It is a very old question with no simple answer. The context is everything. Is the food appropriate to cultural and nutritional needs? Is the food delivered efficiently? Will the food undermine the local economy? If cash (or vouchers for a cash-based alternative), is there anything in the markets to buy and if there is will the injection of demand simply raise prices leading to the traders being the major beneficiaries? Will cash empower men and disempower women? Will cash lead to more conflict? Which is the more politically palatable from the donor perspective?
This issue most recently came up in an interesting exchange between Charles Kenny of CGD and Bill O'Keefe of Catholic Relief Services in Foreign Policy. O'Keefe responded to an article by Kenny entitled "Haiti Doesn't Need Your Old T Shirt" which makes the argument that cash is what is needed in many of humanitarian contexts. O'Keefe provided a defence of the US Food Aid system, saying that it has fed millions of people and that cash is not always better, especially when markets are not functioning, to which Kenny replied, yes it does good work, but it is expensive and inefficient, being tied to US agricultural and shipping suppliers.
A lot has been written about this cash vs food issue, but rarely comparing cash and food within the same programme.
Most recently from IFPRI we have a 2009 report by Akhter Ahmed (pdf) and others on Bangladesh in the context of a number of transfer programmes in Bangladesh. They explored four transfer programmes in Bangladesh, two of which provide participants with mixes of cash and food. For these two programmes they don't attempt to isolate the relative impacts of food and cash on outcomes, perhaps because it is too difficult to get a clean "identification" of the "treatment" (what is driving the receipt of cash vs food) or perhaps because it is too difficult to match time periods with receipt of each type of transfer. (Interestingly they asked participants what they would prefer and the poorer households preferred food.)
IFPRI is currently engaged with WFP in a 5 country study where cash vs food (and various combinations) will be randomly allocated to communities with a range of impacts being assessed.
It will be interesting to finally get a clean assessment of what cash and food are best for, at least in those contexts. The real issue will be how relevant are these internally valid impacts for other contexts--in other words, how strong is the external validity? For this they will need to conduct more qualitative studies to unlock the black box.