Two weaknesses are particularly evident in the now extensive learned helplessness (LH) literature. The first is the lack of fit between operational definitions and the predicted results. The second is the inability of the theory to specify the extent to which the effect should generalize, and a lack of empirical studies to delineate such boundaries. In the recent human literature, these criticisms assume even greater significance. By definition, LH involves the generalization of an expectancy of uncontrollability from one situation to another, separate situation. Because the vast majority of human studies have conducted the post test within the context of the same experimental situation as the pretreatment, the question arises not only ta whether generalization has been demonstrated but whether interference itself has been demonstrated. In an attempt to improve the fit between theory and behavioural outcome, a State-Trait (S-T) Helplessness model was developed, incorporating locus of control and elements from Social Learning Theory, Reactance Theory, and Attribution Theory, in addition to the original LH formulation.

An experiment was conducted to test a part of this theoretical framework. Three groups, equally divided between internals and externals and counterbalanced for sex were exposed to escapable noise, inescapable noise, or no noise. They were then tested on a series of 20 patterned anagrams. Subjective stress self-ratings and peripheral pulse volume and heart rate were the other major dependent variables. In addition to a replication of cross modal interference in man, more complex relations between sex and locus of control were found, indicating that these two trait variables partly determined whether or not interference and mastery effects were found. Increased subjective stress accompanied interference effects on the anagrams. The physiological results were complex, interacting with locus of control and suggesting the presence of bath activation and deactivation among internal subjects on one measure.

A second group received identical pretreatment to the inescapable group already described. But, unlike the first, it received the post treatment phase as part of a separate experiment, conducted by a different experimenter. Anagram performance was similar to that found in the other inescapable group. Subjective stress ratings were slightly lower. These findings indicate not only that generalization occurs but suggests that previous human experiments may have obtained similar results, irrespective of whether the post treatment tests were presented as a part of the same experiment or as a separate experiment. They also raise the possibility of different generalization gradients existing for the different interference components.