Pigeons were given a choice between observing a stimulus source that was uncorrelated with food or one that was informative. The informative source was either positive, in which a stimulus change signalled food, or negative, in which change signalled no food. If observing is supported by the reduction of uncertainty, the negative as well as the positive source should be preferred to the uncorrelated source. On the other hand, if observing requires support by conditioned reinforcement, the negative source should not be preferred to the uncorrelated source. Two keys served as stimulus sources in a discrete trial procedure. The keys were lighted together, remained on for a variable length of time, and went off together. A key could change color 1 sec before going off. In the uncorrelated source, the change occurred equally often on trials ending with or without food. In the positive information source, the change occurred only on food trials, whereas in the negative information source, it occurred only on no-food trials. All stimulus changes and food delivery were response independent. As measured by orientation and autoshaped pecking, the positive information source was preferred to the uncorrelated source. However, the uncorrelated source was preferred to the negative information source. The latter result does not support the view that observing behavior is reinforced by the reduction of uncertainty. The positive and negative information sources reduced uncertainty equally but only the positive source provided a signal that could act as a conditioned reinforcer by virtue of its relation to food.