Dual systems of speech category learning across the lifespan

Abstract

Although categorization is fundamental to speech processing, little is known about the learning systems that mediate auditory categorization and even less is known about changes across the life span. Vision research supports dual-learning systems that are grounded in neuroscience and are partially dissociable. The reflective, rule-based system is prefrontally mediated and uses working memory and executive attention to develop and test rules for classifying in an explicit fashion. The reflexive, information-integration system is striatally mediated and operates by implicitly associating perception with actions that lead to reinforcement. We examine the extent to which dual-learning systems mediate auditory and speech learning in younger and older adults. We examined auditory category learning when a rule-based strategy (Experiment 1) or information-integration strategy (Experiment 2) was optimal, and found an age-related rule-based deficit, but intact information-integration learning. Experiment 3 examined natural auditory category learning, and found an age-related performance deficit. Computational modeling suggested that this was attributable to older adults' persistent reliance on suboptimal, unidimensional strategies when 2-dimensional strategies were optimal. Working memory capacity was also found to be associated with improved rule-based and natural auditory category learning, but not information-integration category learning. These results suggest that dual-learning systems are operative in speech category learning across the life span, and that performance deficits, when present, are attributable to deficiencies in frontally mediated, rule-based processes.

Figure 1

(A) A schematic of one possible conjunctive rule-based problem in which the pitch (High or Low) and duration (Long or Short) dimensions are relevant. (B) A schematic of one possible information-integration problem in which the duration dimension (Long or Short) is irrelevant.

Figure 2

(A) Trials-to-criterion for younger and older adults in the rule-based category learning condition. Standard error bars included. (B) Proportion of younger and older adult learners in the rule-based category learning condition.

Figure 3

(A) Trials-to-criterion for younger and older adults in the information-integration category learning condition. Standard error bars included. (B) Proportion of younger and older adult learners in the information-integration category learning condition.

Figure 4

(A) Scatterplot of all stimuli from Experiment 3. (B) Scatterplot of male-talker stimuli from Experiment 3. (C) Scatterplot of female-talker stimuli from Experiment 3. Stimuli dimensions (Pitch Height and Pitch Direction) were normalized between 0 and 1.

Figure 5

(A) Proportion correct for younger and older adults in the Mandarin tone category learning task. (B) Final block proportion correct for younger and older adult Separators as a function of model type. Standard error bars included.

Figure A1

Scatterplots of the responses along with the decision boundaries that separate response regions from a hypothetical participant using a version of the (A) Striatal Pattern Classifier, (B) Conjunctive rule-based, (C) Uni-Dimensional_Height, and (D) Uni-Dimensional_Direction models as applied to the female talker stimuli shown in Fig 4C.