Abstract The approach used by Cohen and Banerjee to incorporate the effects of the U(1) A anomaly in soliton models leads to a flavour-singlet axial coupling which satisfies a Goldberger-Treiman relation connecting it to the η′N coupling. This is similar to the relation obtained by Efremov et al. assuming η′ and ghost dominance, as in large- N QCD. From this it is shown that the axial coupling calculated from the Cohen-Banerjee current should be identified with the net spin of the quarks in the proton. The usual U(1) axial current of the nonlinear σ model leads to a axial coupling which can be identified with that measured in polarised deep-inelastic scattering. Hence soliton models of this type can give a nonzero value for the net spin of the quarks, and still be consistent with the recent measurement by the EM Collaboration.