Tables 1 and 4 are only available in electronic form at http://www.aanda.org

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

Based on FLAMES observations collected at the European Southern Observatory, proposal 171.B-0588.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... observed

Using different sets of isochrones, e.g. Geneva (Lejeune & Schaerer 2001) or Yonsei-Yale (Yi et al. 2001; Kim et al. 2002), would bring slightly different results, however qualitatively in agreement.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... population

The velocity dispersion values we list are from the weighted standard deviation and do not include the broadening due to measurement errors. The true dispersion $\sigma_{\rm true}$ is approximately $\sigma^2_{\rm true}= \sigma^2_{\rm obs} - \sigma^2_{\rm meas}$ , where $\sigma_{\rm obs}$ is the observed velocity dispersion and $\sigma_{\rm meas}$ is the average measured error in velocity. For our FLAMES data the average error in velocity is $\sim$ 2 km s^-1; the resulting values of the velocity dispersion are thus only marginally inflated, for example if $\sigma_{\rm obs}= 10$ km s^-1 then $\sigma_{\rm true}= 9.8$ km s^-1.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.