A literatura sobre o desamparo aprendido utilizando humanos como participantes de pesquisa, tem encontrado dificuldades metodologicas. O presente experimento teve como objetivo geral estabelecer um procedimento que permitisse controle experimental suficiente para verificacao do efeito de uma historia de incontrolabilidade, com e sem feedback de erro, sobre a aprendizagem de uma nova resposta operante em sujeitos humanos. Estudantes universitarios foram divididos em quatro grupos (n=10), denominados Nao-tratado (N), Controlavel (C), Incontrolavel (I) e Incontrolavel feedback (If). Na primeira fase (tratamento), com excecao dos sujeitos do grupo N, os demais foram expostos a estimulos aversivos (som) que poderiam ser desligados contingentes ao seu comportamento de teclar sequencias variaveis (grupo C) ou independente dele (grupos I e If). No final da sessao, foi perguntado aos sujeitos a que atribuiam o termino do som: as respostas mostraram correspondencia entre a contingencia a qual estavam submetidos e o comportamento verbal para 100% dos participantes do Grupo C, 70% do Grupo If e 60% do Grupo I. Na segunda fase (teste), todos os participantes foram submetidos a resolucao de anagramas: os sujeitos dos grupos C e N apresentaram menores latencias e maior numero de acertos, sendo maiores as latencias e falhas obtidas no grupo If; o grupo I mostrou resultados intermediarios. Esses resultados sao consistentes com o esperado no estudo do desamparo aprendido, sugerindo que o procedimento proposto e adequado para esse estudo. Foram discutidas as contingencias acidentais que possivelmente se estabeleceram no grupo I, a funcao do feedback de erro no controle dessas contingencias, o controle pela regra e as implicacoes metodologicas do procedimento empregado.

The literature on learned helplessness in human has been fraught with methodological difficulties. The present research has the general objective of establishing a procedure that permits sufficient experimental control to verify the effect of a history of uncontrollability with or without feedback of failure on learning a new operant response by humans. College students were divided into four groups (n=10), deemed Non-treated (N), Controllable (C), Uncontrollable (I) and Uncontrollable with feedback (If). In the first phase (treatment), except for Group N, the participants were exposed to an aversive stimulus (tones) that could be controllable (i.e., turned off contingent upon making variable sequences using the keyboard) or independent of behavior (Groups I and If). At the end of the session, participants were asked to what they attributed the end of the tones: the correspondence between their answers and the situation to which they were exposed was 100% for Group C, 70% for Group If and 60% for Group I. In the second phase (test), all participants were asked to solve some anagrams. Groups C and N showed lower latencies and a greater number of correct responses. Latencies and rate of failure were higher in Group If; Group I presented intermediate results. These results were in agreement with the learned helplessness effect, suggesting that the procedure is adequate for its study. The possible accidental contingences established in Group I, the function of the feedback of failure in these accidental contingencies, the role of instructions and some methodological implications are discussed.