Arabic numerals originated in right-to-left languages, where they were written and read starting from the least significant digit on the right. It was when the numbers were imported into left-to-right scripts that the order of digits was reproduced as-is, but read backwards.

> Arabic numerals originated in right-to-left languages,

This is incorrect. They originated in the LTR writing systems of India, which is why the full name is Hindu-Arabic Numeral System.

https://en.m.wikipedia.org/wiki/Arabic_numerals

If you read Arabic, this is clear because the numbers are written in the opposite direction as the words.

> they were written and read starting from the least significant digit on the right.

Only the ones and tens place are read from the right - which many languages do, including English in the 'teens' words:

thir-teen == three and ten

four-teen == four and ten

fif-teen == five and ten

In modern Arabic, the hundreds and above are read left to right.

For example:

42: ithnaan (2) wa (and) 'arba3an (40)

142: maa'a (100) wa (and) ithnaan (2) wa (and) 'arba3an (40)

That's fair. As you note by drawing a distinction, modern Arabic and formal/historical Arabic differ in their digit reading order: https://www.quora.com/Why-are-numerals-written-left-to-right...

I don't know if the original Hindu system read from most significant left digit first, but if it did then the historical transmission into Europe may have entailed two flips of numeric endianness.

> I don't know if the original Hindu system read from most significant left digit first

The Sanskrit language (exemplifying the "Hindu system") had the same pattern as English for the teens, i.e:

11: ekadaśa = eka (1) + daśa (10)

12: dvadaśa = dva (2) + daśa (10)

13: tridaśa = tri (3) + daśa (10)

etc ..

In Sanskrit this pattern (least significant non-zero digit spoken first) continues through to larger numbers. so:

20: vimśati

21: ekavimśati

120: vimśati-śatam

1121: ekavimśati-śatam-sahasra

etc.

But it's important to remember that the spoken rendering of numbers in all these languages precedes the written representation by millenia. People have been counting much longer than they have been writing. Of course later ways of speaking the numbers may have been influenced by writing systems, but the spoken rendering of basic smaller numbers tend to be relatively stable over time.

(And English has historically had more warts than that, e.g. “three score and ten” = 3×20 + 10, “four and twenty” = 4 + 20.)

> Arabic numerals originated in right-to-left languages

But I thought they originated in India, where left-to-right scripts were in use, which is why Arabic is bidirectional with the digits written left-to-right.

So in German the number “21” is spelled “einundzwanzig” which means “one and twenty” compared to “twenty (and) one” in English. If you were to say “einundzwanzig” to me I’d most likely write down the 1 immediately and then put the 2 in front of it, if it’s utmost important to get the number right. This is really common also among the Dutch (they have the same thing) as well.

And one of our most common mistakes when speaking English is saying the number wrong. For example, 85 is pronounced fünfundachtzig, five and eighty, but it would not be uncommon for a German to misspeak in English and say fifty-eight.

I wonder if this has anything to do with what you wrote?

Could be inherited from arabic as cited by another (does all western numbering system derive from hindu-arabic?). What really intrigues me is why specially the tens are reversed. If it were some practical utility or reliability feature (saying the most important first/last?) I'd expect a fixed order throughout. Maybe it's just a random irregularity.

Yeah I figured that was the case for arabic but in hebrew the most significant digit seems to be said first

Wait, so 1,672 in arabic script would have appeared the same way, but would have been read as (something like) "two seventy six hundred and a thousand"?

Historically, yes, exactly. Modern conventions now favor the left-to-right numeric readings for interoperability.