Because of structural uncertainty, an industrial object resulting from mass production can exhibit a large variability in its vibratory and acoustical behaviour. Some previous studies have focused on the variation in noise and vibration between structures with nominally identical design. Sounds emitted by these structures were also modified. Thus it might be thought that sound perception would vary as well because of structural variability. However, that problem has not yet been investigated. The aim of this study was to show how to assess the perceptual consequences of that physical variability. A mechanical model system, on which structural uncertainty could be simulated, was set up for sound synthesis. The sounds "emitted" by three different types of products were then assessed by a group of listeners performing a categorization task. The purpose was to determine if sounds stemming from nominally identical objects were always categorized together or if sounds emitted by nominally different objects could be confused. This experiment enabled to find the optimal partition of the stimulus set. Even though subclasses appeared within stimuli representing one type of object, sounds synthesized with different types of objects were not confused.