Tested several social decision models—or problem-solving models—on groups of different sizes working on a disjunctive task, that is, anagrams of two levels of difficulty. The subjects were 300 ten- and eleven-year-old Swedish boys and girls from Grades 4 and 5, a 2 × 5 × 2 (ability levels high and low × group sizes 1, 2, 3, 4, and 5 × task difficulties hard and easy) design was used. Support was found for the truth-wins model as an explanation of the underlying processes in very small groups (2 and 3), while reasonable explanations for larger groups (4 and 5) were given by the truth-supported wins and the majority if correct, equiprobability otherwise model. Slightly different group processes were inferred on the hard as compared to the easy tasks. The results also extend previous research on the generality of the models, especially the truth-wins model.