In fragmented landscapes the relationship between the probability of occurrence of single species and the amount of suitable habitat is usually not proportional, with a threshold habitat level below which the population becomes extinct. Ecological theory predicts that, although the reduction in species’ occurrence probabilities (and eventually the extinction threshold) is a direct consequence of habitat loss, habitat fragmentation might reduce species’ occurrence probabilities and affect the location of this threshold by reducing its predicted occurrence to lower levels of habitat amount. However, little is known about the validity of this extinction threshold hypothesis. Here, we performed analyses on the relationships between the probability of occurrence of eight tree species and the availability of forest habitat for two different empirical scenarios of low and moderate to high fragmentation. We partitioned the effects of habitat amount versus fragmentation by using two metrics: (1) the percentage of forest cover, and (2) the proportion of this percentage occurring in the largest forest patch. We find that, although decreasing forest cover had negative effects on the occurrence of tree species irrespective of fragmentation levels, forest fragmentation significantly modified the response pattern in six tree species, although only one species confirmed the extinction threshold hypothesis, which we interpret as a consequence of high degree of forest specialism and low dispersal ability. For most species, fragmentation either had positive effects or did not affect significantly the species’ probability of occurrence. This indicates that the effects of habitat fragmentation on tree species are weak relative to the effects of habitat amount, which is the main determinant of the reduction in species’ occurrence probabilities, and eventually species extinction, in fragmented landscapes.