Fig. 4.
Estimates for the representational similarity between sensory and motor representations of the same finger (true value r = 0.5), as a function of the level of noise (σε2) and the covariance of the patterns common to the movement and stimulation conditions (γα). (A) The sample correlation calculated on the mean activation patterns for identical fingers across conditions is strongly influenced by noise and common activation. (B) By subtracting the correlation across conditions for different fingers, the influence of the common activation is eliminated. However, the correlation is underestimated and biased (downwards) by noise. (C) The correlation between patterns for the same fingers, after subtracting the mean pattern for the respective condition accounts partly for the effect of common activation, but is severely biased by noise. (D) The corrected estimate from the covariance-component model is unbiased over a large range of parameter settings.