Methods of development

With the above listed aims and demands in mind, the NEUROGES system has been developed from 1995 to 2012 to its present version. From 1999 to 2013 the NEUROGES project was continuously granted by the German Research Association (DFG LA 1249 / 1-1, 1-2, 1-3). The methods of the development are described below in a quasi chronological order. Actually, different devel-opmental steps occasionally were pursued simultaneously.

(i) Critical review of existing coding systems

First, a critical review was conducted of the existing coding systems and of the reseachers' reports on their experiences with these systems. The review included coding systems used in gesture research, psychology, psychosomatics, psychia-try, psychotherapy, neurology, neuropsychology, linguistics, and anthropology.

The most relevant coding systems are described in Chapter 2. In addition to those, many more systems were studied (e.g. Liepmann, 1908; Krout, 1935; La-ban, 1950; Sainsbury, 1954; Mahl, 1968; Kendon, 1972; Dell, 1979; Haaland &

Flaherty, 1984; Le May et al., 1988; Duffy & Duffy, 1989; Poizner et al., 1990;

Ochipa et al., 1994; Foundas et al., 1995; Blonder et al., 1995; Hermsdörfer et al., 1996; Müller, 1998).

While many classification systems are effective tools for specific research questions, only few systems contain values that meet the aims and demands of the NEUROGES systems (see above). From the different systems, suitable val-ues, e.g. Efron's value deictic, were chosen as pilot values for the NEUROGES development.

(ii) Identification of relevant empirical findings on movement behaviour and its relation to cognitive, emotional, and interactive processes

To compile movement criteria that are potentially linked to cognitive, emo-tional, and interactive processes, the literature was reviewed. Included were empirical studies from gesture research, psychology, psychosomatics, psychia-try, psychotherapy, neurology, neuropsychology, psychomotor and motor con-trol research, linguistics, and anthropology. In addition, the author's empirical research findings were considered. Chapter 2 summarizes the most relevant em-pirical findings for the development of the NEUROGES system.

The review helped to identify movement criteria that are relevant with regard to cognitive, emotional, and interactive functions. An example for such a movement criterion is Freedman's temporal criterion discrete versus continuous.

These movement criteria were used to group movement values and thereby, to define categories.

Furthermore, new, empirically based values were identified that had, thus far, not been noted in the existing coding systems. An example for such a new value is act as a unit. This value was identified as a distinct behavioural unit in the author's split-brain research. Furthermore, if empirical findings strongly sug-gested more fine-grained distinctions of existing values, sub-values were gen-erated. As an example, split-brain patients used their left hand for deictics to themselves, and their right hand for deictics to the external space. This finding suggested that the two forms of deictics are generated in different brain hemi-spheres, and thus, that they constitute different entities. The new values and the sub-values were added to the compilation of pilot values.

(iii) Operationalization of the pilot values and organization of the values in a system

The pilot values that had been adopted from existing coding systems were of-ten insufficiently operationalized. Some of them were defined only by a de-scription of their function rather than a dede-scription of the movement. There-fore, the pilot values were re-defined according to movement criteria that refer to the visual appearance of the body movements.

The newly operationalized pilot values were ordered in pilot categories.

Values that were defined primarily by the same movement criteria, e.g. values referring to where the hand acts, were grouped in a category. The categories were then ordered hierarchically from simple to complex as defined by their number of movement criteria, e.g. the Structure category values (step 2) are defined by the criteria trajectory, dynamics, and phases resulting from changes in these parameters, while the Function category values (step 6) are defined by all criteria of the Structure, the Focus, the Contact, and the Formal Relation categories, as well as by the gesture/action space, the path during complex phase, the orientation, the shape, the efforts, and the body involvement.

(iv) Testing, reviewing, and re-testing of the pilot-system by the author The pilot system was tested by the author by analyzing many video tapes with different individuals in different settings. If the classification of a movement was ambiguous, the definitions of the possible values were elaborated and made more precise and distinct from each other. If a movement could not be classified with the pilot values, a new value was created. The Structure value aborted is the product of such as process.

This process, i.e., the testing of the values in a large sample, the review of the values with an improvement of the operationalization, and the optimiza-tion of the system was repeated several times. Finally, a first version of the coding manual, the NGC⁹ manual, was compiled.

