The principle has finally been accepted that, whenever possible, antimalarial drugs should be deployed in appropriate combinations in endemic areas, in order to minimize the inevitability that monotherapy will, probably sooner than later, select populations of drug-resistant parasites. Which laboratory models can predict the combinations of old or novel compounds that are likely to be of practical value in minimizing this risk? Very few relevant data on the use of Plasmodium falciparum in vitro have been published. Most research has been carried out with one or other strain of chloroquine-sensitive P. berghei or with chloroquine-resistant P. yoelii ssp. NS in mice. The two most widely used procedures to select for resistance are the 'serial technique' (ST), in which drug selection pressure in vivo is gradually increased, and the '2% relapse technique' (2%RT), in which a single, large drug dose is applied at the time of each passage. Both procedures have been used to demonstrate the ability of pairs of drugs (e.g. sulfadoxine with pyrimethamine, cycloguanil with dapsone, pyrimethamine or sulphonamides with chloroquine, mepacrine or mefloquine) or triple combinations (e.g. sulfadoxine-pyrimethamine with chloroquine, mefloquine or pyronaridine) to delay the development of resistance. The relative merits of the ST and 2%RT are discussed and the data obtained by these procedures are compared with the results of deploying drug combinations in man, especially against multidrug-resistant P. falciparum. It is concluded that, even though the rodent malaria models are imperfect, no better alternatives are available at present with which to predict the value of antimalarial combinations for the protection of the individual components. The 2%RT is considered to be the procedure of first choice.