The response of Smylie and colleagues to our Comment, henceforth designated (SJBS), in their Geophysical Journal International paper (Smylie ef al. 1992) presents such a distorted view of the issues we raised that a further attempt at clarification on our behalf is required. Before beginning, we should explain the relevance of what (SJBS) refer to (several times) as ‘Withdrawn Comment’. We had submitted the (original) Comment on one of their earlier papers that led to a response by Smylie and colleagues. After much debate it was suggested (by the editor of Geophysical Journal International) that we resubmit our (original) Comment as an independent paper. We did so, and that paper appeared as Crossley & Rochester (1992) (CR92). That paper contains all the relevant work we had done to substantiate our previous (original) Comment and the latter were certainly not withdrawn due to any assumption of error on our behalf, as (SJBS) are fully aware. In the first paragraph, (SJBS) deliberately mislead the unwary reader to ascribing to us statements we never made and then going on to ridicule these statements. We stated clearly in (CR92) that a variational principle does not exist ‘... for the subseismic description of core dynamics, i.e. with a functional in terms of x along, when the core boundaries are deformable ... the only circumstance in which K (the subseismic wave operator) is Hermitian for elastic boundary conditions, is when the core is neutrally stratified ...’. The underlining clearly delineates the limitations on our conclusion. The (SJBS) interpretation, and the consequences they draw from it, are obviously fallacious and unwarranted. (SJBS) state that ‘Failure to correctly apply the latter condition at the boundaries ... in their withdrawal’. Again this is completely misleading. In fact we could not apply at all the inequality at the boundary since it is violated in realistic earth models because the boundary barely moves! (SJBS) refer to the suppression of sound waves as a desirable feature of the subseismic approximation (SSA). This advantage does not affect a solution of the full system of equations because all the methods in question for solving core dynamics operate in the frequency domain, i.e. they are modal methods. Thus the time-integration problem does not arise in our method, or in that of Wu & Rochester (1990). (SJBS) show plots of y , , y , and p for -& ‘with the Coriolis term’ (Fig. 1). If these eigenfunctions were computed using the method of Smylie (1974), then the Coriolis term has only been rudimentarily taken into account as the self-coupling term. As Crossley (1975), Smith