Geoff Pullum's Five Golden Rules

1. DON'T EVER BEGIN WITH AN APOLOGY. Everyone has seen speakers beginning a presentation by apologizing for how unworthy they are, how little of their work is really conclusive, how they hope people will forgive them and so on. No one has ever seen a case in which this improved the reception of the paper or the mood of the audience. If you're going to be bad, they won't be pleased that they showed up, and if you're not then you are just wasting air time. Pieter Seuren has pointed out to me that the tradition of beginning with an apology is so old that it has a name in Medieval rhetoric: it is called the captatio benevolentiae, the capturing of the audience's good will. My point is that an apology simply doesn't work as advertised. Opening up with an apology is like trying to teach a pig to sing: it wastes your time and annoys the pig. Don't ever do it.
2. DON'T EVER UNDERESTIMATE THE AUDIENCE'S INTELLIGENCE. Few sins are worse than making the audience think you think they are stupid. An audience who sees a presentation somewhat too high-powered for them may still grasp some of it, and at the very least its members will feel that they have been flattered with the assumption that they are smart. But the members of an audience who hear a talk pitched too low for them have both wasted an hour and been treated as if they were dumb. It truly adds insult to injury. So while you should always worry that perhaps you are being confusing, you should worry somewhat less about whether what you are saying is difficult. There are many worse things than a difficult and demanding lecture, and a patronizing and superficial lecture is one of them.

3. RESPECT THE TIME LIMITS. It is sad to be cut off when you are just about to make your major point. Or even a minor one. Plan your time, and don't let it happen. The mood of the audience is not going to improve from seeing someone ramble on when they should have been stopped by now so that questions can begin. A good chair will stop you dead at the agreed time, but don't wait for that: wrap up before the chairperson has to stand up (or the students who are late for their next class have to get up and leave).
4. DON'T SURVEY THE WHOLE DAMN FIELD. You need to make a few assumptions clear before you get going on your main point, but you don't need to begin by summarizing the whole prior content of the discipline, explaining what grammars are, what phonemes are, etc etc. Even in a job talk, where giving your whole dissertation in 55 minutes is the awful temptation, don't do it. Assume a reasonable amount of background, and then present something that can be delivered in a reasonable amount of time. A good rule of thumb if using transparencies is that each one should be up there for three minutes, or at the very least two. Treating each display on a handout as the equivalent of one transparency gives you a rule of thumb for handouts, too.
5. REMEMBER THAT YOU'RE AN ADVOCATE, NOT THE DEFENDANT. It's your idea that's being presented, not you. The reason for not feeling nervous is that you are not what's up for consideration (not even at a job talk; they consider you later!). This isn't about you (that's why you shouldn't begin with an apology: that's about how you feel). It's the ideas that are going to get scrutiny. If those ideas don't survive after today, too bad for them. You can't work miracles. But for today, you're there to do as fair a job as you can for them during their twenty minutes in the spotlight. You're a vehicle, an advocate, a public defender. These ideas might have been unfairly dismissed without a trial. No matter what the ultimate verdict, you will have served the court of scholarly opinion if you defend them effectively.

Finally, though this concerns not the talk but the questions afterward, the sixth of my five rules: during the question session EXPECT QUESTIONS THAT WILL FLOOR YOU. You'd better hope some of the questions to be hard ones. If the combined wits and backgrounds of the audience can't yield a question that really gives you some trouble, or can't come up with any questions at all, you should feel mildly annoyed; they really can't have been seriously thinking about what you said. Or else you're giving talks at events that are way beneath you. It's actually a bit sad to give a presentation so perfect that it leaves no crevice for the critical knife, so that the question period is an embarassing two minutes of silence. It's as if the talk had died. And since it is no great shame to be temporarily flummoxed, it's better if things go in that direction. Listen closely to the question, think, and if it's a great question you had never considered before and you don't know the answer, simply say, "That's a great question that I had never considered before. I don't know the answer."