A two-year field study was conducted to evaluate the potential of two cactus species,  Echinopsis chilensis and Eulychnia acida, to evolve defensive traits against the parasitic mistletoe Tristerix aphyllus (Loranthaceae). The adaptive value of host traits against parasitism was inferred through: (1) identification of the relevant characters of cacti to prevent infection(2) evaluation of the fitness impact of parasitism on cacti, and (3) estimation of the linear and nonlinear selection coefficients on the relevant characters. Different lines of experimental and correlative evidence indicated that spine length was important in preventing individuals of the two cactus species from becoming parasitized. However, the impact of the mistletoe on cactus fecundity was contingent on the species involved. Even though parasitism decreased fruit production, seed number per fruit, and the total seed output in E. chilensis, low and nonsignificant linear and nonlinear gradients of selection were prevalent in this species, indicating absence of directional and stabilizing/disruptive selection for spine length. Additional analysis based on logistic regression however, revealed that long-spined E. chilensis had a higher probability of reproduction than did short-spined individuals. Unlike its effect on E. chilensis, the mistletoe had no fitness impact on E. acida, and the maintenance and evolution of spines in this species could not be attributed to parasite-mediated selection. Even though spines act as a first line of defense against parasitism in the two cactus species, selection was detected only on E. chilensis. These results indicate that inferences on the adaptive value of host traits based only upon their role in preventing infection run the risk of overestimating parasite-mediated selection, and thus the potential for host-parasite coevolution.