I’ve been having an evil thought about the Khirbet Qeiyafa ostracon. It’s not the devil that causes me to have this evil thought but rather the (so far) unintelligible string of glyphs that make up the first four (three and a half, if you prefer) lines of a different ostracon, the ‘Izbet Sartah ostracon. The fifth line of the ‘Izbet Sartah ostracon is an abecedary but the first four lines appear to be random like strings of glyphs. Is it possible that the seemly intelligible strings of glyphs on the Khirbet Qeiyafa ostracon fool us into thinking that the whole thing is or should be intelligible? That’s my evil thought.
While it is too early to claim anything like a consensus, I think the majority of attempts at understanding the Khirbet Qeiyafa ostracon take the first five glyphs, reading left to right, to be intelligible in Hebrew (or something like Hebrew), אל֯ תעש. If you disagree with this for whatever reason and some scholars do, most of following analysis will not mean much to you. However, those holding various alternative views can still use the same general methodology and thought process. To reformulate my evil question: Are the approximately 70 glyphs on the Khirbet Qeiyafa ostracon in a random or near random sequence? First, I will look at the probability that the first five glyphs, as most commonly read, are the result of a random or random like sequence and then I’ll look at the probability that the whole text is a random or random like sequence. You’ll see what I mean by “random like” later.
The probability of finding any given five glyph contiguous string from 22 different reusable glyphs in a randomly generated sequence is 1 in 5,153,632. Now how many intelligible strings of five glyphs are there? I’d guess is that there are over 1000 but less than 10,000. Without seeing each of them, it’s hard to know what to count as intelligible. So for the purposes of estimation, let’s assume the larger of these numbers as the number of five glyph strings that represent intelligible sequences in any of the possible dialects across the most likely dialect continuum. Then the probability of there being one intelligible string in any random string of five letters is about 1 in 50, in other words, about 2% probability. But, if there are, as there seems to be, other intelligible strings on the Khirbet Qeiyafa ostracon, the real probability of the glyphs on the ostracon representing in every place a random or random like string is far less than 2%. Based on this and some simulations (see below) I think it quite unlikely that the approximately 70 glyphs on the Khirbet Qeiyafa ostracon constitute a purely random or random like sequence. Perhaps the probability is not low enough to make it completely perverse to think that some or all of the writing is random or nearly so. But I think such a conclusion would have little justification as a starting hypothesis (not that I know of anyone who does), particularly so if one reads the first five letters as אל֯ תעש and sees other intelligible strings on the ostracon. On the other hand, the probability may still be high enough and the nature of the distribution of glyphs uncertain enough for scholars to eschew overly forced attempts at a coherent end-to-end interpretation.
Looking at this problem from another direction, assuming that the Khirbet Qeiyafa ostracon has 70 glyphs, the number of possible permutations (order matters) of 22 differencing glyphs taken 70 at a time and allowing repeated glyphs is a little more than 9 followed by 93 zeros or many orders of magnitude more than the number of stars in the universe. That’s a lot possible ways to arrange 22 glyphs in groups of 70. Some very small percentage of these permutations would make perfectly good sense in Hebrew or Phoenician or Aramaic or . . . Remember, among these possible strings are the first 70 letters in the book of Genesis and all other intelligible 70 letter Hebrew strings. Many more but likely still a very small percentage of permutations would have multiple glyph strings of various lengths that would seem to make sense. I struggled to find a way of calculating how many glyph strings would be partially or totally intelligible. But I couldn’t devise an algorithm that is defensible or even particularly coherent. I can’t even define what I mean by intelligible precisely. Perhaps one of you can. Some students of such things express the frequency of finding even short intelligible strings of literature is such a massive collection of possible strings in terms of billions and billions of “monkey hours” rather than as probabilities. I tried the next best thing to direct calculation, simulation. There is a random letter sequence generator online. Using the Hebrew alphabet, I asked it to give me five lines of 14 letters each. Here are my first two simulations.
While, given enough time and imagination, I might see a few seemingly intelligible things here or there, I don’t see any that extend over five letters. And none of them comes close to even the most limited attempts to see intelligible strings in the Khirbet Qeiyafa ostracon. They look more like the first four lines of the ‘Izbet Sartah ostracon.
I did about 100 simulations and never found any embedded strings that seemed to contain anything like the level of seeming intelligibility of the glyph strings on the Khirbet Qeiyafa ostracon. Of course, I may have missed some. 100 is a very small sample of the total possible number. But it does tend to show that intelligible Hebrew strings are not extremely common in sets of randomly generated strings of letters. By the way, the same applies if one uses the Latin alphabet and looks for strings that read as English, French, German, etc. Such things appear to be of very low frequency indeed. It’s a good thing that finding intelligible strings among random letters is extremely infrequent. Otherwise, we could never be sure we were reading the work of humans rather than that of monkeys typing at a keyboard. Out of human hubris, I naturally assume there is an important difference.
People are not nearly as good at generating random strings of glyphs as computers are. When asked to make random strings of letters or numbers, a person tends to make judgments about what is and is not random and therefore introduces non-randomness. In addition, other, all too human, factors like aesthetic considerations render human attempts at randomness less than random. Even if the glyphs on the Khirbet Qeiyafa ostracon or the ‘Izbet Sartah ostracon do not constitute completely intelligible strings, I doubt they are truly random even if they might be random like.
The few rather certain intelligible strings combined with the infrequency of such things in simulation makes me reasonably confident that the text of the Khirbet Qeiyafa ostracon was intended to be intelligible. Whether it is (in toto) or not is another question. Enough evil thoughts for today.