We questioned the different interpretations of ecological sexual segregation from a novel perspective, i.e., by carrying out diverse temporal and spatial scale analyses within a long-term study (1984–2003). Thus we combined spatial (small/large) and temporal (small/large) scale analyses to identify the factors generating sexual segregation in fallow deer in San Rossore, Italy. The study site was divided into an eastern sector characterized by human disturbance (DS) and a western undisturbed sector (US). According to census data, human presence increased in DS from 1984, and while females gradually abandoned it, males remained—thus supporting the predation risk hypothesis (large spatial and temporal scale)—and actually increased their presence in DS, where they seemingly benefited from a lower female density. This supported the indirect competition hypothesis. The analysis of data on a large temporal and small spatial scale confirmed that intersexual competition, in particular for grass, was higher in a crowded pasture in US. Observations by means of radio-telemetry of 23 adult females and 25 adult males (1997–2001, reduced temporal and large spatial scale) showed that large scale segregation was relevant during the day and disappeared at night, when disturbance was absent and also the females reached DS. This also supported the predation risk hypothesis. Moreover, sexes showed different habitat choices inside DS at night, thus supporting the forage selection hypothesis (small spatial and temporal scale). In conclusion, failure to address the whole set of combinations of spatial and temporal scale analyses would have led to monocausal explanations of ecological sexual segregation.