The levels of analysis revisited.

The term levels of analysis has been used in several ways: to distinguish between ultimate and proximate levels, to categorize different kinds of research questions and to differentiate levels of reductionism. Because questions regarding ultimate function and proximate mechanisms are logically distinct, I suggest that distinguishing between these two levels is the best use of the term. Integrating across levels in research has potential risks, but many benefits. Consideration at one level can help generate novel hypotheses at the other, define categories of behaviour and set criteria that must be addressed. Taking an adaptationist stance thus strengthens research on proximate mechanisms. Similarly, it is critical for researchers studying adaptation and function to have detailed knowledge of proximate mechanisms that may constrain or modulate evolutionary processes. Despite the benefits of integrating across ultimate and proximate levels, failure to clearly identify levels of analysis, and whether or not hypotheses are exclusive alternatives, can create false debates. Such non-alternative hypotheses may occur between or within levels, and are not limited to integrative approaches. In this review, I survey different uses of the term levels of analysis and the benefits of integration, and highlight examples of false debate within and between levels. The best integrative biology reciprocally uses ultimate and proximate hypotheses to generate a more complete understanding of behaviour.