This paper reports on a judgment experiment intended to test a prediction of Anagnostopoulou's (2003) seminal proposal concerning locality of A-movement in languages with theme-goal orders in active and passive contexts. Results from an experiment with native speakers of UK English show a significant cross-speaker correlation between scores for theme-goal orders in active contexts and theme passives. The results nevertheless indicate a richer inventory of grammars than the two-dialect distribution entailed by a single parametric difference assumed in locality approaches. We propose that these data are best expressed on a modified version of GB case-based approaches to passive symmetry that models the availability of theme-goal orders in both passive and active contexts in terms of the merged position of the probes responsible for agreement on the internal arguments.