The development of the relationship between otolith and body size in Atlantic salmon (Salmo salar) between hatching and emergence was examined by repeatedly measuring individually identified fish. Otolith growth increments were deposited daily in the period between hatching and emergence. Comparison of back-calculated otolith size and standard length using least squares regression analyses revealed a weak relationship between these variables at each of the 5-day sampling intervals. However, when data sets were pooled among intervals, variation in otolith size accounted for 98% of the variation in alevin length. A computer simulation demonstrated that levels of measurement error similar to those documented in our study resulted in the failure of regression analyses to detect strong relationships between otolith and fish size. Mortality that occurred during the experiment was strongly size selective. This truncated the size ranges of fish in cross-sectional data sets and thus reduced the ability of regression analysis to detect relationships between otolith and fish size. We propose that the weak relationship between otolith and fish size at emergence recorded in previous studies was an artifact of measurement error and the truncation of size ranges in regression analyses. Differences in alevin size at emergence were present at hatching and had been propagated by growth.