There's a brilliant article by Govert Schilling in April's 'Sky and Telescope' magazine on the prospects for MOND, a modified theory of gravity. MOND rejects the now commonly accepted postulate that most of the matter in the universe is composed of dark matter.
The orbital velocity of the stars in a spiral galaxy, expressed as a function of the radial distance from the centre of the galaxy, cannot be explained using the standard inverse-square law of gravity if one assumes that the distribution of matter coincides with the distribution of luminous matter. Dark matter accepts the inverse-square law, but proposes that there is extra non-luminous matter, of an as-yet unspecified constitution. Modified Newtonian Dynamics (MOND) proposes that the inverse square law of gravity becomes an inverse linear law over larger length scales. MOND has been very successful at explaining the rotation curves of spiral galaxies, but a cosmological version has yet to be developed. In contrast, dark matter provides a successful cosmological model as well as a galactic model.
An advocate of MOND, Stacy McGaugh makes the excellent point, "Suppose Newton, upon studying the solar system, would have formulated an inverse-cube law, claiming that the observed inverse-square law is the result of some mysterious, undetected dark matter. That would clearly be less satisfying than simply adopting the observed force law."
Amongst the advocates of dark matter quoted in the article is Joel Primack, who calls MOND "an irrelevant theory," and asserts that "If other cosmologists want to waste time on it, that's great - it means less competition for me." With respect to dark matter (and dark energy) cosmology, Primack claims that "We now have what appears to be a totally reliable theory. As far as I know, there is no single piece of contradicting evidence. It's a continuous success story."
I'm rather dubious about the claims made in the article by Primack, and also by David Spergel, that dark matter explains the large-scale distribution and clustering of galaxies. It seems to me that the dark matter (and dark energy) models simply have adjustable parameters built into them, which are tuned until agreement is obtained with observation. For example, consider this from an article on galaxy formation in the 10th March 2007 New Scientist:
[David Hogg's team] maintain that the cold dark matter model explains the Sloan [galaxy distribution and clustering] data quite accurately. For that to be true, however, Hogg's team have to put a number called a bias parameter into their equations. It reflects the difference between the distribution of matter in computer simulations of the cold dark matter model and the observed distribution of luminous matter…In the case of the Sloan survey, the bias is 2: the visible galaxies are clumped twice as densely as the predicted total distribution of matter in the universe. (p32-33)