We tested the hypothesis that leaves of broad-leaved tree species are more highly defended at low latitudes than at high latitudes. We used canonical discriminant analysis to compare tree species from Panama (9°N, 39 species), Missouri, USA (38°N, 37 species), and southern Ontario, Canada (44°N, 34 species) with respect to two structural and five nutritional traits, taking into account each species' tolerance to shade. Trees from the three locations differed significantly, with Panamanian species the most distinct. Defenses of shade-tolerant species were significantly greater than those of shade-intolerant species, but only for the Panamanian sample, which is consistent with the low latitude/high defense hypothesis. Because we sampled many of the same tree species from Missouri and southern Ontario, and many tree species in the same taxonomic families in Missouri and Panama, we were able to control for the potential confounding effects of phylogeny. Overall defense levels, calculated by summing the z-scores for individual traits in each location, were significantly higher for Panama compared to Missouri, and marginally so for Missouri compared to southern Ontario, again consistent with the low latitude/high defense hypothesis. Traits contributing to these differences were mostly structural factors (e.g., fiber) and to a lesser degree nutritional traits, while secondary compounds made no independent contribution to differences in overall defense levels (four traits compared between Panama and Missouri). Contrary to our expectation, the number and types of secondary compounds per species reported in the literature for our species did not differ between temperate and tropical locations, while the diversity of these compounds was greater for the temperate species. Overall, our results provide some support for the hypothesis that leaf defenses against herbivory are better developed in tropical than in temperate trees, but the differences were due to structural and nutritional factors rather than secondary compounds.