The oral interaction hypothesis, proposed by Long and investigated by Pica, in second language (L2) acquisition is critiqued. The interaction hypothesis advances two major claims about the role of interaction in L2 acquisition: (1) comprehensible input is necessary for L2 acquisition; and (2) modifications to the interactional structure of conversations that take place in the process of negotiating a communication problem help make input comprehensible to an L2 learner. Changes that have taken place since the initial formulation of the hypothesis are reviewed, as is evidence cited in support of the hypothesis. In view of theoretical arguments and empirical counter evidence, a revised version of the hypothesis is proposed. The revisions suggest a theoretical account of how input made comprehensible through interactional modification results in acquisition, and also takes into account the basic procedures of noticing, comparison, and integration. Contains approximately 75 references. (LB)