In the first condition six hens responded on concurrent schedule with two variable interval 60-s schedules leading to 2 s of access to wheat. This gave estimates of bias resulting from any uncontrolled differences between the two schedules (i.e., inherent bias). Then the same hens responded on concurrent chain schedules of reinforcement for two different durations of access to food: 2 s and 8 s. The initial links were both variable interval 60-s schedules and the terminal links were equal fixed ratio schedules. Over Conditions 2 to 8, the response requirement in the terminal links was varied over 1 to 128. Response and time ratios in the initial links showed responding and time allocation was greater on the key associated with the 8 s duration reinforcer, showing a preference for that terminal link, even when any inherent bias was removed. These preferences were generally greater the larger the response requirement. Preference followed the 8-s reinforcer when the key associated with that reinforcer was changed in Conditions 9 and 10, and was greater the larger the response requirement in these conditions. Thus increases in response requirement increased preference for the longer duration reinforcer. The terminal-link length increased equally for both food durations with increases in response requirement up to fixed ratio 32 for most hens. With fixed ratios of 64 and 128 there were some unequal length terminal links and these were associated with preference changes, with greater inequality between the terminal link lengths giving greater preference for the shorter terminal link. The study cannot separate out the effects of terminal link length and response requirement on preference, both most likely contributed to the preference changes. In the terminal links there were higher overall response rates and shorter post-entry pauses with the shorter duration of reinforcer (2 s) at the smaller fixed ratio values. These data follow the pattern seen in studies of fixed ratio performance, where less preferred reinforcers have given higher overall response rates and shorter post-reinforcement pauses at smaller fixed ratio values. These data indicate that differences in overall response rate in do not reflect the preference seen in the initial links.