When attention matters: Domain-specific disruption of explicit but not implicit memory in statistical learning

Statistical learning (SL), defined as the ability to extract regularities in the environment, plays a key role in speech segmentation and language acquisition. An important yet still unresolved question is whether top-down attention influences SL, and if so, whether this depends on domain-specific or domain-general attentional resources. To address this question, participants listened to syllables adhering to a statistical structure (forming trisyllabic "words") under different attentional manipulations. In one (domain-general) condition, participants either engaged with or passively viewed visuospatial stimuli, while in the second (domain-specific) condition, participants engaged with or passively viewed competing linguistic stimuli. SL was assessed using an explicit familiarity-rating task and an implicit reaction-time based syllable detection task. On the explicit measure, participants who engaged in a competing linguistic task performed more poorly than control participants, whereas performance was similar in the two visuospatial conditions, regardless of attentional condition. In contrast, all participants showed comparable and robust performance on the implicit measure of statistical learning. These results suggest that explicit memory for statistical regularities is impaired if, and only if, domain-specific attention is reduced, whereas the implicit expression of SL is robust against reductions in attentional resources of any kind. Moreover, both explicit and implicit performance remained above-chance under conditions of divided attention, even when the concurrent task required linguistic processing. These results have practical implications for language learners, who may be able to extract basic statistical properties of language input while engaging in a concurrent, unrelated task.