In lager beers the intensity of "estery" aroma character is regarded as an important component of sensory quality, but its origins are somewhat uncertain. Overall "estery" aroma intensity was predicted from capillary gas chromatographic (GC) data following solid phase micro extraction (SPME) of headspaces. Estery character was scored in 23 commercial lagers using rankrating, allowing assessors (13) constant access to a range of appropriate standards. From univariate data analysis, all assessors behaved similarly and lagers fell into three significantly different groups: low (1), high (1) and intermediate (21). The quantification of 36 flavour volatiles by SPME of headspaces was reproducible and principal component analysis explained 91% total variance. Multiple linear regression could utilise only a restricted (26) set of flavour volatiles, whereas partial least square regression, that considered all flavour components, showed significant differences and improved prediction. However, an artificial neural network that could compensate for nonlinearities and interactions in ester perception gave the most robust prediction at R 2 = 0.88.