To determine the cement bond strength and marginal leakage of castings cemented to implant abutments.Fifty-six titanium abutments and castings were divided into 7 groups (n=8), 1 for each cement. Castings were cemented to abutments using 1 of 3 resin-based cements (RES, RES-B, and RES-B-P), a resin-modified glass ionomer (GI), a polycarboxylate cement (PCB), an acrylic urethane cement (UDM), or a zinc phosphate cement (ZP). Specimens were placed in 100% humidity at 37 degrees C for 24 hours. Specimens were subjected to compressive load cycling followed by thermal cycling; they were then immersed for 24 hours in 0.5% basic fuchsin. Castings were removed with an Instron universal testing machine with a crosshead speed of 0.125 cm/min. Leakage was visually graded from 0 (no leakage) to 2 (leakage extended beyond the lower half of the internal surface of the casting). Failure load (FL) was analyzed with analysis of variance and Scheffe's test (alpha = .05). Chi-square was used to analyze leakage (alpha = .05).Cements were categorized by FL into 4 statistically unique groups: (1) RES-B-P (351 N) and GI (337 N); (2) ZP (245 N) and RES-B (241 N); (3) PCB (107 N); and (4) RES (63 N) and UDM (55 N). Leakage was greater for the PCB group than for the other groups (7 of 8 specimens demonstrated leakage; P < .01). Three ZP specimens demonstrated leakage. UDM and RES each had 1 specimen with leakage. RES-B-P, RES-B, and GI showed no leakage.Luting agents designated by the manufacturer as provisional cements demonstrated lower resistance to removal, regardless of material type. Luting agents described by manufacturers as "permanent" differed in resistance, with resin cements being most resistant, followed by zinc phosphate and polycarboxylate cements. Provisional cements demonstrated leakage comparable to higher-strength materials.