Emerging Interfaces

How Do You Feel a 3D Shape Through a Flat Touchscreen? Researchers Tested Two Ways.

A new study compares a single height map against a stack of binary slices for encoding 3D geometry onto 2D haptic displays — and finds both work well enough to let users reassemble real 3D shapes by touch alone.

Most surface haptic displays — programmable pin arrays, vibrotactile touchscreens, friction-modulating screens — are fundamentally flat: a 2D grid of points you can feel. Most of the shapes people actually need to understand by touch, from CAD models to map topographies, are not flat at all. A new study, posted to arXiv as an unreviewed preprint, tests two different ways of squeezing that third dimension into a 2D tactile display, and measures whether people can actually use either one.

Two strategies, two display types

Researchers MacKenzie C. Harnett and Rebecca F. Friesen compared two encoding methods. The first is a single tactile height map: one pass over the display surface gives a coarse but quantifiable sense of depth, the same idea as a raised-relief map. The second is a series of binary layers — sequential slices through the 3D model, each rendered as a simple on/off pattern — which can represent more complex geometry a single height map can’t, including overhangs and cavities that a one-shot depth map would simply flatten out.

They tested both encoding strategies across two different display hardware types: a friction-modulating display, which simulates texture by rapidly varying the friction between the screen and a fingertip, and a pin array, which physically raises and lowers an actual grid of pins.

Both worked — with one hardware edge that didn’t quite matter

In a user study built around a 3D assembly task — essentially, feel the encoded shape, then reconstruct it — participants hit high accuracy across all four combinations of strategy and display. The pin array edged out the friction-modulating display on speed and on a (not statistically significant) user-experience rating for the binary-layer task specifically. But none of that translated into a meaningful difference in the metric that actually matters for a tactile graphic — whether users could correctly reassemble the shape. Both display types, and both encoding strategies, got people to the right answer about equally well.

Why this matters beyond the lab

This is accessibility research first: the clearest beneficiaries of a solved “3D shape over a flat haptic surface” problem are blind and low-vision users who currently have few options for exploring 3D geometry — a CAD model, a data visualization, a map’s terrain — without a bespoke and often expensive physical model. Harnett has worked on related tactile-interaction problems before, including a 2025 study (with Friesen and Pijuan Yu) on freehand contour drawing on a refreshable pin display, built on the same observation that most existing tactile tech for this purpose is “primarily restricted to niche and expensive products.” A result showing that a standard, already-shipping display type can represent real 3D complexity — overhangs and cavities included — without needing exotic new hardware is a meaningfully practical finding, not just an HCI curiosity.