The Nerd Corner

Inverse-T History

Here's how it happened. Please understand that this all happened in 1981 or 82, and I'm working from memory nearly 20 years later, so I may make minor errors.

The decision to use the inverse-T layout was made at a meeting of the LK201 keyboard committee, a group composed of engineers and researchers from all across the company. At the time I was working in a group called Human Engineering Research (HER -- over time the name evolved from HER to SHE: Software Human Engineering to SUE: Software Usability Engineering, as we moved from R&D to Small Systems, in the Maynar mill, to the VMS group in Nashua, NH).

The previous three keyboards (for the VT05, VT50/52, and VT100 generations) had been based on then current common industry practice, which in turn basically followed IBM's lead (compare DEC keyboards to the IBM Selectrics for instance). IBM was constantly refining their keyboards, so "common practice" was constantly shifting. This time around we were determined to base our design on sound human factors, the needs of software and the experience of hardware and software engineers. This worked -- witness the fact that this time IBM and the industry followed our lead.

The committee consisted of representatives from the hardware, software and research communities. There were lots of opinions and positions, not a lot of initial agreement and thus a lot of wrangling. But everyone was very dedicated to a sound engineering process, so opinions often collapsed in the face of evidence and research.

In the case of the inverse-T, the breakthrough in the analysis and the decision to adopt the layout all happened during a single meeting in a matter of just a few minutes. There had been several proposed arrangements rattling around for weeks or months, each with its proponents and detractors. I remember at least four:

The first one, done with 4 keys arranged in a 2x2 square and then rotated 45 degrees to make a diamond of diamonds was my personal favorite. The third one had a "home" key in the middle of a cross-shape. The two cross-shaped designs were the general favorites. The only thing anyone agreed on was that four keys side by side, the way they were arrabged in the VT100, was out.

VT-100 Arrow Key Arrangement

My group had conducted a number of experiments which gave us a huge collection of key-stroke logs for various text editors. Mike Good and I'd done a run of key-transition analysis on that data and I had a fresh pile of largely unread reports with me at the meeting.

While the Nth week of wrangling over the best choice was going on I started looking up the transitions that involved arrow keys and found first that transition from one arrow key to itself was at the top of each of the lists, especially runs of left and right arrows.

One of the next most common transitions (and the next one involving arrow keys) was from down arrow to left arrow. Up and down to right was also common (the exact order differed among the editors) as were transitions between down and up and vice versa. I started drawing the more common pairs out:

     

I quickly saw that the inverse-T put three of the four most common transactions next to each other.

->

Further, the three most common used non-alphanumeric keys in editors were the arrow keys (or their equivalents) and delete. In the ones that used arrow keys, the order of use was: left, right, down and finally up. Inverse-T put the three most used arrow keys all on one row.

I made a couple of sketches showing the relationships and all the favorite arrow key layouts and showed it to Dr. John Whiteside, our chief psychologist, during the next pause in the meeting. He agreed with me and we showed the rest of the committee. Everyone bought the analysis and we immediately moved on to one of John's hot buttons, the need to have small clusters of editing keys. He almost immediately sold us all on a set of six just above the inverse-T. The meeting then bogged down on issues around line feed/return backspace/delete and escape key issues that took weeks to resolve (and which came up again over the next few years).

Since within a few minutes of doing the first analysis everyone on the keyboard committee bought in on it, I never did a full write up specifically of it. We shifted our attention to the controversial issues. The data and general conclusions, though, did appear in a couple of papers put out by our group in the next year or so. The keystroke analysis for one of the editors was included in "How Do People Really Use Text Editors?" by my colleagues John Whiteside, Norman Archer, Dennis Wixon, and Michael Good. Mike also outlines the reasons behind the inverse-T in "Software Usability Engineering" and lists the editors we had logging data for in "The Use of Logging Data in the Design of a New Text Editor".

All of these papers and several others that came from our group can be found at Mike's sites (once known as "Good Paper" now at https://michaelgood.info/publications/).

Recently, a reader, Michael Walden, asked me for some additional details, such as when DEC first shipped the inverse-T arrangement and what the first IBM or Apple product to copy it was.

As far as I recall, the LK201 first shipped with the VT220 terminal in early 1983 or possibly late '82, and the IBM 3161 terminal was the first IBM product with the inverse-T arrangement (and interestingly, the DEC- and Apple-like 18-key numeric keypad rather than the IBM-style 17-key arrangement). The 3161 came out in 1985. An IBM web page says that they introduced their 101 key keyboard in 1984, but doesn't say on what product. Perhaps there was something earlier than the 3161, or it shipped earlier than I recall. Apple was the late comer to the party. Steve Jobs believed that the mouse should be used and not arrow keys.