A distant topic, yet not distant at all. Science, as a process, creates trustworthy knowledge; it has proven this over and over, and much of our economy is based upon it. But scientists? As individuals, they are just as human as anyone else: they make mistakes, harbor biases, believe things without much evidence, overstate their cases, and do all sorts of problematic things. Some are a lot more reliable than others, and often the unreliable ones are the ones who take their scientific case to the media (the reliable ones often not seeing the need to do so). Consequently, media stories on science are on average somewhat less reliable than the ones scientists produce and read.
So when a brand new science story makes the headlines — say, the possibility of faster-than-light neutrinos from the OPERA experiment at the Grand Sasso laboratory (withdrawn before publication due to discovery of a technical problem), or the bacteria substituting arsenic for phosphorus as claimed by a group of scientists who first predicted the possibility and then claimed to find it (here’s the abstract of the published paper in Science magazine, and a popular New York Times article about it) — caution is advised. [Cold fusion, anyone? Utah’s governing institutions should perhaps have been a bit more patient…]
Multiple arguments against the arsenic-incorporating bacterium result have been made from the beginning. Now two new papers (#1, #2) have hit Science magazine, and the press. Both argue (among other things) that the bacterium in question is still dependent on phosphorous and that there is no sign of arsenic in its DNA.
Which is all to say that reproducibility lies at the core of scientific research. It is the process of confirmation and cross-checking of a result that allows a media story that begins “Scientists say…” to be transformed into a textbook that begins “Science says…” When you read a story in the press, you should always look for the key indications that multiple and separate groups of scientists agree on the result — not necessarily that there is broad consensus, but at least that the scientists making the claim aren’t entirely out on their own. [For the recent discovery of the Higgs particle, we have two different experiments making the claim.] If those indications are missing, I’d advise you to file the result under “unconfirmed,” and not to bet on it.