Table 3.

Information transfer estimates for consonant features in 20 cochlear implant users from Donaldson and Kreft (2006). Next to each feature in parentheses is the number of feature categories. Avg: average estimate per user, N=285 samples per matrix; Pool: estimate obtained from data pooled in one matrix with N=5700 samples; and diff=Avg-Pool. Without the small-sample bias, diff should be close to zero.

$I\widehat{T}$ (%)	∕Ca∕			∕Ci∕			∕Cu∕
	Avg	Pool	diff	Avg	Pool	diff	Avg	Pool	diff
Voicing (2)	59	56	3	63	60	3	67	65	2
Place (3)	35	33	2	23	21	2	27	25	2
Manner (4)	60	54	6	59	54	5	56	52	4