Making sense in music
An enquiry into the formal pragmatics of art (part 11)

by Jos Kunst (1978)

Appendix V

A Concept is a Recognitional Capacity


Suppose we have a robot R which causally reacts to certain events in its environment. Let the occurrence of a lightningfiash be one such event. Let, in case a lightning-flash occurs, the reaction of R consist in, among other things, the inscription on a tape which is part of it, of


I take it then that the occurrence of :: on the tape is just as 'exterior' to R as the lightning-flash which caused it. In fact, for R, they are one and che same event.


Now suppose further that R has a scanning and matching program (SMP) that is triggered into action each time something is written on the tape. If it finds that something occurs for the second time on the tape, it labels the second inscription with a special sign, e.g., if the lightning-flash occurred again,


where we take the natural number to represent the bookkeeping of the number of repeats found by R's SMP. This procedure may then give rise to configurations like the following occurring on the tape:


which says that (is the trace of the fact that) £ occurred for the first time, :: for the sixth time, and ::;% for the fourth time. In this way, it is recorded that four out of six times the lightning-flash occurred in (some, and in part by R specifiable) conjunction with the event(s) causing R to write ;% on its tape.


R is a concept forming machine. A distinction can now be made between the lightning-flashes and the :: inscriptions: whereas it may be the case that the lightning-flashes present no differences, we have now differences, produced by R and detectable by its SMP, not only between the unlabeled inscription and the labeled ones, but also between the different labelings. These inscriptions differ in a way to which no corresponding differences can be found in the events that caused them. The machine has developed a sense of time.


The SMP mentioned above contains criteria for the identity of inscriptions which, if concepts in the above sense are safely established, amount to a rule-governed syntax deciding which inscriptions are well-formed and which are not. Obviously, rules are then formulated with respect to that set of well-established concepts.


In practice, R's adequacy as a model would be greatly enhanced if it also incorporated a troubleffhooting program (TSP). This TSP (in which the capacity of self-reference certainly would not come amiss) would then take on to try out possible corrections of "almost well-forme " inscriptions — it would take on to "jump to conclusions", to "filter out noise". The above-mentioned syntax will then have to play a key role.


Truth conditions can also be made sense of, with respect to inscriptions on the tape which have been found to be, or have been made, well-formed. An inscription may then be called "true" iff the identity construed on the tape is matched by some identity of the events which caused R to make the inscriptions in the first place. If R has also action eomponents, and if the results of its actions register in some way on its tape, it will have a use for this semantics: it will enter into contact with the outer world, and its actions will create a feedback on its conceptual constructs. It will test its concepts, and develop a "model of the world". Let, by way of example, the shutter before its window be part of such an action system; its TSP may then stumble onto the fact that if the shutter is closed, ;% occurs without :: preceding it (if you wish: thunder is heard without lightning having been seen).


Lastly, every inscription considered as a fact occurring will, given R's semantics, be pragmatically describable in terms of the change brought about by it in R's "model of the world". Thus (let the shutter have been smashed in), the sequence

shutter closed ::,% (with appropriate labelings added)

will have a very strong pragmatical relevance to it.

Natural language

Suppose now that there are more than one of those machines; suppose further that their actions co-determine the others' inputs; that they have one action system designed to do not much else besides co-determining the others' (and, naturally, their own) inputs; we will then have, I think, the minimal conditions for explaining syntax, semantics and pragmatics of natural language as a subcase of the inscriptions' (cf 0.0.3). R's TSP will be of the utmost importance here, as it will have to take on to "jump", on the basis of previously tested conceptual associations, from "word" inscriptions to possible percepts.


Genetically, syntax, semantics and pragmatics can only "get off the ground" together, as an integrated system; in this context it is interesting to note that newborns come into the world armed with a number of fully activated action systems, which may plausibly be taken to play an essential role in the starting up of the conceptualization process (cf Piaget).