Sink-source balance for carbohydrate at the cotton plant level involves the fruit shedding process as the main mean for the cotton plant to adjust his demand to the offer. Obviously cotton crop simulation in a material balance approach needs to take in account the fruiting sites abscission process, The COTONS® model simulates cotton plant squares, blooms and bolls shedding as (i) a discrete process in a "plant population" simulation mode, and as (ii) a continuous process in an "average plant" simulation mode. This study compares field observations with the output data respectively in -average plant" and in "plant population" simulation modes. The model "is /ate" in both simulation modes: later model growth cut-out Compared to observed data. This results from an over estimation of square abscission and under estimation of boll abscission by the model. In preferring square abscission the model favors vegetative growth because of higher sink force for boll compared to square. This is mainly the case in plant average simulation mode with the right simulated number of open bolls; even the green bolls simulated number is widely under estimated. At the opposite, the total number of squares and bolls abscissions in plant population simulation mode is well estimated in spite of a widely over estimated number of open bolls, Simulating the abscission process as a discrete process in plant population simulation mode seems to decrease the differences in cut-out delay, in green bolls number and in abscission number estimations. To improve cotton growth and development simulations, especially in fruit abscission process, the COTONS® model needs to be calibrable in plant population simulation mode respecting the discrete character of the simulated process.