A good mechanic will often use the following reason: 90% of cars with symptom x problem Y, so that is what I will check first.

Then, he will discover whether Y is actually the problem (ETA: for this particular car), and if not, discard that hypothesis and look for something else. This essential step is missing from all papers in psychology reporting statistical results. The fault is even worse when those results are reported in terms such as (to take a recent example) "willpower is a scarce resource".