Consolidating skill learning through sleep

Highlights

• We review current research on how sleep benefits skill learning by consolidating memory.

• Sleep restores memories from learning a skill that are lost through interference.

• Sleep protects memories from learning a skill against future forgetting from interference.

• Sleep consolidation of learning may reorganize memories in order to stabilize memory.

• Sleep promotes the generalization and abstraction needed for effective skill learning.

Skill learning depends on retaining memories of skill-use experiences over time. These memories need to be robust against interference and therefore depend on consolidation. Further, skills must generalize beyond the learning experiences to be useful in novel but related situations. We review the role of sleep in the consolidation of skill learning, along with research findings that sleep: (1) reduces the effects of interference on skill learning, (2) protects against future interference with skill learning, (3) aids in the abstraction and generalization of skill learning. We discuss theories of sleep consolidation in terms of putative neural mechanisms and describe the key paradigms and questions in sleep research.

Introduction

We sleep (if lucky) for a substantial portion of our lives, but the role of sleep in mind and biology is still not well understood [1]. Although there has been a longstanding interest in the role of sleep in learning, ranging from studies of frozen cockroaches [2] to soccer skills [3] to studies of dreaming (cf. [4]), there are many unanswered questions, such as how sleep aids learning and memory. One of the primary theories of learning and memory is that the learning of one thing makes it harder to learn something different but similar afterwards, and learning something new can interfere with the prior learning — called proactive and retroactive interference (see [5]). Although both proactive and retroactive interference can be classified as either informational (via the overlap of content: A maps to B and A also maps to C) or processing oriented (via mental exertion or memory formation) [5], for the purposes of the present discussion such a distinction is less relevant than the recognition that (1) prior learning and subsequent experiences can have adverse effects on learning a task and (2) that sleep plays a role in mitigating such interference [6]. However, it has been demonstrated that conscious framing — i.e., the way the participant understands the task or expectations about the task — can affect both interference effects in motor sequence learning and sleep consolidation [7]. Understanding the mechanisms of interference may ultimately be linked to a clearer understanding of sleep consolidation.

An early view of the function of sleep is that it aids memory passively, simply because sleeping reduces opportunities for interference from subsequent experience via the lack of consciously directed activity [8]. Research rejected this simple view of sleep (e.g., [9]) and replaced it with an alternative view that sleep actively consolidates memories [10]. However, there are new theories proposing different possible mechanisms operating during sleep to protect or consolidate memory [11]. One broad neural network view holds that sleep operates to consolidate immediate encodings of experience based on some specific brain regions (e.g., hippocampus) into more stable neocortical representations [12] whereas other theories focus on changes in the synaptic connections among neurons [13].

A paradigmatic shift from the historical focus on explicit memories encoding information about specific events or experiences (remembering word lists or lunch yesterday) to more implicit memories demonstrating learned performance, such as playing golf or piano [14] has been important to understanding the role of sleep in learning. As in other approaches to memory research such as neuropsychological patients like HM [15] or non-human animal research [16] the distinction between declarative memory (i.e., describable) and procedural or non-declarative memory (i.e., skill learning, task learning) has provided different kinds of results. This shift from explicit memory for experiences to non-verbal procedures (e.g., perceptual discrimination, motor sequences) however, has changed a number of aspects of the learning situation. These aspects primarily arise because procedural or non-declarative memory tasks marked a shift from learning many things (e.g., lists of words) to learning a single simple task (e.g., repeated finger movements [17] or a single visual pattern discrimination [18]). Given that non-verbal, procedural skills like tennis or telegraphy develop from experiences over time — over days or even years [19], memories of these experiences must accrue to support learning. Moreover, for skills to generalize to new but related situations, skills must be abstracted from specific experiences. Our view is that a skill is not simply the learning of a single motor pattern (i.e., a sequence of finger movements) or a single visual discrimination, but instead is best conceptualized as a generalized capability to perceive new patterns that share properties or to produce new movement patterns under novel conditions ranging from riding a bicycle to playing chess to becoming a radiologist. But it is also important to note that while declarative and procedural (skill-like) memories have many identified differences, declarative memories are not studied directly. The formation and retrieval of declarative memories is strategic [20] and may be conceptualized as procedural [21], and thus may itself be a kind of skill.