The 123 on the bottom makes some sense for calculators because calculator keypads, unlike telephone keypads, are often used to enter data, physical constants, and mathematical constants.
Those quite often follow Benford's Law [1]. That means if you are using scientific notation, where numbers consist of one digit to the left of the decimal point, the decimal point, and then a fractional part, then an exponent for the power of 10 the number is to be multiplied by, then you are much more likely to be entering a small digit before the decimal point than a large digit.
123, therefore, arguably belongs near the decimal point on the keypad on calculators.
I don't even think it's that technical, though I agree Benford's law does help.
The numeric keypad is largely based off of ribbon tape desk calculators [0], which is why the + is double size, the enter (=) is double size, and the 0 is double size. The + and Enter are designed to be used by the right hand pinky, while the 0 is designed for use with the thumb. The numeric keypad is designed for five finger entry. Leaving it in that configuration means that people in an office who are used to using a ribbon calculator can more easily adapt to the PC keyboard. That's what office workers expect to find on a machine that sometimes is just a calculator.
If you go into a business or accounting office today, you'll still see people using these stupid calculators (stupid because they waste paper and ink). And if you've ever seen someone use one that's good with them, they can be extremely fast. Faster than you'd ever need to be entering a phone number (unless you need to enter the IT Crowd's emergency number).
When you use a telephone, however, you only use one finger. You're only entering 7 to 11 digits, typically.
The article mentions the 1960 Bell System user study of different keypad layouts. I can't speak to the other layouts, but I think in hindsight it's clear why the "touch tone" layout (with 123 at the top) won over the the "calculator" layout (with 789 at the top).
Everyone or nearly everyone in the user study was surely a telephone user. At the time, that meant being accustomed to the rotary dial layout, where the order of the digits is 1234567890. 0 is the last digit, after 9.
Both the calculator and touch tone layouts had 0 at the bottom. So the calculator layout has 0 next to 123. The touch tone layout has 0 next to 789 (and following the 9), just like a rotary dial.
This must have accounted for some of the preference between these two layouts.
The numeric keypad on computers has stayed remarkably consistent given how adept manufacturers are at moving the other keys around the keyboard. It is not a given that Page Up/Down, cursor keys, the delete key and others will be where you expect them to be by convention, these keys can be moved around for 'designer reasons' even on keyboards from the same manufacturer (yes, you, Logitech).
I have yet to see the numeric keypad on the left for left handed users that type in lots of numbers, e.g accountants, data entry clerks etc. We also never had desktop computers integrated into landline phones with a 'phone lock' button to switch over to 'dial someone' mode. At all times the numeric keypad has opted to follow convention as defined by the original IBM PC circa 1981.
It's a considerable inconvenience for me now to encounter a phone-style input. I so rarely dial numbers that aren't already in my contacts that it is unusual to encounter, and many phone number inputs just use numeric layouts. I have to stop and think or my 1s are all 7s.
This reminds me of the two opposing conventions for drawing memory maps --- do you put "lower" (numerically lesser) addresses at the bottom and increase to "higher" addresses at the top, or start with 0 at the top and "increase" downwards? Just like with the Endianness Wars, it seems that multiple conventions often prevail when there isn't a clear and obvious advantage of one.
One of the articles cited is available for free: <https://archive.org/details/bstj39-4-995>. I find it interesting to note that the engineers also made variable force profile pushbuttons to test different springinesses as well.
> In 1902, the Dalton went on to become one of the most popular 10-key adding machines of the time, rendering multi-column calculators obsolete.
Multi-column calculators survived much longer, being king of the hill for high-accuracy multiplication and division. I had a job using one until the end of 1969.
On rotary phones, zero was usually at the bottom of the rotar, just after the after nine. When touch tone phones came out, I suspect many people preferred the zero at the bottom of the keypad since it was near the nine and better aligned with their existing muscle memory
Those quite often follow Benford's Law [1]. That means if you are using scientific notation, where numbers consist of one digit to the left of the decimal point, the decimal point, and then a fractional part, then an exponent for the power of 10 the number is to be multiplied by, then you are much more likely to be entering a small digit before the decimal point than a large digit.
123, therefore, arguably belongs near the decimal point on the keypad on calculators.
[1] https://en.wikipedia.org/wiki/Benford%27s_law