Fluid Abilities and Rule Learning: Patterning and Biconditional Discriminations

Abstract

Previous experience with discrimination problems that can only be solved by learning about stimulus configurations enhances performance on new configural discriminations. Some of these effects can be explained by a shift toward increased configural processing (learning about combinations of cues rather than about individual elements), or by a tendency to generalize a learned rule to a new training set. We investigated whether fluid abilities influence the extent that previous experience with configural discriminations improves performance on subsequent discriminations. In Experiments 1 and 2 we used patterning discriminations that could be solved by applying a simple rule, whereas in Experiment 3 we used biconditional discriminations that could not be solved using a rule. Fluid abilities predicted the improvement on the second training set in all experiments, including Experiment 3 in which rule-based generalization could not explain the improvement on the second discrimination. This supports the idea that fluid abilities contribute to performance by inducing a shift toward configural processing rather than rule-based generalization. However, fluid abilities also predicted performance on a rule-based transfer test in Experiment 2. Taken together, these results suggest that fluid abilities contribute to both a flexible shift toward configural processing and to rule-based generalization.

Transfer of associability and relational structure in human associative learning.

“…Humans and other animals typically find biconditional discriminations more difficult to acquire than patterning discriminations (Harris & Livesey, 2008; Harris, Livesey, Gharaei, & Westbrook, 2008). The relations between cues in a biconditional discrimination are more complex and difficult to describe via a simple heuristic like the opposites rule (see Baetu, Burns, Yu, & Baker, 2018). Thus, it is debatable whether learning the relations involved in the structure of the biconditional task is as useful as in patterning, where it may be advantageous for overcoming the deleterious effects of summation on negative patterning, in particular (e.g., see Livesey, Thorwart, & Harris, 2011; Thorwart, Uengoer, Livesey, & Harris, 2017).…”

“…Although Mitchell, Griffiths, Seetoo, and Lovibond (2012) found that the learned predictiveness in one experiment was completely reversible by instructions, others have reported persistent resistance to instructed reversal (Don & Livesey, 2015;Shone, Harris, & Livesey, 2015). Baetu et al (2018) found that improvements in performance completing the second of two successive biconditional discriminations were predicted by measures of fluid ability. Rather than concluding that this effect was based on rule-learning, they argued that this demonstrates flexible use of configural processing among individuals with higher cognitive capacities.…”

Relational rule discovery in complex discrimination learning.

“…Both human participants and laboratory animals find learning biconditional discriminations relatively difficult; they are acquired more slowly than linearly solvable discriminations that are matched for presentation of compound cues (Livesey, Thorwart, de Fina, & Harris, 2011;Saavedra, 1975) but also more slowly than positive and negative patterning discriminations Harris, Livesey, Ghareai, & Westbrook, 2008). Baetu, Burns, Yu & Baker (2018) recently argued that biconditional discriminations cannot be solved using a rule. However, there is a relational structure within a biconditional discrimination that can be abstracted and applied to novel stimuli.…”

“…Baetu, Burns, Yu, and Baker (2018) recently argued that biconditional discriminations cannot be solved using a rule. However, there is a relational structure within a biconditional discrimination that can be abstracted and applied to novel stimuli.…”

“…Similarly, Wills, Graham, et al (2011) found that participants who scored higher on a measure of complex working memory (OSPAN; Turner & Engle, 1989) were more likely to display rule transfer in the patterning task. Baetu et al (2018) measured participants’ performance on a wider range of fluid abilities and also gave them either patterning or biconditional discriminations to solve sequentially. They found that a number of fluid abilities (most consistently, reasoning ability as measured by RPM and the Comprehensive Abilities Battery–Induction, Hakstian & Cattell, 1975) were related to both rule transfer in the patterning task and performance on the second of two complex discriminations that shared the same relational structure as the first.…”