Chemistry does not need much of it (depends on the particular issue you are studying, though). For me, it is either the reaction works or not. If I can isolate or identify the product I was looking for, it works, If I can't it doesn't work. Medicinal Chemistry is something more complicated, since there are biological tests and you can have some variability in the data, but at the end you focus on the media and you only look at the standard deviation when the activity of the compound under testing is not so great and you are wondering if the activity data is important enough to be taken into account. Which brings the question: Is statistics only important when the phenomena under study is too weak and it is difficult to tell whether the cause under study is the responsible of the studied effects? Let me give you a couple of examples taken from off-topic threads. * Second-hand smoking. VRVCAgent opened this thread with the following numbers: Those exposed to second-hand smoke: 12.5 deaths per 1,000,000 Those not exposed to second-hand smoke: 10 deaths per 1,000,000 So, some people (CarlosMM and others) say that there is a 25 % increment, so it is highly significant. Others, like Nanocyborgasm or ChrTh, say that the increment is not significant. I don't know what the standard deviation is in this particular example, and with those small numbers (2.5 deaths in one million), I tend to think that is not a big issue, but I am not sure. What do you guys think about that? <EDIT> (I recall reading somewhere that The Lancet does not publish anything in statistical medicine with less than a two fold increase i don't know if that is true, Does anybody know if it is true?) </EDIT> * The whole global warming thing, with fractions of degrees of change and pretty big standard deviations, plus a lot of possible causes and positive and negative feedbacks between causes and effects. I am not sure about that either. Plus, the 100 year climate computer simulations that some people take as proven facts. That would make me laugh if they weren't using the simulations to dictate policies.