<b><i>Aims/Background:</i></b> The Edinburgh Feeding Evaluation in Dementia (EdFED) scale has been shown to have good psychometric properties using a range of methods including Mokken scaling. We aimed to study the Italian version of the EdFED using Mokken scaling. <b><i>Methods:</i></b> Data were gathered at 7 time points from 401 nursing home residents affected by dementia in the course of a 6-month intervention study using analysis of variance, Mokken scaling, and person-item fit statistics. <b><i>Results:</i></b> The properties of the EdFED-I scale were stable over the course of the study with 4 items showing invariant item ordering at all time points. Some items behaved differently at different levels of difficulty in the scale and also depending on the mean level of feeding difficulty. The test information function showed a dip in the mid-range of difficulty scores.