Only few factors influencing pest populations can be studied in the laboratory, but many populationdriving factors interact in the field. Therefore, complementary laboratory and field approaches are required for reliable predictions of real-world patterns and processes. Laboratory and field
experiments with the red spider mite, Oligonychus ilicis McGregor (Acari: Tetranychidae), and the coffee leaf miner, Leucoptera coffeella Guérin-Méneville (Lepidoptera: Lyonetiidae), on coffee plants, Coffea arabica L. (Rubiaceae), were combined to study the relative importance of biotic interactions, including resource preferences and natural-enemy impact, and habitat factors, such as agroforestry type and management intensity, on coffee pest densities. In the laboratory, leaf discs cut from undamaged coffee plants were significantly preferred by red spider mites over those from plants infested with
conspecific mites, leaf rust pathogens [Hemileia vastatrix Berkeley & Broome (Uredinales)], or coffee leaf miners, resulting in higher reproductive success. Similarly, undamaged plants were preferred by coffee leaf miners over red spider mite-infested plants. However, in the field, red spider mite densities
were positively correlated with coffee leaf miner and leaf rust densities, thereby contrasting with laboratory predictions. Hence, our study suggests that the importance of resource preferences and fitness expected based on laboratory experiments was suppressed by environmental conditions in the field, though other unassessed biotic interactions could also have played a role. Furthermore,
intensified agroforestry was characterized by higher red spider mite densities, whereas densities of its major natural enemy, the predatory mite Amblyseius herbicolus Chant (Acari: Phytoseiidae), were not related to agroforestry management. Densities of coffee leaf miner and its main natural enemy,
a eulophid parasitoid (Hymenoptera), were not affected by management practices. In conclusion, patterns found in the laboratory did not hold for the field, emphasizing the difficulties of extrapolating small-scale experiments to larger spatial scales and the need to combine both approaches.