Tree diversity is one of the drivers of forest resistance to herbivores. Most of the current understanding of the diversity resistance relationship comes primarily from short-term studies. Knowing whether tree diversity effects on herbivores are maintained over time is important for perennial ecosystems like forests. We addressed the temporal dynamics of the diversity resistance relationship by conducting a 6-year survey of pine attacks by the pine processionary moth Thaumetopoea pityocampa (PPM) in a tree diversity experiment where we could disentangle tree composition from host density effects. During the first years after planting the trees, PPM attacks on maritime pine Pinus pinaster were reduced in the presence of birch Betula pendula, a fast-growing non-host tree (i.e. associational resistance). This effect was maintained but faded with time as the pines eventually grew taller than neighbouring birches. The number of repeated attacks on individual pine trees also decreased in mixed pine-birch stands. Pine density had a positive effect on stand colonisation by PPM and a negative effect on the proportion of trees that were attacked. Pines were less likely to be repeatedly attacked as pine density increased, with attacks being spread over a larger number of host trees. Collectively, these results unravel the independent contribution of tree species composition and host density to tree resistance to herbivores. Both processes had directional changes over time. These results indicate that the resistance of planted forests to herbivores can be improved by carefully choosing the composition of mixed forests and the timing of species planting.