A wealth of evidence supports the idea that the circuitry of the developing nervous system is gradually elaborated under the instructive influence of neural activity. Some care must be taken, however, not to handicap the acceptance of this useful perspective with too onerous a theoretical superstructure.

We argue that on logical grounds the constructivist algorithms mentioned by Quartz & Sejnowski (Q&S) do not resolve the learning paradox. In contrast, a neural network might acquire a more powerful structure by means of phase transitions. The latter kind of developmental mechanism can be in agreement with the constructivist manifesto.

Quartz & Sejnowski (Q&S) find flaws in standard theories of neural selection, which they propose to repair by introducing Lamarckian mechanisms for anatomical refinement that are analogous to directed mutation in evolution. The reversal of cause and effect that these mechanisms require is no more plausible in an explanation of cognition than it is in an explanation of evolution.

We suggest that neither selectionism nor constructivism alone are responsible for learning-based changes in the brain. On the basis of quantitative structural studies of human brain tissue it has been possible to find evidence of both increase and decrease in tissue mass at synaptic and dendritic levels. It would appear that both processes are involved in the course of learning-dependent changes.

Is (some) innate cognitive modularity consistent with a lack of innate neural modularity? Quartz & Sejnowski's (Q&S's) implicit negative answer to his question fuels their antinativist and antimodular cognitive conclusions. I attempt here to suggest a positive answer and to solicit discussion of this crucial issue.

Activity-dependent processes play an active role in shaping the structure of neuronal circuitry and therefore contribute to neural and cognitive development. Neural constructivism claims to be able to account for increases in the complexity of cognitive representations in terms of directed growth of neurons. This claim is overstated, rests on biased nterpretations of the evidence, and is based on serious misapprehensions of the nature of somatic variation and selection.

The strong correlation between the geometry of the dendritic tree and the specific function of motoneurons suggests that their synaptic contacts are established on a selective stochastic basis with the characteristic form of dendrites being the source of selection in the frog. A compromise is suggested according to which specific structures may have evolved on a selective stochastic basis and “constructive learning” could be the source of selection in the cortex.

As the commentaries reveal, cognitive neuroscience's first steps toward a theory of development are marked by vigorous debate, ranging from basic points of definition to the fine details of mechanism. In this Response, we present the neural constructivist position on this broad spectrum of issues, from basic questions such as what sets constructivism apart from other theories (particularly selectionism) to its relation to behavioral theories and to its underlying mechanisms. We conclude that the real value of global theories at this stage of cognitive neuroscience is not just their answers but the new set of research questions they pose.

Despite close scrutiny in recent years, the traditional properties of LTP are holding up remarkably well, and they remain a credible influence on the belief that LTP has something to do with learning and information storage.

Long-term potentiation (LTP) is operationally defined as a long-lasting increase in synaptic efficacy following high-frequency stimulation of afferent fibers. Since the first full description of the phenomenon in 1973, exploration of the mechanisms underlying LTP induction has been one of the most active areas of research in neuroscience. Of principal interest to those who study LTP, particularly in the mammalian hippocampus, is its presumed role in the establishment of stable memories, a role consistent with “Hebbian” descriptions of memory formation. Other characteristics of LTP, including its rapid induction, persistence, and correlation with natural brain rhythms, provide circumstantial support for this connection to memory storage. Nonetheless, there is little empirical evidence that directly links LTP to the storage of memories. In this target article we review a range of cellular and behavioral characteristics of LTP and evaluate whether they are consistent with the purported role of hippocampal LTP in memory formation. We suggest that much of the present focus on LTP reflects a preconception that LTP is a learning mechanism, although the empirical evidence often suggests that LTP is unsuitable for such a role. As an alternative to serving as a memory storage device, we propose that LTP may serve as a neural equivalent to an arousal or attention device in the brain. Accordingly, LTP may increase in a nonspecific way the effective salience of discrete external stimuli and may thereby facilitate the induction of memories at distant synapses. Other hypotheses regarding the functional utility of this intensely studied mechanism are conceivable; the intent of this target article is not to promote a single hypothesis but rather to stimulate discussion about the neural mechanisms underlying memory storage and to appraise whether LTP can be considered a viable candidate for such a mechanism.

This commentary argues that (1) arousal is not sufficient to induce LTP in the hippocampus, (2) learning can profoundly modulate synaptic plasticity in a state-dependent manner without affecting baseline synaptic efficacy, and (3) unilateral, synapse-specific LTP induction triggers an interhippocampal communication manifested as bilateral increases in gene expression at multiple sites in the hippocampal network.

Shors and Matzel's evaluation of approaches to the behavioural function of LTP is welcome, and should encourage a widening of conceptual approaches to this problem. In addition to their call for increased sophistication in thinking about LTP, there needs to be a parallel increase in sophistication in the study of behaviour. Changes in emotional state or tone may be a better function for LTP than attention mechanisms.

Since learning emerges from a circuit mediating behavior it is unrealistic to require LTP to be a sufficient explanation of it; however, LTP is a necessary component of some learning circuits. The properties of that component bear a closer formal relationship to the acquisition of associative strength than to the modulation of attention.

Gene targeting has generated a great deal of data on the molecular mechanisms of long-term potentiation and its potential role in learning and memory. However, the interpretation of some results has been questioned. Compensatory mechanisms and the contribution of genetic background may make it difficult to unequivocally prove the existence of a causal (genetic) link between LTP and learning.

Several types of amygdala-dependent learning can be blocked by local infusion of NMDA antagonists into the amygdala. This blockade shows anatomical, pharmacological, temporal, and behavioral specificity, providing a pattern of data more consistent with a role for NMDA receptors in learning than in arousal or attention, and supporting the contention that an “LTP-like” process is a neural substrate for memory formation.

Examples of how LTP and LTD can control adaptively-timed learning that modulates attention and motor control are given. It is also suggested that LTP/LTD can play a role in storing memories. The distinction between match-based and mismatch-based learning may help to clarify the difference.

LTP is thought to be an experimental model for studying the cellular mechanism of learning and memory. Shors & Matzel review some contradictory data concerning the linkage between LTP and memory and suggest that LTP does not underlie learning and memory. LTP is a cellular and synaptic process and cannot be a memory mechanism. In fact, it is a cellular information storage mechanism.

The hypothesis that there is a 1:1 correspondence between LTP and learning is simplistic, and the correlation approach to testing it is therefore too limited. The alternative hypothesis that LTP plays a role in arousal is consistent with activity-dependent neuromodulation, but ignores the Hebbian properties of LTP. LTP may involve both types of mechanisms, suggesting a possible synthesis of the two hypotheses.

The problem of neural memory storage is discussed, based on the results of studies of memory impairment after hippocampal lesions, motor learning, and electrophysiological research on “spinal memory.” I support Shors & Matzel's major statements. The absence of reliable evidence on the LTP memory storage function and other data cast doubt on the synaptic theory of memory.

Shors & Matzel provide compelling arguments against a role for hippocampal long-term potentiation (LTP) in mammalian learning and memory. As an alternative, they suggest that LTP is an arousal mechanism. I will argue that this view is not a satisfactory alternative to current conceptions of LTP function.