Shaping behavioral patterns

Pigeons were rewarded for distributing eight pecks across two keys (L and R) in various patterns. The simplest pattern was at least one switch between the two keys (LR or RL) anywhere during the sequence; the next simplest was at least one instance of LLRR or RRLL anywhere during the sequence; the next was LLLRRR or RRRLLL; the most complex was LLLLRRRR or RRRRLLLL. Note that each more complex pattern contains the simpler ones within it. Initially, all patterns were reinforced but amount of reinforcement varied directly with complexity of pattern. The pigeons typically began the eight-peck sequence by pecking on their dispreferred key and then switched to their preferred key during the sequence. In subsequent conditions, simpler patterns were progressively unreinforced until finally only the most complex pattern (exactly four pecks on one key followed by exactly four pecks on the other) was reinforced. Three of the 4 pigeons tested maintained responding under this contingency; responding of the 4th pigeon extinguished. A second group of 4 pigeons was exposed immediately after training to extinction of all patterns except the most complex one. Three of the pigeons failed to maintain responding and the 4th maintained responding at a very low level. These results are evidence that response patterns can be shaped directly without building them up from a sequence of individually reinforced responses. The results may serve as a model of how self-controlled and altruistic behavior can arise through reinforcement.