Rudnick & Jones (1982) first pointed out that the GPS sources tend to have low centimeter-wavelength polarization. Further observations (Pearson & Readhead 1988; Rusk 1988; O'Dea et al. 1990b; Aller, Aller, & Hughes 1992; Stanghellini et al. 1998b) have firmly established that GPS sources have very low integrated polarization (~ 0.2% at 6 cm). At present, there are not many high-frequency ( 8.4 GHz) measurements, so the frequency dependence of the polarization is not known. High-frequency polarization measurements of the GPS sources would be very useful. The low integrated polarizations could be due to (1) cancellation of polarization due to vector averaging of polarizations with different orientations in different components, (2) a very tangled magnetic field, and (3) large Faraday depths in or around the radio source. Cawthorne et al. (1993a, 1993b) have obtained polarization VLBI measurements of several GPS sources from the Pearson-Readhead sample. They find that the galaxies 0108+388, 0153+744, 2021+614, and 2352+495 (even the bright core component) are unpolarized; however, the quasar 0711+356 is 2.5% polarized. Stanghellini et al. (1998a) report 15 GHz VLBI observations, which reveal that OQ 208 and 0500+019 are unpolarized. The lack of high polarization in individual components in the sources argues against (1) as an explanation. Cawthorne et al. (1993a) also favor high Faraday depths as the explanation for the low polarization in the GPS galaxies.
The fractional polarizations of the CSS sources tend to be higher than those of the GPS sources (~ 1%-3% at 6 cm) and higher still (~ 6%-7%) at higher frequencies (van Breugel et al. 1984a, 1992; Saikia, Singal, & Cornwell 1987; Mantovani et al. 1994; Sanghera et al. 1995; Akujor & Garrington 1995). Some sources show a polarization asymmetry where the polarization tends to be higher on the side with the jet in the quasars and the side farthest from the nucleus in the galaxies (Mantovani et al. 1994; Akujor & Garrington 1995; Lüdke et al. 1996), consistent with the Laing-Garrington effect (Laing 1988; Garrington et al. 1988).
The fact that the polarization increases with increasing frequency (in the CSS sources) suggests that large Faraday depths are responsible for the depolarization between 15 and 5 GHz rather than magnetic field geometry.
There is some evidence that the CSS quasars are more highly polarized than the galaxies at 6 cm (Saikia et al. 1987; Sanghera et al. 1995). However, Akujor & Garrington (1995) find that quasars and galaxies have a similar distribution of fractional polarization at 3.6 cm. These conflicting results may mean that the final answer is not yet in, or it may mean that the galaxies depolarize more than the quasars. In this case, at higher frequencies, when the sources are more Faraday thin, the differences in polarization should become smaller.