Drawing on an example from artificial grammar learning, I present the case that similarity processes can be computationally identical to rules processes, but that participants in an artificial grammar learning experiment may use different processing modes to classify stimuli. The number of properties and other representational differences between rule and similarity processes are an accidental consequence of strategies used.

Are rules processes or similarity processes the default for acquisition of grammatical knowledge during natural second language acquisition? Whereas Pothos argues similarity processes are the default in the many areas he reviews, including artificial grammar learning and first language development, I suggest, citing evidence, that in second language acquisition of grammatical morphology “rules processes” may be the default.

In most cases, rule-governed relations and similarity relations can indeed be distinguished by the number of relevant features they require. This criterion is not sufficient, however, to explain other properties of the relations that have a more dichotomous character. I focus on the differential drive for consistency by inferential processes that draw on the two types of relations.

There is a continuum between prototypical cases of rule use and prototypical cases of similarity use. A prototypical rule: (1) is explicitly represented, (2) can be verbalized, and (3) requires that the user selectively attend to a few features of the object, while ignoring the others. Prototypical similarity-use requires that: (1) the user should match the object to a mental representation holistically, and (2) there should be no selective attention or inhibition. Neural evidence supports prototypical rule-use. Most models of categorization fall between the two prototypes.

It is difficult to see how one can support the continuum between rules and similarity, as Pothos proposes. A similarity theory could dispense with the rules end of the continuum. The only thing that we need is one (or more than one) theory of similarity that goes beyond the stimulus-carrying information and behavioristic restrictions that have usually been attributed to similarity theories.

The Simplicity and Power (SP) theory (Wolff 2003a) provides support for Pothos's proposals by illustrating how the effect of “rules” and “similarity” may be achieved within an integrated model that makes no explicit provision for either concept. The theory is described here in outline with simple examples to show how rules and similarity can emerge as properties of the system in learning, reasoning, categorization, and the parsing of language.

This response to the open peer commentary discusses what should be the appropriate explanatory scope of a rules versus similarity proposal and accordingly evaluates the Rules versus Similarity one. Additionally, coherence, goals, and commitment are presented as inferential notions, fully consistent with the Rules versus Similarity distinction, that allow us to predict when Rules would be preferred to Similarity.

Research in the neurosciences continues to provide evidence that sleep plays a role in the processes of learning and memory. There is less of a consensus, however, regarding the precise stages of memory development during which sleep is considered a requirement, simply favorable, or not important. This article begins with an overview of recent studies regarding sleep and learning, predominantly in the procedural memory domain, and is measured against our current understanding of the mechanisms that govern memory formation. Based on these considerations, I offer a new neurocognitive framework of procedural learning, consisting first of acquisition, followed by two specific stages of consolidation, one involving a process of stabilization, the other involving enhancement, whereby delayed learning occurs. Psychophysiological evidence indicates that initial acquisition does not rely fundamentally on sleep. This also appears to be true for the stabilization phase of consolidation, with durable representations, resistant to interference, clearly developing in a successful manner during time awake (or just time, per se). In contrast, the consolidation stage, resulting in additional/enhanced learning in the absence of further rehearsal, does appear to rely on the process of sleep, with evidence for specific sleep-stage dependencies across the procedural domain. Evaluations at a molecular, cellular, and systems level currently offer several sleep specific candidates that could play a role in sleep-dependent learning. These include the upregulation of select plasticity-associated genes, increased protein synthesis, changes in neurotransmitter concentration, and specific electrical events in neuronal networks that modulate synaptic potentiation.

Based on brain state-dependent behavioral changes, consolidation of sensorimotor memories has been posited to evolve in two different functional stages. Only the second of these stages requires sleep and leads to performance benefits. Recent results, however, suggest that sleep is not always crucial for the expression of delayed behavioral gains but might be critical for enhancing automaticity in the absence of attention, another expression of memory consolidation.

