Scientific knowledge generation can be conceptually divided into two complementary, distinct, and equally important phases: exploratory, in which hypotheses are generated to represent data patterns, and confirmatory (refutational), in which predictions based on these hypotheses are tested by collecting new data to provide evidence of practical/clinical utility. However, scientific publishing weights confirmatory research more heavily than exploratory research. The devaluation of exploratory research pressures authors to disguise their exploratory work as confirmatory, leading to questionable research practices (QRPs), including p-hacking (massaging data to yield statistically significant results), HARKing (hypothesizing after results are known), cherry-picking (selective outcome reporting), and fishing expeditions (data dredging). These bad research practices increase false-positive rates and jeopardize the credibility of scientific findings, leading to one of science's worst failings: the replication crisis. Radical reform in how science is conducted and published, including preregistration, redefining statistical thresholds, advancing research training, providing more room for exploratory research, and changing the wrong incentives in scientific practice, may partially solve the problem of blending and its associated QRPs. In conclusion, the distinction between exploratory and confirmatory research is crucial for better bench-to-bed translation of preclinical evidence. Tentative exploratory results must be confirmed before establishing clinical utility.