Towards A Good Samaritan World

Thursday, September 07, 2006


In the course of doing some analysis work here at my job, I think I might have just made an intriguing philosophical discovery: Wittgenstein's private language argument is wrong!

My job was to analyze a bunch of data and then write a report. I had to include in the report many statistics based on the data, and tables and figures. I generated the statistics mainly with an econometric program known as Stata. No one else involved with the project knows how to use Stata. I used text files (CSV format) to import data into Stata. Also, Stata's output was typically in a text-equivalent format. The data was initially provided into Excel, and later on I put some statistical tables from Stata back into Excel in order to create figures (i.e., graphs/charts). The report, of course, was in Microsoft Word, and tables (as opposed to figures) I usually developed in Microsoft Word. Some day was also stored in Microsoft Access.

In the course of this analysis, which took about three months, I wrote thousands of lines of Stata code, manipulating about 1,400 data points (observations) and in some cases upwards of 5,000 variables; I generated dozens of Stata command files (do-files) and data files, as well as Word documents (drafts etc.) and Excel files and text files. Ultimately I'm responsible for being able to back up the claims made in the Report, which are based on the analysis. Also I sometimes want to change the claims, based on discovering errors, or changing my mind about the most relevant way of reporting certain information based on new evidence, or other considerations. For both of these purposes I need to know what I did.

Because of this, a recording system, a system of signs, a private language, as it were, is necessary. What I realized this morning, which provoked this post, is that the private language that I've developed is inadequate. I cannot, without great effort, carefully scrutinizing my work, figure out everything that I did. Not that I didn't try: I recorded a lot of notes in the Stata command files, I tried to give the files informative names, I made sure only to do analysis in do-files rather than in the normal command window because do-files stay on the computer, an "electronic trail," as it were. But it's still very hard. For one thing, there are a lot of dead-end files, files where I was exploring some question that turned out not to be interesting, but which I never deleted because, at the time I stopped pursuing it, it wasn't yet clear whether it would be interesting or not in the context of the broader analysis (or merely whether there would be space for it in the report) and I thought I might return. Now those files are a source of confusion: I try to decipher them, and there's nothing of relevance there, yet I keep thinking, "Maybe they connect to something," and of course a file that seems irrelevant sometimes does turn out to be the source of some statistic, so I must be careful about deleting something.

In short, I had a private language, but it wasn't sufficiently sophisticated or systematic. Now I have a better idea what the difference between myself and a more experienced analyst consists of.

Why do I describe this as a private language? It's a language because it's a system of signs, external to the mind/thoughts (I see this as the essential point), for the purpose of communication. Why is it private? Because the only intended author and audience of this language (i.e., both of the private language that I used, and of the more sophisticated private language that I wish I had developed) is myself.

It is perhaps not in principle private. Another analyst with Stata skills and a familiarity with the history of the data and the issues involved might be able, with effort, to decipher the data and make connections. I am aware of this and it was a small consideration motivating me to write the file, although I certainly don't expect that anyone else will go over my do-files and check my work, or that they would be successful (without my help) in deciphering it, if they did.

Anyway, why is the fact that my language might be intelligible to someone else in principle is decisive. One could imagine a solitary hunter-gatherer on a desert island who develops a sort of script to help him in his hunting-gathering tasks. He has positive and negative experiences with plants: some are useful, others poisonous. He has trouble remembering which are which, so on the wall of his cave, using charcoal, he draws a representation of the good and the bad. There's no great need for the pictures to be realistic: he draws, instead, stylized pictures, which will serve their mnemonic function. Or again, he marks the places where birds like to nest in spring-- to find their eggs-- with one sign, and the trail to a blueberry patch with a different sign. This person is, let us say, totally unaware of the existence of any other human being: he was born in the dark and placed in a basket which washed up on this solitary island. (Never mind how he survived infancy...) Of course, it's true that another person, an archeologist or psychologist, might be able, after coming across this person's haunt, be able to decipher his language through careful inference. But surely it is unreasonable for believers in the (anti-)private language argument to demand that believers in a private language prove a negative: that the private language could never be deciphered, i.e. to prove a negative. If you define a private language as a language of which there is only one intended author and audience, and they are the same person, a private language is certainly possible, and even useful and normal-- and I would be in less of a pickle at the moment if I had been more skillful in designing mine!

