“…In defence of this step, now clearly seen to have been too optimistic, we should point to the studies of the linear stability by Eltayeb and Roberts (1970) and Eltayeb (1972Eltayeb ( , 1975, which gave results qualitatively insensitive to the relative orientations of g, fi and Bo, and which initially encouraged us to believe that geometrical considerations would not be crucial. The discovery of the anti-Taylor inequalities for the present situation, was already a foretaste of an impending disappointment, for it demonstrates that our model behaves at finite amplitude quite differently from the plane layer Boussinesq model in which rotation is parallel to gravity, a configuration extensively studied by Roberts andStewartson (1974, 1975), Soward (1980Soward ( , 1986 and Skinner and Soward (1989), and in which Taylor states occur.…”