(v) Testing of the system by independent trained raters

When the author reliably classified all movements, the next step was the test-ing of the system by independent raters. Several raters were trained with the NGC coding manual. They tested the system on a large sample of different individuals in different settings. The independent raters gave direct feed-back to the author concerning the intelligibility of the definitions and the user-friendliness of the system and the procedures. Discussions with the trainees of NGC / NEUROGES training workshops, which started in 2007 at the Berlin

9 The earlier version of the NEUROGES system was termed NGC system.

Gesture Center (http://www.berlingesturecenter.de/ seminare/seminare.html), further helped to improve the system.

During this process, the variety in subjective interpretation of the NEURO-GES values could be successfully reduced more and more. Furthermore, the user-friendliness of the system was improved. This was achieved especially by developing coding algorithms that guide the raters through the coding process.

(vi) Development of a statistical method to calculate the interrater agree-ment for moveagree-ment behaviour segagree-mentation

Another source of feed-back for the system was the statistically obtained inter-rater agreement. As it is a specific feature of the NEUROGES coding procedure that it comprises segmentation and coding, the interrater agreement has to refer not only to the value that is given to a unit but also to the segmentation of the behaviour into units. Raters may agree on the value of a unit, but not on the segmentation of the behaviour into units, e.g. if a motion constitutes a unit, and when exactly the unit begins and when it ends and the next unit starts. Vice versa, raters may agree on that a new unit starts, but they may give different val-ues to the unit.

In the stream of behaviour, the rater identifies a movement with a specific value and (s)he tags the beginning and the end of the movement with the specific value. This procedure im-plies the segmentation of the behaviour. As an example, in a one-minute lasting video-clip of hand movement behaviour, a shift, then a rest position, and then an irregular movement are identified. The rater then searches and tags the beginning of the shift, the end of shift - which is also the beginning of the rest position - , the end of the rest position - which is also the be-ginning of the irregular movement - , and the end of the irregular movement.

The perceptual process may, however, also happen the other way round. The rater first identifies changes in the behaviour, such as a change in the trajectory. Accordingly, (s)he segments the behaviour into units, and then gives values to the units.

Most likely, in coding both types of perceptual processes are ongoing.

Since not many movement behaviour researchers have segmented the ongoing stream of behaviour into natural units, until recently, no statistical procedure was available to calculate the interrater agreement concerning the segmentation of behaviour into units. A major project during the development of the NEU-ROGES system was, therefore, to create a statistical method that enabled to cal-culate the raters’ agreement concerning the segmentation of behaviour. The re-cently developed algorithm by Holle & Rein, the modified Cohen's Kappa, is described in detail in Chapter 15.

(vii) Review of the values based on the modified Cohen's Kappa scores After the successful development of the modified Cohen's Kappa, several studies with trained raters were conducted to further test the objectivity and the reliability of the segmentation process. The modified Cohen's Kappa score

for each value helped particularly to identify those values for which it was dif-ficult to identify their units in the stream of behaviour and to determine the beginning and end of these units. As an example, raters often had difficulties to identify the beginning and end of continuous irregular movement units.

The criteria for the segmentation of behaviour in each category and the op-erationalization of the values were improved. This process was repeated sev-eral times until a substantial agreement for segmentation and coding was achieved for many trained raters with different educational backgrounds.

(viii) Finalization of the NEUROGES coding manual

Over time, the increasing operationalization of the categories and values and the improvement of the instruction for the coding process resulted in a comprehen-sive coding manual. The 230 pages comprising Coding Manual is forthcoming as a book, supplemented by an interactive CD. The manual consists of seven chapters, each of which contains the coding manual for one category. The seven coding manuals for the categories Activation, Structure, Focus, Contact, Formal Relation, Function, and Type are always set up in the same way:

1. The category is defined.

2. The units or unit segments are specified that are submitted to the analysis with the respective category (i.e., not all units are further classified in all catego-ries).

3. The criteria for the segmentation of the behaviour and for the definition of the values are defined.

4. The values are defined. After a short definition that orients the reader about the content of the value a more detailed definition of the value follows. For each criterion, the qualities are explained that may occur in the value, e.g. what kinds of hand shape may occur in a deictic. Furthermore, for each value examples are provided. Finally, it is reported how to distinguish the value from other values or phenomena that are similar regarding the visual appearance of the movement.

NEUROGES trainees reported the latter information to be particularly helpful.

5. The technical procedure is described for coding the category with the NEU-ROGES-template.

4.5 Development of the modules, categories, and values, and of