A bacteria-free halibut larval rearing system was used to test 20 bacterial isolates, from British halibut hatcheries, for their toxicity towards halibut yolk-sac larvae under microbially controlled conditions. The isolates tested spanned a range of genera and species (Pseudoalteromonas, Halomonas marina, Vibrio salmonicida-like, Photobacterium phosphoreum and V. splendidus species). A pathogen of turbot, Scophthalmus maximus, V. anguillarum 91079, and 2 isolates from adult halibut were also included. Isolates were inoculated, at a concentration of 5 x 10(2) cfu ml(-1), into flasks containing 25 recently hatched axenic halibut larvae, using a minimum of 3 flasks for each treatment. Control survivals to 38 d post-hatch for the 3 experiments averaged 84, 51.5 and 49%, respectively. With the exception of V. anguillarum 91079, which was highly pathogenic towards halibut yolk-sac larvae, there was no statistically significant difference in survival between the controls and the different treatments. This suggests that most of the bacteria routinely isolated from halibut hatcheries are not harmful to yolk-sac larvae, even though most flasks contained in excess of 5 x 10(6) cfu m(-1) of the inoculated organism when the experiments were terminated. Three organisms previously shown to inhibit growth of bacteria in vitro were tested for their ability to protect halibut yolk-sac larvae against invasion by V. anguillarum. In 4 separate challenge experiments none of the test isolates, a Pseudoalteromonas strain and 2 Carnobacterium-like organisms, showed any protective effect. To investigate how particular bacteria influence their start-feed response, larvae were fed axenic and gnotobiotic Artemia colonized with a range of different Vibrio spp., and examined after 8 d. There were no statistically significant between-treatment differences in the proportion of Artemia-containing larvae, indicating that bacterial contamination of the live food does not appear to influence initiation of the feeding response.