient for the cognitive systems and argues that this imperfection is remedied by interface strategies. These strategies take the form of reference-set computation: "the reference set consists of pairs <d, i> of derivation and interpretation. A given <d, i>-pair is blocked if the same interface effect could be obtained more economically-in other words, if there is a better <d, i> competitor in the reference set" (p. 36). The application of this strategy comes with some observable processing cost because it involves constructing alternative derivations and comparing their semantic interpretations.

The book consists of an introduction, five chapters, notes, references, and author and subject indices. In "Introduction: Optimal design", Reinhart briefly traces the history of reference-set computation introduced in the early 1990s and argues that this computation should be available to the CS, as it interacts with the other cognitive systems. Reinhart is also concerned with the relation of the human parser with the CS. She notes that, given the minimalist hypothesis mentioned above, the parser should be transparent with minimum parser-specific adjustments.

Chapter 1, "Reference-set computation", surveys the development of the reference-set computation in the Minimalist Program based on a case study of superiority. Reinhart introduces here the notion of interpretation-dependent computation (Golan 1993) which proposes that the CS determines the most economical derivation relative to interpretive goals. This is illustrated in (1-4).¹

1. 1.

   1. a. Who e knows what who bought e?

   2. b. Who e knows who e bought what?                                       (p. 26)

2. 2.

   1. a. For which <x, y>, x knows what y bought.

   2. b. For which <x, z>, x knows who bought z.                               (p. 27)

3. 3.

   1. a. *What did who buy e?

   2. b. Who e bought what?                                                     (p. 27)

4. 4. For which <x, y> x bought y.                                                  (p. 27)

The sentence (1a) is grammatical only with the higher scope construal of who, as in (2a). This interpretation is not available for (1b), which has the interpretation in (2b). (1a) is acceptable because it is the most economical derivation to achieve this interpretive goal. (3a) is rejected, however, because there is a more economical derivation (3b) without a superiority violation that produces the same interpretive effect (4) as (3a).

The rest of the book demonstrates that this reference-set computation proves useful in the areas of scope shift, focus shift, and anaphora resolution (Chapters 2-4) and provides acquisitional evidence for this computation in these areas (Chapter 5).

Chapter 2, "Scope-shift", illustrates the reference-set computation with Quantifier Raising (QR). The bulk of this chapter is devoted to showing that many cases of scope that have heretofore been analyzed via QR (the wide scope reading of existential quantifiers, among others) are captured in situ by choice functions. Reinhart then reanalyzes Fox's (2000) data in (5) as support for her reference-set computation.

(5)

1. a. A doctor will examine every patient.

   (ambiguous)

2. b. A doctor will examine every patient, and Lucie will [ ] too.

   (only narrow scope for every patient)                                  (p. 107)

(6)

The problem to be resolved here is why the ambiguity in (5a) disappears in (5b). This problem is solved under Reinhart's system. To obtain the wide scope reading, the parallelism requirement on ellipsis requires that QR applies to (5b), as in (6a). However, there is a more economical derivation (6b) without QR that has the same interpretive effect as (6a). Thus, QR is blocked.

Chapter 3, "Focus: The PF interface", discusses focus shift. Adopting Cinque's (1993) syntactic theory of stress and focus, Reinhart proposes that each derivation is associated with a set of possible foci in a given context. The focus set...