We have prepared cytochrome-c oxidase from bovine heart (using a modification of the method of Kuboyama et al. (1972) J. Biol. Chem. 247, 6375-6383) which binds cyanide rapidly, shows no kinetic distinction between the two haems on reduction by dithionite, has a Soret absorption maximum above 424 nm, and has a negligible 'g' = 12' EPR signal. On incubation at pH 6.5 this 'fast' oxidase reverts to the 'slow' ('resting') form characterised by slow cyanide binding, slow reduction of haem a3 by dithionite, a blue-shifted Soret maximum and a large 'g' = 12' signal. Incubation of 'fast' oxidase with formate produces a form of the enzyme with properties almost identical to those of 'slow' oxidase. The kinetics of formate binding to 'fast' oxidase are found to be biphasic, revealing the presence of at least two 'fast' subpopulations in our preparations. Evidence is presented that there is an equilibrium mixture of high-spin and low-spin forms of haem a3 in both 'fast' subpopulations at room temperature. Incubation of 'fast' oxidase with chloride or bromide at pH 6.5 produces forms of oxidase with much lower rates of cyanide binding. Our working hypothesis is that formate mimics a binuclear centre ligand which is present in the 'slow' form of cytochrome oxidase. Although we show that chloride and bromide can also be ligands of the binuclear centre, possibly onto CuB, we can rule out either of these being the ligand present in the 'slow' enzyme. We will argue that the 'fast' and 'slow' forms of oxidase are equivalent to the 'pulsed' and 'resting' forms of oxidase, respectively.