Two mechanisms for solar wind plasma transfer across Earth's magnetopause are expected on theoretical grounds: magnetic reconnection and particle diffusion. Of these, reconnection has been confirmed observationally beyond any doubt, while diffusive entry has never been demonstrated altogether convincingly. The occurrence of reconnection naturally depends strongly on the relative orientation between the interplanetary and terrestrial magnetic fields at the magnetopause, and there is evidence that the process can be patchy and time-varying. Diffusion, on the other hand, is often assumed to be constantly operating, producing a boundary layer that does not depend on solar wind conditions. If diffusion were indeed occurring over the entire frontside magnetopause, a boundary layer would result that gets progressively thicker the further one moves away from the subsolar point. An ideal vantage point is therefore the flanks of the magnetopause where the accumulative effect of diffusion should be most easily observable. We have analyzed AMPTE-IRM crossings in these region and have found cases where no boundary layer of solar wind plasma could be observed at all. And this in spite of the fact that the magnetopause current layer itself lasted approximately 60 s and thus was well-resolved in the measurements. In one particular case we could infer from the measured plasma flow speed along the magnetopause normal direction that the magnetopause was moving at less than 20 km/s. The magnetopause was therefore no more than 1200 km thick. The time resolution of our measurements was 4.4 s. Taking 5 s as an upper limit for its true duration, we conclude that the boundary layer could not have been thicker than 100 km in these cases. Taking the measured plasma density and bulk speed values at the inner edge of the magnetopause as representative of the boundary layer, we can compute the average boundary layer particle flux per unit height. From the known distance from the subsolar point, the magnetopause area across which this flux must have entered can be computed, and this fixes the average diffusive particle flux across the magnetopause. This flux should equal , where D is the diffusion coefficient and is the density change across a diffusion layer of thickness h. Taking the measured densities and observing that h = 1200 km in this case, we obtain an upper limit for the diffusion coefficient of 5·107m2s-1. This is a factor of 20 lower than the canonical value of 109m2s-1 that is often used in the literature when estimating the contribution of diffusive entry. It is interesting to note that in the case discussed no signatures of reconnection were observed either. Thus it seems entirely possible that at times very little solar wind plasma is traversing the magnetopause.