Gap-Fluorescence Recovery After Photobleaching (gap-FRAP) is a technique used to estimate functionality of intercellular connections in biology. Such a technique could potentially be involved in the diagnostic of normal/cancer cells. Discrimination of cell types may be performed directly, by comparing plots of fluorescence kinetics or indirectly by statistical testing applied to model parameters. This paper focuses on the latter model-based approach. Up to now, more than ninety percent of the models used to fit gap-FRAP responses have been derived from diffusion equations (partial differential equation). We propose to simplify the modeling procedure by using behavioral models derived from system identification techniques used in control engineering. To assess in practice the relevance of this concurrent method, two human head and neck carcinoma cell lines (KB and FaDu) were used. The former (KB) expresses connexin proteins (positive line) while the latter (FaDu) does not (negative line). Moreover, each cell line was tested on spheroid (3-D) and monolayer (2-D) slices and in vitro assays were repeated six times. System identification algorithms of the CONTSID Matlab toolbox were used to estimate the model parameters from the in vitro data sets. Results have particularly emphasized there is no need to use complex models to fit the observed gap-FRAP responses. We show that the static gain of the estimated transfer functions is able to discriminate cell types used in this study, which corroborates the relevance of system identification techniques for diagnostic applications based on gap-FRAP analysis.