This study analyzed the bond strength (BS) of HEMA versus HEMA-free self-etch adhesive systems to dentin. The occlusal surface of 20 third molars was removed and the dentin was abraded with 600-grit silicon carbide paper. The teeth were randomly assigned to 4 groups (n = 5) and restored with GO, Adper SE Plus, OptiBond All-In-One, and Clearfil 3S Bond adhesive systems and Filtek Z350 composite. After light curing (600 mW/cm²), the teeth were stored in distilled water at 37 ºC for 24 h and were sectioned in the mesiodistal and buccolingual directions to obtain sticks (0.8 mm²). The sticks were subjected to tensile force using a universal testing machine (0.5 mm/min), and the modes of failure were analyzed by scanning electron microscopy and classified as adhesive, cohesive, or mixed. The BS data (in MPa) were treated using one-way ANOVA and Tukey's tests at 5% significance. GO presented the lowest mean bond strength value (10.57 ± 3.72) and differed significantly from the other materials (p = 0.001), which, in turn, presented statistically similar results (p>0.05) among themselves: Adper SE Plus (29.08 ± 8.93), OptiBond All-In-One (28.36 ± 6.49), and Clearfil 3S Bond (28.62 ± 6.97). Mixed fractures were the most prevalent. It was concluded that the influence of HEMA on BS to dentin was material dependent.