OP, you opened one hell of a can of worms!
There's a number of things being said in this thread, the mainstream, and in the physics community, many of which are false, debatable, or misleading regarding this issue. The short answer to your question is that no one knows. This has been one of if not the largest open problems with non-relativistic quantum mechanics since its inception:
http://en.wikipedia.org/wiki/Measurement_problem
Now for a longer answer:
The person that has come closest to saying the right thing was StevieTNZ:
A quantum system interacting with another quantum system cannot collapse anything.
Now this is wrong to the extent to which everything is a quantum system in some sense (if we believe QM is 100% correct), and we indeed "observe" wavefunction collapse in the lab (in whatever sense), so all we have are interacting quantum systems and we know they collapse. However, this is right in some sense as strictly speaking quantum mechanics offers no method for wavefunction collapse without introducing the measurement postulate (which is extremely ad hoc).
Strictly speaking, non-relativistic quantum mechanics says that the state of a quantum system unitarily evolves according to Schrödinger's equation with the the proper hamiltonian for the system. This only applies to a closed quantum system, so in the case of measurement you would have to look at the state space formed by the tensor product of the measurement device's state space and that of the system being measured. In this context, the state will
deterministically evolve in time by rotating around in the state space according to Schrödinger's equation.
According to this, measurement devices will just get entangled with the state of the system being measured, as will everything else that interacts with it. In short, you can treat the entire universe as a closed quantum system (if you assume that there is nothing outside of our universe that pokes in and interacts in a non-quantum way) in which case the state of the universe should continuously and deterministically evolve, and there should be no collapses.
The issue, of course, is that in practice we don't see this happen when we introduce large enough systems like measurement devices which are composed of tons of individual particles. Instead, we can "make measurements" that cause a collapse. Not only is it contradictory that this is a discontinous collapse, but moreover its
random! So how can quantum mechanics, a deterministic theory at heart, cause random results!?
No one has come up with a universally satisfying answer, and every approach seems to have some flaws or oversights. Some answers are that the universe isn't closed, in which case there would have to be a non-quantum external interaction (otherwise you could expand your system to include the post-universe). Another answer is the many worlds interpretation which says that there is no collapse and that the universe remains in a superposition of every possible outcome (not quite the physical splitting into parallel universes that is popularized), although this still doesn't really explain how our observation chooses just one element of the superposition to see (amongst other issues with the interpretation).
There is also the concept of decoherence which does away with collapse and instead predicts a continuous process by which states apparently collapse and superpositions decay into mixtures. This is more or less inherent in QM already if you accept the fact that systems are getting entangled with the environment and that when we make observations we are looking at a marginal state (i.e. a subsystem that doesn't include every single particle that interacted with the system). But, this doesn't solve the actual problem of what happens when we observe a system and how we actually see randomness. In fact, decoherence doesn't even really
try to explain measurement in the first place.The whole "consciousness causes wavefunction collapse" has a huge list of issues. First of all, if we are quantum systems then consciousness is explainable by QM and so it doesn't solve the problem. So unless consciousness is post-quantum (and perhaps contains ways of the universe interacting with something external ;) ) then it doesn't solve the problem. The idea of the consciousness being post-quantum isn't totally farfetched even if it is not external to our universe, because in reality QM is incomplete since it doesn't take relativity into account. Moreover, relativistic quantum field theory(ies) makes the wavefunction a much less clear concept, and so far can only solve a number of pretty specific problems. Furthermore, to date it doesn't explain gravity so it too is an incomplete theory. Even in light of that, it isn't a theory of general wavefunction dynamics in the same way that non-relativistic quantum mechanics is. Not to mention that it has a number of non-understood issues of its own.
The biggest issue of all with the arguments for and against consciousness-caused collapse is that we can't even define consciousness! We can't explain it biologically let alone physically. We would need a definition/explanation in terms of physical interactions and processes before we could even attempt to involve it in any theory of physics. So it can't even be discussed meaningfully (yet).
The most likely scenario, I believe, is that we will need a new theory of physics to answer this question, because it doesn't appear to be answerable in quantum mechanics.
and it might even be possible to provide a full quantum description of it without invoking the mystical "observation" (one can always hope). 
