Purpose: To determine if collagen fibrils on the dentin side of failed resin-dentin interfaces undergo mechanical disruption during microtensile bond testing. Materials and Methods: Extracted, caries-free human third molars were divided into four groups. The occlusal enamel was removed, leaving a flat dentin surface for bonding. Resin composite buildups were made after the acid-conditioned dentin was bonded with either Single Bond {Si or One-Step (0}, and using either moist bonding (M) or air drying for 5 s (DJ. After storage in water for 24 h. the teeth were vertically sectioned into an array of 0.9 >■ 0.9 mm resin ccmposite-dentin beams They were stressed to failure using the nontrimming version of the microtensile bond test. Fractured dentin and resln composite sides of representative beams from each group that exhibited adhesive failures under tight microscopy examination were prepared for descanning (SEM) and transmission electron microscopy (TEM). Results: A two-way ANOVA showed that moist bond strengths were significantly higher than those made to dry dentin {M > 0; p < 0.001), but that there was no difference between the adhesive (S vs 0; p = 0.547). SEM analysis showed the presence of loose collagen fibrils within fractured hybrid layers ¡n the dry groups but not in the moist groups. TEM examination of the dry-bonded groups revealed collagen fibrils that wore thinner and exhibited abnormally wide interfibrillar spaces when hybrid layers were intact. Within dry-bonded fractured hybrid layers, broad mechanical disruption zones could be seen, consisting of fibrils that were devoid of cross banding, defibrillation of the subfibrillar architecture, and gross disaggregation into microfibrils. In the moist-bonded groups, only short mechanical disruption zones were found along the torn edges of the collagen fibrils. The rest of the fibrils beyond the fracture site were intact and retained their periodicity. Mechanical testing of demineralized matrices.