We investigate the plausibility of sympatric speciation through a modelling study. We built up a series of models with increasing complexity while focussing on questioning the realism of model assumptions by checking them critically against a particular biological system, namely the sympatric benthic and limnetic species of threespine stickleback in British Columbia, Canada. These are morphologically adapted to their feeding habits: each performs better in its respective habitat than do hybrids with intermediate morphology. Ecological character displacement through disruptive selection and competition, and reinforcement through mating preferences may have caused their divergence. Our model assumptions include continuous morphological trait(s) instead of a dimorphic trait, and mating preferences based on the same trait(s) as selected for in food competition. Initially, morphology is intermediate. We apply disruptive selection against intermediates, frequency-dependent resource competition, and one of two alternative mating preference mechanisms. Firstly, preference is based on similarity where mating preference may result from “imprinting” on conspecifics encountered in their preferred foraging habitat. Here, speciation occurs easily—ecological hybrid inferiority is not necessary. Hybrid inferiority reinforces the stringency of assortative mating. Secondly, individual preferences exist for different trait values. Here, speciation occurs when linkage disequilibrium between trait and preference develops, and some hybrid inferiority is required. Finally, if the morphology subject to disruptive selection, frequency-dependent competition, and mate choice, is coded for by two loci, linkage disequilibrium between the two loci is required for speciation. Speciation and reinforcement of stringency of choosiness are possible in this case too, but rarely. Results demonstrate the contingency of speciation, with the same starting point not necessarily producing the same outcome. The study resulted in flagging issues where models often lack in biological realism and issues where more empirical studies could inform on whether assumptions are likely valid.