Abstract : q-Gaussian behavior is often encountered in quite distinct settings. A possible explanation is here given with reference to experimental scenarios in which data are gathered using a set-up that performs a normalization preprocessing. We show that the ensuing normalized input, as recorded by the measurement device, will always be q-Gaussian distributed if the incoming data exhibit elliptical symmetry, a rather common feature. Moreover, we find that in these circumstances, the value of the associated parameter q can be deduced from the normalization technique that characterizes the device. Finally, one concludes from the above remarks that great care should be exercised when empirically detecting q-Gaussian behavior. As an example, Gaussian data (the most common situation) will appear, after normalization, in the guise of q-Gaussian records. (c) 2008 Elsevier B.V. All rights reserved.