Synfire chains with conductance-based neurons: internal timing and coordination with timed input

Synfire chain models store and retrieve hetero-associative sequences of firing patterns, thereby explaining basic aspects of the neuronal processing of temporal information. Existing models were based on McCulloch-Pitts or integrate & fire neurons and therefore neglect most physiological properties of real neurons. Here, we study a model with conductance-based neurons and both, hetero- and auto-associative couplings which support synfire vs. attractor activity, respectively. We show that the speed of synfire recall is influenced by slow neuronal variables and is sensitive to the ratio between auto- and hetero-associative synapses while quite insensitive to background activity. We then propose a bidirectional synfire model where the duration of states in a synfire chain is variable and can be coordinated by a timed but otherwise unspecific external signal.