Most current process-based crop models use "de Wit"representation, whereby different types of stress reduce a potential plant development pattern that solely depends on the variety, temperature and radiation. That approach was largely validated for conventional agriculture and the corresponding range of environmental conditions. The aim of our study was to test the robustness of that representation under early sub-optimum growth conditions. Investigations were carried out on cotton in southern France under controlled glasshouses and field conditions. We analysed the effects of light interception, water supply and planting density on the node appearance pattern of the plant's main stem.Early stress on seedlings and a variation in planting density induced a plant reaction that could not be properly taken in account by common "de Wit" representation. First, delays in potential pattern establishment after emergence were observed with PARi values under 0.15 mol plant-1 d-1. Furthermore, early low PARi and drastic water stress (FTSW < 0.25) affected the node initiation rate once 10 nodes had already been initiated. Subsequently, given node appearance rate (NAR) dependency on the initiation rate, the indirect impact of past conditions on seedlings might affect the potential development pattern during transition from the vegetative to the reproductive phase. Lastly, under stress-free conditions, planting density was shown to affect potential development patterns.