Wittgenstein’s views on ‘private language’ have been concisely summarized by Radden (2011, p.70):
“… he insists that an ideosyncratic ‘private language’ could not be a proper language. Meaning and significance are tied to how words are used, and such use occurs within some linguistic community. Only a mistaken conception of meaning could permit us to envision the possibility of a ‘language’ for one person only.” [This not quite what Wittgenstein said, but the point is useful.]
Complication 1: Suppose the inventor of Esperanto had never convinced anyone else to learn that language. Would Esperanto then be a private language in Wittgenstein’s sense? Well, no, you might say, because ‘in principle’ Esperanto could be taught to thousands of people and used fluently, which it in fact has been.
But how would such a ‘principle’ be formulated? Validated? Applied to Esperanto? What are the criteria for determining that Esperanto would or would not be a ‘real’ (non-private) language in principle, if no more than one person were ever to speak it?
Complication 2: A program written in a computer language can be understood, obviously enough, by a suitable computer program (interpreter, compiler, or assembler, for instance). And we know it’s been understood because the resulting program can execute (whether or not it executes exactly as intended). Is that computer language a ‘language’ simpliciter, or is that only a metaphor?
Complication 3: A computer language is a language if programs written in it can be read and turned into executable code by a suitable computer program. Suppose there is only one such computer program (a compiler, for example). Is that language then a private language? Or would the existence of instances of that compiler on many different machines count as making the language a ‘real’ language, when otherwise it would not be? Why?
** Actually, the intelligibility and use of programs written in the Algol language was quite independent of their translatability by machine. Many short Algol programs were published to be read by people, not compiled and executed.
Appendix: An Example of Computer Language Archaeology
A very long time ago, I was assigned to modify a user’s Honeywell H-400 assembly-language program. Digging into the code, I found its structure rather strange. Digging further, I found not only that it was an almost line-for-line translation from IBM 1401 assembly language, but that the 1401 code was itself a literal translation from wiring of the system where it had originally been developed: Not a computer, but an IBM 407 PCAM (punch-card accounting machine, or ‘tab machine’). There are many examples where human-language writings have undergone an analogous recoding process. This is called translation, of course, and the number of language layers can reach an arbitrary depth.
