The Mouse

What the mouse actually measures

Every pointing device makes a choice: does it report where it is, or how far it has moved? A digitising tablet reports absolute position — lift the stylus, put it down elsewhere, and the cursor jumps to match. The mouse reports relative displacement. It has no fixed origin. It knows only that it moved three millimetres to the left since the last reading, and it hands that delta to the operating system, which updates the cursor accordingly. This is a profound architectural difference, and it shapes everything the mouse is good and bad at.

Douglas Engelbart's team at the Stanford Research Institute demonstrated the device publicly in December 1968, in what has since been called "the Mother of All Demos." Engelbart's mouse was a small wooden block with two perpendicular wheels on its underside; each wheel drove a potentiometer that reported rotation on one axis. The X and Y displacements were read separately and combined by software into cursor movement. The patent Engelbart filed in 1967 describes it plainly as an "X-Y position indicator for a display system" — a name that quietly overstates it, because the device never knows its absolute X-Y position at all. It knows only motion.

The optical-mechanical ball mouse, which replaced the wheel design and dominated for decades, works on the same principle. A rubber ball rolls against the desk; two rollers inside the housing, set at ninety degrees to each other, pick up the rotation on each axis and feed it to encoders. The ball was always a liability — it collected grit, and the rollers needed cleaning — but the sensing geometry was identical to Engelbart's. When optical mice arrived and replaced the ball with an LED and an image sensor doing frame-by-frame correlation, the geometry stayed the same: still delta-X and delta-Y, still relative, still homeless without a screen coordinate to anchor it.

Why relative sensing is wrong for drawing

Relative displacement is excellent for navigation. You can pick the mouse up, reposition it, set it down, and the cursor does not move. You can cross a wide screen with a short physical gesture by increasing the sensitivity ratio — the transfer function that multiplies raw counts into pixel movement. None of this is available to an absolute device.

But drawing demands something the mouse cannot give: a direct spatial correspondence between hand and mark. When a draughtsman traces a curve, the pen's position on the paper is the mark's position. A mouse user moves the cursor to a position, then clicks or drags, but there is no persistent mapping between the hand's location and the image coordinate. Lift the mouse, reposition it, and the relationship breaks. This is why, when Richard Shoup built SuperPaint at Xerox PARC in the early 1970s — the first system to let a person paint directly into a framebuffer — the input device was a tablet, not a mouse. Pressure, angle, contact: all of these require a device that knows where it is.

The mouse arrived at Xerox PARC shortly after Engelbart's demo. Xerox engineers replaced the two-wheel design with a ball for smoother movement across arbitrary surfaces, and the PARC mouse fed the graphical interfaces that would eventually reach Apple and then the broader market. But PARC's paint research kept using tablets. The bifurcation was already established by the mid-1970s: mouse for the interface, tablet for the image.

What survived and why

The relative device won the desktop because desktops are mostly navigation. Opening files, selecting text, clicking controls — none of these require absolute correspondence between hand and screen. The mouse is faster at these tasks, cheaper to manufacture, and requires no special surface. An optical mouse today samples its sensor at rates above a thousand times per second and interpolates sub-pixel motion, but its fundamental measurement is unchanged since 1968: how far did it move?

The stylus survived because images are not navigation. A curve needs to land where the hand says it should land, and Bézier control handles need to be placed, not approached. Pressure needs to modulate line weight in real time, which requires knowing contact force as a continuous signal. The mouse cannot supply any of this, and no transfer function compensates.

Both devices are still in production. They measure different things, they serve different purposes, and sixty years of use has not blurred the line between them.

This is a profound architectural difference, and it shapes everything the mouse is good and bad at.