A: observers successfully identified the debaters' true attitudes every
    time

    Incorrect.
In section, group 1 was assigned to argue that the prison study was ethical, group 2 was assigned to argue that the prison study was unethical and group 3 was given the task of observing the debate to ascertain the true attitude of the members of groups 1 and 2.  Group 3 did not successfully identify the debater's attitudes every time.  In fact, they believed that the debaters actually believed the positions they were arguing.  They believed this, even though the members were assigned to the groups randomly (e.g. equal chance of being chosen for groups 1,2 and 3), and in doing so, committed the fundamental attribution error, which is overestimating dispositional causes (e.g., person in group 1 believes the Stanford Prison Study was ethical)  of a behavior, and underestimating situational causes (e.g., person in group 1 was assigned to argue the Stanford Prison Study was ethical).  This particular aspect of the fundamental attribution error (e.g., an observer believing that a person arguing a particular position, actually supports that position, even when the observers knows that that person was assigned to argue that position) is termed the Correspondence Bias.  Since members of the observer group (group 3) did commit this error, answer A is not correct.