The Newell Test as it is proposed by Anderson & Lebiere (A&L) has the disadvantage of being too positivistic, stressing areas a theory should cover, instead of attempting to exclude false predictions. Nevertheless, Newell's list can be used as the basis for a more stringent test with a stress on the falsifiability of the theory.

Cognitive architectures, like programming languages, make commitments only at the implementation level and have limited explanatory power. Their universality implies that it is hard, if not impossible, to justify them in detail from finite quantities of data. It is more fruitful to focus on particular tasks such as language understanding and propose testable theories at the computational and algorithmic levels.

Modelling human reasoning characterizes the fundamental human cognitive capacity to describe our past experience and use it to form expectations as well as plan and direct our future actions. Natural language semantics analyzes dynamic forms of reasoning in which the real-time order determines the temporal relations between the described events, when reported with telic simple past-tense clauses. It provides models of human reasoning that could supplement ACT-R models.

The Newell Test is an important step in advancing our understanding of cognition. One critical constraint is missing from this test: A cognitive architecture must be self-contained. ACT-R and connectionism fail on this account. I present an alternative proposal, called Distributed Adaptive Control (DAC), and expose it to the Newell Test with the goal of achieving a clearer specification of the different constraints and their relationships, as proposed by Anderson & Lebiere (A&L).

Although we agree with Newell and Anderson & Lebiere (A&L) that a unified theory of cognition is needed to advance cognitive science, we disagree on how to achieve it. A hybrid system can score high in the Newell Test but may not offer a veridical and coherent theory of cognition. A multilevel approach, involving theories at both psychological and brain levels, is suggested.

We draw an analogy between Hilbert's program (HP) for mathematics and Newell's program (NP) for cognitive modeling. The analogy reveals that NP, like HP before it, is fundamentally flawed. The only alternative is a program anchored by an admission that cognition is more than computation.

The criterion of computational universality for an architecture should be replaced by the notion of compliancy, where a model built within an architecture is compliant to the extent that the model allows the architecture to determine the processing. The test should be that the architecture does easily – that is, enables a compliant model to do – what people do easily.

The commentaries on our article encourage us to believe that researchers are beginning to take seriously the goal of achieving the broad adequacy that Newell aspired to. The commentators offer useful elaborations to the criteria we suggested for the Newell Test. We agree with many of the commentators that classical connectionism is too restrictive to achieve this broad adequacy, and that other connectionist approaches are not so limited and can deal with the symbolic components of thought. All these approaches, including ACT-R, need to accept the idea that progress in science is a matter of better approximating these goals, and it is premature to be making judgments of true or false.

An action-oriented theory of embodied memory is favorable for many reasons, but it will not provide a quick yet clean solution to the grounding problem in the way Glenberg (1997t) envisages. Although structural mapping via analogical representations may be an adequate mechanism of cognitive representation, it will not suffice to explain representation as such.

This commentary examines Glenberg's characterization of “suppression” in light of negative priming and related phenomena. After offering a radically different slant on suppression, an attempt is made to weave this alternative version into Glenberg's provocative discussion of embodied memories.

Arthur M. Glenberg omits discussion of motivation and this leads him to an underestimation of the part played by pleasure and pain and desire and fear in both the clamping and the updating of percepts. This commentary aims at rectifying this omission, showing that mutual correction plays an important role.

Standard models of cognition are built from abstract, amodal, arbitrary symbols, and the meanings of those symbols are given solely by their interrelations. The target article (Glenberg 1997t) argues that these models must be inadequate because meaning cannot arise from relations among abstract symbols. For cognitive representations to be meaningful they must, at the least, be grounded; but abstract symbols are difficult, if not impossible, to ground. As an alternative, the target article developed a framework in which representations are grounded in perception and action, and hence are embodied. Recent work (Glenberg & Robertson 1999; 2000; Glenberg & Kaschak 2002; Kaschak & Glenberg 2000) extends this framework to language.

Data from magnetic resonance imaging (MRI), autopsy, endocranial measurements, and other techniques show that: (1) brain size correlates 0.40 with cognitive ability; (2) average brain size varies by race; and (3) average cognitive ability varies by race. These results are as replicable as one will find in the social and behavioral sciences. They pose serious problems for Rose's claim that reductionistic science is inadequate, inefficient, and/or unproductive.