RESUMEN
One-hundred 3-person groups and 300 individuals solved 2 letters-to-numbers problems, requiring identification of the coding of 10 letters to 10 numbers by proposing an equation in letters, receiving the answer in letters, proposing a hypothesis, and receiving feedback on the hypothesis on each trial. There were 5 instruction conditions: (a). standard, (b). use at least 3 letters on all equations, (c). use at least 4 letters on all equations, (d). number 1 known before beginning problem, and (e). number 9 known before beginning problem. The groups had fewer trials to solution, proposed more complex equations, and identified more letters per equation than the best individuals. Performance was best under instructions to use at least 4 letters and with the number 9 known.