According to some research handbooks, nothing goes wrong. Every study has neatly balanced samples, no pupil is ever awkward, or, for that matter, absent. Teachers obediently follow instructions. Statistical tests provide brilliantly clear conclusions.
The writers of these clinically clean, paragon tomes must never have ventured into a real classroom, dragged muttered responses syllable by syllable out of the adolescent acne brigade, been distracted by gruesome tales of a dead rabbit from some five-year-old with a very runny nose and a livid scab, or witnessed the terrifying sight of data disappearing for ever from a computer screen.
I've been conducting educational research for more than 30 years. Does long service confer immunity from pitfalls and hazards? Does it hell. It just makes you better prepared and fills your head with first-hand tips for the unwary. That said, I think all teachers should carry out research into their own school, class, or the teaching and learning in which they are involved. If keen GPs can monitor diabetics, investigate young patients with glandular fever, or systematically test new treatments, why can't equally curious teachers study their immediate world?
Research is a fundamental human and professional activity, despite the potential pratfalls. Anyway, we expect our pupils to do it as part of their lessons or coursework.
The insider story is fascinating. Sara Delamont of Cardiff University once told how she had to wear white gloves and carry a handbag while she was researching a fee-paying girls' school in Scotland, otherwise she would never have got past the ferocious headmistress.
My Exeter colleague Ron King was known as "The man in the Wendy house" when he carried out his seminal study of infant schools. As a strictly non-participant classroom observer, Ron had to escape the curious eyes of young children by hiding in the Wendy house, so he could scribble up his notes. If children were playing in it, he retreated to the gents' toilet instead. The staff thought he had a weak bladder.
I once supervised a teacher who said he wanted to do 800 interviews, as that was the number on roll in his school. What is more they would be "in depth". No one ever does shallow interviews, of course; all interviews claim to be "in depth". The first rule of bog-up avoidance is: pilot what you are going to do first. Even one "in depth" interview, administered, transcribed and analysed, would reveal how many lifetimes were needed to do 800. Only masochists sign up to the undeliverable.
Another huge error is to use leading questions, as these automatically discredit the answers. A few years ago, one pressure group came up with a blinder: "Is there too much sex and violence on television and what are you going to do about it?" Er, yes indeed. Not much doubt where you're coming from, sunshine.
Experimental research is notoriously hazardous in education. Most experiments work, because they are conducted by enthusiasts for the programme being tested, who unwittingly or intentionally manipulate the conditions. When psychology PhD students testing out behaviourist learning theory used to run rats through mazes, some of the more unscrupulous were observed giving them a prod: "Turn left, you little bastard - my thesis is at stake."
Carrying out a true experiment involves controlling the conditions under which the experimental and control groups operate. Some researchers have tried scripted lessons (you can write questions for teachers to ask but, sadly, not prescribe the answers), or have produced what they claim are "teacher-proof" materials. Some of these are about as teacher-proof as kerosene is fire-proof.
One experiment, where children filled in specially written story booklets designed to make them think more imaginatively, failed ignominiously. The teacher administering the supposedly teacher-proof programme was Mr Dreary, who constantly told the class how bored they looked. The control group was given to Mr Fizzbang, who had just been on a creative drama course, to do what he liked with. When the post-tests were given, the control group won hands down.
The biggest minefield can be interviewing the family Frankenstein at home.
Poisonous snacks, scalding or freezing tea, long-winded interviewees, are the minor problems. Ferocious or - even worse - amorous dogs; unruly toddlers smearing half-chewed lollies over you; granddads smoking what smells like, and probably is, shredded horse manure, are a much greater health hazard. "Yes, we do read with our child at home - when we're not actually killing each other."
Finally there is writer's block. The data analysed, the computer screen clear. No more wretched hourglass and meaningless triangular warnings about memory, illegal acts and insufficient flobbabytes. Now you just have to write up the results.
Hmmmmm. One more cup of coffee? One more game of solitaire?
Bugger it. I'll do it tomorrow.