Insights into the role of sleep in the molecular mechanisms of memory consolidation may come from studies of activity-dependent synaptic plasticity, such as long-term potentiation (LTP). This commentary posits a specific contribution of sleep to LTP stabilization, in which mRNA transported to dendrites during wakefulness is translated during sleep. Brain-derived neurotrophic factor may drive the translation of newly transported and resident mRNA.

It is suggested that Walker's consolidation-based enhancement of memory during REM sleep corresponds to the simulated annealing technique used for function optimization, and that robotic and AI design could benefit from inclusion of a deliberate REM-like memory optimization phase.

Walker's target article proposes a refinement of the well known two-stage model of memory formation to explain the positive effects of sleep on consolidation. After a first stage in which a labile memory representation is formed, a further stabilisation of the memory trace takes place in the second stage, which is dependent on (REM) sleep. Walker has refined the latter stage into a stage in which a consolidation-based enhancement occurs. It is not completely clear what consolidation-based enhancement implies and how it can be dissociated from a stage for memory-stabilisation. A more serious consideration, however, is whether a second stage in memory consolidation that is solely dependent on sleep, is really necessary. The classical, passive, interference theory is able to explain adequately the findings related to the effects of sleep and memory, and can lead perhaps better to an understanding of the highly variable data in this field.

Following Karni's seminal work, Walker and other researchers have recently provided gradually convincing evidence that sleep is critical for the consolidation-based enhancement (CBE) of motor sequence learning. Studies in our laboratory using a motor adaptation paradigm, however, show that CBE can also occur after the simple passage of time, suggesting that sleep effects on memory consolidation are task-related, and possibly dependent on anatomically dissociable circuits.

We address the notion of integration of new memory representations and the potential dependence of this phenomenon on sleep, in light of recent findings on the lexicalization of spoken words. A distinction is introduced between measures tapping directly into the strength of the newly acquired knowledge and indirect measures assessing the influence of this knowledge on spoken word identification.

Learning procedural skills involves improvement in speed and accuracy. Walker proposes two stages of memory consolidation: enhancement, which requires sleep, and stabilization, which does not require sleep. Speed improvement for a motor learning task but not accuracy occurs after sleep-dependent enhancement. We discuss this finding in the context of computational models and underlying sleep mechanisms.

In this target article, Walker seeks to clarify the current state of knowledge regarding sleep and memory. Walker's review represents an impressively heuristic attempt to synthesize the relevant literature. In this commentary, we question the focus on procedural memory and the use of the term “consolidation,” and we consider the extent to which empirically testable predictions can be derived from Walker's model.

I discuss how the work in Walker's article adds to the considerable body of research on dreaming, sleep, and memory that appeared in the early days of modern sleep research. I also consider the issue of REM-independent and REM-dependent kinds of learning. This requires including emotional issues in our discussion, and therefore emphasizes the importance of studying and understanding dreams.

We argue that by neglecting the fact that procedural memory may also have episodic qualities, and by considering only a systems approach to memory, Walker's account of consolidation of learning during subsequent sleep ignores alternative accounts of how sleep stages may be interdependent. We also question the proposition that sleep-based consolidation largely bypasses hippocampal structures.

The progressive multistage stabilization of memory (consolidation) relies on post-acquisition neural reorganization. We hypothesize that two processes subserve procedural memory consolidation and are reflected in delayed post-acquisition performance gains: (1) synaptic consolidation, which is classical Hebbian, and (2) in some tasks, concurrently or consequently, “system consolidation,” which might in some skills be sleep-dependent. Behavioral interference may affect either type of consolidation.

This commentary reviews electrophysiological research suggesting that oscillatory electrical potentials recorded by the EEG could have function at cellular and DNA levels. Evidence supporting the potential functional significance of sleep-state-specific frequencies includes psychoactive neurochemical alteration of CNS electrophysiology, and sleep-state-specific alteration of dreaming. As Walker proposes, physiologic electrical fields are likely to have a functional role in the consolidation of memory.