The behavioral theory of the firm predicts that poor performance relative to aspirations levels leads to increased effort and search for innovations while high performance relative to aspirations allows search without risk of performance falling below aspirations. The classic behavioral arguments do not specify, however, whether increased search leads to incremental or significant innovations, how low and high performance searches differ, or whether performance relative to aspirations has different effects when those aspirations reflect comparison to historical performance or comparison to relevant others. We argue and present evidence that, holding R&D spending constant, poor and high performance lead to decreased rates of overall patenting but increased the rates of breakthrough patenting. Maximum overall patents and minimum breakthrough patents appear for firms near their reference points. Taken together, breakthroughs are most likely to be invented by firms that are doing very well, relative to their industry, and at the same time, very well or very poorly relative to their own historical performance.