Some general aspects of the concept of imaginal discs in the Holometabola are reevaluated. Their monolayer character and continuity with the surrounding epidermis are confirmed. Studies on the imaginal discs of the silkworm (Bombyx mori) and data from the literature show that the discs and their peripodial cells produce cuticle during larval life, as well as at metamorphosis. In B. mori it is demonstrated that adult and larval antennae are produced by the same cells or their progeny. The results also suggest that segments of the typically three-segmented larval antenna of Holometabola are not scape, pedicel, and one-segmented flagellum; at least segments 2 and 3 are of flagellar origin. Based on these and some additional facts it is argued that: (1) No larval organs are "replaced" at metamorphosis, but strict "sequential homology" is always maintained. (2) Imaginal discs are not undifferentiated structures destined to form the adult after larval breakdown, cannot be unambiguously defined, and do not represent qualitatively different epidermal structures. Classical imaginal discs (invaginated and present also in pre-final larval instars) arose several times independently and were not present in the larvae of ancestral Holometabola. (3) Since the disc cells are not undifferentiated and "embryonic" (if these words have a defined meaning at all), it is unreasonable to expect that the processes taking place in discs at metamorphosis would differ fundamentally from those occurring in other diploid metamorphosing epidermal cells.