lessly or cautiously, but not irresponsibly, since he is subject to correction from his informants so long as he has contact with his informants. The promise in transform grammar is projected in the realm of final analysis. Neither the prediction of open corpus field work nor the promise of transform grammar can by-pass the problem of eliciting, though they can handle the problem differently [see below]. It turns out that however they handle the problem of eliciting, both emerge with statements of structure of classes of morphemes and sequence classes ('grammaticalness'), irrespective of restrictions on individuals within the classes. But it is precisely in the stance that each takes upon facing the restrictions and special privileges within classes ('selection') that open corpus preliminary analysis differs from transform grammar final analysis. The former characteristically turns to cultural domains, as body parts, or artifacts; for Hopi, we say that all body parts may combine with a set of possessor affixes; for Fox, Bloomfield says that all manufactured articles (artifacts) are inflected for the same gender, and so on. It is certainly possible to approach the problem of 'selection' through cultural domains, but not without loose-ends and intersection (as detailed in Hopi Domains, IUPAL Memoir 14, 1957). In effect, the promise of transform grammar is to solve the problem of 'selection'; to generate all possible utterances from rules which leave neither loose ends nor uneconomical intersections.