The empirical investigation of information structure

The underdetermination of theory by empirical data is an inherent property of scientific research, also beyond the study of language. For instance, it was widely held at the time of Copernicus that his theory and Ptolemaic theory do not necessarily make different predictions concerning the available astronomical data at that time (see Newton-Smith 2000: 532). Nevertheless, it is virtually impossible to carry out a scientific investigation that is free from theoretical considerations. First, even if empirical data exist independently of theories, their relevance for the scientific knowledge can only be judged in the light of a particular theory (see Feyerabend 1983: §3). Second, it is not possible to report on observations without reference to background assumptions that are derived by inductive procedures and are necessarily theory-specific. Hence, an astronomer adopting a geocentric theory such as Ptolemy and an astronomer adopting a heliocentric theory such as Copernicus may observe the same empirical phenomenon, i.e., the perceived position of the sun, but keep track of different observations. The challenge of the present article is to report about the methods of investigating information structure abstracting away from the theoretical possibilities that are used in order to interpret the empirical data. The empirical question is what are the linguistic forms and the information-structural concepts that constitute the building blocks of a research paradigm. For example, accent as a formal property was presumably only established by Paul (1880), and it seems that Dretske (1972) was first in claiming a specific category of contrastive focus. Given a set of linguistic forms and a set of information-structural properties, we can address the question whether there is a correlation between form and function, such as ‘movement to the preverbal position’ ~ ‘focus’ or ‘left dislocation’ ~ ‘topic’. The observation of such a correlation only allows for weak statements concerning the form-function correspondence. Such a correlation raises the next empirical question: is the functional concept a necessary and/or sufficient condition for the