32 Moreover, recently it has been shown that the excitation–inhibition balance strongly modulates the magnitude of these trial-by-trial variations

(N. Haroush, personal communication, 2011). Thus, it seems that there is no “elementary” input-output function of these networks – rather they exhibit unstable patterns with step transitions between modes and long-term correlations in the firing statistics. Figure 3 Latencies to population responses. A: Population post-stimulus time histogram (pPSTH). A total of 52 electrodes in Inhibitors,research,lifescience,medical which spikes were detected in >15% of the stimuli were considered for this analysis. The number of spikes recorded in a time window … MAPPING THE CONCEPT OF LEARNING TO THE NETWORK PREPARATION Once the aim is to study neural mechanisms of learning, it is important to be clear about what exactly one means Inhibitors,research,lifescience,medical by “learning”. Learning can be loosely defined as a process of changing behavior in order to achieve a growing success in any a-priori task within a fixed environment. With this definition in mind, we map the concept of learning to the network preparation: The behavior, we assume, may be represented by temporal structures described in terms of associations between neuronal activities. The network is required to modulate associations between neuronal activities such that it noticeably increases the efficiency with which Inhibitors,research,lifescience,medical an input stimulus is processed and a desirable spatiotemporal firing pattern is reached. The learning

process can be artificially divided into two overlapping phases – one of exploration, that is a search in the space of possible input–output relations, and a second phase of recognition or Inhibitors,research,lifescience,medical consolidation once

the “appropriate” response pattern has been reached. In the past years, there have been many publications regarding different protocols to induce plasticity in these networks33,34 (and references therein). All of these methods are based on the hypothesis that certain patterns of activation by stimulation can induce lasting changes in the network’s functional connectivity or activation pathways. What these studies mainly show is that such changes can indeed be achieved, but there are no Inhibitors,research,lifescience,medical simple “plasticity rules” at the network level, such as those discovered for single synapse in the sense of long-term potentiation (LTP), long-term depression (LTD), or spike-timing-dependent plasticity (STDP). By using measures such as conditional firing probability (CFP)33 or association pairs,35 the changes in the functional connectivity between thousands of neuronal pairs from can be quantified and monitored over time. It seems that stimulation drives changes in connectivity, but the direction and amplitude of change is not easily predicted and varies between different protocols and laboratories.30,34,36 It does seem, however, that the “harder” the stimulation drive, the larger the change. Using these observations, buy INNO-406 Shahaf and Marom a decade ago developed a protocol for achieving learning in these networks.