Rosehip oil was extracted from milled seeds with supercritical carbon dioxide (SC-CO2). The effects of sample pretreatment and interstitial solvent velocity (0.40-2.45 mm/s) on kinetics of extraction (oil yield versus time) were evaluated at 40degreesC and 300 bar, and at 50degreesC and 400 bar in a one-pass screening system (laboratory scale). Seeds were grinded in a hammer mill or a roller mill with a gap of adjustable thickness, and samples from the hammer mill were further size-classified. Data was fitted to a two-stage model, with extraction rate controlled by oil solubility initially, and by mass transfer in the solid phase at the end. The mass transfer coefficient and axial dispersion of the solute in the supercritical phase, and the solubility of the oil in SC-CO2 were estimated as a function of the extraction conditions using literature correlations, and the "free" oil content (X-f) and mass transfer coefficient in the solid substrate (k(s)) were used as fitting parameters. Both X-f (0.026-0.030 g oil/g substrate) and k(s) (0.6-0.9 x 10(-9) m/s) depended on sample pretreatment but not on the assayed extraction conditions, as expected. Free oil represented 33-41% of the total content, and effective diffusivities of oil in the solid were 330-710 times smaller than binary diffusivities of oil in SC-CO2. Based on best-fitting parameters for kinetic data at the laboratory-scale, extraction kinetics was simulated for a process development unit with solvent recycle (scale-up factor of 30 for the volume of the extraction vessel). It was observed that extraction was slower at the pilot plant than laboratory scale due probably to flow heterogeneity in the extraction vessel, increased dispersion of solute between the extraction and separation vessels, entrainment of oil droplets in recycled gaseous stream, or a combination of these three effects. (C) 2003 Elsevier B.V. All rights reserved.