Specular highlight

A specular highlight is not shading. Shading describes how a whole surface is turned relative to a lamp, and it changes slowly across the form. A specular highlight is the reflected image of the lamp itself, appearing only where the surface happens to bounce that light straight at the viewer. Because it is a reflection of one object, it carries two pieces of information and no others. Its size and hardness report how rough the surface is: a polished surface reflects the lamp almost whole, so the highlight is small and its edge is crisp, while a rougher surface scatters the same light into a broad, faint smear. Its position reports where the lamp is, which is why a highlight moved to the other side of a shape instantly reads as a different scene rather than as a different material.

Interfaces do not render this, they draw it. A radial gradient stopped short of the edge is a highlight; a linear wash across the top half with a hard stop is gloss; an inset white blob under an over-rounded corner is what makes claymorphism look inflated; a bright line clinging to a rim is how liquid-glass says a panel has thickness. Aqua is the best documented case, since it put the same pill of light on buttons, scrollbars and progress bars until the highlight became the operating system’s signature. What bevel adds is a different claim: a bevel is geometry, an angled face catching light along an edge, while a specular highlight is a spot of reflection that needs no edge at all. One useful detail of the physics survives into the drawing. A highlight takes the colour of the light rather than the colour of the object, so it is the one place a saturated fill drops most of its chroma, which is why a tinted highlight reads as plastic and a white one reads as glass.

Three habits separate a convincing highlight from a smudge. Keep it off-centre, because a highlight centred on a shape reads as a glow coming from inside the object instead of light landing on it. Keep one per object, since a second highlight implies a second lamp and the eye notices the contradiction before it can name it. And keep the direction consistent across a set: a row of icons lit from four different angles looks like a row assembled from four different products, which is the most common way a highlight goes wrong in practice.

The cost is contrast. A highlight is the brightest region of a control, so white label text crossing it is measured against near white, and a button that passes when averaged can fail in the band where the text actually sits. Measure the light region, not the mean. The related trap is maintenance: a highlight has to be tuned per fill colour, which is exactly the per hue work that flat-design was invented to escape. Both are arguments for spending the effect once, on the thing that most needs to look touchable, rather than on every surface in the product.