What can be inferred about neural population codes from psychophysical and neuroimaging data?