QTL for a complex trait like yield tend to be unstable across environments and show QTL by environment interaction. Direct improvement of complex traits by selecting on QTL is therefore difficult. For improvement of complex traits, crop growth models can be useful, as such models can dissect a target trait into a number of component traits. QTL for the component traits are assumed to be more stable across environments. The target trait can be reconstructed from its component traits together with environmental inputs. Instead of observed component traits, QTL fits for component traits may be used when QTL explain a reasonable proportion of the variation in the components. We applied this dissection approach to the target trait total shoot biomass for a population of 149 recombinant inbred lines from the intraspecific cross of Capsicum annuum ‘Yolo Wonder’ and ‘Criollo de Morelos 334’. A simple LINTUL-type simulation model was used, with rate of change of leaf area index and light use efficiency as genotype specific component traits. These two component traits were determined in four phenotyping experiments (spring and autumn cultivations in the Netherlands and Spain), and subjected to QTL analysis. Seven QTL were found for both component traits. For leaf area index development rate 40 to 50% of the observed variance was explained by the QTL, while this was slightly lower for light use efficiency (23-39%). Using the QTL fitted values of the component traits following QTL analysis, the crop simulation model explained 27- 43% of the observed variation in total shoot biomass, which was higher than the variation explained by the QTL for total shoot biomass itself for most experiments. The approach of dissecting a complex trait into its component traits is therefore a promising one. Next step is to extend the model with biomass partitioning