Statement of problem: Whether polyvinyl siloxane impressions are capable of reproducing 5-μm changes on natural freeform enamel and potentially enabling clinical measurements of early surface changes consistent with wear of teeth or materials is unclear. Purpose: The purpose of this in vitro study was to investigate and compare polyvinyl siloxane replicas with direct measurements of sub-5-μm lesions on unpolished human enamel lesions by using profilometry, superimposition, and a surface subtraction software program. Material and methods: Twenty ethically approved unpolished human enamel specimens were randomized to a previously reported cyclic erosion (n=10) and erosion and abrasion (n=10) model to create discrete sub-5-μm lesions on the surface. Low viscosity polyvinyl siloxane impressions were made of each specimen before and after each cycle and scanned by using noncontacting laser profilometry and viewed with a digital microscopy and compared with direct scanning of the enamel surface. The digital maps were then interrogated with surface- registration and subtraction workflows to extrapolate enamel loss from the unpolished surfaces by using step-height and digital surface microscopy to measure roughness. Results: Direct measurement revealed chemical loss of enamel at 3.4 ±0.43 μm, and the polyvinyl siloxane replicas were 3.20 ±0.42 μm, respectively. For chemical and mechanical loss direct measurement was 6.12 ±1.05 μm and 5.79 ±1.06 μm for the polyvinyl siloxane replica (P=.211). The overall accuracy between direct and polyvinyl siloxane replica measurements was 0.13 +0.57 and −0.31 μm for erosion and 0.12 +0.99 and −0.75 μm for erosion and abrasion. Surface roughness and visualization with digital microscopy provided confirmatory data. Conclusions: Polyvinyl siloxane replica impressions from unpolished human enamel were accurate and precise at the sub-5-μm level.