Interpreting gradient acceptability judgments: evidence from word order alternations in English
English & Linguistics, Purdue University
What conclusions can we draw from gradient patterns of acceptability judgments about speakers’ implicit grammatical knowledge? In this talk, I highlight some of the main arguments from my forthcoming book, Gradient Acceptability and Linguistic Theory, in the context of several studies from our lab at Purdue.
Among linguists working within various syntactic frameworks, there is a consensus that many factors in addition to syntactic constraints can affect the responses obtained in an acceptability judgment task (Schütze 1996, 2011). These include semantic, pragmatic, and prosodic constraints as well as general cognitive factors. However, each syntactic framework comes with certain assumptions that influence how linguists tend to interpret judgments in relation to speakers’ knowledge of grammar. Here, I identify two such assumptions, as elaborated in (1) and (2).
(1) Strict form-meaning isomorphism vs. flexible mappings: In Minimalism/P&P, there is a strong assumption of isomorphism between syntactic structure and meaning, especially for event structure (Hale & Keyser 1993; Marantz 2013) and discourse information structure (Rizzi 1997). This assumption is not shared by constraint-based theories such as Construction Grammar (Goldberg 2006; Michaelis 2012) and Simpler Syntax (Culicover & Jackendoff 2005). As a result, constraint-based theories show a more flexible set of form-meaning mappings, reducing the role of syntax in accounting for gradient judgments that involve contrasts in meaning.
(2) Probabilistic vs. categorical constraint application: In contrast to Minimalism/P&P, standard Optimality Theory, and some constraint-based theories, Stochastic Optimality Theory (Bresnan et al 2001) and Linear Optimality Theory (Sorace & Keller 2005) allow for ‘soft constraints’ – grammatical constraints that contribute to statistical preferences rather than strict ungrammaticality. Similar ideas are also proposed in usage-based versions of Construction Grammar (Goldberg 2006, 2019).
In this presentation, I show how different theoretical assumptions lead to different interpretations of the same judgment data, using previous and ongoing studies from our lab as examples. These studies investigate two word order alternations in English: relative clause extraposition (Francis & Michaelis 2014, 2017; Weirick & Francis in prep) and the dative alternation (Weirick 2021; Weirick & Francis in prep). I argue that while the patterns of judgments shown in these studies are compatible with various theoretical interpretations, they can most plausibly be interpreted as involving soft constraints within the grammar. Finally, using the same studies as examples, I show how data from additional methods (corpus studies, production tasks, and comprehension tasks) and from second language speakers of English from different L1 backgrounds can shed further light on competing theoretical accounts.