In 237 low birth weight for gestational age term infants (birth weight less than or equal to 10th percentile; gestational age less than or equal to 37 wk) percentage deviations of birth weight and crown-heel length for gestational age were correlated (a) in a series of linear regressions of three subsequent groups formed according to the magnitude of weight retardation (b) in the pooled material by non-liner regression. Linear regressions yielded equivocal results as to the existence of concomitant length retardation (no correlation in group 1, significant correlation in group 2 and 3). The significant non-linear correlation (i = 0.66 ; p less than 0.001) and the exponential type regression curve (Y) % length deficit = 2.45 1.04 X' % weight deficit) gave an explanation of the controversial findings. If mild to moderate weight retardation is included in studies comparing weight and length, a greater variability and, possibly, a conclusion of non-existence of concomitant retardation will result while, if severe or extreme weight retardation are screened out using stricter criteria, (e.g. 5th, 3rd percentiles, -2 SD) one would find more infants with significant length retardation, and the conclusions would be the opposite, i.e. an obligatory length retardation following malnutrition. In anthropometric-statistical studies the results should never be generalized and extrapolated below, or beyond the cut-off points chosen.