OBJECTIVE: The bonding potential of 'mild' self-etch adhesives may be compromised due to smear interference, as they may not dissolve/penetrate the smear layer effectively due to their relatively low acidity. We observed that the thickness of the dentin smear layer differed depending on the surface-preparation methodology used. METHODS: The interaction of an (ultra-)mild self-etch adhesive (Clearfil S3 Bond, Kuraray Noritake) with human dentin, prepared either using a medium-grit diamond bur ('thick', clinically relevant smear layer) or 600-grit SiC-paper ('thin' smear layer), or just fractured (smear-free), was evaluated using high-resolution transmission electron microscopy (TEM). Non-demineralized/demineralized 30-100nm interfacial cross-sections were prepared following common TEM-specimen processing and diamond-knife ultra-microtomy. RESULTS: The adhesive did not dissolve the bur-cut, nor the SiC-ground smear layer, but impregnated it. Within this 'resin-smear complex', hydroxyapatite was abundantly present. At fractured dentin, this complex was not present, while the actual layer of interaction of the adhesive was limited to about 100nm. Non-demineralized 'ultra-thin' (30-50nm) sections confirmed the interfacial ultra-structure to differ for the three surface-preparation methods. An electron dense band was consistently disclosed at the adhesive interface, most likely representing the documented chemical interaction of the functional monomer 10-MDP with Ca. SIGNIFICANCE: The dentin surface-preparation method significantly affects the nature of the smear layer and the interaction with the ultra-mild self-etch adhesive.