Nevertheless, a worthwhile subject of study - too long ignored. (see also Evolutionary Psychology)
However, there's various approaches to the study of evolutionary linguistics that do not depend on fossil record. One of them is the route taken by CSL/Paris (http://www.csl.sony.fr/Language/Research/Field/index), headed by prof. Luc Steels in collaboration with the Artificial Intelligence Laboratory of the Vrije Universiteit Brussel, and their description of their approach is the following:
"Research in evolutionary linguistics: A good way to test a model of a particular phenomenon is to build simulations or artificial systems that exhibit the same or similar phenomena. We are applying this approach to the problem of the origins of language and meaning by developing simulations, both in software and (in collaboration with the Artificial Intelligence Laboratory (http://arti.vub.ac.be) of the Vrije Universiteit Brussel) physically grounded on robotic agents. These simulations are based on the notion of language games - interactions in which speakers and hearers use language to identify or talk about features and events in the real or simulated world that they inhabit. Over the course of repeated interactions, linguistic phenomena similar to those seen in the real world may be observed. The goal of our research is to study the mechanisms that give rise to 'realistic' phenomena and to try to trace the processes and structures involved. In this way we hope to shed light on a number of important topics, including questions about the origin of language, the process of language change, the role of social factors in language development (and vice-versa) and child language learning."
Their definition of evolutionary linguistics is: "evolutionary linguistics is the study of language viewed as a complex adaptive system. Among the issues addressed are questions of how languages arise, how they change over time, and how they can be learnt. Language is seen as the product of local interactions between language users from which the global characteristics of the language emerge by selectionist mechanisms. Language is not created by some central agent; it is an emergent phenomenon that arises spontaneously. (http://www.csl.sony.fr/Language/Research/Field/index)."
The approach above is certainly a viable one, yet relies heavily on computer models -artificiality, and that could be a severe drawback since such models are invariably the result of parameters fixed by the programmers/researchers.