Architectural models are frequently judged by how faithfully they predict a completed building. Under this view, a model is a miniature promise: the closer its surfaces, colors, and proportions come to the eventual structure, the more successful it is. When a finished building departs from the model, commentators accordingly describe the model as misleading. This standard makes sense for some presentation models shown to clients near the end of design. In that setting, resemblance can protect a client from discovering too late that a promised facade or room has quietly changed. The standard therefore has a legitimate, but bounded, consumer-protection role. Applied to models generally, however, it mistakes one species for the genus. Architects also build study models to think with. Such models may omit windows, exaggerate a stairwell, or represent neighboring buildings as plain blocks. These departures are not necessarily shortcuts. By suppressing some features and amplifying others, a model can isolate a problem—how light reaches an interior court, how pedestrians approach an entrance, or how separate volumes bear on one another—that would be harder to perceive in a comprehensive simulation. The model operates less like a scaled-down building than like a diagram with mass. Critics sometimes object that even study models are dangerously seductive. Because viewers readily imagine themselves inside a three-dimensional object, the model's selective world may acquire an authority its maker never intended. A blank block standing for an adjacent apartment building can make that building's residents disappear from consideration; an immaculate white surface can conceal the maintenance demands of the proposed material. This objection identifies a genuine risk, but it does not justify demanding more complete models. Completeness is itself selective: adding realistic trees and tiny people may intensify the illusion of neutrality while leaving consequential assumptions unexamined. The better safeguard is to make the model's operations explicit. An architect presenting a study model should state what question it was built to investigate, which features have been suppressed, and what evidence would have to come from other models or methods. A sequence of deliberately partial models may then provide a fuller inquiry than one highly finished object. One might test circulation, another shadows, and a third the relation to surrounding streets; discrepancies among them become prompts for judgment rather than defects to be polished away. Thus the proper criticism of a model is not simply that it differs from the building eventually constructed. We should ask whether its simplifications were suited to the inquiry it claimed to conduct and whether audiences were equipped to recognize their limits. Visual resemblance can be useful, but for much architectural thinking, declared and disciplined incompleteness is not a failure of representation. It is the condition that allows the model to function as an instrument of analysis.