Table 5.

Plasticity accelerates the discovery of a new optimal genotype network

N	c	d	Sample size	Mean t*, control	Mean t*, plast	p-value
8	0.4	0.125	498	164.93	54.45	< 2.2 × 10-16
		0.25	498	160.3	52.86	< 2.2 × 10-16
		0.5	497	167.58	46.95	< 2.2 × 10-16
	0.3	0.125	495	104.92	42.82	< 2.2 × 10-16
		0.25	498	103.84	38.03	< 2.2 × 10-16
		0.5	498	136.33	40.53	< 2.2 × 10-16
16	0.25	0.125	420	2744.8	2319.81	0.00059
		0.25	435	2792.44	2299.51	3.9 × 10-5
		0.5	421	2832.02	2173.54	2.2 × 10-7
	0.2	0.125	464	2510.68	1700.95	1.4 × 10-11
		0.25	468	2483.48	1867.23	1.2 × 10-6
		0.5	473	2400.1	1758.22	1.4 × 10-7
20	0.2	0.25	154	3961.17	3303.32	0.006869

The number of generations that a population takes to 'discover' a circuit in a new genotype network is significantly lower when we allow plasticity (t_{*, plast} <t_{*, control}), according to a Wilcoxon signed-rank test.

The value of d is that of the old genotype network. We analyzed 500 pairs of evolving populations for each combination of N, c and d. We discarded population pairs in which any of the populations had not reached the new genotype network by the end of the simulation (t = 10⁴). Thus, our actual sample size was lower than 500 populations. The probability α ofgene-activity perturbation in s₀ equaled 0.05 per gene when N = 8, 0.025 when N = 16, and 0.02 when N = 20. Population size M = 1000; μ = 0.5; ω^native = 0.5.