OBJECTIVES: To test marginal and internal adaptation of five different adhesive composite core systems under load. METHODS: 30 human premolars were endodontically prepared and obturated with an epoxy sealer and vertically condensed gutta-percha. Thereafter the entire clinical crown was removed. The teeth were randomly assigned to five different composite core groups, all using the same fiber reinforced radicular post (DT White). Gr. 1: Optibond solo plus/Nexus II/Prodigy; Gr. 2: Scotchbond 1/Rely X Arc/Filtek P60; Gr. 3: EBS multi/Compolute applicap/Pertac II; Gr. 4: ART Bond/Parapost cement/Synergy; Gr. 5: Superbond C&B catalyst S+polymer/Metafil CX. Polyvinyl-siloxane impressions of the external margins of the cores were readied before and after 1,200,000 cycles of mechanical loading with max. 100 N at 45 degrees to the longitudinal axis of the tooth. After loading, 5 of the 6 samples of each group were cut longitudinally and the sixth sample was cut transversally to be able to take replicas for evaluation of internal adaptation after loading. RESULTS: Percentages of external "continuous margins" ranged from 97.9+-4.6 to 66.5+-7.8 before and from 87.4+-25.0 to 5.8+-12.5 after loading. Internal adaptation ranged from 96.4+-8.0 to 17.1+-20.4 after loading for the core adaptation and from 89.8+-12.2 to 65.9+-14.3 for the dentin-luting composite interface and from 99.4+-1.2 to 88.6+-9.4 for the composite-post interface. CONCLUSIONS: Surprisingly, the best materials' combination for the adhesive composite core was a self-etch light cured adhesive with a chemically cured luting agent.