Late-successional riparian forests regulate autotrophic production in low order streams through shading provided by adjacent and over-hanging canopies. However, few studies have directly linked forest structure with periphyton biomass in adjoining stream reaches. Our hypotheses were: 1) heterogeneous horizontal structure in old-growth forests creates a more spatially variable below-canopy light environment over streams compared to mature forests; and 2) variability in light availability over streams correlates with the spatial distribution of periphyton biomass. We surveyed 15 low-order stream reaches in old-growth and mature northern hardwood-hemlock forests of the Adirondack Mountains, New York, U.S.A.. Riparian forest structure was inventoried within 6-10 sampling plots per site.  Below-canopy light environment was measured every 25 m over the center of neighboring stream channels using hemispherical canopy photography. We measured the chlorophyll a density of periphyton sampled from artificial substrates in late summer at the same locations where light was measured. Multivariate analyses (AIC_c and CART) were performed to describe the below-canopy light environment as a function of forest structure and topography, and periphyton biomass as a function of the light environment. CART results indicated that below-canopy light over streams decreased as forest structural characteristics progressed with stand development, but spatial variability in available light increased. Multivariate models of horizontal forest structure (e.g. gapiness) and topography showed strong relationships with the below-canopy light environment of late-successional forests, specifically regarding mean light intensity and spatial variability. Chlorophyll a density was highly correlated with the spatial distribution of light availability.  These results indicate that headwater streams within late-successional temperate forests include a heterogeneous mix of high and low light locations, possibly corresponding to a mosaic of heterotrophic (shaded) and autotrophic (lighted) micro-habitats along low-order stream reaches.  This may contrast with the prevailing assumption that organic production in forested low-order streams is driven almost entirely by allochthonous inputs.