Magnitude estimation of temporal intervals.

Abstract

Data are presented from a magnitude estimation study of temporal intervals. The individual subject data do not support the lognormal model of error. Many of the response distributions are negatively skewed,and, in general, the variance of the log responses decreases with stimulus duration. The fit of the two-parameter power function, obtained using nonlinear estimation, is compared with the fit obtained using linear least squares in log-log coordinates. Nonlinear estimation provides a better description of the individual subject data and yields estimates of the exponent which are closer to 1.00. The addition of a third parameter brings the exponents very close to 1.00. It is concluded that, since the value of the power function exponent depends upon the curve-fitting procedure used, magnitude estimation data do not provide compelling evidence against a linear psychophysical function for time.