Fluvial suspended sediment typically consists of a variety of complex, composite particles referred to as ¯ocs. Floc characteristics are determined by factors such as the source, size and geochemical properties of the primary particles, chemical and biological coagulation processes in the water column and shear stress and turbulence levels in the stream. Studies of ¯oc morphology have used two contrasting methods of sampling and analysis. In the ®rst method, particles settle on a microscope slide and are observed from below using an inverted microscope. The second method uses ®ltration at no or low vacuum and particles deposited on the ®lter are observed with a microscope. Floc morphology can be quanti®ed using fractal dimensions. The aims of the present study were to examine the eect of the two sampling methods on the fractal dimensions of particle populations, and to evaluate for each method how well the fractal dimensions at the various sampling sites re¯ect basin conditions. Suspended solids were collected in triplicate on inverted microscope slides and on 0 . 45 mm Millipore HA ®lters in two southern Ontario streams with contrasting riparian zones during a minor runo event resulting from the melt of a freshly fallen snowpack. An image analysis system was used to determine area, longest axis and perimeter of particles. The morphology of the particle population of each sample was characterized using four fractal dimensions (D, D 1 , D 2 and D K ). Systematic dierences in fractal dimensions obtained with the two methods were observed. For the settling method, outlines of larger particles were frequently blurred because of the distance between the focal plane (the top of the inverted microscope slides) and the plane of the particle outline. In this method, the blurring of large particles can cause an increase in the projected area and length of the particle. The eect on the particle perimeter is unpredictable because it depends on the amount of detail lost through blurring and its eect on the apparent increase in particle size. Because of blurring, D and D 1 tend to be systematically lower for the settling method, whereas the net eect on D 2 is unpredictable. Particle size distributions derived from settling are typically coarser because small, low density particles may remain in the water column and all particles may not deposit on the slides. This loss of ®nes results in systematically lower D K values for the settling method compared with the ®ltration method. Fractal dimensions and particle size distributions obtained with the ®ltration method were sensitive to and clearly indicated dierences between drainage basins and between sites within each basin. These dierences were explained by basin characteristics and conditions. Fractal dimensions and particle size distributions obtained with the settling method were less sensitive to drainage basin characteristics and conditions, which limits their usefulness as process indicators.