Some Thoughts on WordPrint

Just a quick post on the current kerfuffle over wordprint studies. Wordprint studies are a type of stylometry that look at certain connective words that aren’t main words in a sentence. The claim is that they can determine the authorship of a text. Now I’ve always been skeptical of this, even back in its heyday in the 90’s.

The main problem is of course that depending upon how you slice up the text you get very different answers. More significantly with the text from Mosiah through Mormon the author is primarily Mormon. It’s basically impossible to tell, even if a figure is speaking first hand, what is Mormon summarizing in his own words versus what the original speaker said. I’ve also always have in the back of my mind the worry you see in econometrics. There sometimes the data is sliced and resliced until a desired result appears with an appropriate p value. Of course this isn’t quite the same, but in the back of my mind that’s long been my worry. There’s a lot of subjectivity to most of these studies of the Book of Mormon.

I’ve thus been a strong skeptic of wordprint studies back when they were so popular among apologists. Every few years a critic does a wordprint study and I remain critical for the same reasons.

The problems go well beyond the problem who who wrote what. There are good reasons to assume that the underlying text is highly compressed and possibly consists of ideograms. Joseph Smith and others at the time thought this was true of Egyptian but Egyptian was actually primarily phoenetic. That doesn’t mean whatever mysterious cipher Nephi and Mormon are writing with isn’t more ideographic in nature. Indeed Joseph appears to think they are. The descriptions in the text of the Book of Mormon at minimum suggest a high degree of textual compression and plausibly ideograms. The other huge problem is that more and more are coming around to Brant Gardner’s model of a very loose translation of the underlying text.[1] There are good reasons for this such as apparently using KJV and other phrases to translate a similar idea on the underlying plate. However that mean that if we have quotations used frequently to translate that the connective words would be highly biased from the KJV and other phrases.

The biggest problem is that problem of ideograms though. I don’t see how it is solvable. Admittedly that the underlying text is made of ideograms of some sort (including mnemonic shortands) is a speculative theory. But it seems a theory not easily dismissed. The problem is that with such a text there are no connective words for wordprint to operate on. They would be added in according to the aesthetics and methodology of whatever process is doing the actual translation.

There’s an easy test for all this of course. Test the methodology with texts based upon ideograms and see what survives translation using wordprint analysis. It’s telling that neither apologists nor critics have made this obvious test. Compare two translations of say the Tao Te Ching. Or better yet the Analytics comparing passages that scholars have concluded have different authors.[2] Getting translations of the Analytics along with authorship lists should be easy. I’ll go out on a limb and say the wordprint methods won’t work. If they won’t work there then I think we should dismiss them relative to the Book of Mormon.[3]

 

  1. To be clear, since people confuse this all the time, the translation can be loose while the control Joseph uses with the Urim & Thummim or seer stone can be tight. They are orthogonal claims.
  2. To clarify Chinese is not a pure ideogram system. So it almost certainly contains more information than I suspect the Book of Mormon does. My point is just when you haven’t even done those cases it’s hard to take claims towards the Book of Mormon seriously.
  3. Want to argue about Shakespearean authorship? Go for it. There are least there some ground for the methods. I’m still extremely skeptical but at least there’s a level of plausibility. That said I’d have far more confidence if we had more scientific testing. The claim is that wordprints are like fingerprints. However again I’m pretty skeptical. I should note that identify or rejecting 19th century authors might be doable. As I said I’m skeptical but my main target is authors within the text.

8 comments for “Some Thoughts on WordPrint

  1. Statistical work that starts with one’s own special sauce of how the variables are formed is always suspect. Lots of degrees of freedom there that can lead to invalid statistical tests. The psych people have been hammered by this the last few years, but the problem is widespread.

  2. If I’m understanding things correctly, BoM Central is saying that the diversity in the voices of BoM authors is even greater than what experts have found between characters created by notable 19th century novelists. Now, assuming that the BoM was tested in the same way as the fictional works were — and remember, all of the texts in the comparative study were written originally in English — how is it that, with regard to voice diversity, the BoM surpasses them as a *translated* document? Either Joseph Smith was literary a super genius or the BoM is a product of multiple authors–real people whose voices would likely be unique enough to maintain some of their distinction after going through a process of translation.

    That said, it seems to me that the results of the study either, 1) provide evidence for multiple authentic voices in the BoM or, 2) prove that the study itself is so severely flawed as to be completely untrustworthy.

  3. I think it is fair to request a control document for a method and to remain skeptical until such a control is provided… What is not fair is too claim that since no control was provided the method is obviously wrong…
    Call it an interest unsubstantiated method, for sure, but you didn’t go out on a limb, you nose dived onto the Rocky field below… Good luck recuperating from that spiritual head injury…

    Lack of evidence is not evidence of lack…

  4. I’m not sure a control document gets at the types of critiques that the method is open to. My main criticism isn’t just about a control document but is that the underlying text most likely doesn’t contain the data that this method depends upon. Now where you might be getting a bit confused is that I said there might be some plausibility if they could demonstrate this method on translation from texts closer to ideograms. (Again the Analytics aren’t ideograms – but they’re closer than European languages) That they haven’t even attempted this undermines the plausibility.

    This isn’t an “absence of evidence” argument at all but is an argument based upon the nature of the purported source languages/scripts.

    I believe they have done an attempt of a control document but I believe it was a German to English translation which tells us nothing about ideograms to English. English and German are closely related and presumably the markers used for identification would be in some form in both language.

    An bigger concern is how much data fitting is going on. i.e. to what degree are the identifying markers chosen and tested on documents independent of Mormon documents prior to doing the Mormon documents. Otherwise it’s very easy to even subconsciously fit the method to the desired structures. (This is why double blinds are so important in science)

  5. It’s interesting to me that “wordprint analysis” and “wordprint” are almost exclusively reserved for Book of Mormon studies. You don’t find those terms associated with any studies from the AML. The idea of peer review doesn’t seem to apply at all.

Comments are closed.