Three plausible hypotheses about developmental commitments in the Drosophila embryo propose that: (1) a micromosaic of localized determinants in the egg trigger somatic commitments; (2) monotonic anterior-posterior and dorsal-ventral gradients in the egg specify positions by a series of threshold values; (3) sequential subdivision of the early embryo into ‘anterior’ or ‘posterior’ ‘middle’ or ‘end’, ‘dorsal’ or ‘ventral’, ‘odd’ or ‘even’ compartmental domains encodes the somatic commitment in each region in a combinatorial epigenetic code. Evidence in favour of such a combinatorial code includes its capacity to account for major features of transdetermination and for many single and coordinated homoeotic transformations. In particular, both these metaplasias often cause transformations between ectodermal tissues such as antenna and genitalia, whose anlagen lie far apart on the blastoderm fate map. This phenomenon is not naturally explained by monotonic gradient models. In contrast, not only transformation between distant regions of the fate map, but also the observed geometries of compartmental boundaries on the wing, and probable ones in the early embryo, are naturally explained by reaction-diffusion models. These systems form a discrete succession of differently shaped monotonic and nonmonotonic eigenfunction gradient patterns of the same morphogens, as the tissue containing the chemical system changes in size and shape, or in other parameters. The successive mirror symmetries in non-monotonic gradients predict that distant regions of the embryo make similar developmental commitments, and also predict specific classes of pattern mutants forming mirror symmetric structures along the embryo on a variety of length scales. Finally, reaction diffusion systems spontaneously generate transverse gradients of the underlying chemicals when more than one eigenfunction is amplified at once, and therefore specify two-dimensional positional information within domains. Although it is attractive, no feature of the combinatorial code hypothesis is verified. Current data relating to whether the sequential formation of compartmental boundaries actually reflects the commitment of the two isolated ‘polyclones’ to alternative fates, whether any genes act continuously to maintain disc commitments, and whether homoeotic mutants actually ‘switch’ disc determined states, are assessed.