Phase sequences in balanced recurrent networks

However, transient synchronization and precise sequential firing has been observed during certain perceptual tasks or behavioral states. The neural mechanism behind these activity patterns has not yet been resolved. Here we address the problem by modeling a phase sequence embedded into a random recurrent network with sparse connectivity p_rand. The phase sequence as originally proposed by Donald Hebb [1] is a series of activity of cell assemblies connected in a feed-forward fashion. In our model, neurons representing one assembly are connected with recurrent probability p_rc>p_rand while neurons in subsequent assemblies are connected with feed-forward probability p_ff>p_rand. Each assembly has a corresponding inhibitory subpopulation to which it is recurrently connected with probability p_rc (see Figure 1A). Additionally, we apply an inhibitory plasticity rule that balances excitation and inhibition [2]. The role of such inhibition is twofold: it maintains asynchronous irregular firing of the excitatory population, and it enhances the response of an excitatory assembly through balanced amplification [3]. In the extreme case of no recurrent connectivity (p_rc = 0), the network resembles a synfire chain: a feed-forward network with convergent-divergent connections between subsequent groups of neurons [4].