An Analysis of the Standard English Keyboard

Yuzuru Hiraga, Yoshihiko Ono, Hisao YamadaCOLING 1980

TL;DR

This paper analyzes the QWERTY keyboard's efficiency by examining typing motion and error patterns. It reveals that typing is a highly mental process influenced by context, and the QWERTY layout's inherent inefficiencies are masked by typist adaptation, leading to potential health issues and inaccurate performance predictions.

What Problem Does This Paper Address?

For decades, researchers have questioned the efficiency of the standard QWERTY keyboard layout, which originated from a need to prevent mechanical typewriter jams rather than optimize typing speed. While alternative layouts like the Dvorak Simplified Keyboard (DSK) have been proposed, the QWERTY layout remains dominant. This paper aims to provide a deeper understanding of typing behavior and keyboard efficiency by analyzing the physical and mental aspects of typing, moving beyond simple key-pair analysis to consider the influence of context and typist adaptation. The authors seek to explain why QWERTY, despite its known inefficiencies, persists and why performance predictions based on isolated key analysis might be misleading.

Methods

The study employs a two-pronged approach:

  1. Time Interval Analysis: The core of the physical analysis involves measuring the time intervals between consecutive key strokes. This data is then factorized to understand the influence of various factors like hand transition, row transition, and finger transition on typing speed. A multi-variate linear regression model is used to quantify these effects.
  2. Error Analysis: The paper categorizes and analyzes common typing errors (omission, insertion, replacement, interchange, chattering) to understand their root causes, particularly those related to mechanical motion rather than cognitive lapses.

A novel aspect is the investigation into the "levelling effect," where the context of preceding and succeeding key pairs influences the timing of the current key pair, smoothing out variations and creating a more consistent typing rhythm.

Datasets

The study utilizes typing data collected from a professional English typist (Japanese, female) using the "Superwriter" system on an H-10 computer at the University of Tokyo. The input texts were:

  • The latter part of "Alice in Wonderland" by Lewis Carroll.
  • The entire "Through the Looking-glass" by Lewis Carroll (excluding verses).
  • "Information Processing" by M. Bohl (a computer textbook).

The combined dataset consists of 302,392 strokes (142,262 from Alice texts and 160,130 from the computer text). Time intervals were measured down to milliseconds, with values exceeding 500 ms being eliminated as likely due to external factors.

Key Insights and Ideas

  • Typing is Highly Mental: The study strongly suggests that typing is not just a mechanical process but a highly mental one, where context significantly influences productivity.
  • The "Levelling Effect": A crucial finding is the "levelling effect," where typing rhythm is smoothed out by the influence of surrounding key pairs. This adaptation helps create a consistent rhythm, potentially masking underlying inefficiencies.
  • Contextual Adaptation Masks Inefficiency: The levelling effect explains why QWERTY's inefficiencies might not translate directly into proportional speed differences compared to optimized keyboards. Typists adapt by smoothing out awkward sequences, which can lead to increased tension and fatigue.
  • Performance Predictions are Tricky: Analyzing isolated key pairs is insufficient to predict performance on different keyboard layouts. The contextual influence and typist adaptation mean that real-world performance can deviate significantly from predictions based on simple metrics.
  • QWERTY's Overload: The analysis of stroke distributions on QWERTY reveals an overload on the left hand, excessive use of the upper row, and poor finger loading relative to dexterity.

Key Results

  • Hand Transition Dominates: Alternate hand stroking is significantly faster than same-hand stroking, with a difference of approximately 40 ms.
  • Row and Finger Transitions: In same-hand stroking, row transition causes a slowdown of around 20 ms. Finger transitions show a less clear-cut picture, with adjacent finger stroking generally being slower than remote finger stroking.
  • Error Dominance: Omission errors constitute the largest category (65.2%), particularly in sequences where the text transitions from alternate to same-hand strokes or vice-versa.
  • Levelling Effect Magnitude: The levelling effect is most pronounced in alternate hand stroking, with a significant pull towards smoothing out time intervals in convex and concave situations.
  • QWERTY Inefficiencies: The study highlights that QWERTY has a high rate of awkward sequences (9.6% hurdles, 8.2% reaches) and an overloaded left hand.

Limitations and Future Work

  • Single Subject: The study acknowledges that error patterns are highly individual, and findings are based on a single subject. This limits the generalizability of the error analysis.
  • Unfamiliar Text: The subject was not accustomed to the input texts, which may have influenced typing behavior and error rates.
  • Keyboard Specifics: The influence of factors like key reactive force and the arrangement of special characters were not fully controlled for.
  • Future Work: The authors suggest that a deeper analysis of the psychological aspects of word recognition and chunking is needed to fully understand typing behavior. They also imply that future research should focus on how the levelling effect, while beneficial for rhythm, can lead to fatigue and potentially health issues, and how this impacts the comparison between keyboard layouts.

Summary updated

BibTeX
@inproceedings{hiraga1980analysis,
  author = {Yuzuru Hiraga and Yoshihiko Ono and Hisao Yamada},
  title = {{An Analysis of the Standard English Keyboard}},
  booktitle = {International Conference on Computational Linguistics},
  year = {1980},
}

Loading related content...

Figure