This paper analyses the gender wage gaps by education throughout the wage distribution in Spain using individual data from the ECHP (1999). Quantile regressions are used to estimate the wage returns to the different characteristics at the more relevant percentiles and a suitable version of the Oaxaca-Blinder decomposition is then implemented to estimate the component of the gender gap not explained by different characteristics. Our main findings are two-fold. First, in contrast with the steep pattern found for other countries, the flatter evolution of the gap in Spain hides a composition effect when the sample is split by education. On the one hand, for the group with college/tertiary education, we find a higher unexplained gap at the top than at the bottom of the distribution, in accordance with the conventional glass ceiling hypothesis. On the other, for the group with lower education, the gap is much higher at the bottom than at the top of the distribution. We label this novel pattern as glass floors and argue that it is due to statistical discrimination exerted by employers in view of the low participation rate of women in this group. Such a hypothesis is confirmed when using the panel structure of the ECHP.