This manuscript addresses errors of measurement that exist in cross-sectional research, but that are not readily recognized unless longitudinal data are collected. Both specific and transient errors of measurement are defined, a method of identifying them if they exist is presented, a method of correcting for them is presented, and their implications for longitudinal data analysis are briefly discussed.It is well known that when any two variables, say X and Y, are measured validly, the cross-sectional correlation between them, r xy , is attenuated by any error of measurement in either or both variables. 2 Moreover, the means of correcting for this underestimate produced by this source of error is also well known-the correction for attenuation due to error of measurement or r 0 xy ¼ r xy =ðr xx Þ :5 ðr yy Þ :5 , where r 0 xy is the corrected correlation coefficient, r xx is the estimate of the reliability of the measure of X, and r yy is the estimate of the reliability of the measure of Y. 3 Correcting correlations for attenuation does not, however, come without cost. There is more sampling error associated with a correlation corrected for attenuation due to error of measurement than that associated with its uncorrected counterpart