Applying an Exemplar Model to the Artificial-Grammar Task: Inferring Grammaticality from Similarity

Abstract

We present three artificial-grammar experiments. The first used position constraints, and the second used sequential constraints. The third varied both the amount of training and the degree of sequential constraint. Increasing both the amount of training and the redundancy of the grammar benefited participants' ability to infer grammatical status; nevertheless, they were unable to describe the grammar. We applied a multitrace model of memory to the task. The model used a global measure of similarity to assess the grammatical status of the probe and captured performance both in our experiments and in three classic studies from the literature. The model shows that retrieval is sensitive to structure in memory, even when individual exemplars are encoded sparsely. The work ties an understanding of performance in the artificial-grammar task to the principles used to understand performance in episodic-memory tasks.

An Instance Theory of Semantic Memory

“…More generally, our demonstration shows that an established and classic theory for memory that has previously been applied to understand a suite of behaviors including (a) recognition memory (Hintzman 1984), (b) frequency judgment (Hintzman 1988), (c) cued recall (Hintzman 1986), (d) classification (Hintzman 1986), (e) function learning (Kwantes and Neal 2006), (f) judgment and decision (Dougherty et al 1999;Thomas et al 2008), (g) speech normalization (Goldinger 1998), (h) confidence/accuracy inversions in eyewitness identification (Clark 1997), (i) language processing (Rosch and Mervis 1975), (j) false memory (Arndt and Hirshman 1998), (k) memory dissociations in aging (Benjamin 2010), (l) implicit learning (Jamieson and Mewhort 2009a, 2010, (m) speeded choice (Jamieson and Mewhort 2009b), (n) associative learning (Jamieson et al 2010b, (o) the production effect in recognition memory (Jamieson et al 2016a), and (p) selective memory impairment in amnesia (Jamieson et al 2010a;Curtis and Jamieson 2018) can also be used to understand semantics. The cross-lab and cross-domain effort represents the way that science ought to progress-by developing a general account of memory and its processes in a working computational system to produce a common explanation of behavior rather than a set of labspecific and domain-specific theories for different behaviors (Newell 1973).…”

The effects of ordinal load on incidental temporal learning

“…This speaks in favour of a single (implicit-explicit) memory system, possibly showing different outputs depending on the means used for probing acquired knowledge (Jamieson & Mewhort, 2009). A different, though more speculative approach might be based on the fact that temporal learning was combined with explicit learning in the two groups of our study.…”

Implicit learning of conjunctive rule sets: An alternative to artificial grammars

“…Because of the inherent structure of grammatical strings, larger chunks can be used to process them at test compared to nongrammatical strings, resulting in them having higher familiarity and more likely to be classified as grammatical. Other frameworks and models have been proposed to account for AGL that also fundamentally rely on superficial knowledge of chunks rather than rule abstraction (e.g., Dienes et al, 1991;Jamieson & Mewhort, 2009a;Knowlton & Squire, 1994;Redington & Chater, 1996; although see Higham, 1997a). …”