The emergence of behavioural isolation between populations under divergent selection can be crucial for ecological speciation, but the mechanisms underlying such isolation are poorly understood. Several experimental evolution studies have shown that positive assortative mating (preference for similar mates) can arise rapidly in Drosophila laboratory populations reared in different stressful conditions, while other studies failed to confirm this effect. Here, we present the results of an evolution experiment in which outbred lines of Drosophila melanogaster were reared for 1-2years on one of the three different diets (standard, starch based or high salt). We show that nonrandom mating arose in some, but not all lines, and that the manifestations and possible interpretations of this nonrandomness depend strongly on the type of tests used to assess mating preferences. More specifically, multiple-choice four-fly tests revealed positive assortative mating (prevalence of homogamic matings) in some starch-adapted and salt-adapted lines when paired with a control line reared on the standard diet, but competitive three-fly tests rather revealed competitive advantage of control males and females over the flies reared on stressful diets. The results imply that divergent adaptation can result in differences in mating propensity or competitive ability, which, in turn, may either facilitate or hamper speciation depending on the relative frequency of high- vs. low-competition settings in natural habitats of the diverging populations. The results also emphasize the importance of using diverse tests for assessing mating structure in natural and laboratory populations.