A Merge-Based Approach to the Syntax-Semantics Interface

These notes are based on a career of looking at issues in the syntax-semantics interface, including argument structure, anaphora, negation, quantifiers, coordination, and ellipsis.

Philosophy

1.

Be open to syntax.

Be open to radically syntactic approaches to traditionally semantic and pragmatic issues, no matter how authoritative or how widely adopted the semantic or pragmatic approach might be.

2. 

Semantic operations are defined in terms of syntactic structure. 

In other words, there is no semantics without syntax. As a consequence, all semantic analyses involve important assumptions about syntax. 

3.

Hypothesis: The correct syntactic analysis will have a semantic interpretation that is simple, transparent and uniform (in the sense clarified below in 4-8).

Specific Assumptions

Here are some specific assumptions that help to articulate the above hypothesis.

4.

Semantic operations are simple and limited in type (simplicity).

Semantic operations are as constrained as possible, perhaps only function application and predicate abstraction. Powerful semantic operations such as type shifting, function composition, existential closure and semantic coercion are not part of UG. If an analysis uses such operations, try to think about how to replicate their effects syntactically (e.g., using Merge) and then test the consequences.

5.

Semantic values are simple (simplicity).

The semantic values of lexical items are simple. The more complex a semantic value is, the greater the chance that it should be broken into several different morphemes each with its own simple semantic value. These morphemes will them be merged together in the syntactic derivation.

6.

Variables are represented syntactically (transparency).

If there is a variable x in the semantic representation, then there is a DP in the syntactic representation which is interpreted as the variable x. This constraint forces variables to be introduced into both syntax and semantics via Merge. A simple example is the passive. Since there is a variable representing the agent in the short passive, there must be a DP in the syntactic representation interpreted as that variable. See Collins 2024 for a discussion and references.

7.

Semantic interpretation is uniform (uniformity).

Semantic interpretation is uniform in the sense that a single syntactic structure is not interpreted in two different ways. A single syntactic structure has a single interpretation. 

For example, a DP in a position P, cannot be interpreted optionally as specific or non-specific. To get the two interpretations, something must be different, either the internal syntactic structure of the DP, or the syntactic position of the DP, or both.

8.

Semantic relations and predicates are directly represented in syntactic structure to the greatest extent possible (transparency).

For example, the scope of one quantifier phrase QP1 over another QP2 is represented syntactically in terms of asymmetric c-command at some point in the derivation (not in terms of type shifting one quantifier phrase to given it semantic scope over another). Another example is Rizzi 1997 use of FocP to syntactically represent the semantics of focus.

Of all the assumptions concerning the syntax-semantics interface, (8) is the vaguest and hardest to pin down since the notion of “semantic relations and predicates” is very broad. Furthermore, it may be that (8) follows from the other assumptions in (4-7).

Method

9.

Work out the truth conditions, entailments, presuppositions, scope and other semantic properties. Pay close attention to how the semantic properties vary with varying syntactic configurations. Are there important generalizations about how differences in syntax affect changes in interpretation? 

10.

Work out the syntactic analysis using Merge-based framework. 

(9) and (10) feed into one another. Semantic properties may suggest a particular syntactic analysis (in conformity with (4-8). And the syntactic analysis may suggest investigating certain semantic properties.

11.

Work out the compositional semantics based on assumptions 4-8.

Depending on the audience and venue, it is not always necessary to give the complete compositional semantics. But you should be confident about what the semantic properties of your proposed syntactic analysis are. Your syntactic representation should capture the semantic properties of the construction.

Evaluation

12.

Justify your syntactic analysis using traditional syntactic tests.

You don’t just get to propose syntactic structure free of charge. Since the assumptions in (4-8) are a hypothesis, you need to prove that your proposed analysis is correct or at least better than alternatives. Whatever representation you ultimately come up with, it should be consistent with standard syntactic tests (e.g., constituent structure tests, Binding Theory, islands, distributional tests, agreement, etc.). This is your opportunity to dig very deeply into the syntax of the construction.

13.

Justify your syntactic analysis using cross-linguistic comparisons.

For some topics, it might be easier probe syntactic structure and its semantic interpretation in a different language. In some languages, the morphosyntax of a construction clearly reflects its semantic properties. Investigating such languages can give important clues about the best syntactic analysis cross-linguistically. For example, it is much easier to investigate the syntactic and semantic properties of logophoricity in Ewe, which has a morphologically distinct logophoric pronoun.

14.

Look for the smoking gun.

Does the syntactic approach yield insights into any particular aspect of the problem? Are there interesting empirical predications? Is there a deep theoretical consequence? Try to find something that would convince a disinterested party, or even a skeptic. 

Acknowledgments: I thank Richard Kayne, Paul Postal and Gary Thoms for comments on a draft of this blog post.