Abstract: | In an experiment on attitude attribution, subjects were instructed to estimate the “true attitude” of a target person after reading an essay in which he took one or the other side of a controversial issue (legalization of marijuana). Four independent variables were manipulated: the direction of the essay (pro- versus anti-legalization), its extremity (strong versus weak), freedom to choose position versus assignment to position (choice versus no choice), and prior expectancy (expect pro versus expect anti). All experimental predictions were confirmed by the results. When the essay was strong: (1) attitudes were attributed more in line with behavior under choice than under no choice conditions; (2) even under no choice conditions, the target person was seen as believing to some extent in the arguments of his essay; (3) the role of choice was especially prominent when the position of the essay was unexpected. In both strong and weak essay conditions, (4) the behavior was ignored in favor of prior expectancy in no choice conditions, but a contrast effect was observed in the choice conditions. When the essay was weak: (5) subjects attributed the opposite attitude under no choice conditions; under choice conditions, the weak essay was construed as moderate endorsement. As in previous experiments, there was considerable variability in those conditions in which the target person wrote an essay under no choice instructions endorsing the position he presumably opposed. In a second experiment, attempting to determine what produced this variability, it was found that subjects with generalized expectancies of internal control (Rotter, 1966) were more sensitive to variations in choice than were subjects whose expectancies were externalized. Under no choice conditions, “internals” were more inclined to ignore the essay and go by their prior expectancies in attributing attitude; “externals” seemed more impressed by the essay itself. |