Short term trials with surrogate measures instead of hard outcomes are often used to study chronic diseases. The effects of an intervention may, however, take time to develop and persist after discontinuation, producing wash-in and washout effects that threaten trial validity. This problem, especially involving carryover effects in crossover trials, is well recognized in the pharmacology and statistics literature but commonly disregarded in some areas of non-pharmaceutical research. Dietary trials, including feeding studies, are highly susceptible to bias because physiological adaptation to a major change in nutrients may take several weeks or longer. This article describes the nature and extent of this bias in nutrition research, as an important and illustrative case; considers implications for various other interventions; and proposes measures to strengthen causal inference.