The error that influenced me most was my graduate thesis, "The Habituation of the Peripheral Vasoconstrictive Response to Noise" - basically seeing how skin bloodflow, which is responsive to noise and is a measure of stress, gets accustomed to hearing it at intervals over 5 days. I had to recruit fellow students as subjects, have them come and sit, listening to noise, in a tiny sound chamber while the blood flow in a finger was measured (non-invasively), for half an hour a day over five consecutive days... I never realised how hard it would be just to get that commitment from a few people.
When all the testing was finished, and I'd spent days analysing the data, it was all over the place - seemingly completely random. It wasn't until I read up in detail about what I'd been measuring, that I discovered that it wasn't just sensitive to noise, it was even more sensitive to alcohol, nicotine, caffiene, exercise, yawning, stress, quality of sleep, and a host of other influences - and I'd been measuring it on students, whose activities (in those days) were organized around all those things, more often than not to excess...
I had no sensible data for my write-up, so I switched it to a critique of all the things that went wrong with my 'research', and how they could be fixed. My naive idea of making some useful contribution to the field ended with me explaining why it had crashed and burned.
In the end, the thesis was well received, and I got good marks (apparently everyone else on the course had turned in rather unlikely impeccable research results, confirming current understanding), but it was a salutary lesson for a young would-be scientist - preparation is the most important and most difficult part of the work.
