The aim of the present study is to understand how different definitions of snacking influence the estimated probability of obesity in the presence of concurrent risk factors. Factors influencing obesity were evaluated by reviewing the relevant literature through a PUBMED search. Six different modalities to define snack consumption were identified. A Bayesian network model in which nodes represent the variables that the retrieved studies indicate as affecting the probability of obesity was implemented and used to estimate the individual risk of developing obesity taking into account the concurrent effect of the considered risk factors. For a subject with a given profile of factors, the probability of obesity varies according to the chosen definition of snacking, up to maximum of 70%. The variability of the probability of obesity attributable to the chosen definition of snacking is very high and may threaten any conclusion about the effect of snacking, which may be related to the specific definitions adopted in the study.