This detection was possible because the sample used was sufficiently large to overcome problems with small number statistics. Concerns have been raised that such a large sample with unrestrictive (and in fact unknown) selection effects might be biased, although this does not appear to be the case (Goldschmidt & Rephaeli 1993). One problem is that while the control sample contains mostly background sources, the test sample has a much greater proportion of cluster members. This does not appear to bias the result unacceptably, and unfortunately using only background sources reduces the significance of the RM detection (although it is still present). Until a larger sample of RMs are available, we prefer to use all the available data.
Converting the excess RM into a magnetic field strength in the cluster requires knowledge of the tangling scale of the field. Unfortunately, RM observations tell us very little about this scale length as we usually probe a few (at best) independent sightlines through any cluster. RM images of cluster central sources (e.g. Dreher, Carilli & Perley 1987; Taylor et al. 1990; Perley & Taylor 1991; Ge 1991) give ordered RM on scales of a few kpc. The tangling scale of the field at large radii is constrained to be about 15 kpc or less from the absence of large surface brightness fluctuations in the Coma radio halo (Tribble 1991).
Strong Faraday rotation is also detected from central cluster galaxies (Dreher et al. 1987; Taylor et al. 1990; Perley & Taylor 1991; Ge 1991). These are sources embedded in cooling flows at the very centre of the cluster where the gas density is highest. The inflow also compresses the magnetic field, leading to very large RM values. Maps of RM indicate that the RM is ordered on scales of a few kpc. If the inflow were perfectly smooth, the magnetic field would be combed into a radial pattern and the RM would increase very strongly towards the centre (Soker & Sarazin 1990). This does not appear to be observed - the RM variation is less steep (Tribble 1992), consistent with isotropic compression of the field. This is probably because the inflow is rather slow, and motions in the gas can isotropize the magnetic field distribution on much shorter timescales (Tribble 1993).