At all wavelengths, the polarized intensity and the degree of linear polarization have well-defined minima near both edges of both jets, but (away from the brightest knots) neither jet exhibits much excess polarization on-axis. These "rails" of depressed polarized intensity are near the half-power points of both jets, as illustrated for the main jet in Figure 2. The depth of the ``rails'' is typically from 20% to 40% of the total intensity of the jet at their locations, and they are not seen anywhere there is no jet emission.
We believe that these pairs of polarization "rails" indicate a systematic misalignment between the dominant magnetic field components on the edges of the jet and in surrounding (lobe plus sheath) emission. Four-frequency fits to the Faraday rotation measure across 3C353 show that the apparent (synchrotron-emissivity weighted, line-of-sight averaged) magnetic field in the emission around both jets is predominantly, but not exclusively, perpendicular to the jet axes. The ``rails'' can therefore be accounted for by the crossed-field configuration that results if the apparent field near the edges of the jets is predominantly axial, while that near their centers is disordered, or is dominated by toroidal or radial components. Such a configuration might result if a velocity shear converts radial field components to axial field in the outer layers of a quasi-cylindrical jet. We suggest that if 3C353's jets could be studied in isolation, they would be limb-brightened in polarization, with the apparent magnetic field direction parallel to the jet axis along both edges of both jets.