Information theory and artificial grammar learning: inferring grammaticality from redundancy

Abstract

In artificial grammar learning experiments, participants study strings of letters constructed using a grammar and then sort novel grammatical test exemplars from novel ungrammatical ones. The ability to distinguish grammatical from ungrammatical strings is often taken as evidence that the participants have induced the rules of the grammar. We show that judgements of grammaticality are predicted by the local redundancy of the test strings, not by grammaticality itself. The prediction holds in a transfer test in which test strings involve different letters than the training strings. Local redundancy is usually confounded with grammaticality in stimuli widely used in the literature. The confounding explains why the ability to distinguish grammatical from ungrammatical strings has popularized the idea that participants have induced the rules of the grammar, when they have not. We discuss the judgement of grammaticality task in terms of attribute substitution and pattern goodness. When asked to judge grammaticality (an inaccessible attribute), participants answer an easier question about pattern goodness (an accessible attribute).

AGSuite: Software to conduct feature analysis of artificial grammar learning performance

“…Several research groups have approached the exploration of features via experimental design, whereby the features of strings are crossed factorially and their relative contributions are gauged by group-level comparisons between the conditions (e.g., Jamieson, Nevzorova, et al, 2016;Johnstone & Shanks, 2001;Kinder, 2000;Kinder & Lotz, 2009;Vokey & Brooks, 1992). Others have made use of alternative statistical techniques that offer more fine-grained analyses.…”

“…First-Order Redundancy (Jamieson, Nevzorova, et al, 2016) BHow predictable are the bigrams (two letter strings) in this particular test string?T he test string is first decomposed into its bigrams and the probability of each bigram appearing in the string is computed. First-Order Redundancy measures the predictability of the bigrams in the test string, by computing the complement of the average information relative to the maximum information possible defined as a string of the same length made up of…”

“…Analysis of the features that guide performance is not in itself a new idea. Indeed, the AGL literature is replete with examples of factorial experiments designed to tease apart the contributions of various string features, and thereby distinguish between the several accounts of the phenomenon (e.g., Jamieson, Nevzorova, Lee, & Mewhort, 2016;Johnstone & Shanks, 2001;Kinder, 2000;Kinder & Lotz, 2009;Vokey & Brooks, 1992). Although the value of such analyses cannot be denied, they are limited by the a priori nature of their experimental design.…”

Designing Improved Teams for Crowdsourced Competitions

“…Optimal scrum team size is based on Miller's number (7±2) [33], the average number of information chunks that a person can hold in short-term memory [34]. It is now used in information theory and user interface design [35,36]. Miller's (1956) work indicates a natural human tendency to chunk, or group, information so that it is manageable for our short-term memory.…”

Designing Improved Teams for Crowdsourced Competitions

“…It is now used in information theory and user interface design [35,36]. Miller's (1956) work indicates a natural human tendency to chunk, or group, information so that it is manageable for our short-term memory.…”