Stanovich & West's dual-system represents a major development in an understanding of reasoning and rationality. Their notion of System 1 functioning as a computational escape hatch during the processing of complex tasks may deserve a more central role in explanations of reasoning performance. We describe examples of apparent escape-hatch processing from the reasoning and judgement literature.

Two additional sources of evidence are provided in support of localist coding within connectionist networks. First, only models with localist codes can currently represent multiple pieces of information simultaneously or represent order among a set of items on-line. Second, recent priming data appear problematic for theories that rely on distributed representations. However, a faulty argument advanced by Page is also pointed out.

The present description of the Merge model addresses only auditory, not audiovisual, speech perception. However, recent findings in the audiovisual domain are relevant to the model. We outline a test that we are conducting of the adequacy of Merge, modified to accept visual information about articulation.

Numerous studies have replicated the finding of mentation in both rapid eye movement (REM) and nonrapid eye movement (NREM) sleep. However, two different theoretical models have been proposed to account for this finding: (1) a one-generator model, in which mentation is generated by a single set of processes regardless of physiological differences between REM and NREM sleep; and (2) a two-generator model, in which qualitatively different generators produce cognitive activity in the two states. First, research is reviewed demonstrating conclusively that mentation can occur in NREM sleep; global estimates show an average mentation recall rate of about 50% from NREM sleep – a value that has increased substantially over the years. Second, nine different types of research on REM and NREM cognitive activity are examined for evidence supporting or refuting the two models. The evidence largely, but not completely, favors the two-generator model. Finally, in a preliminary attempt to reconcile the two models, an alternative model is proposed that assumes the existence of covert REM sleep processes during NREM sleep. Such covert activity may be responsible for much of the dreamlike cognitive activity occurring in NREM sleep.

Rolls defends a higher-order thought theory of phenomenal consciousness, mapping the distinction between conscious and non-conscious states onto a distinction between two types of action and corresponding neural pathways. Only one type of action involves higher-order thought and consequently consciousness. This account of consciousness has implausible consequences for the nature of stimulus-reinforcement learning.

Stanovich & West (S&W) have two goals, one concerned with the evaluation of normative models, the other with development of prescriptive models. Individual differences have no bearing on normative models, which are justified by analysis, not consensus. Individual differences do, however, suggest where it is possible to try to improve human judgments and decisions through education rather than computers.

We consider the motivation for the principle of bottom-up priority and its consequence for information flow in Merge. The relationship between the bottom-up priority rule and inhibitory effects is also discussed, along with data that demonstrate inhibitory influences in phoneme monitoring.

In his target article Page proposes a definition of the term “localist.” In this commentary I argue that his definition does not serve to make a principled distinction, as the inclusion of vague terms make it susceptible to some problematic counterexamples.

The argument that Broca's area is preferentially involved in specific syntactic operations is based on a strong assertion regarding patterns of sentence comprehension found among patients with Broca's aphasia. This assertion is shown to be largely inconsistent with the available evidence from published studies, which indicates that only a subgroup of Broca patients demonstrate the target pattern.

We present evidence disputing the hypothesis that memories are processed or consolidated in REM sleep. A review of REM deprivation (REMD) studies in animals shows these reports to be about equally divided in showing that REMD does, or does not, disrupt learning/memory. The studies supporting a relationship between REM sleep and memory have been strongly criticized for the confounding effects of very stressful REM deprivation techniques. The three major classes of antidepressant drugs, monoamine oxidase inhibitors (MAOIs), tricyclic antidepressants (TCAs), and selective serotonin reuptake inhibitors (SSRIs), profoundly suppress REM sleep. The MAOIs virtually abolish REM sleep, and the TCAs and SSRIs have been shown to produce immediate (40–85%) and sustained (30–50%) reductions in REM sleep. Despite marked suppression of REM sleep, these classes of antidepressants on the whole do not disrupt learning/memory. There have been a few reports of patients who have survived bilateral lesions of the pons with few lingering complications. Although these lesions essentially abolished REM sleep, the patients reportedly led normal lives. Recent functional imaging studies in humans have revealed patterns of brain activity in REM sleep that are consistent with dream processes but not with memory consolidation. We propose that the primary function of REM sleep is to provide periodic endogenous stimulation to the brain which serves to maintain requisite levels of central nervous system (CNS) activity throughout sleep. REM is the mechanism used by the brain to promote recovery from sleep. We believe that the cumulative evidence indicates that REM sleep serves no role in the processing or consolidation of memory.

