This blog post doesn’t really add anything beyond what the original paper says, though at least it links to it, though the OCR font replacement in the version he links isn’t my cup of tea. The original paper is thorough and well worth reading for anyone interested in designing keyboard-type input devices.<p>Here’s a link to a scan where the OCR doesn’t clobber the image <a href="http://mrserge.lv/assets/human-factors-engeneering-studies-of-the-design-and-use-of-pushbutton-telephone-sets.pdf" rel="nofollow">http://mrserge.lv/assets/human-factors-engeneering-studies-o...</a><p>The lowest errors and fastest keying rate were actually with the standard circular telephone arrangement (most familiar), followed by the two-row five-column layout, and those were also the test subjects’ preferred layouts. In the final test between the 3x3+1 grid vs. the 5x2 grid, the vote tally was 12-3 in favor of the 5x2 gris.<p>I assume the company’s final choice was based on some engineer or manager’s aesthetic preferences or non-human-factors technical criteria (e.g. it fit better on a particular phone body design), though I’ve never seen any concrete explanations of how the final grid layout was chosen. Or in other words, we still don’t have a solid answer to the question “Why are the numbers on a telephone keypad arranged the way they are?”
My school had us enter our IDs into a keypad to buy lunch, so I did this everyday. Then, I had to enter it in a computer and got annoyed as the system wouldn't take it. Turns out the differences was that the lunch keypad was set up like a phone with 1 on the top while computers were set up like calculators with 1 on the bottom. It had become muscle memory and I was swapping two digits because of the change in entry device. Ever since then I've thought the incompatibility baffling, as well as surprising that it doesn't cause more problems. I hadn't noticed it at all until then.
Why the difference between numeric keypads and phone keypads?<p>Numeric keypad layouts mimic calculator layouts, which in turn, mimicked the original ‘adding machines’. The inventors of these early devices did not perform extensive user based testing (as it often the case in technology; early adopters end up creating standards that others follow, good or bad).<p>The whole point of the article is the Bell labs undertook an extensive user based study to investigate which layouts might be better (Pretty much the first time this testing had been done on this issue). They tested both canonical forms of the (3x3)+1 layout and selected what we now know today as the telephone layout. (If I had to speculate as to why this layout performs slightly better than the calculator layout, I might suggest that, in the West, we read left->right and top->bottom and so this is the more natural numbering scheme).<p>Had extensive testing been performed by the original adding machine developers, maybe both devices would use the same layout.<p>Interestingly, the most creative benefit I’ve heard for keeping the calculator layout as it is, is because calculators are often used to sum up real world measurements, and these fall under to purview of Benford’s Law* Because of this, the keys which are more likely to be used as the leading digits are closer to the at-rest position of the hand and require less energy to move and press … yeah … I’m not convinced either, but it sounds good <p>* <a href="http://datagenetics.com/blog/march52012/index.html" rel="nofollow">http://datagenetics.com/blog/march52012/index.html</a>
Here's a fun tidbit. DTMF supports ０１２３４５６７８９＊＃ as you probably know. But it <i>also</i> supports Ａ, Ｂ, Ｃ and Ｄ - not present on modern phones. These tones were added for use with remote computers and such, think today's phone menus.<p>The lack of them means there's a whole DTMF "column" gone. DTMF works on a grid. When you press a key, the tone for the row and the tone for the column are played simultaneously:<p><pre><code> (Hz) 1209 1336 1477 1633
-------------------------
697 | 1 | 2 | 3 | A |
-------------------------
770 | 4 | 5 | 6 | B |
-------------------------
852 | 7 | 8 | 9 | C |
-------------------------
941 | * | 0 | # | D |
-------------------------
</code></pre>
Given the DTMF grid matches the typical layout on dialing pads, I wonder if early DTMF phones would wire the pad in the same grid to generate the corresponding tones?
There's two things I found surprising. That they tested so many circular arrangements, and that there was a measurable difference between the two 3x3+1 orientations.<p>If the author of the post is reading this, the image of the two vertical columns is wrong. On the site it's essentially two columns, in the paper it's five rows. That is rows 15 26 37 etc compared to 12 34 56 etc. It makes for a surprisingly large difference, at least to me. The former looks completely unusable, the latter seems quite ok :-) So maybe I should not be surprised that the two 3x3+1 setups didn't perform the same...
