Herrnstein's
Rank-Order Statistic, Rho
In order to
determine discrimination performance is such a daily session we rank
the peck rates counted in every trial. We then calculate the
probability that a randomly selected positive exemplar will be ranked
higher than a randomly selected negative exemplar by using the
non-parametric Mann-Whitney U test. The resulting U value is then
linearly transformed into Herrnstein's (Herrnstein et al, 1976;
Herrnstein, 1979) rho measure of discrimination. It makes no
assumptions about the underlying distributions of response rates. When
rho is 0.5, discrimination is absent; and when rho is 1.0,
discrimination is perfect. With 20 positive and 20 negative stimuli, a
rho of 0.68 indicates that discrimination is significant at the 0.05
level.