Page's target article makes a good case for the strength of localist models. This can be characterized as an issue of where new information is integrated with respect to existing knowledge structures. We extend the analysis by discussing the dimension of when this integration takes place, the implications, and how they guide us in the creation of cognitive models.

A comparison of Merge, a model of comprehension, and WEAVER, a model of production, raises five issues: (1) merging models of comprehension and production necessarily creates feedback; (2) neither model is a comprehensive account of word processing; (3) the models are incomplete in different ways; (4) the models differ in their handling of competition; (5) as opposed to WEAVER, Merge is a model of metalinguistic behavior.

Threat themes are clearly over-represented in dreams. Threat is, however, not the only theme with potential evolutionary significance. Even for hypnagogic and hypnopompic hallucinations during sleep paralysis, for which threat themes are far commoner than for ordinary dreaming, consistent non-threat themes have been reported. Revonsuo's simulation hypothesis represents an encouraging initiative to develop an evolutionary functional approach to dream-related experiences but it could be broadened to include evolutionarily relevant themes beyond threat. It is also suggested that Revonsuo's evolutionary re-interpretation of dreams might profitably be compared to arguments for, and models of, evolutionary functions of play.

[Revonsuo]

Rolls's book depends significantly on a definition relating emotion to reward and learning. This definition confuses two separable concepts, and may result in the exclusion of notions of emotion and motivation from lower animals that may possess limited learning capacities. A more useful definition might revolve around the notion that emotions are states that function to optimize the performance of behavior.

Stanovich & West analyze individual differences with respect to response output (e.g., participants' numerical estimates). They do not analyze the underlying cognitive processes that led to the outputs; they thereby probably misclassify some non-normative responses as normative. Using base rate neglect and overconfidence as examples, I demonstrate the advantages of analyzing cognitive processes further.

Page's definition of localism is inspired by the instance-based paradigm. However, the locality of representations is not necessary for a model to be instance-based and, on the other hand, explicit featural representations are generally considered local. The important distinction is between instance-based and noninstance-based paradigms and not between distributed and local representations as Page claims.

There are reasons for reservations with respect to the postulated function of Broca's area. Evidence for the psychological reality of the relevant traces does not exist. In addition, because the syntax of non- (or partly) configurational languages is not described in terms of empty categories, no receptive agrammatism should be observed in these languages. Aphasia should not be examined in isolation from its cognitive components.

Norris, McQueen & Cutler argue that there is no need for feedback in word recognition. Given the accumulating evidence in favor of feedback as a general mechanism in the brain, I will question the utility of a model that is at odds with such a general principle.

Approaches to model evaluation in terms of Occam's razor or principles of parsimony cannot avoid judgements about the relative importance of aspects of the models. Assumptions about “core processing” are usually considered more important than those related to decision components, but when the decision is related to a central feature of the processing, it becomes extremely difficult to tease apart the influences of core and decision components and to draw sensible conclusions about underlying architecture. It is preferable, where possible, to use experimental procedures that avoid the necessity for subject decisions related to critical aspects of the underlying process.

Examples of terror generated by an aircraft disaster, of human courtship behaviour, and of the application of laboratory techniques to the commercial training of animals suggest (1) that emotion is simply the subjective counterpart of (objective) motivation (so that separate brain mechanisms would be an embarrassment) and (2) the apparent involvement of reward and punishment is a consequence of the excessively narrow range of experimental procedures used and has no foundation in the design of the brain.

The data provided by Grodzinsky demonstrating a syntactic comprehension deficit in Broca's patients provide no evidence for the theoretical concepts of movement, trace or “trace deletion.” The comprehension deficit data can be more economically accounted for with traditional grammatical concepts that are less theory-internal and more empirically based.

Information processing theories of memory and skills can be reformulated in terms of how categories are physically and temporally related, a process called conceptual coordination. Dreaming can then be understood as a story-understanding process in which two mechanisms found in everyday comprehension are missing: conceiving sequences (chunking categories in time as a higher-order categorization) and coordinating across modalities (e.g., relating the sound of a word and the image of its meaning). On this basis, we can readily identify isomorphisms between dream phenomenology and neurophysiology, and explain the function of dreaming as facilitating future coordination of sequential, cross-modal categorization (i.e., REM sleep lowers activation thresholds, “unlearning”).

[Hobson et al.; Nielsen; Solms; Revonsuo; Vertes & Eastman]

The finding of a correlation between normative responses to judgment and reasoning questions and cognitive capacity measures (SAT score) suggests that the cause of the non-normative responses is computational in nature. This actually is consistent with the rational competence view. The implications of this finding for the adaptation view of cognition are discussed.

Page argues that localist models can be applied to a number of problems that are difficult for distributed models. However, it is easy to find examples where the opposite is true. This commentary illustrates the superiority of distributed models in the domain of artificial grammar learning, a paradigm widely used to investigate implicit learning.

The representation of events, in primates at any rate, is a separate process from their emotional evaluation. The same holds for cognitive evaluation. Here too representation and evaluation are separate operations. Acknowledging the symmetry leads to the notion of free representation.

Grodzinsky interprets linguistic differences between agrammatic comprehension and production symptoms as supporting the hypothesis that the mechanisms underlying grammatical encoding (sentence formulation) and grammatical decoding (syntactic parsing) are at least partially distinct. This inference is shown to be premature. A range of experimentally established similarities between the encoding and decoding processes is highlighted, testifying to the viability of the hypothesis that receptive and productive syntactic tasks are performed by the same syntactic processor.

The validity of dream recall is discussed. What is the relation between the actual dream and its later reflection? Nielsen proposes differential sleep mentation, which is probably determined by dream accessibility. Solms argues that REM sleep and dreaming are double dissociable states. Dreaming occurs outside REM sleep when cerebral activation is high enough. That various active sleep states correlate with vivid dream reports implies that REM sleep and dreaming are single dissociable states. Vertes & Eastman reject that REM sleep is involved in memory consolidation. Considerable evidence for this was obtained by REM deprivation studies with the dubious water tank technique.

[Nielsen; Solms; Vertes & Eastman]

Examining the implications of a localist model for linguistic performance, I show the strengths of the P-graph, a network of elementary units of meaning where utterance results from relaxation through the operation of a dynamics of affect values. A unit of meaning is stored in a synaptic connection that brings together two words. Such a model, consistent with the anatomy and physiology of the neural tissue, eschews a number of traditional pitfalls of “semantic networks”: (1) ambiguity ceases to be an issue, as similar uses of a word are automatically clustered together; (2) faster retrieval of words acquired early is explained by the larger number of their instances. In addition the P-graph takes advantage of a plausible form of information storage: the local topology of the neural tissue.

The target article focuses exclusively on System 2 and on reasoning rationality: the ability to reach valid conclusions from available information, as in the Wason task. The decision-theoretic concept of coherence rationality requires beliefs to be consistent, even when they are assessed one at a time. Judgment heuristics belong to System 1, and help explain the incoherence of intuitive beliefs.

As a proponent of phoneme-like units in speech perception, I am very sympathetic to Merge's use of phoneme-oriented input. However, in the absence of any known way to provide input in exactly the form assumed, further consideration needs to be given to how the variation in the details of the specification of this input might affect Merge's (and Shortlist's) overall behavior.