The fossil record, even with its severe limitations and uncertainties, provides the only direct access to the chronological and paleogeographical dimension of biological events and processes that took place in ancient insular environments in completely natural conditions.
Terrestrial mammals, unlike other land vertebrates such as reptiles, have a high metabolic rate that requires large amounts of food and water. This would lead one to expect that land mammals could not cross extended expanses of deep water. Differing abilities in colonising islands in large versus small mammals are also expected. One can roughly distinguish three groups of models to explain the occurrence of mammals on islands:
1- Insular mammals are the relics of a previous population (vicariance);
2 - Mammals reached the island through filtering barriers of various intensity (dispersals);
3 - A third possible dispersal mechanism, is passive transport on floating islands or by strong marine currents. These mechanism is poorly realistic for large sized mammals while empirical observations on how this mechanism might operate for small mammals are scant.
Nevertheless, in general, these considerations pose major constraints to the processes responsible for the presence of land mammals in insular domains, and thus to the models that can be employed to explain them.
When the population of islands originates by dispersal, it is likely that those dispersals originated from lands close to the islands, either across direct temporary land connections or by overcoming moderately extended barriers. This model predicts a pattern in which dispersals are not scattered in time but concentrated in precise time zones corresponding to suitable palaeogeographic conditions separated by phases of isolation. Such a pattern can be corroborated or tested by the study of the fossil record and, as will be shown later, can be observed in most cases of populating the Mediterranean islands.
After having traced a synthetic outline of palaeogeographic evolution of the Central Mediterranean area two case histories, Mio-Pliocene Gargano insular domain, and Quaternary Sicilian archipelago are analysed, basing on literature data, searching for confirmation of the expected pattern.
The results can be outlined in a number of issue. A first consideration, is that the complexity of real cases cannot be reduced to classical models (dispersive and vicariant) without unacceptable simplifications. Secondly, the concept of 'polyphasic' populations appears to have great value when applied to unravelling the history of island populations. This term means a faunal population in which, in a given time zone, faunal elements derived from more ancient population phases coexist with elements of more recent events, often with different modalities and processes. Thirdly, there is actually an evident difference in behaviour between large sized and small sized mammals with respect to island colonisation. For small mammals the 'floating islands' model may be important. Fourth, the ecological conditions on islands apparently plays an important in controlling biodiversity and extinction in insular environments and deserve further and deeper investigations. Last but not least, more investigation on new fossil localities is needed to enhance the quality of models and testing With regard to model development, model testing and interpretation of case histories, it is evident from the present work that a "cross checking" strategy is the most effective: acquiring as much information as possible from as many sources as possible, and using multidisciplinary information to constrain models and results.
Keywords: vertebrate palaeontology, islands, Neogene-Quaternary, Mediterranean, models, case histories.