Acquisition and generalization of conditional choice value in pigeons and humans

When organisms choose option A instead of B, it is impossible to know whether they did so because they desired A, because they avoided B, or both. A popular way to identify which of those possibilities explained the choice of A involves tests in which either A or B is replaced with a novel option of unknown value. In this dissertation work, I used this technique to better understand the choices of pigeons and humans. In 6 different experiments, I taught pigeons and humans to choose between many options in a task in which the value of each option was determined, not by itself, but by another stimulus. Most importantly, the tasks were designed to encourage subjects to base their choices on either the selection or rejection of a single option. Initial findings suggested that pigeons and humans had learned about the structural regularities present in the task. Further findings suggested that pigeons were more effective than humans in learning these regularities. A final set of findings confirmed that the behavior of pigeons was instead the result of them also using the value of each option by itself. I then built this idea into mathematical models that varied in how they represented value and how they used it in novel situations. An evaluation of all models showed that one of the simplest models was the best at capturing the dynamics of human and pigeon choice.