The amygdaloid complex was investigated in 36 serially sectioned staged human embryos, including 20 impregnated with silver. This is the first such account based on graphic reconstructions, 28 of which were prepared. Significant findings in the human include the following. (1) The medial (first) and (then) lateral ventricular eminences arise independently at stages 14 and 15, and unite only at stage 18 to form the floor of the lateral ventricle. (2) The future amygdaloid region is discernible at stage 14 and the amygdaloid primordium at stage 15. (3) The anterior amygdaloid area and the corticomedial and basolateral complexes appear at stage 16. (4) These three major divisions arise initially from the medial ventricular eminence, which is diencephalic. (5) Individual nuclei begin to be detectable at stages 17–21, the central nucleus at stage 23 and the lateral nucleus shortly thereafter. (6) The ontogenetic findings in the human embryonic period accord best with the classification used by Humphrey (J Comp Neurol132, 135–165, 1968). (7) The lateral eminence, which is telencephalic, contributes to the cortical nucleus at stage 18. (8) The primordial plexiform layer develops independently of the cortical nucleus. (9) Spatial changes of the nuclei within the amygdaloid complex and of the complex as a whole begin in the embryonic period and continue during the fetal period, during the early part of which the definitive amygdaloid topography in relation to the corpus striatum is attained. (10) The developing amygdaloid nuclei are closely related to the medial forebrain bundle, which has already appeared in stage 15. (11) Fibre connections develop successively between the amygdaloid nuclei and the septal, hippocampal and diencephalic formations, constituting the beginning of the limbic system before the end of the embryonic period. Although the nucleus accumbens also appears relatively early (stage 19), connections between it and the amygdaloid complex are not evident during the embryonic period. (12) Influence of the olfactory bulb and tubercle on initial amygdaloid development, as postulated for rodents, is unlikely in the human. The findings exemplify the necessity of beginning developmental studies with the embryonic period proper.