of positive findings might temporarily create a halo in which negative findings might be more readily entertained by journals for a short period of time. Furthermore, and inconsistent with expectations based on statistical power, Duncan and Keller found that the larger the cGxE sample, the less likely it was to be significant. This trend would not be expected if cGxE findings were valid. These observations are consistent with widespread publication bias in the cGxE literature. Publication bias can arise when authors, editors, and reviewers believe that positive findings are more worthy of publication than are negative or null results. There are many uninteresting ways for an empirical test of a hypothesis to fail to support it---the hypothesis was implausible to start with, the power of the test is inadequate, the operationalization of the variables is not valid, etc. As a consequence, the greater interest in positive results is understandable, especially when they bring an insightful increment to, or even transformation of, our understanding. Yet, interest in positive findings is clearly misplaced if they are false, as they can (mis)guide research efforts and funding priorities.
