I can't say I have taken part in anything quite as interesting as the ketamine experiment mentioned in Comment is free's recent open thread, but some have been hair-raising. I once had to stay on a hospital bed for 10 hours with three IVs in my right arm and two IVs in my left, one of which was used to take regular blood samples. The experiment involved taking a chemical that suppressed some liver functions, and left me showing mild diabetic symptoms for a week afterwards – this annoyed me greatly as I thought the experiment had paid very little for rather a lot of discomfort.
The best and easiest experiments were psychophysiological ones, as I was paid quite a bit for just 90 minutes' attendance. One involved lying on a comfortable recliner chair in a soundproofed room, in which I had to concentrate on not falling asleep. If I did, the researchers would yell through an intercom; they could easily see when a subject was drifting away by the constant EEG recording. That experiment had only one drawback, which was the sounding of single short and very loud noises at random times. The test was about shock reactions and, yes, a nasty and savagely loud noise heard at random can be very shocking indeed.
Once I had got much further in university, myself and others started running experiments ourselves. A friend of mine ran psychological experiments measuring emotional reactions: for the pleasant condition, he used holiday snaps he had taken himself, and for the shock condition, he used a set of slides he had got from the local police – photographs of murder victims and scenes in that city. There is a real immediacy about photos of genuine murder scenes that leaves Hollywood in the shade and is electrifyingly unpleasant.
Murder is sometimes part of the picture, too. Apart from the infamous Tuskegee syphilis study (an observational study of the long-term course of untreated syphilis), there was also the abhorrent experiments run by US doctors in Guatemala in the 1940s, where subjects were deliberately infected with syphilis without consent. In the same vein, novelist John le Carré wrote a thriller, The Constant Gardener, based on rumours of underhand pharmaceutical experimentation in Africa. And of course, everyone also knows about the Nazi and Japanese medical experiments run on prisoners during the second world war period, and the Soviet Union experiments in alleged "psychiatry" run on captive political dissidents. Clinical trial can also unexpectedly develop into lethality: back in 2006 six volunteers for a medical trial in London became life-threateningly ill.
Should such experimentation be carried out on humans? At some stage it must be, because computer and animal models are simply not enough for final testing of drugs designed for humans. We are always experimenting on ourselves: possibly the largest group of self-experimenters are those who are chronically ill or suffer from chronic pain, most of whom are only too willing to try out something new (in private, non-scientific ways) to see if it can alleviate their suffering.
Should volunteers be paid for taking part in such experiments, thereby encouraging what may be seen as vulnerable segments of the population to undertake risk? Not paying volunteers would be the far worse option; it's difficult enough to get volunteers without refusing them recompense for their time, discomfort and risk.
While medical researchers often experiment on themselves too, for true results you need quite large groups of subjects. I believe that university students being the favoured group for such processes is a lot better than using villagers in Africa: the students are western consumers who benefit most from new pharmaceuticals, and it seems fair that they should be the experimental test rabbits – plus the income is often of great help to them. They are also the ones most able to complain loudly and get proper treatment and compensation should anything go wrong. And, well, it gives them something to talk about at parties.
All in all, it's most often a win-win solution – except when things go horribly wrong.