Museum labels were once written as compact inventories: artist, title, date, medium, provenance. Their impersonal tone implied that the label merely transferred settled facts from scholarship to visitor. Yet even the shortest label directs attention. Calling a portrait “courtly” invites scrutiny of status; identifying a vessel as “ritual” makes wear marks look ceremonial rather than domestic. Selection begins before interpretation is openly announced.

Some curators conclude that labels should disappear so viewers can encounter objects without mediation. But an unlabeled gallery is not unmediated. Architecture, lighting, sequence, and the decision to classify an object as art have already framed the encounter. Removing text chiefly withholds one frame while leaving less visible ones intact. It can also favor visitors who arrive with the background knowledge needed to recognize a dynasty, technique, or disputed attribution.

The alternative is not to burden every object with an essay. Layered labeling can separate observation from inference while making both available. A first line might identify material and date; a second can state the interpretation guiding the display; a digital layer can present competing accounts and the evidence for each. Such labels do not eliminate curatorial authority, but they make some of its operations inspectable.

The goal should therefore be accountable mediation, not imagined neutrality. A good label gives visitors enough orientation to see why an object matters while marking where knowledge is uncertain or contested. Its success is measured not by disappearing from the visitor's awareness, nor by dictating a single response, but by enabling informed attention and disagreement.

Layering creates its own risks. A digital supplement can become a warehouse that buries uncertainty beneath links few visitors open, while a confident first line continues to govern the encounter. The order of information is itself interpretive. If an attribution is disputed, presenting one artist's name as fact and relegating doubt to a third screen does not make the label accountable. Important uncertainty must appear at the level where the corresponding claim first appears.

Nor is more explicit interpretation always more honest. A curator can overwhelm an object with every scholarly dispute and leave visitors unable to distinguish consequential disagreement from minor technical variation. Accountability requires editorial judgment about significance, followed by enough evidence for visitors to understand and question that judgment. This is why the layered model is not a mechanical recipe. It is a discipline of matching the prominence of a claim to the strength and importance of its support. The label remains an argument about attention, but an argument whose structure a visitor can begin to see.