In an essay I posted ten years ago (link), I quoted a letter count in the Torah. A representative of the Celerus Corporation contacted me to correct the count. The company provides digital Torah databases for developers and book publishers. Below are the corrected counts based on the company’s computer calculation.
I don’t see any theological significance in this revision. A manual count is prone to error. I publish this only because it seems like information worthy of publication. (For the record: I have not verified this new count.)

See “Proofiness” by charles Seife for a compelling discussion how, at a certain point, people cannot be exact in counting. He applies it (and it’s based on, in the main) elections and counting ballots, but would, I think, apply here as well.
To prove that point. According to the chart above, there are 16 letters in which Celerus has a higher number than Even Shshan (about 100+ letters). Yet the total is the same. Something’s wrong.
Since the total numbers are the same, and only the letters differ, might it suggest another reason for the miscount other than human error?
Joseph – pay attention to the צ – there’s nearly 100 more in the Even Shoshan than by Celerus which makes up for it. But that overestimation/count is very perplexing.
“To prove that point. According to the chart above, there are 16 letters in which Celerus has a higher number than Even Shshan (about 100+ letters). Yet the total is the same. Something’s wrong.”
You’re right. Specifically, though, Celerus says that there are higher numbers of 16 letters than Even Shoshan, but 90 fewer tzadis (and that’s the only one with fewer). Seems like a drastic over/undercount, but I don’t have the data they’re working with.
Tzadi and Ayin can be easily confused, not sure about the other letters though.
That can’t be it. They have the same number of ע, more or less. That there would be exactly the same number of letters, and that there is that huge a discrepancy on one letter to account for the others, strikes me as bizarre.
Doesn’t Ibn Shoshan count final forms separately? What was done about them here? (Note that Tzadi is one of them.)
Also: Yemenite would be off by about eight, right?
Just checked. Yemenite has six less and two more, for a total of four less. Note that the Breuer editions follow the Yemenite, with a note.
A error of ten on “yud” raises eyebrows statistically, but an error of nearly 100 — and an overstatement of number at that– seems of the scale of probability for Even Shoshan or any other professional publisher. I would question that….perhaps both are incorrect and the answer lies in the middle.
Apologies, the error of nearly 100– and as an overstatement rather than understatement.– relates to the “tzadik” and presumably the “tzadik sofi”.
On review, I was mistaken to quote this from the Even Shoshan Concordance. It is the Tanach Yehoash (a Yiddish translation). I got confused last night at the last minute and mistakenly changed it.
(How) do you count the backwards nuns?
I would have expected differences between “look alike” letters. For example, kav and beit, or vav and zayin. To have Tzadik off by 90 and the others off by 1’s and 2’s means that when they did the original count, when they saw a tzadik they counted it as a beit or a kaf or a mem? Doesn’t feel right.
Of course I have no idea what happened but I have seen accountants who when faced with a discrepancy between the whole and the sum of the parts, when they are sure the whole is correct, say “just throw the delta into category X”
KT
By the way, the letter is named Tzadi not Tzadik.
http://he.wikipedia.org/wiki/%D7%A6
I did look into this in 2000 based on some Chumash databases I had found. I have just checked back on this, and I get exactly the same figures as in the Celerus column (which probably just means that the Chumash text I had was produced by them).
There is a list eg at the end of the standard Torah Temimah of the numbers of each letter. This differs significantly for a few letters, eg final ץ – there are 1035 in my count, but 1067 in the Mesorah count – this is by far the largest difference. OTOH there are 13 too few נ and 8 ך
I also had a download of BHS (Leningrad codex) but it appeared that that had an extra 56 letters overall.
Rav Saadia Gaon has a letter count for all of Tanakh.