Perhaps one of the most important reasons for understanding the structural information of these networks is studying plasticity and its role in information processing and storage (memory). Of course the best method to validate Granger causality would be to compare the structural information derived from our analysis with that of the
actual structure of the network on the MEA. However, given the complexity of these networks it is extremely difficult to gather this information across the thousands of neurons and tens of thousands of synapses. One method to roughly estimate the network structure is to systematically probe the network at the 60 different stimulation sites provided by the MEA and measure the response. Each stimulation typically produces a burst of activity across the entire network lasting for hundreds of milliseconds. The early phase of this response (< ~20 ms) is thought to represent the direct activation of pathways and their postsynaptic targets near the stimulating electrode (Jimbo et al., 2000) since action potentials during this period are typically precisely timed with very small jitter. In contrast, the late phase (> 20 ms) of the response is characterized by widely fluctuating spike timings and likely represents activity spreading and reverberating throughout the network. Hence, the first 20 milliseconds of the response might provide at least a crude estimate of the connectivity at least from the 60 locations stimulated using the MEA.
Following this reasoning, Jimbo and others (Jimbo, 1999; Tateno et al., 1999) have shown that by probing the network before and after a tetanization procedure in which one channel was stimulated rapidly to induce changes in plasticity (i.e., synaptic weights) across the network. Differences between the response at each site before and after tetanization would then provide a rough estimate of which of the synaptic connections evoked from that site had been modified.
Figure13_Jimbo_plasticity_ourlab_exampleThe figure shows an example of the plasticity produced by this procedure in my laboratory. The horizontal axis represents the location of the stimulation probe and the vertical axis is the response across channels to that probe before and after tetanization. The magnitude of the change is color coded with red to blue representing increased to decreased number of action potentials, respectively. Tetanization of one of the 60 channels resulted in both increases and decreases in the number of action potentials produced across the MEA. Further, each stimulation site showed an overall increase, decrease, or no effect across channels. For example, the stimulation site of channel 40 showed an overall increase in activity on most channels (vertical axis), while stimulation site 10 showed consistent decreases. This would be true if stimulation of each site evoked only localized activity in neurons or pathways near that channel. In contrast, the direction of those changes were dependant on which channel was being probed. Neurons near channel 45 (horizontal) appear to be enhanced when measured by probes of channel 40 but were depressed during probes of channel 10. This result would make sense if it were the pathways (reflected in the vertical axis) and their associated synapses that were being modified and not the individual neurons (horizontal axis). In other words, the activity of each neuron can be both enhanced and depressed, dependent on which particular pathway is stimulated.

Jimbo, Y., Tateno, T., and Robinson, H. P. C. (1999).
Simultaneous Induction of Pathway-Specific Potentiation and Depression in Networks of Cortical Neurons. Biophysical Journal, 76, 670-678.
Tateno, T., & Jimbo, Y. (1999).
Activity-dependent enhancement in the reliability of correlated spike timings in cultured cortical neurons. Biol Cybern, 80(1), 45-55.