Brain size represented by cranial capacity (CC) is one of the most frequently analysed characters of hominids. Accuracy of individual CC estimates depends on completeness of specimens and methods used for reconstruction and measurement. A file of published estimates of CC of hominids dated from 3.2 Ma (million years) to 10 Ka (thousand years) including 606 estimates for 243 specimens was compiled. In the file, 75 specimens are available with multiple values (3 to15) obtained by various methods and/or by various authors. Using individuals as classes in ANOVA, intraclass variation, which represents «error» of estimates, was calculated. For the total sample of multiple estimates (N = 382) the error variance is 5315 ml2. The error standard deviation is 73 ml (coefficient of variation (CV = 7.3%), quite large in comparison to the actual variation in CC in modern humans, SD = 157 ml (CV = 11.6%). This large error makes us wonder whether any fossil can be reliably placed with respect to a particular «cerebral Rubicon» between palaeospecies. Recent discussions concerning cranial capacity of Stw505 are a case in point regarding errors in CC estimation. In actual repeated 30 time measurements on a research quality cast we obtained with various methods (water, seeds, plasticine) CC estimates ranging from 484 to 586 ml. The range of estimates in the literature is from 515 to 625 ml. When hominid CC by taxon with date as a covariate is subjected to ANOVA, taxon is responsible for 5% of the variance while date is responsible for the main portion, (89%). The relationship between CC and date is best characterised as a gradual time trend. It is proven by the ANOVA test for linearity, by G test for trend and by ASReml fitting of a linear function. The line of best fit to this time trend is a double exponential curve which explains 90% of the total variance in CC: CC = 306.63 (4.830.9995DATE) Essentially the same curve fits subsamples of CC dated at less than 1 Ma and at 3.21.0 Ma. This has several implications for the nature of the Darwinian process to be reconstructed.