Neurophysiologists are often faced with the problem of evaluating the quali
ty of a code for a sensory or motor variable, either to relate it to the pe
rformance of the animal in a simple discrimination task or to compare the c
odes at various stages along the neuronal pathway. One common belief that h
as emerged from such studies is that sharpening of tuning curves improves t
he quality of the code, although only to a certain point; sharpening beyond
that is believed to be harmful. We show that this belief relies on either
problematic technical analysis or improper assumptions about the noise. We
conclude that one cannot tell, in the general case, whether narrow tuning c
urves are better than wide ones; the answer depends critically on the covar
iance of the noise. The same conclusion applies to other manipulations of t
he tuning curve profiles such as gain increase.