The current procedure resulted from design interaction between the following steps or approaches.

1. Activity word list

Since the preoccupation of international organizations extends beyond the ranges of the specialized thesauri noted above, one point of departure was to extract (by computer) all significant keywords from the names of organizations listed in the current edition of the Yearbook of International Organizations. To these were then added words extracted from the multi-disciplinary publication, Encyclopedia of World Problems and Human Potential (70), resulting, after suppression of prepositions and other non-essential words, in a total of some 20,000 words, including 1,000 word pairs (see below, 'Procedural revision'). A particular merit of this list is its comprehensive coverage of active concerns of the international community, whether problem, discipline or value oriented. The computer system is designed so that this list can be re-extracted at any time to capture new words associated with new organizations or preoccupations.

2. Interrelating major classes

The various international thesauri noted above were used to isolate major classes (eg science, religion, etc) which have traditionally proved to be a practical basis for grouping concepts. Particular attention was however paid to 'awkward' classes which did not fit naturally into such groupings (eg standardization, design, and systemology are treated as 'general' or 'interdisciplinary' classes in the case of the Unesco Thesaurus). Also of interest were classes that had for convenience been forced within other classes even though they represented a relatively distinct concern.

3. Elaborating a matrix of distinctions

Using the major classes derived above in the light of the variety reflected in the extracted word list, considerable time was spent in juggling items into some sort of matrix form. This process, as an exercise in design, was very much a blend of science and art as described in Christopher Alexander's Synthesis of Form (18). The matrix was not perceived as being a purely logical clustering of fields of knowledge but rather a pattern of activity domains in which the degree and quality of objectivity varied. The constraining factors which emerged as useful in this process include the following:

• a) The avoidance of entrapment in a purely linear sequence by somehow including a non-linear patterning feature. This was achieved by considering neighbouring columns and rows of the matrix as functional complements of a mutually counterbalancing nature, rather than simply as members of a logically defined set.
  
• b) The perception of matrix cells as representing functional domains of which only some might have a cognitive emphasis. The words that can currently be placed 'in' such a 'semantic cell' do not therefore necessarily exhaust the meaning that may come to be associated with that cell. The words are indicators of significance but they do not delimit it.
  
• c) Following Dahlberg's approach, the use of rows of the matrix to distinguish different functional 'levels'. The order is then such that the 'lower' or more fundamental levels must first 'emerge' prior to the 'higher' levels for which they provide a foundation. The succession of levels thus constitutes a developmental sequence.
  
• d) At any given level, the representation by the cells of the row in question of a set of interdependent functional domains whose interaction is essential to the stability of that level, in effect the expression of one evokes the expression of the others.
  
• e) The ordering of the cells of the matrix, in the light of the previous points, to go some way towards reflecting the attitudes and behaviour of those associated with them as in: the 'pecking order' of the sciences; the 'non-scientific' nature of certain domains; the less 'concretisable' characteristics of some domains.
  
• f) When appropriate, the ordering of the cells to reflect the order of 'emergence' of functions, either as they become explicit in a community (in roles or programmes, for example) or as they can be explained in the stages of some coherent educational pro gramme.
  
• g) In contrast with the usual practice in classification schemes, the avoidance of grouping everything associated with a given subject into a class which primarily reflects the expression of some intellectual discipline (eg political science, sociology). When appropriate, words associated with such distinct orientations as social praxis, material conditions, theoretical approaches, value expression and modes of awareness should be separated into different levels, although possibly in the same column. Thus 'love', and 'sex' would not necessarily be grouped under 'psychology' (as is done in the Unesco Thesaurus).
  
• h) Just as the previous point stresses the need to counteract the tendency in favour of a theoretical emphasis, so attention would be given to counteracting an anthropocentric emphasis (eg 'fish' as a sub-class of 'agriculture' in the OECD Macrotheasaurus) or a legalistic emphasis (eg 'prostitution' as a sub-class of 'crime' in the Unesco Thesaurus).
  
• i) Distinction would be made between levels constrained by nature or patterns of behaviour, those at which category boundaries were called into question, and those at which the initiation of change or development was emphasized. This offers a means of separating functions concerned with analyzing or reacting to the human environment from those concerned with various forms of development, whether individual or social.
  
• j) With regard to the levels related to social praxis, the cells would each be associated with characteristic institutional features of society such as: government ministries or portfolios (in simpler and more developed administrations), university faculties and functionally specific building (eg hospital, factory, military base, school, laboratory, etc).
  
• k) The size of the matrix needs to be constrained by its comprehen sibility, as determined by man's difficulty in dealing with more than approximately seven categories unless extensive patterning features are incorporated as mnemonic coding devices (32). There is an obvious practical advantage in computer processing if the cells can be defined in terms of the decimal system, as in the case of Dahlberg's proposal.
  
• l) Although the pattern of matrix cells is conceived as being com plete, the representation of the content of those cells should be open to continuing development. Thus the range of words reflecting the significance of each such cell may change (aside from the possibility that words may be allocated to more appropri ate cells). In particular the cells corresponding to more existential or value-related concerns should be open to future clarification (possibly in the light of the very extensive Eastern reflection on such categories). As noted earlier, it is the words signifying dimensions awkward to associate with the earlier cells which raise the possibility that they should be associated with some other cell to which few words have been previously allocated. In this sense, it is the 'earlier' portion of the matrix which is 'complete', whereas the open-endedness is primarily associated with the 'higher' levels.

The process of distinguishing qualitative attributes and their analogies to one another bears an interesting resemblance to the documented history of the manner in which chemical elements were slowly juggled into a meaningful periodic pattern (16). As in that case, part of the problem lies in the fact that words often refer to qualitative 'compounds' of two or more elements although the distinction between an element and a compound may well be unclear.