Can sentence meaning be built up locally from the meanings of words and their modes of combination, without positing a separate intermediate “logical form” to do the interpretive work?
My links: https://linktr.ee/frictionphilosophy.
1. Guest
Polly Jacobson is Professor of Linguistics at Brown University, where she has been a professor for many years. Her work focuses on linguistics and formal semantics, and especially on the formal tools needed to model the syntactic and semantic systems of natural languages.
2. Interview Summary
After the introductions, Polly Jacobson (Brown University) and the host of Friction move straight into compositionality: the idea (roughly) that sentence meanings are determined by the meanings of their parts plus the rules for putting those parts together. Jacobson treats some compositionality as non-negotiable, but emphasizes that the real controversies concern how much compositionality natural language exhibits, what the composition rules are, and how the topic developed into a live research program rather than a slogan.
From there, the interview centers on Jacobson’s “direct compositionality” approach by contrasting it with an “indirect” picture where syntax first generates an intermediate representation (often described as a kind of “logical form”) and semantics interprets that representation afterward. The direct-compositional alternative aims to avoid that intermediate interpretive layer and keep syntax and semantics working in tighter tandem. This leads into the “degrees” or “types” of direct compositionality: at the strong end, syntax is (in a sense) “blind” to structure and semantics does the interpretive work; at weaker ends, syntax is allowed more operations—without collapsing back into the idea that it’s merely producing a logical form for semantics to read off.
A recurring payoff claim is that direct compositionality helps preserve local interpretation—making it clearer how meaning can be built up step-by-step without global “repair.” In that spirit, they discuss why an operation like infixation (letting substrings/phrases combine in more flexible ways) might be needed to prevent a syntax–semantics mismatch where some structural piece “should” get an interpretation but can’t under stricter locality constraints—while also noting how quickly things could become unwieldy if the system allowed unlimited iterations of such operations. The conversation also links the view to processing considerations: if humans don’t wait until the end of a sentence to interpret it, a direct-compositional framework can look like a natural fit, whereas denying it may require extra theory to explain incremental understanding. The interview closes with Jacobson on open problems and on the broader value of theorizing about language—even if models idealize a messy phenomenon, there’s real intellectual and explanatory value in trying to understand complex systems better.
3. Interview Chapters
00:00 - Introduction
00:45 - Direct compositionality
09:45 - Misconception
12:46 - Direct vs. indirect compositionality
22:02 - Cross-serial dependencies and infixation
32:07 - Example with infixation
35:48 - Motivation for direct compositionality
40:07 - Language without direct compositionality?
51:26 - Language processing
56:48 - Complicated syntax and direct compositionality
1:00:15 - Degrees of direct compositionality?
1:01:00 - What more might be needed?
1:06:41 - Donkey sentences
1:11:58 - More motivations for direct compositionality
1:15:08 - Analogy to logic
1:17:26 - Idioms and locality
1:26:13 - Challenges to view
1:28:17 - Value of linguistics
1:32:47 - Idealization
1:34:54 - Conclusion
