The question of recombination control has been the subject of intensive investigation over many years. This paper is restricted to an evaluation of the information obtainable from cytological studies of chiasmata, since other approaches are well represented in other papers in this volume. The advantages and disadvantages of chiasmata as means of studying recombination are discussed. Cytological observations reveal many constraints on the distribution of chiasmata, which are interpreted as outward expressions of controls operating at different levels, namely between cells, between bivalents and within bivalents. Detailed descriptions of these constraints are presented as a preliminary to discussing models and mechanisms of control. They include, at the between-bivalent level, over-dispersal of chiasma number per bivalent and the formation of an obligatory minimum of one chiasma per bivalent; and at the within-bivalent level, chiasma localization and chiasma interference. The genetical basis of these constraints is emphasized and evidence presented which suggests that chiasma distribution at different levels is subject to both separate and coordinate genetical controls. These observations lead to the consideration and evaluation of general stochastic models of chiasma control based on the modification of site probabilities for chiasma formation. While these models are a useful approach to thinking about the control of recombination, they do not give a fully satisfactory explanation of all aspects of chiasma control. Finally, specific mechanisms for chiasma distribution control are considered. In a few exceptional species, localized pairing and synaptonemal complexes lead to localized chiasma formation; details of two such cases are presented. In the majority of species, pairing is not localized or restricted and in these species the control of chiasma distribution is more conjectural. Arguments are presented which suggest that even in these cases of full pairing, the pattern of pairing initiation and progression and perhaps the timing of these events could be important factors controlling the distribution of chiasmata.