Harking

HARKing, short for Hypothesizing After the Results are Known, is a questionable research practice in which a researcher presents a post hoc finding as if it were hypothesized before data collection. In effect, a hypothesis is formulated after inspecting the data and then described as if it had been predicted in advance. This misrepresents the evidential status of the test.

The term was coined by Kerr in 1998 in discussions of research practices in psychology and related fields. HARKing relies on a retrospective framing of hypotheses, so standard significance testing does not reflect the true probability of observing the result under the stated hypothesis.

Examples include reporting a significant effect and stating that a particular variable was hypothesized to predict the outcome all along, or reinterpreting exploratory patterns as if they were predicted and tested a priori.

Implications include inflated Type I error rates, reduced study validity, and poorer replicability. When researchers present post hoc hypotheses as confirmatory, it becomes difficult to distinguish theory-driven results from data-driven findings.

Prevention and response involve preregistration of hypotheses and analysis plans, registered reports, transparent reporting of exploratory analyses, and clear labeling of confirmatory versus exploratory results. Replication and open science practices also help mitigate HARKing.

Relation to other practices: HARKing is related to but distinct from p-hacking or data dredging, which focus on manipulating data to achieve significance; HARKing specifically concerns misrepresenting the status of hypotheses.