Figure 5.

Comparing the sensitivity of synaptic changes to spike-pair correlations and firing rate variations. A, Illustration of two neurons firing with correlation coefficient ρ < 1. The presynaptic neuron is made to fire a Poisson spike train, and each presynaptic spike evokes a postsynaptic spike at a fixed time lag Δ with probability p = ρ (see Materials and Methods). A fraction 1-p of uncorrelated postsynaptic spikes is added to keep the firing rate independent of p. The cross-correlation function is given by a δ function with an amplitude proportional to ρ (right). B–D, Change in synaptic strength as a function of firing rate for several values of the correlation coefficient ρ and time lag Δ in the calcium- (with linear calcium dynamics, B), the triplet- (C), and the pair-based (D) models. Five cases are shown: (1) uncorrelated Poisson spike trains, ρ = 0 (black); (2) pre-post pairs with Δ = 10 ms at ρ = 0.2 (light red) and ρ = 0.4 (dark red); and (3) post-pre pairs with Δ = −10 ms at ρ = 0.2 (light blue) and ρ = 0.4 (dark blue). E–G, Sensitivity of synaptic changes to spike-pair correlations in the calcium- (E), triplet- (F), and pair-based (G) models. See Equation 25 for the definition, and red arrow in C for an illustration of sensitivity to correlations at 20 spk/s. The red triangle in F represents the corresponding value. H, I, Sensitivity of synaptic changes to firing rate changes in the calcium- (H) and triplet-based (I) models. See Equation 26 for the definition. The gray arrow in C indicates sensitivity to firing rate changes for a baseline rate of 20 spk/s, and an increase of 15.3 spk/s. The gray triangle in I represents the corresponding value. Color plots represent the synaptic change as function of baseline firing rate (x-axis) and increase in firing rate (y-axis). Light and dark red lines indicate the firing rate increase evoking the same synaptic change as spike-pair correlations with ρ = 0.2 and ρ = 0.4, respectively.