Although one can endorse the complexity of the data and processes that Phillips & Singer (P&S) review, their mathematical suggestions can be compared critically with cases in nonlinear psychophysics, where the theoretician is faced with analogous problems. Owing to P&S's failure adequately to recognise both the intricate properties of nonlinear dynamics in networks and the constraints of metabolic demands on the temporal generation of patterns in biological nets their conclusions fail to meet the problems they properly address.

Functional roles for cortical synchronization in self-organizing neural networks are described. These properties are best understood by models that link brain to behavior. Synchrony can express itself differently in cortical circuits that perform different behavioral tasks. Cortical temporal properties that seem inexplicable by synchrony are also mentioned.

Synchronizing oscillations may be just one case of integration and/or coding, one which explains associations by concurrence. Understanding the sequencing of neural/behavioral events requires a clock mechanism that imposes structure behind mere associations, and may be best served by dissociating oscillations and synchronization in terms of physiologic and computational mechanisms.

The neural substrates of context effects in word perception are still largely unclear. Interhemispheric priming phenomena in word recognition, typically observed in normal subjects, are absent in commissurotomized patients. This suggests that callosal fibers may provide contextual integration. In addition, certain characteristics of human frontal cortical fields subserving sensorimotor learning, as investigated by positron emission tomography, provide evidence for contextual integration not confined to the visual system. This supports the notion of common aspects of cortical computations in different cerebral areas.

Recent experimental work suggests that the concept of contextual fields should be generalized to allow the modulation of local information extraction by both external and internal context. The external context relates to the coherent information of the stimulus; the internal context refers to the parts of this information which are relevant for behavior. This dual interaction, present at every level of the hierarchy, requires a fundamental unit of processing more complex than a single neuron appears today. We argue that the cortical column supplies the required mechanisms for a separate top-down and bottom-up processing and allows for interaction without destructive cross-talk. This flexibility might explain why the same basic architecture of a cortical column is used in the whole isocortex.

Although Phillips & Singer's proposal of commonalities seems sound, information theory and artificial neural network modeling omit important detail. An example is given of a distributed neural transformation that has been characterized mathematically and found to have both overall commonalities and differences of detail in different regions. P&S's contextual field is compared to inclusive regions in a formalism relevant for modeling bodily-kinaesthetic intelligence.

In field computing a topologic organization of CFs is necessary to support sensorimotor planning. A simple model of cortical dynamics can exploit such topologic organization.

We argue that the separation between CF (contextual field) and RF (receptive field) in relation to the NMDA (N-methyl-D-aspartate) system is empirically questionable and that it is functionally unnecessary. In addition, the proposed suppression of unexpected information will in many cases be counterproductive.

Well-posed questions about information processing may require physiologically based, quantitative models of large scale neocortical dynamic function. “Synchronization” of this dynamics can be viewed in different contexts of the binding problem.

We briefly review the long-standing ideas about the use of synchronicity in the brain, which rely on Donald Hebb's views on cell assemblies and synaptic plasticity. More recently the distinction among several timescales in the description of neural activity has become a focus of theoretical discussion. Phillips & Singer's target article is criticized mainly because it does not distinguish these timescales properly and hence does not really address the questions so intensely debated today.

Phillips & Singer's compelling presentation is weakest in its demonstration of commonalities between sensory plasticity and higher forms of learning and behavior. We propose that available data on schizophrenia can provide such evidence, because of the presence of impairments in a number of functions central to their model, and strong relationships between these dysfunctions and behavior.

This commentary describes some extensions of the Phillips & Singer model of contextual interactions to cater for the contrast- and extent-dependency of surround facilitation and suppression. I also comment briefly on some semantic problems with what exactly the role of synchronization might be: input preprocessing or output postprocessing?

Simple hypotheses are intrinsically attractive, and, for this reason, need to be formulated with utmost precision if they are to be testable. Unfortunately, it is hard to see how Phillips & Singer's hypothesis might be unambiguously refuted. Despite this, the authors have provided much evidence consistent with the hypothesis, and have proposed a natural and powerful extension for information theoretic approaches to learning.

A series of intriguing experiments has emphasized the need to understand synchronization effects. Information theory is proving to be an important tool for understanding the computations performed by cortical circuits. Nevertheless, Phillips & Singer's target article fails to provide a convincing correlation between synchronous firing and information transmission, which could well prove to be anticorrelated.

The distinction between receptive field and conceptual field is appealing and heuristically useful. Conceptually, it is more satisfactory to distinguish between information from the environment and from the brain. We emphasize here a selectionist view that considers information transmission within the brain as modulated by a stimulus, rather than information transmission from a stimulus as modulated by the context.

For scene analysis, it is important to ask the question of how synchronized population codes – the basic representation employed in the target article – are generated. Recent computational advances have resolved the critical challenges of rapid synchronization with local coupling and rapid desynchronization. The synchronized codes make real contributions to tackling the problem of scene analysis.

Simulations and analytic considerations show that synchronisation occurs in delay neural networks at the surrounds of externally driven sites. In the synchronous fields, network transmission has a static transfer function and H(X|R,C) is minimal. But when autonomous local states with attractor dynamics develop in the network, H(X|R,C) is not minimized. Physiological realism may therefore require some modifications in application of coherent infomax.

The usefulness of contextually guided processors is investigated a little further. A more general use for binding V1 cell responses than the one in Phillips & Singer's target article is proposed, which takes into account that strong responses of these cells can mean more than the presence of lines and edges. The possibility of different grouping depending on the activities of neighboring cells is essential to the approach.

The additional data, perspectives, questions, and criticisms contributed by the commentaries strengthen our view that local cortical processors coordinate their activity with the context in which it occurs using contextual fields and synchronized population codes. We therefore predict that whereas the specialization of function has been the keynote of this century the coordination of function will be the keynote of the next.

To describe phenomena that occur at different time scales, computational models of the brain must incorporate different levels of abstraction. At time scales of approximately 1/3 of a second, orienting movements of the body play a crucial role in cognition and form a useful computational level – more abstract than that used to capture natural phenomena but less abstract than what is traditionally used to study high-level cognitive processes such as reasoning. At this “embodiment level,” the constraints of the physical system determine the nature of cognitive operations. The key synergy is that at time scales of about 1/3 of a second, the natural sequentiality of body movements can be matched to the natural computational economies of sequential decision systems through a system of implicit reference called deictic in which pointing movements are used to bind objects in the world to cognitive programs. This target article focuses on how deictic bindings make it possible to perform natural tasks. Deictic computation provides a mechanism for representing the essential features that link external sensory data with internal cognitive programs and motor actions. One of the central features of cognition, working memory, can be related to moment-by-moment dispositions of body features such as eye movements and hand movements.