The corpora pedunculata, or mushroom bodies, are paired lobes of neuropile present in the protocerebrum or dorsal brain of all insects. They are divisible into three parts: calyx, stalk and roots. The latter usually comprise two simple lobes, the a and ft lobes. The corpora pedunculata of a variety of Lepidoptera were examined. All had a double calyx-cup. Each ‘cup-cavity’ is composed of ‘globuli’ cell bodies. The broad stalk, a tract of fibres and neuropile, leads from the calyx to the complex ‘roots’—a, g and y lobes. A third group of globuli cells near the calyx gives rise to a tract leading to a second lobe-system—the tripartite Y-lobe—in the roots. As neither the Y tract nor the Y lobe has been described before in any insect, their possible homologues are unknown. The two lobe systems in the roots are closely intertwined, yet have no interaction except in the y lobe. A number of different neuron types with branches in the mushroom bodies has been described from Golgi preparations. Some (intrinsic cells) divide in the calyx and again in the roots, but do not pass out of the mushroom bodies. Others (extrinsic cells) branch in the mushroom bodies and in other areas of the brain, thus connecting two regions. Intrinsic cells arise from cell bodies in the calyx-cups or posterior to them. There are two types: one has extensive spine-covered branches in the calyx, while the second has claw-like terminals covering a narrow cylindrical field. Processes from these cells run to the a, j>and y lobes via the stalk. A wide-field accessory cell, which arises from the third group of globuli cell bodies, also has claw-like endings in the calyx. A process of this cell runs in the Y-tract to the Y-lobe. Extrinsic terminals in the calyx arise from cells branching in the antennal lobe, in an accessory optic area in the protocerebrum, in the ‘undifferentiated’ protocerebral neuropile, or in the suboesophageal lobes. The antennal terminals in the calyx are knob-like. It is proposed that they form the centre of the ‘glomeruli’ typically present in calycal neuropile. The claws of the bunched intrinsic and accessory cells probably fit around these knobs. Within the stalk, different subvarieties of intrinsic cells have been distinguished on the basis of the distribution of the side-branches and spines which they bear. The stalk is thought to be the site of extensive postsynaptic interaction between intrinsic cells. Fibres in the stalk run in bundles or groups. All the fibres in one bundle are of the same subvariety. In the roots, the subvarieties of intrinsic cells have different branching patterns. The a and B lobes are not homogeneous, but are divided into sublobes. Extrinsic fibres ramify only within one sublobe generally, though some have very large fields. The connexions of the roots are obscure. Some extrinsic fibres branch again in the ‘undifferentiated’ protocerebral neuropile; others, from the B lobe, may run to the suboesophageal lobes. There are profound differences between the internal organization of the mushroom bodies in Hymenoptera (Kenyon 1896; Goll 1967) and Lepidoptera. The functional implications of the Lepidopteran form are discussed.