How many of Bani's offspring returned from Babylon?

What follows is one scholar’s way of reading this disagreement — a critical-historical view that often concludes the texts genuinely disagree. It draws on the peer-reviewed sources listed below and stands as one position among the answers above, presented for transparency, not as GEBible’s position.

How critical scholarship weighs this: Widely regarded as a genuine contradiction

This summarizes where critical scholarship tends to land on this tension — a report on the field’s center of gravity, not GEBible’s own ruling.

Ezra 2 and Nehemiah 7 are not two independent rosters. They are two transmissions of a single document — the registry of the first wave of returnees who came back from Babylon under Sheshbazzar and Zerubbabel. The frame is explicit: Nehemiah 7:5 reports that Nehemiah found "the book of the genealogy of those who came up at the first" (sefer hayyachas) and copied it into his own record. Source-critically, then, both chapters depend on a common underlying list (often labeled the Golah list in the scholarship), and the question is not whether two events produced two numbers but why one document, copied twice, produced two different numbers for the same family. For Bani/Binnui the divergence is doubled: Ezra 2:10 reads 642, Nehemiah 7:15 reads 648, and the patriarch is named differently — Bani in Ezra, Binnui in Nehemiah.

The standard apologetic response, articulated in Gleason Archer's Encyclopedia of Bible Difficulties, is that one list counts those who originally enrolled to leave and the other counts those who actually arrived, with deaths and defections en route accounting for the gaps. A more candid wing of the conservative tradition simply concedes scribal error in the numerals, which is the more defensible move. The two-census theory does not survive contact with Nehemiah 7:5: Nehemiah is not conducting a second enrollment, he is explicitly reproducing an earlier register. If he were updating the figures to reflect later arrivals, the text would say so; instead it claims faithful transmission of the older document.

The Binnui/Bani alternation is the most telling detail. This is one family name preserved through two textual streams, with each stream stabilizing on a different form. Bani and Binnui (Hebrew bny / bnwy) differ by a single internal vowel-letter (waw), the sort of variation that arises routinely in transmission and that recurs elsewhere in Ezra-Nehemiah's own lists (compare the Binnui of Nehemiah 10:9 and 12:8). That is the signature of manuscript drift, not deliberate theological revision. Williamson, Blenkinsopp, and Klein all treat these divergences as transmissional rather than substantive: the underlying figures are unstable across the Masoretic Text, the Septuagint, and 1 Esdras. 1 Esdras 5 — a Greek text reproducing essentially the same list — supplies yet a third set of numbers for several families and aligns now with one Hebrew witness, now with the other. Where three witnesses disagree about a number, the parsimonious inference is corruption, not inspired contradictory accounting.

A frequent supporting claim is that ancient Hebrew used alphabetic numeral signs (aleph for one, and so on) that scribes easily miscopied. That mechanism is real in some Second Temple documentary contexts, but the Masoretic numbers in these lists are written out as words, not signs, so the corruption is better explained by ordinary copying slips in numerical material — a category notoriously prone to error precisely because numbers carry no semantic redundancy. A misread name can often be caught from context; a misread number cannot. Emanuel Tov catalogs exactly this vulnerability in the transmission of numerals across the Hebrew Bible.

For the inerrantist the structural problem is acute, because both chapters claim to reproduce the same source. One cannot rescue the totals by appealing to differing perspectives or differing dates — the text itself forecloses that route. Either the original register carried one figure and one of the two transmissions corrupted it, or both drifted from an archetype we no longer possess. Either way the doctrine of a perfectly preserved text cannot accommodate the data without special pleading.

Where the critical reading lands: genuine discrepancy, almost certainly scribal corruption in numerical transmission, with a parallel onomastic variant in the family name.