There's an interesting post at Dynamics of Cats about the difficult task of detecting the signature of life on extra-solar planets:
This is a hard question: for Earth-like carbon base, photosynthesising life on a low mass planet around a main sequence star, we think we have robust biosignatures - we want to see water vapour, oxygen (or ozone as a more easily detected proxy), carbon dioxide and methane.
However, as the post goes on to point out, early life on earth did not generate this signature, with significant atmospheric oxygen possibly only being present from about 1.5 billion years ago (which leaves a good 2-3 billion years with a different, and poorly constrained, atmospheric chemisty). And that's before you consider non-carbon based life.
The obvious limitation on detecting life outside the solar system is that spectroscopy is pretty much our only tool - the hope is that whatever life is out there will generate a 'weird' (out of equilibrium) atmosphere which we can actually detect. But astrobiologists don't need to look so far before they encounter problems. There's still some argument over the results of the Viking Lander experiments, for example (see here or here), not to mention the whole ALH84001 thing. In both cases the cautious consensus is that there is no conclusive evidence of Martian life, but sorting out life from non-life is clearly not trivial even when we can literally get our hands dirty in the search; no wonder trying to make the determination from light years away is 'hard'.