Past attacks in¯uence host selection by the solitary bark beetle Dendroctonus micans MARIUS GILBERT, 1,3 GE  RARD VOULAND 2 and JEAN-CLAUDE GRE  GOIRE 1,4 1 Laboratoire de Biologie animale et cellulaire, Universite  Libre de Bruxelles, Belgium, 2 Institut National de la Recherche Agronomique, Laboratoire de Recherches Forestie Áres Me Âditerrane Âennes, Avignon, France, 3 Fonds pour la Formation a Á la Recherche dans l'Industrie et dans l'Agriculture, Brussels, Belgium and 4 Fonds National de la Recherche Scienti®que, Brussels, Belgium Abstract. 1. A spatio-temporal study of host selection and local spread of a solitary bark beetle attacking live spruce Dendroctonus micans (Kugelann) was carried out using a combination of standard statistical methods, geostatistical analyses, and modelling. The study was based on data from three plots (150±300 trees, 0.3±1 ha) from 1978 to 1993. All trees were mapped and successful and abortive bark-beetle attacks on each tree were counted annually. Because the attacked trees usually survived, temporal attack patterns as well as spatial patterns could be analysed. 2. The distribution of successful insect attacks on the trees was slightly aggregative, indicating some degree of choice rather than totally random establishment. 3. The level of yearly individual attacks per tree was very stable, suggesting that D. micans usually leave the host in which they develop. 4. The attacked trees were distributed randomly in the plots; at the study's spatial scale, the insects dispersed freely throughout the plot (no spatial dependence). 5. On the other hand, time dependence was strong; some trees were attacked repeatedly while others were left untouched. 6. Among a choice of scenarios (random attack, ®xed variability in individual host susceptibility, induced host susceptibility following random attack), the best ®t was obtained with the model involving induced individual host susceptibility. This type of relation to the host tree contrasts strongly with patterns generally described in host±plant relationships (including gregarious, tree-killing bark beetles), where local herbivore damage results in induced resistance. 7. These results suggest that the ®rst attacks in a new stand are made at random, that all or most of the beetles emerging from a tree disperse and resample the stand, and that they settle preferentially on trees that were colonised successfully by previous generations. Key words. Dendroctonus micans, dispersal, geostatistics, insect host±tree relationship, Scolytidae, tree susceptibility. Introduction For most primary bark-beetle species, successful establishment on a host tree requires pheromone-mediated mass attack and inoculation of fungal pathogens to overcome the tree's defences (Coulson, 1979; Wood, 1982). Convergent evidence suggests that the pioneer beetles responsible for initiating mass attack land on the trees at random and subsequently make their choice at very close range using olfactory or gustatory clues (Raffa & Berryman, 1980; Moeck et al., 1981). Contrasting with these species, Dendroctonus micans (Kugelann) is a solitary coloniser of live spruce, which is not Correspondence: M. Gilbert, Laboratoire de Biologie animale et cellulaire, CP 160/12, Universite  Libre de Bruxelles, 50 av F. D. Roosevelt, B-1050 Brussels, Belgium. E-mail: mgilbert@ulb.ac.be # 2001 Blackwell Science Ltd 133 Ecological Entomology (2001) 26, 133±142 Ecological Entomology (2001) 26, 133±142