So why did Wittgenstein think that a private language is impossible? Here's Wikipedia:

Wittgenstein goes on to argue that a private language is impossible, which is known as "the private language argument". The exact nature of the argument is disputed, but it can be roughly assembled from several points that Wittgenstein makes. First, he seems to argue that a private language is not really a language at all. This point is intimately connected with a variety of other themes in his later works, especially his investigations of "meaning". For Wittgenstein, there is no single, coherent "simple" or "object" that we can call "meaning". Rather, the supposition that there are such things is the source of many philosophical confusions. Meaning is a complicated phenomenon that is woven into the fabric of our lives. A good first approximation of Wittgenstein's point is that meaning is a social event; meaning happens between language users. As a consequence, it makes no sense to talk about a private language, with words that mean something in the absence of other users of the language.

Another point that Wittgenstein raises is that one couldn't possibly use the words of a private language.[17] He invites the reader to consider a case in which someone decides that each time she has a particular sensation she will place a sign S in a diary. Wittgenstein points out that in such a case one could have no criteria for the correctness of one's use of S. Again, several examples are considered. One is that perhaps using S involves mentally consulting a table of sensations, to check that one has associated S correctly; but in this case, how could the mental table be checked for its correctness? It is "[a]s if someone were to buy several copies of the morning paper to assure himself that what it said was true", as Wittgenstein puts it.[18] Ultimately, Wittgenstein argues that, while one may have direct or privileged access to one's current mental states, there is no such infallible access to identifying previous mental states that one had in the past. That is, the only way to check to see if one has applied the symbol S correctly to a certain mental state is to introspect and determine whether the current sensation is identitcal to the sensation previously associated with S. And while identifying one's current mental state of remembering may be infallible, whether one remembered correctly is not infallible. Thus, for a language to be used at all it must have some public criterion of identity.

Well. Of course if language is simply defined as interpersonal then the argument against private language is tautological. But if Wittgenstein thinks that "meaning happens between users" then he is neglecting that the persons on the two ends of this between-ness may be my past self and my present self, or my present self and my future self; nor is this some arcane hypothetical but is actually the rationale for a very large amount of perfectly normal language use. I may write something down to remember it far in the future; or I may write it down as a means of communication, as it were, between myself of now and myself of ten seconds from now. Thus, I doubt I could have written this entire blog post wholly in my head; having a keyboard to type on makes a difference.

As for the infallibility thing, what the heck is he talking about. No, a private language can't be infallibly checked for its correctness. Neither can the utterances that occur in a public language, most of the time.


  • I don't believe that Wittgenstein's definition of private language is usually understood to be quite so broad - it could only refer to private experiences that were essentially unknowable from an outsider's perspective. As soon as one's referents became public objects, then the signified half of the sign becomes publicly accessible even if the signifier remains (contingently) private.

    I believe that most would allow your old self talking to your current self to count as a social exchange and thus an instance of private language at least in attempt. However, if the referents are entirely private experiences, then you have only your own memory on which to judge content, rendering the verdict always indeterminate in a way that a public language is supposedly not.

    Personally, I am not quite sure of the importance of the private language discussion since I have a less rigid, essentialist conception of what is public and private, internal and external. In fact, I might even suggest that there's more in common with your example of Stata notes and human episodic memory than almost anyone in Wittgenstein's time would have appreciated, casting doubt on the coherence of the proposed public/private dichotomy.

    By Blogger Nato, at 10:57 AM  

Post a Comment

<< Home