The steps (Pt{111) x {111} and Pt{001} x I I I 11) and terraces (Pt{ 100) and Pt{ I 11)) of a 5% Pt/graphite catalyst have been identified by cyclic voltammetry and their contributions to rate and enantioselectivity in high-pressure ethyl pyruvate hydrogenation assessed. Bi preferentially adsorbed from solution onto the platinum surface of the catalyst at step and { 100) terrace sites. Further increasing Bi coverage led to the gradual occupation of [ I I I] terrace sites, followed by the formation of Bi multilayers. In contrast, S adsorbed from solution onto terrace and step sites simultaneously but with the I I I I) x { I 111 step sites being strongly disfavoured. Pt/graphite, Bi-Pt/graphite, and S-Pt/graphite catalysts have been modified by cinchonidine and used to catalyse the enantioselective hydrogenation of ethyl pyruvate to ethyl lactate at 30 bar and 293 K. The effect of increasing Bi coverage at step sites was to increase activity substantially but reduce enantiomeric excess from 43%(R) to 17%(R), whereas the effect of increasing S adsorption at terrace sites was to decrease activity and increase enantiomeric excess to 52%(R). These unexpected contrary effects on activity and enantioselectivity were confirmed for Bi adsorption by repeating the experiments using the standard reference catalyst 6.3% Pt/silica (EUROPT-1), for which enantiomeric excess fell linearly from 73%(R) to 20%(R) as Bi loading was increased. The well-documented rate enhancement associated with catalyst modification by cinchonidine has been reassessed in the light of this further rate enhancement by Bi adsorption, and its origin has been attributed to inhibition of ethyl pyruvate dimerisation/polymerisation by the strongly basic alkaloid. Rate enhancement is now attributed to reaction occurring at a normal rate at an enhanced number of sites, not (as previously proposed) to a reaction occurring at an enhanced rate at a constant number of sites. The opposing effects of Bi and S on rate and enantioselectivity are consistent with (i) preferential initiation of pyruvate polymerisation at step sites, (ii) inhibition of propagation of this polymerisation by alkaloid, and (iii) a higher enantiomeric excess in cinchonidine-modified near-step sites than in similarly modified terrace sites. These results have important implications for future catalyst design. (c) 2005 Elsevier Inc. All rights reserved.