Contra Shepard we argue, first, that his presentation of a three-dimensional representational (psychological or phenomenal) colour space is at odds with many results in colour science, and, second, that there is insufficient evidence for Shepard's stronger claim that the three-dimensionality of colour perception has resulted from natural selection, moulded by the particulars of the solar spectrum and its variations. [Shepard]

In conducting experiments, economists take more care than psychologists in specifying characteristics of people and tasks and are clearer about the conditions to which results can be generalized. Psychologists could learn much from this practice. They should also think in terms of programs of research – involving laboratory and field studies – that will allow them to test the generality of their theories.

We criticise Shepard's notions of “invariance” and “universality,” and the incorporation of Shepard's work on inference into the general framework of his paper. We then criticise Tenenbaum and Griffiths' account of Shepard (1987b), including the attributed likelihood function, and the assumption of “weak sampling.” Finally, we endorse Barlow's suggestion that minimum message length (MML) theory has useful things to say about the Bayesian inference problems discussed by Shepard and Tenenbaum and Griffiths. [Barlow; Shepard; Tenenbaum & Griffiths]

The mathematical approach to such essentially biological phenomena as perseverative reaching is most welcome. To extend these results and make them more accurate, levels of analysis and neural centers should he distinguished. The navigational nature of sensorimotor control should be characterized more clearly, including the continuous dynamics of neural processes hut not limited to it. In particular, discrete conditions should be formalized mathematically as part of the biological process.

Models of biological and political systems share in common an irreconcilable tension. They must extract the essential features of the system to make its workings comprehensible to the human observer, and yet the omitted underbrush is essential to the workings of the system. A good model accommodates both the workings of the system and the cognitive makeup of the observer.

Stoffregen & Bardy's arguments against separation of the senses fail to consider the functional differences between the kinds of information potentially available in the structured energy arrays that correspond to the traditional senses. Since most perception/action research pursues a strategy of information perturbation presupposing differential contributions from the various ambient arrays, the global array hypothesis can only be extended and tested by analyses that consider the functional aspects along which the senses can, in fact, be separated.

After critical appraisal of mathematical and biological characteristics of the model, we discuss how a classical hippocampal neural network expresses functions similar to those of the chaotic model, and then present an alternative stimulus-driven chaotic random recurrent neural network (RRNN) that learns patterns as well as sequences, and controls the navigation of a mobile robot.

The use of high hypothetical payoffs has been justified by the realism and relevance of large monetary consequences and by the impracticality of making high cash payments. We argue that subjects may not be able to imagine how they would behave in high payoff situations.

Thelen and her colleagues' model overemphasizes the role of action in cognitive development. Recent research has shown that infants do not have to be trained to reach for a hidden object. By 7.5 months of age, infants can recall the location of a hidden object with no practice trials. Thelen at al.'s goal to design a parsimonious account of A-not-B behaviors was successful, but at the expense of focusing primarily on implicit and ignoring explicit memory.

There is a discrepancy between Stoffregen & Bardy's concept with experimental work on human self-motion perception. We suggest an alternative: (1) higher brain centers are informed by a given sensory cue in a direct and rapid way (direct perception), and (2) this information is then used to prime and shape a more complex mechanism that usually involves several cues and processing steps (inferential).

Shepard's concept of internalization does not suggest mechanisms which help to understand how the brain adapts to changes, how representations of a steadily changing environment are updated or, in short, how brain learning continues throughout life. Neural mechanisms, as suggested by Barlow, may prove a more powerful alternative. Brain theories such as Adaptive Resonance Theory (ART) propose mechanisms to explain how representational activities may be linked in space and time. Some predictions of ART are confirmed by psychophysical and neurophysiological data. [Barlow; Shepard]

Tsuda's article suggests several plausible concepts of neurodynamic representation and processing, with a thoughtful discussion of their neurobiological grounding and formal properties. However, Tsuda's theory leads to a holistic view of brain functions and to the controversial conclusion that the “binding problem” is a pseudo-problem. By contrast, we stress the role of chaotic patterns in solving the binding problem, in terms of flexible temporal coding of visual scenes through graded and intermittent synchrony.

Animals detect and acquire resources through a sequence of shape changes. This process is tightly coupled to the sensory and mechanical ecology of the animal. Building physical models allow us to prescind from modeling these aspects of the environment, which may not yet be described or suitably abstracted. The significance of this hybrid of physical modeling and experimentation to the acquisition of scientific knowledge is discussed.

The practices of economists increase experimental reproducibility relative to those of selected psychologists but should not be universally adopted. Procedures criticized by Hertwig and Ortmann as producing variable data are valuable, instead, for generating questions. The procedure of choice should depend on the theoretical goal: measure a known factor or learn what factors are important and need to be measured.

The procedures used to ensure reliable occurrences of the A-not-B error distort and miss essential features of Piaget's original observations. A model that meshes a mental event, highly restricted by testing procedures, to the dynamics of bodily movement is of limited value. To embody more than just perseverative reaching, the formal model must incorporate Piaget's essential features.

I question whether chaotic itinerancy is anything new or different to existing research on heteroclinic cycles (cycling-chaos), and blow-out bifurcations (attractor-bubbling) that provide more detailed and better definition for nonlinear phenomena occurring in neural systems. I give a brief description of this research for comparison and expansion, and see it as an important component in dynamical models of neural activity.

A view is put forward, according to which various aspects of the structure of the world as internalized by the brain take the form of “neural spaces,” a concrete counterpart for Shepard's “abstract” ones. Neural spaces may help us understand better both the representational substrate of cognition and the processes that operate on it. [Shepard]

Decision making can be studied using hypothetical payoffs because it is hypothetical to its very core. However, the core process can be influenced by contextual features. As there is no theory for these contextual features, a “do-it-both-ways” rule amounts to a waste of money. If we had such a theory, doing it both ways would be unnecessary.

Before one can talk about global arrays and multimodal detection, one must be clear about the concept of information: How is it different from energy and how is it detected? And can it come to specify a needed movement? We consider these issues in our commentary.

Although robots can contribute to the understanding of biological behavior, they fail to model the processes by which humans cope with their environment. Both development and learning are characterized by complex relationships that require constant modification. Given present technology, robots can only model behaviors in specific situations and during discrete stages. Robots cannot master the complex relationships that are the hallmark of human behavior.