Heyes is right that behavioural tests able to distinguish mentalistic from nonmentalistic alternatives should be sought, but the theoretical issue is less about the passing of behavioural tests than it is about the internal mechanisms which allow the passing of the tests. It may be helpful to try to assess the internal mechanisms directly by measuring brain activities.

I focus on the logic of the goggles experiment, which if it as watertight as Heyes argues, should clearly support ape theory of mind if positive, and clearly reject it if negative. This is not the case, since the experiment tests for only one kind of mindreading, “experience projection”: but it is an excellent test for this, given adequate controls.

Heyes discounts findings of imitation and self recognition in nonhuman primates based on flimsy speculation and then indicates that even positive findings would not provide evidence of theory of mind. Her proposed experiment is unlikely to work, however, because, even if the animals have a theory of mind, a number of assumptions, not directly related to theory of mind, must be made about their reasoning ability.

The target article argued that there is currently no reliable evidence of theory of mind in nonhuman primates and proposed research methods for future use in this field. Some commentators judged the research proposals to be too chauvinist (in danger of falsely denying that primates attribute mental states), but a majority judged them to be too liberal (in danger of falsely affirming theory of mind in primates). The most valuable comments from both camps exemplified “experimental thought,” the obverse of “thought experiments,” and recommended specific alterations and alternatives to the studies I proposed. This Response evaluates these recommendations and presents a revised version of the proposals that appear in the target article. Other valuable commentary cast doubt on the assumption that people have a theory of mind, aired the possibility that language may be a prerequisite for either possession or detection of a theory of mind, questioned the notion of critical experiments, and emphasized the distinction between attribution of sight and belief. In addition to commenting on these issues, I respond to objections to my interpretation of existing research on self-recognition, imitation, and deception.

We argue that probability effects on P300 amplitude are the product of an automatic frequency detector not subject to voluntary control and relatively inaccessible to consciousness. “Expectancies” related to P300 therefore appear to be passive, perceptual ones. If probability-based expectancies do become conscious, they are inversely related to P300, supporting the view of Donchin & Coles (1988).

In this response, I will first summarize my 1988 argument on the issue of expectancy. Second, I will re-evaluate that argument in the light of the findings made by Sommer et al. and by others. Third, I will describe some other problems with the 1988 views. Fourth, I will outline the direction in which I feel that a hypothesis on the P3 complex should be pursued further.

We concur with Sommer et al. that the processing manifested by P300 is not necessarily conscious. We also note that Verleger is yet again adducing evidence that contradicts an assertion we never made.

This commentary makes a case for a connection between the hierarchically organized skills emphasized in Greenfield's (1991t) target article and rhythmic skills utilized in music. It also links hierarchical organization with automated processing. Implicit is the notion that lower levels of a hierarchy become automatic, as they go under control of higher levels of organization.

Greenfield (1991t) fails in an attempt to defend her own original synthesis of cognitivist and nativist accounts of language development. The proposed synchronous stages of object and phoneme combination are not supported by the empirical data she presents. The functional specification of hypothetical neural circuits is almost entirely speculative. Nor is it likely that new data could save her model, since it is formulated in a simplistic information processing framework that is now of little more than historical interest.

The target article presented a model to stimulate further research and ultimately, a more definitive theory of the ontogeny and phylogeny of hierarchically organized sequential activity. Methodologically, it was intended to stimulate methods for integrating data from different neuropsychological techniques. This response to Givon and Swann focuses on several substantive areas: (1) the role of automaticity in hierarchically organized activity and its neural substrate, (2) the neural ontogeny of planning, (3) cognitive and neural architecture for language functions, and (4) the role of environmental input and interaction in the ontogeny and phylogeny of language, tools, and brain.

The basic issue is the measurement of memory accuracy and how correspondence should be assessed. Problems of completeness and variations in the degree and type of inaccuracy make measurement intractably difficult. This difficulty places severe limitations on generalizing across tasks, domains, and individuals and, consequently, on theory construction.

Koriat & Goldsmith (1996t) present two different programs for memory research. Different though they are on the methodological level, on the substantive level the two programs are based on the same view, according to which memory consists of represented information that is permanently stored in the mind (or brain). This view is, I think, wrong. One can support the methodological pluralism Koriat & Goldsmith advocate, but on the substantive level pluralism is not admissible.

In response to Cohen, we point out that many of the assessment difficulties raised by the correspondence metaphor stem from the assessment of memory in meaningful, real-life contexts rather than from the assessment of memory accuracy per se; these difficulties are equally troublesome for the assessment of memory quantity in such contexts. Moreover, the need to focus on particular aspects of memory performance – correspondence-oriented or quantity-oriented – does not preclude the development of useful and general theoretical models. In response to Shanon, we argue that (1) the distinction between the correspondence and storehouse metaphors of memory is metatheoretical, not substantive or methodological, (2) the correspondence metaphor is compatible with both a “representationalist” view of memory and a more “direct” view, and (3) as an epistemological strategy, metaphorical pluralism is both acceptable and desirable.