Sharper is not larger.

Two changes often arrive together: more pixels to draw with, and larger labels. This small experiment separates them. Each panel contains the same line, points and label, in a 280 × 120 CSS-pixel box.

Compare left to right for backing resolution; compare top to bottom for font size. On a narrow screen the panels stack in reading order. The scale is deliberately fixed at 1 or 2, so the experiment stays the same on different displays.

What stayed the same?

Doubling the backing scale gives the canvas four times as many stored pixels. With a matching drawing transform and a fixed displayed box, it does not double the label’s CSS size. The live measurements below report the text advance in drawing coordinates, normalized to CSS pixels. They do not measure how well someone can read the text.

What this does—and does not—test

This is a geometry and rendering example, not a user study. Your display, browser, zoom and font affect the appearance. A 2× backing store is a comparison condition, not a universal production setting. This example does not determine a minimum readable font size, a reading-error rate, or mathematical correctness.

For a reading study, a next step would use several matched labels and a specific question, vary the order, record actual display conditions, and measure errors separately from preference. No such study has been run here.

Reproduce it

Save this HTML page and open it locally: it uses no external script, font, network request or account. The four conditions and drawing code are in the page source. You may copy and adapt Reed’s example with attribution.

Context: a response to the public Visual Quality discussion. This does not audit the notebook described there. References: HTML canvas standard and MDN on device pixel ratio.