Print issue
Current | Past
Experiments that do not arrive at their expected conclusion are more frequent than previously thought, and disregarding them can generate biases, study shows