“…However, these models usually utilize an alternative to such an excitatory wave, such as prominent sensory cue inputs that strongly stimulate a subset of neurons at the onset of each sequence (Tsodyks et al, 1996; Lisman et al, 2005), selective amplification of already strong synaptic connections that promotes activation of a subset of cells (Fiete et al, 2010), or very strong connectivity, which gives rise to a distinct region of high activity that perpetually activates itself due to asymmetry of connections, dynamic thresholds, or dynamic strength of synapses (Kleinfeld, 1986; Sampolinsky and Kanter, 1986; Horn and Usher, 1989; Hopfield, 2010; Itskov et al, 2011). We propose that synchronized waves of excitability are required for the generation of sequences in networks with no sensory inputs, with relatively weak connections and with relatively sparse activity (Wang et al, 2015; Romani and Tsodyks, 2015). Without such synchronized waves, the activity of the network is dominated by non-specific feed-forward inhibition that prevents activation of specific sequences.…”