Axe 7: Experimental Methods and Epistemology

Axe 7: Experimental Methods and Epistemology
Resp. B. Hemforth (LPNCog, Paris Descartes) and S. Auroux (HTL, Paris 7)
One of the central goals of the current project is the integration of scientific approaches from different disciplines such as linguistics, psychology, computer science, and neuroscience, in order to arrive at an empirically as well as theoretically well-founded description of a variety of languages. While extensive data bases such as those developed in the resource strand allow for a detailed analysis of factors influencing many linguistic phenomena (see for example Bresnan & Ford, 2010), a considerable number of questions remains to be studied with different methodological approaches, notably the direct interaction with linguistic informants in more or less controlled situations. Reasons for the need to approach theoretical questions with this family of techniques are multifold. We will only provide a small subset in the following paragraphs.
Detailed corpus analyses can only be carried out for well-described languages for which annotated corpora exist. This is only the case of a small subset of the languages spoken in the world. Even for well-described languages, the particular phenomenon of interest may not be available in the corpora, either for more technical reasons (because the relevant annotations are missing) or because it is not sufficiently frequent in spontaneous language production (sparse data problem). These cases ask for more or less controlled paradigms, which allow eliciting the constructions of interest.  One example of such an eliciting technique is the Map Task Paradigm, where an instruction giver and an instruction follower have to interact in order to reconstruct paths in only partially overlapping maps (Anderson et al., 1991).  Tasks of this kind are only available for a small group of languages. An important advancement in linguistic research could be reached by constructing tasks which are easily transferable to different languages and which are constructed jointly by researchers from different fields and disciplines such that they become exploitable for many research questions.
Corpus data moreover often provide us with a series of factors relevant for a particular phenomenon covarying in a context. Often enough, it is not possible to identify the particular contribution of a single factor or its particular interaction with one or more other factors. To isolate the contribution of a subset of factors controlled experiments are needed, with a variation of only the factors under investigation.  This is in particular true for speech perception and language comprehension in general, since corpus data mostly provide production preferences, which can or cannot coincide with comprehension preferences. Controlled experiments are, however, more often than not as necessary in production as in comprehension in order to arrive at a precise and fine-grained picture of the processes under investigation (Hemforth, 2006, Hemforth & Konieczny, 2006, Konieczny & Hemforth, 2002).
Neurophysiological data and brain imaging techniques add knowledge about the integration of behavioral data and the underlying neural circuits, which can be particularly useful for disentangling the implication of different kinds of linguistic (syntactic, semantic, ..) or general cognitive (working memory, cognitive control) processes.
The design of experiments as well as the statistical analysis of the data will play an important part in the joint training of students at the master's as well as at the PhD level. Very importantly, all students and researchers pursuing empirical studies with human participants will have to follow an intensive training on the ethics of experiments following the IRB guidelines (http://extranet.fhcrc.org/EN/sections/iro/irb/guidelines.html). This is a necessity for collaborative research with international institutions. Even though our studies are generally non-invasive and do not need individual approval of the ethics committees of the universities, we consider it as highly important that researchers in our laboratory are aware of potential violations of experimental ethics even in this kind of studies (violation of privacy, obligation of participants in hierarchical relations such as participants in courses, etc).  A lab internal ethics committee will survey experimental studies run within the program.
A series of important questions will, however, have to be tackled with respect to the epistemological status of experimental data in linguistics. This historically informed research will analyze the empirical nature of linguistics compared to psychological and neurocognitive approaches.
The research labs implicated in this project are provided with extensive technical equipment for the different kinds of studies proposed in strands 1 to 5. A short description follows for the different sites.
Experimental platforms available
Several mutualized technical platforms are currently available at different sites.
·       At the LPP-Saint-Pères site, there is a full infant laboratory, made up of different test booths for all major behavioral techniques (HPP: Headturn Preference Procedure; IPLP: Intermodal Preferential Looking Paradigm; Switch task; Tobii Eyetracker) and an electrophysiological method (Evoked-Response Potentials) that can be used to test infants from about 4 months of age up; it is also equipped for testing adults. Furthermore, the team has an antenna at the Port Royal maternity for testing newborns using the NIRS (Near Infra-Red Spectroscopy) brain imaging technique.
·       The LPNCog-Boulogne-Billancourt site is provided with the technical equipment necessary for most kinds of psycholingustic experiments , including several soundproof rooms, eyetracking labs (Eyelink II, Eyelink 1000, Dr Bouis Eyetracker, Dual Purkinje Image Eyetracker), an EEG lab for evoked-response potentials. Neuroimaging studies can be carried out in collaboration with the hospital Saint Anne, Paris.
·       The LPP-Paris 3 is particularly well-equipped with the technological means necessary for studies on phonetics and phonology necessary to study the coordination of the articulatory organs, in particular the vocal folds, the velum, the tongue, the jaw and the lips. The technical platform of the LPP-Paris 3 includes static and dynamic microphones, an EVA2 station, an electroglottograph, a non invasive  photoglottograph, a photonasograph, an electropalatograph, an echographe, different cameras and a system for motion capture.