Abstract
The degree of ecological specialization plays a crucial role in shaping the structure and functioning of communities. However, comparing specialization within and among groups of organisms is complicated by both methodological issues and conceptual and terminological inconsistencies. Environmental predictability has been considered a key determinant of specialization though empirical evidence is still limited. Fungi and their insect consumers provide a poorly studied but promising system to measure host specialization and test the predictability hypothesis. In this study, we systematically sampled mushrooms in North European boreal forest, and reared total samples of fungivores colonizing the fruitbodies. Due to the unpredictable nature of mushrooms as a resource, low levels of host specialization can be predicted for these insects, which have indeed widely been considered polyphagous. Contrary to expectations, the majority of the studied fungus gnats were found not to exploit their host taxa indiscriminately. Not only were some mushroom taxa never colonized, the infestation rate of acceptable hosts also differed in most of these fungivores. Gnat species themselves formed continua with respect to the estimates of the degree of specialization, derived from parametric individual‐based analyses of presence‐absence data. In most cases, host use was best explained by models in which the hosts were classified at genus level, with limited support to specialization to particular host species, families, or orders. Indeed, most of the common fungivores appeared to preferentially use various species from one or a few mushroom genera while occasionally feeding on members of other host taxa. This pattern has likely evolved as a compromise between selective forces stemming from host unpredictability and taxon‐specific chemical profiles of the mushrooms. Our study highlights the multidimensional nature of ecological specialization: a high number of acceptable hosts does not preclude considerable discrimination among members of the available resource pool. Such situations can only be revealed by individual‐based analyses capable of capturing differences in partner‐to‐partner interaction intensities.