We propose a mechanistic modelling approach for studying the indoor dispersion of aroma compounds which are released from, for instance, food products. The approach combines the indoor velocity field with a release model for aroma compounds. The release mass flux is expressed as a function of key variables such as mass transfer and gas-liquid partition coefficients, and the source geometry. The transport properties of ambient air are assumed to be independent of the aroma concentration; hence release and dispersion problems can be solved separately. First, the velocity field is obtained as solution of the fluid flow problem through computational fluid dynamics (CFD). The turbulent velocity field is then used to predict the time evolution of concentration of an aroma compound released by a constant rate source, in an initially aroma-free environment. These results are interpreted in terms of a step response function. The aroma concentration as a function of time is finally estimated by convolving the possibly time-varying release mass flux and the response function associated with the position of interest. The modelling approach is flexible and computationally effective, since different release models as well as the release of distinct aroma compounds can be directly studied by taking into account a same velocity field, without any additional CFD simulation. The validity of the approach is assessed from measurements of aroma concentration in a 140 m(3) room, under constant release mass flux. The approach is also illustrated for a case where the release mass flux is not constant in time.