Nature Physics, in its June Editorial (4, 429; 2008), adds its own advice to recent articles on presenting talks, summarized and referenced here. From the Editorial:

Presenting your research to an auditorium of peers can be a daunting prospect, particularly for those at the start of their careers. But with a little thought and preparation, it needn’t be.

We editors go to a lot of meetings, and have listened to a lot of talks. To hear a good talk can give you a reason for being. To hear a bad talk can make you wish you’d never left your hotel room. But even if your results won’t earn you a trip to Stockholm (yet), there is no reason they shouldn’t be the seeds for a lively discussion. And lively discussion is what it’s all about. We’ve put together a collection of do’s and don’ts to delivering a talk that will move your audience, not put them to sleep.

Once you’ve cut, tightened and improved the content, deliver it again, listen to it again, and fix any remaining weaknesses, again. Then again. And again. Until you are so familiar with its structure and content that you could give it in your sleep. Familiarity fosters confidence, and a confident talk is a compelling talk.

As it is with writing papers, so it is with giving conference talks — if your research is worth being presented to your colleagues, it’s worth being presented to them well.

Read the rest of the Editorial at Nature Physics.