Revisiting the GRAIL Handwriting Recognizer: A Look Back at the Future (2016)

An Active Essay by Jack Schaedler

Half a century ago, the RAND Corporation pioneered a software system known as the Graphical Input Language, or GRAIL.

According to a 1966 RAND memorandum, the core mission of GRAIL was to:

“Investigate methods by which a user may deal directly, naturally, and easily with [their] problem.”

Unlike the workstations we use today, GRAIL dispensed with the usual peripherals. It featured no mouse, no keyboard, no buttons, and no joysticks. Instead, users interacted with the machine via a tablet and a pen-like stylus. Whenever a user needed to input text, they simply wrote it by hand.

Even by modern standards, demonstrations of this system possess a certain elegance and a sense of magic. To borrow a sentiment from Tony Hoare, GRAIL is, in many respects, superior to almost everything that followed it.

---

The Philosophy of Natural Interaction

Gabriel Groner, in 1966, noted the primary success of the system:

“This recognition scheme meets its primary objective of enabling any user to communicate naturally with a computer. A user is not distracted by any operational mechanics but, rather, may concentrate on [their] problem.”

![Image Description: A historical photo from the Computer History Museum showing Tom Ellis, one of the lead inventors, using the RAND tablet to manipulate an image on a screen during the 1960s.]

The Contribution of Gabriel Groner

This essay focuses on the work of Gabriel Groner, who engineered a sophisticated program capable of identifying:

• Handwritten letters
• Numerical digits
• Punctuation marks
• Geometric shapes

The system was remarkably efficient, running in real-time on an IBM System/360. It was also highly reliable, achieving a 90% accuracy rate for symbols drawn by people using the system for the first time. Groner detailed these findings in his 1966 memorandum, Real-Time Recognition of Handprinted Text.

Learning Objectives

By the end of this exploration, you should be able to:

• Understand the mechanics of Groner’s recognition scheme.
• Develop an intuition for the various phases of the algorithm.
• Explore the interactive implementation of these heuristics.

Note: This essay is designed as a dynamic companion to the original memo. Verbatim quotes from Groner's text will be rendered in $\textsf{a distinct sans-serif font}$.

While the original system was versatile, the version presented in this essay is a simplified implementation focusing exclusively on uppercase letters and numbers.

• Source Code: Available on [GitHub].
• Feedback: Please use the issue tracker for bugs or contact the author via Twitter.
• Deep Dive: The original IBM 360 Assembly code written by Groner is also available for review.

---

The Hardware: The RAND Tablet

The interface relied on a specialized piece of hardware: the RAND Tablet. The philosophy behind it was expressed by M.R. Davis and T.O. Ellis (1964):

“Present-day user-computer interface mechanisms provide far from optimum communication... It was felt that exploration of man’s existent dexterity with a free, pen-like instrument on a horizontal surface, like a pad of paper, would be fruitful.”

Technical Specifications

The Input Process

The following diagram illustrates the flow of data from the user's hand to the CPU:

Simulating the Experience

To help you understand this process, this essay provides a virtual tablet. However, there are some differences in scale:

1. Resolution: While the original was $100 \text{ lpi}$, the virtual version is $\approx 20 \text{ lpi}$.
2. Scaling: Imagine the virtual tablet as a small section of the original, magnified by roughly $500\%$.

Instructions for Use:

• Placement: Draw your characters so they occupy roughly $70\%$ of the tablet's height.
• Action: Click and hold the mouse button to simulate the "pen-down" action.
• Movement: Drag the pointer to create marks.
• Practice: Try drawing continuous-stroke characters such as 2, 3, 6, C, M, or S.

Once you have mastered the feel of the virtual tablet, you may proceed to the next section.