Barley (Hordeum vulgare L.) is often grown in sites with low rainfall and high temperature during grain filling. Because spike architecture is one of basic footprints of barley domestication, the importance of spikes in adaptation to different environments or abiotic stresses can be hypothesised. In order to compare different barley spike types in terms of kernel growth and yield components, we tested 15 two-row and 10 six-row winter genotypes in eight environments where terminal drought was simulated by defoliation at 7 days after heading (7 DAH). Control plants were grown intact. On average, two-row genotypes outyielded six-row genotypes by 17% under control conditions and 33% under simulated late drought. Observations of kernel dry weights from 7 DAH through to harvest maturity at 5-day intervals were regressed onto a measure of thermal time. After preliminary evaluation of four nonlinear (S-shaped) models for kernel dry-weight accumulation, the ordinary logistic model was deemed the most appropriate in most cases and was finally applied to all plant-growth curves. Four parameters were estimated from the logistic model. Whereas two earliness estimators (inflection point and thermal time needed to reach maximum kernel weight) were similar for the two barley types, maximum kernel weight (Y_max) and mean rate of kernel growth (RG) were higher (P < 0.05) in two-row than in six-row barleys. Differences in Y_max and RG among six-row barley genotypes were greater between control and defoliation treatments than between years, whereas among two-row barley genotypes, differences between years were greater, suggesting better stability of six-row types and better drought tolerance of two-row types in the tested barley set.