Identifying factors that influence variation in light availability within forested ecosystems represents an important component in our understanding of the complex determinants of tree seedling regeneration. We assessed the influence of forest structure and canopy tree architecture on spatial heterogeneity of understory light availability in three old-growth and three second-growth forests in lowland Costa Rica. Forest structure and understory light availability were measured within forest types using contiguous 10 × 10 m quadrats along three 130-160 m transects in each stand. Two 20 × 60 m plots in each forest type were sampled more intensively, including vertical profiles of light availability from 1 to 9 m height. Mean diffuse light transmittance increased from 2% at 1 m height to over 10% at 9 m height and did not differ significantly between forest types at any height. However, the relationships among height classes differed between forest types. Second-growth plots showed a negative spatial autocorrelation for light measurements separated by vertical distances over 4 m Differences in the vertical distribution of light and foliage suggest that old-growth and second-growth stands differ in vertical organization of the vegetation. The most pronounced structural differences between forest types were found in trees between 10 and 25 cm in diameter at breast height (dbh). In second-growth stands, trees in the 10-25 cm dbh size class were more abundant and differed in allometry. They were taller for a given stem diameter and had narrower crowns for a given height than old-growth trees. Within forest types, we did not find strong relationships between measures of forest structure and light availability, although the strength of these relationships differed between forest types. In both old- and second-growth forest, understory light availability at 0.75 m decreased with increased sapling and shrub density, but was not significantly influenced by local tree density or basal area. From 1-m to 9-m heights, tree density was a significant, but weak, predictor of light availability in old-growth plots. In second-growth plots, tree density showed little or no influence on light availability at heights below 9 m. Our findings challenge the view that, within a forest, canopy and subcanopy vegetation directly influence light transmittance near the forest floor. Instead, we argue that spatial patterning of the light environment occurs through complex interactions among canopy, subcanopy, and understory vegetation.