In this discussion of recent ideas concerning the mechanism of formation of classical antibodies, the attention is focussed on the structure of a general conceptual framework rather than on particular processes that may represent important links in the stimulus-reaction chains. The antibody-forming system in an animal is analogous to an electronic translation machine, the parts of which are scattered among other devices in a factory exercising a large variety of functions. In trying to understand the design and structure of this translating machine, we might obtain useful clues by studying parts such as bits of magnetic tape and photographic film. But it would be equally important to analyse the general performance of the machine, in order to deduce what essential operations have to be provided for, and to try to induce the machine to make mistakes that would reveal limitations of its complexity. Having observed that the machine translates a foreign language into English, we might say: The machine produces English, but recognizes only Foreign. We would realize that both English and the foreign language are composed of the same alphabet, and that single letters cannot be the units that are recognized by the machine. We might also feel sure that a sentence cannot be recognized as a whole because the number of possible sentences would surpass any tolerable complexity. We would prob ably conclude that an important feature of the machine would have to be an ability to recognize single foreign words, and that the mechanism by which it functions must, in some form or other, include the consultation of a For eign-English dictionary.

In the interesting analysis of this electronic device by Delavenay (21), several other analogies with the antibody-forming system suggest themselves to the imagination. The idea of pre-and postediting is reminiscent of Burnet & Fenner's (8) modified enzyme proposals, whereas the possibility of insert ing an interlanguage between entry and final translation reminds one of