How curious, I had no idea american dials went from 1 to 0 with 0 the longest to dial! Where I'm from, they went from 0 to 9, with 0 being the shortest number to dial, like this:<p><a href="http://www.skrotnissen.se/prylshop/common/pic/whitegoods/dsc06666.jpg" rel="nofollow">http://www.skrotnissen.se/prylshop/common/pic/whitegoods/dsc...</a><p>(This is why the emergency number is 911 in the US, whereas it used to be 90000 where I'm from)<p>Anecdote number two is that when I was working in Denmark in 2004 or so, I noticed that all the number pads on credit card readers were calculator-style, not phone-style, which was a legacy from when Dankort was rolled out in the early 80's, <i>just before</i> keypad phones surpassed rotary dial phones.<p>(In 2010 or around there, most retailers switched their card readers to phone-style layout, so all stores had to have small signs apologizing for it, but mentioning that the rest of the world had phone-style layout, so let's just all get in line with the program...)
Kind of ignores how DTMF phones worked. The tones are actually two tones played simultaneously. Each row and column are separate tones. Laying them out in a circle could work but its not very practical.
I worked as a telephone repairman in the dial era. Impatient diallers would often "push" the dial back rather than let it unwind naturally. Switch gear in the central office was sensitive to mis-timed pulses so wrong numbers often resulted. Also, as mentioned elsewhere, the early keypads sent pulses rather than DTMF, which made for an annoying clicking sound in one's ear.
I'm curious about the difference between the calculator and phone layouts.<p>I expect that the bias to the winning phone design came because of the bias in phone numbers - that is, because rotary dial phones would dial lower numbers more quickly, lower numbers were more sought after, and were allocated to higher population areas. This was both for personal convenience, and because it tied up the phone lines for less time with the dialing pulses.<p>I expect that this asymmetry also introduced a bias toward the 1-on-top dialing pattern that won, though I don't see analysis to that effect. (The original paper simply stated "that the arrangement frequently found in ten-key adding machines...was not the best [in its group]...the same geometric configuration with a different numbering scheme [one on top] was superior in keying performance... However the performance differences between the two were small".<p>Of course, the two arrangements primarily differ in which keys are obscured by your hand, and the proximity of the 0 key, which shouldn't be commonly used, so maybe this is just post-hoc reasoning...
Isn't there a simpler reason - a square design is easier to implement with a 3x3 matrix switch and the number layout ensures that the letters associated with the digits are in alphabetic order.
Related to this I recall reading (somewhere) about going from a keypad back to a rotary dial and how infuriating it is even though there's only a tiny difference in speed.<p>We humans get used to the speed as little as it is and having to go back to the old way and it's only 1/2 or even 1/10 second slower drives us crazy.
What I wonder about is why they didn't choose the calculator layout, or, conversely, why don't calculators use phone dial layout. In both cases speed and accuracy are very important. So, why the difference.
The circular dial pads had me look up a Nokia oddity.<p><a href="https://en.wikipedia.org/wiki/Nokia_3600/3650" rel="nofollow">https://en.wikipedia.org/wiki/Nokia_3600/3650</a><p>Note btw that it ran Series 60, meaning Symbian. Meaning that for all intents it was a smartphone.<p>And while not circular, Siemens also had a oddity:<p><a href="https://en.wikipedia.org/wiki/Siemens_SX1" rel="nofollow">https://en.wikipedia.org/wiki/Siemens_SX1</a><p>And yep, that one is also running Symbian with the Series 60 UI (licensed from Nokia).
I worked in a call center where people where simultaneously using a phone and keyboard number pad accurately and not even noticing. Subtle differences between the two but still interesting.
Did the "calculator" layout yield much worse results than the winning inverted layout? If so, I'm really curious about the reasons for the difference in performance.<p>Should actual calculators adopt the phone layout, too? I've always been curious why not all devices use the same order (PIN entry pads, phones, calculators, etc.)
There is a numberphile video about at&t's article. <a href="https://youtu.be/kCSzjExvbTQ" rel="nofollow">https://youtu.be/kCSzjExvbTQ</a>
'We still "Hang up" when we want to terminate a call, and this refers to the action of placing the receiver back onto the hook of a phone ...'<p>oh my god. I'm old.
The final winner three-by-three plus one is the only one suitable for mobile phone. Perhaps because in mobile phone screen is used as reference; in the phone keyboard the receiver used as reference. In the calculator and adding machine are not moved around and the table is used as reference.