When rates of return on bonds are computed over extremely short holding periods, the ex post cross-sectional relationship between realized return and risk is linear. It is therefore possible, at any time, to extrapolate the cross-sectional relationship to a zero risk level, and thus to determine the implied instantaneous riskless rate of interest. We apply this technique to French bond price data. Using a rather unique data set in which prices are sampled daily, we are able to compare the overnight rate implied in bond price data to the actual overnight money market rate. We conclude that the two rates are significantly different, which is evidence of segmentation between the two markets. The institutional set-up prevailing in France during the sample period explains the segmentation result.