Although conditional stimulus (CS)/unconditional stimulus (US) contingency awareness appears to be necessary for human Pavlovian autonomie conditioning, only a selective review of the literature and the forgetting of certain basic, brute facts can allow the cognitive conclusion that awareness causes, or even is important for, conditioning. That conclusion is theoretically barren for explaining the phenomenon and is also of little potential practical use.

We outline fresh findings that show that our macroscopic electrocorticographic (ECoG) simulations can account for synchronous multiunit pulse oscillations at separate, simultaneously activated cortical sites and the associated gamma-band ECoG activity. We clarify our views on the approximations of dynamic class applicable to neural events at macroscopic and microscopic scales, and the analogies drawn to classes of ANN behaviour. We accept the need to introduce memory processes and detailed anatomical and physiological information into any future developments of our simulations. On the issue of intrinsic cortical stability and the role of extrinsic fibre systems in maintaining stability, we argue that this position is not in extreme contradiction to those of our commentators, and that the mechanisms implicit in our simulations' properties imply rich computational possibilities. We discuss some of the reasons for and against the existence of significant global resonances in the brain and explain why such behaviour appears absent in our simulations. Last, we discuss other phenomena, such as rhythmic driving of the cortex, which have not yet been introduced into our models, and indicate lines for future development of the simulations.

Immediate imitation is likely to be a major, direct input to Barresi & Moore's level 2 competence, but deferred imitation is unlikely to play a key role in the transition to level 3, because (1) the attribution of first person knowledge is neither a necessary cause nor an obvious consequence of deferred imitation, and (2) deferred imitation does not correlate phylogenetically with capacities that more plausibly either yield or reflect a concept of intentional agency.

Epstein et al. attribute second language learning to the forces of transfer and language universals. They show that transfer is minimally involved in certain types of learning and therefore conclude that universals are involved. However, they forget to consider the important role of learning in second language acquisition.

If Epstein et al. are right that adult second language learners have full access to UG, then all of the following should be true: adults should be able to consciously transform their I-Language; adults should be able to transform pidgins into Creoles; adults should be as likely as children to restructure their grammars on the basis of “functional” pressure. All the foregoing are false, however, which seriously calls into question the correctness of their hypothesis.

Shanks and St. John (1994a) suggest that “human learning is almost invariably accompanied by conscious awareness (of what is being learned).” From the viewpoint of a computer scientist who tries to construct learning systems, that claim seems rather implausible. In this commentary I wish to suggest why, in the hopes of shedding light on the relationship between consciousness and learning.

Barons's target article approaches errors in decision-making by defining three kinds of models: normative, descriptive, and prescriptive. Baron's prescriptive model is at the center of this commentary. From a theoretical perspective, is Baron's prescriptive model a set of rules through which one can move from the descriptive to the normative? Or is it a practical goal one can achieve as opposed to a normative unachievable theoretical ideal? Delineating an efficient prescriptive account for decision making necessarily depends on a very specific normative model. However, Baron's normative consequentialism model is too general to enable adequate prescriptive accounts.

Barresi & Moore provide a welcome focus on children's abilities to integrate first and third person information about intentional relations but they pay insufficient attention to the origins of children's understanding of the nature of subjective orientations vis-à-vis a shared world and the potential significance of such understanding as a source (rather than an outcome) of domain-general information-processing capacities.

Epstein et al.'s theory of SLA is heavily dependent on assumptions about both the nature of the acquisition device and the grammar that it produces. This commentary briefly explores the consequences of an alternative set of assumptions, focusing on the possibility that the acquisition device does not include UG and that syntactic representations do not contain functional projections.

The “intentional schema” seems identical to or dependent upon kinesthetic–visual matching, both of which account for similar empirical findings. The intentional schema, however, fails to account for variability in children's understanding of false belief and differences in children's understanding of self and other in pretense.

In the original target article (Shanks & St. John 1994), one of our principal conclusions was that there is almost no evidence that learning can occur outside awareness. The continuing commentaries raise some interesting questions, especially about the definition of learning, but do not lead us to abandon our conclusion.

Baron has provided some examples of nonconsequentialism in decision making and describes them as biases; these may be the remnants of the biological origin of decision making. One may argue that decisions are made on the basis not of rationality but affective processes. Behavior follows the trend toward maximizing pleasure. This mechanism might explain apparent nonconsequentialism.

The intentional schema may not be sufficiently characterized to make questions about its role in individual and species development intelligible. The idea of metarepresentation may perhaps give it enough content. The importance of metarepresentation itself, however, can be called into question.

The range of hypotheses considered is surprising in that the most arguably plausible one is not included: the invariant principles of language are available for life, while the parameters of variation cannot be set after puberty. This hypothesis provides a better explanation than the author's for both the deep similarities and the vast differences between child “language growth” and adult language acquisition.

Baron assumes that ethical decision-making can be evaluated without specifying more general features of the cognitive system within which it occurs. It is suggested that ethical principles are heuristics employed during goal-oriented action planning. Heuristics are bound to generate suboptimal decisions in some cases. It is rational to replace a particular heuristic only when the cost of associated error exceeds the cost of constructing and installing a more successful alternative.

A formal language to implement a computational-level theory of memory cannot be based on standard (Zermelo-Fraenkel) set theory. For correct models of a memory database to exist, non-well-founded sets must be permitted. This is illustrated by modifying an example from Barwise & Moss (1991).

Curiously, two central areas are unaddressed by Epstein et al.: (i) L1A–L2A differences; (ii) L2 development. Here, findings relevant to (i) and (ii) – as well as their significance – are discussed. Together these form the basis for contesting Epstein et al.'s “Full Access” approach, but nonetheless analyses of the L2 data argue for UG-constrained L2A. Also discussed is the inadequacy of accounts (like Epstein et al.'s) without an explicit and prominent role for the L1.

Near the end of his target article, Baron argues that we need to address the question of how to conduct education in consequentialist decision making. However, recent trends in education have deemphasized and denigrated decentered and decontextualized thought. It is argued here that perspective decentering and decontextualized thinking are absolutely essential to the development of consequentialist reasoning.

Colonius suggests that, in using standard set theory as the language in which to express our computational-level theory of human memory, we would need to violate the axiom of foundation in order to express meaningful memory bindings in which a context is identical to an item in the list. We circumvent Colonius's objection by allowing that a list item may serve as a label for a context without being identical to that context. This debate serves to highlight the value of specifying memory operations in set theoretic notation, as it would have been difficult if not impossible to formulate such an objection at the algorithmic level.

The intentionality schema is an abstraction that relates phylogenetic and ontogenetic sequences of social understanding, but it also obscures the differences between humans and other primates. In particular, it ignores human social developmental and communicative history and the important roles that language plays in human understanding of others' intentional states.