This paper attempts an explanation of the behaviour of the regional earnings struicture through time with particular emphasis on the underlying mechanism that holds the structure together. Earnings data that relate to the British engineering industry are used as the basis for statistical tests. Earlier analysis of these data (Hart and MacKay, 1974) revealed that, for the interwar and postwar periods, the regional earnings hierarchy appears to be stable, with some tendency for regional differentials to narrow over time and no consistent relationship between local unemployment conditions and local earnings changes. Of the four hypotheses that have been advanced so far to explain the behaviour of the regional structure (see Thirlwall, 1970, for a useful summary), our emphasis is on the one that is popularly termed the earningsspread hypothesis. The simplest version of this hypothesis takes the form that earnings changes in a leading market(s) are a function of the excess demand for labour in that market, and that earnings changes in the leading market are passed on, in whole or in part, to lagging markets. It is worth observing that such an hypothesis can be interpreted in two very dissimilar fashions. Earnings adjustment can be regarded as essentially an institutional process where coercive comparisons, established relativities or some similar manifestation of inter-group wage rivalry causes employees to bargain for and obtain increases in earnings similar to those obtained by employees in other areas irrespective of differences in regional labour market conditions (Cowling and Metcalf, 1967; Thirlwall, 1970; Lerner and Marquand, 1963). Alternatively, and this argument in the neoclassical tradition has escaped undetected by the British literature, the existing structure of earnings differentials, and its adjustment through time, can be given an equilibrium interpretation. Taking this view, the existing structure of wage differentials may reflect non-pecuniary elements, regional differences in the cost of living, the cost of movement and other barriers to geographical mobility. Hence, changes in regional earnings differentials would be interlinked because of competition, real or potential, between firms operating in different labour markets (Brechling, 1973). As suggested above, the models used to test the earnings-spread hypothesis have incorporated a simple, one-way transmission mechanism in order to represent the transference of earnings increases from leading to