Smolensky and Goldrick (2016): Gradient Symbolic Representations in Grammar: The case of French Liaison

Comments welcome!

Direct link:

ROA: 1286
Title: Gradient Symbolic Representations in Grammar: The case of French Liaison
Authors: Paul Smolensky, Matthew Goldrick
Length: 37 pages
Abstract: Longstanding theoretical debates about whether structure A or structure B is the correct analysis of phenomenon X are commonplace. For example, at the juncture of two words W1 and W2, French liaison consonants alternate with zero. Theories of French phonology have long debated whether the consonant is associated with W1 or W2. In this work, we argue for an alternative approach. Phenomena X is not accounted for by either A or B, but rather a conjunctive blend of structures A and B. This notion of ‘blend of structures’ is formalized using Gradient Symbolic Representations, symbol structures in which a particular position is generally occupied by a sum of gradient symbols, each symbol having a partial degree of presence: its activity. The grammatical consequences of a Gradient Symbolic Representation are the sum of the consequences of all the symbols blended to form it; the consequences of a symbol – e.g., the costs of constraint violations – are proportional to its activity. The proposed grammatical computation consists of optimization with respect to a numerical weighting of familiar phonological constraints from Optimality Theory and Harmonic Grammar, straightforwardly extended to evaluate Gradient Symbolic Representations. We apply this general framework to French liaison consonants, blending together elements of previous proposals to give a single analysis that covers a wide range of data not previously explicable within a single theory.
Type: Paper/tech report
Keywords: gradient symbolic representations, probabilistic harmonic grammar, French phonology, liaison



2 thoughts on “Smolensky and Goldrick (2016): Gradient Symbolic Representations in Grammar: The case of French Liaison

  1. Joe Pater Post author

    Very happy to see this paper! As I’ve already mentioned to Matt and Paul (who expressed gratitude for the reminder and interest in adding more to the paper), there is an analysis of French liaison (and other phenomena) in terms of another approach to lexical gradience, UR constraints, in Brian Smith’s 2015 dissertation:

    I should also mention that I’ve long thought that “lexically conditioned variation” is a good argument for some form of lexical gradience. There’s a discussion of this in section 5.1 of my variation overview paper with Andries Coetzee, and a survey of some possible approaches in sec. 5.2:

    This also comes up in section 4 of some work on learning and UR constraints I did with Brian, Karen Jesney and Robert Staubs:

  2. Matt Goldrick

    Thanks for this, Joe! As Paul mentioned to you offline, one of the things we want to do in producing the next version of that paper is to collect as many neglected references as possible. We really appreciate any help Phonolisters can provide in this area–as well as any other comments on the work!


Leave a Reply

Your email address will not be published. Required fields are marked *