In my experience, there has been no instance where this has been more true than in the implementation of statistics in research.
I've always been keen on math, briefly considering a career as a high school math teacher. I really understood math, and appreciated that their was rarely an ambiguity in what was correct. When I received the assignment to be a teaching assistant for research methods and statistics in psychology my second year of graduate school, I was absolutely thrilled.
My enthusiasm about using statistics in my professional field was much more subdued, however. During my first year of graduate school, I took the required statistical courses and was actually delighted at how difficult they were. The following year when I began to write up my thesis project, my implementation of statistics as I learned in my statistician-lead class was chastised. In short, my mentor reminded me of the grey bridge, and insisted that I adopt the statistical methods accepted within our subdiscipline of neuroscience.
In fact, an article passed around Twitter yesterday illustrated how statistics can be - and often is - erroneously used in research (http://www.sandernieuwenhuis.nl/pdfs/NieuwenhuisEtAl_NN_Perspective.pdf). And this only details one way in which statistics are incorrectly employed! In truth, the interest of most (neuro)scientists lie in the scientific variables relevant to their field, and not the statistics that ultimately assess their success. As a (neuro)scientist myself, I both completely understand and find the situation reprehensible.
This is a topic of both incredible breadth and interest to me, so expect to hear plenty on the topic in posts to come.