Our commentary outlines a number of arguments questioning an autonomous model of word recognition without feedback. Arguments are presented against the need for a phonemic decision stage and in support of a featural level in a model including feedback.

Despite problems with statistical significance, ancillary hypotheses, and integration into an overall view of cognition, Grodzinsky's demotion of Broca's area to a mechanism for tracking moved constituents is intrinsically plausible and fits a realistic picture of how syntax works.

Stanovich & West interpret errors in syllogistic reasoning in terms of computational limitations. I argue that the variety of strategies used by reasoners in solving syllogisms requires us to consider also performance errors. Although reasoners' performance from one trial to another is quite consistent, it can be different, in line with the definition of performance errors. My argument has methodological implications for reasoning theories.

Page proposes a simple, localist, lateral inhibitory network for implementing a selection process that approximately conforms to the Luce choice rule. I describe another localist neural mechanism for selection in accordance with the Luce choice rule. The mechanism implements an independent race model. It consists of parallel, independent nerve fibers connected to a winner-take-all cluster, which records the winner of the race.

Motivation and emotion are not clearly defined and differentiated in Rolls's The brain and emotion, reflecting a widespread problem in conceptualizing these phenomena. An adequate theory of emotion cannot be based upon reward and punishment alone. Basic mechanisms of arousal, agonistic, and prosocial motives-emotions exist in addition to reward-punishment systems.

Several theories claim that dreaming is a random by-product of REM sleep physiology and that it does not serve any natural function. Phenomenal dream content, however, is not as disorganized as such views imply. The form and content of dreams is not random but organized and selective: during dreaming, the brain constructs a complex model of the world in which certain types of elements, when compared to waking life, are underrepresented whereas others are over represented. Furthermore, dream content is consistently and powerfully modulated by certain types of waking experiences. On the basis of this evidence, I put forward the hypothesis that the biological function of dreaming is to simulate threatening events, and to rehearse threat perception and threat avoidance. To evaluate this hypothesis, we need to consider the original evolutionary context of dreaming and the possible traces it has left in the dream content of the present human population. In the ancestral environment human life was short and full of threats. Any behavioral advantage in dealing with highly dangerous events would have increased the probability of reproductive success. A dream-production mechanism that tends to select threatening waking events and simulate them over and over again in various combinations would have been valuable for the development and maintenance of threat-avoidance skills. Empirical evidence from normative dream content, children's dreams, recurrent dreams, nightmares, post traumatic dreams, and the dreams of hunter-gatherers indicates that our dream-production mechanisms are in fact specialized in the simulation of threatening events, and thus provides support to the threat simulation hypothesis of the function of dreaming.

Distributed representations can be distributed in very many ways. The specific choice of representation for a specific model is based on considerations unique to the area of study. General statements about the effectiveness of distributed models are therefore of little value. The popularity of these models is discussed, particularly with respect to reporting conventions.

I respond to Norris et al.'s criticism of Gaskell and Marslen- Wilson (1997). When the latter's network is tested in circumstances comparable to the Merge simulations in the target article, it produces the desired pattern of results. In another area of potential feedback in spoken word processing, aspects of lexical content influence word recognition and our network provides a simple explanation of why such effects emerge. It is unclear how such effects would be accommodated by Merge.

Recent work on functional brain architecture during dreaming provides invaluable clues for an understanding of dreaming, but identifying active brain regions during dreaming, together with their waking cognitive and cognitive functions, informs a model that accounts for only the grossest characteristics of dreaming. Improved dreaming models require cross discipline apprehension of what it is we want dreaming models to “explain.”

[Hobson et al.; Neilsen; Revonsuo; Solms]

Data from published case and group studies bear on the trace deletion hypothesis. The deficit-lesion correlational literature does not support Grodzinsky's claim that lesions in and around Broca's area inevitably lead to comprehension deficits specifically related to coindexation of traces or his claim that other lesions spare this function.