We examined the use of nonlinear transformation of variables in a random sample of 323 articles published in six top journals during 2012-2017. Coding categories included the number of transformed variables, the type of transformation, the kinds of variables transformed, reasons provided for transforming variables, how transformed results were reported, and pre- and posttransformation analysis of variables. Common problems include insufficient justification for transforming variables, overreliance on log transformations, failure to report important information on the effects of transformation, and incomplete reporting and discussion of transformed results. Perhaps most importantly, there was frequent misalignment between statements of hypotheses, typically stated in terms of nontransformed variables, and the transformed data used to test them. We discuss the implications of these problems for science and practice, offer recommendations for addressing the issues, and provide illustrations of how to implement the recommendations.