1. The fractal dimension of natural habitats may influence both numbers and body size distribution of arthropods. Plants with highly complicated and interrupted leaf shapes, i.e. whose fractal dimension is high, may have more space available for small animals than for large ones. 2. Morse et al. (1985) argued that scaling of body size with number of individuals may be predicted by combining assumptions of population densities as proportional to metabolic rate-1 and vegetation fractals. 3. The hypothesis by Morse et al. (1985) was tested in a field experiment in which spiders were allowed to invade artificial plants placed in oak and spruce habitats. Comparisons between the natural spider populations in oak and spruce habitats were also made. 4. The fractal dimensions were estimated in the artificial plants and in the natural vegetation, and the following sequence, from low to high fractal dimension, was obtained: plastic broad-leaf plant, oak, plastic 'spruce', spruce. 5. The prediction about scaling of body size with number of individuals based on vegetation fractals was supported qualitatively by the experimental results. However, some discrepancy between data and the quantitative predictions may be due to (a) the need of an improved theory, and/or (b) methodological artefacts. 6. Field data suggested that qualitative predictions about spider mean size based on vegetation fractals may hold within a habitat. 7. It is concluded that the general hypothesis that habitat structure influences body size distributions in arthropods was supported.