Moscovici and Zavalloni (1969) suggest that both risk shifts and attitude shifts after group discussion are examples of a general group tendency to polarize opinions. In the present experiment, using both attitude and risk items, group discussion did not make individual opinions more extreme; only the group average became more extreme. This group extremity increase was not simply a more general way of conceptualizing the directional shifts in attitude and risk; group extremity increase appeared to be an effect of discussion that was independent of the risk and attitude shifts.
Also, subjects in the co-working pretest of the standard risk-shift paradigm were found to be less extreme and more ‘agreeing’ than pretest subjects who were truly alone. This co-working/alone difference persisted after discussion and was not related to group extremity increase. On both attitude and risk items, group extremity increase was strongly correlated with group opinion convergence. It is argued from this correlation that group extremity increase may be an effect of some aspect of conformity influence.