During training, some mice failed to sample both choice alternatives. Specifically, failure to sample both alternatives was observed in adolescent D2 (n=3), adult D2 (n=1), and adult B6 (n=1) mice. Mice that did not sample both alternatives during training did not participate in the testing phase and their data were excluded from the analysis of the training segment, yielding 11-14 mice per group. Mice that sampled both options during training sessions continued to do so during each testing session, as has been commonly reported for the across-session discounting task (Koot et al., 2009
; Adriani & Laviola, 2006
; Adriani et al., 2004
Over the course of training, a preference for the large volume of milk emerged in both strains shows the mean percentage of large-volume choices for each session of training. A mixed-model ANOVA was used to analyze the data. The model included two between-subject factors (Age and Strain) and one within-subject factor (Session). The data were log-transformed prior to the analysis of delay discounting measures. The ANOVA revealed a significant main effects of Age (F(1,43) = 9.2, p < .004) and Session (F(6,258) = 12.8, p < .0001). There was no main effect of Strain, nor of any interaction term. The effect of age appeared to be due to greater preference for the larger volume in the choices of adult mice compared to those of adolescent mice. Age-related differences likely reflect different rates of acquiring preference, rather than any absolute difference in preference, as the data in show that values for all strains converged by the end of training. As we were most interested in the data immediately prior to testing, we examined choice proportions from session 7 of training in greater detail. First, choice of the larger volume of milk exceeded 50% for all groups by session 7 (Adolescent D2, t = 3.6, df = 10, p < .002; Adolescent B6, t = 5.4, df = 13, p < .0003, Adult D2, t = 3.3, df = 10, p < .002; Adult B6, t = 6.3, df = 10, p < .003, Bonferroni corrected t-tests). Second, post-hoc analysis of the data from session 7 failed to identify differences among any of the groups (p > 0.05 for all comparisons).
Mean (SEM) percentage of choice of large milk volume across training sessions.
In addition to changes in preference, the absolute number of choices made changed with training; in general, mice made more choices in later training sessions than in earlier sessions. shows mean number of choices made during each session for each group. A mixed-model ANOVA revealed a significant main effect of Session (F(6,258) = 21.1 p < .0001) and main effect of Strain (F(1,43) = 20.3, p < .0001). Additionally, there were significant Strain × Session (F(6,258) = 5.7, p < .0001) and Session × Age (F(6,258) = 2.6, p < .016) interactions. Primarily, our concern was that groups might be different at the end of training, and so we were interested in comparisons among groups on the 7th training session. Post-hoc t-tests showed that adults of both strains and adolescent B6 mice made more choices than adolescent D2 mice (p's ranged from p < .047 − p < .005, uncorrected t-tests), but the number of nose pokes did not differ significantly among the remaining groups. Collectively, the data show that during training preference came under control of the large milk volume to a similar extent in all strains, though adolescent D2 mice made fewer absolute choices.
Mean (SEM) number of choices made during each session across training and testing phases. The number in parenthesis by each day indicates the delay to the large volume of milk; the delay to the small volume was always 1 s.
shows the effects of increasing the delay to the large milk volume on preference. As shown in the figure, increasing the delay to the large volume resulted in a shift in preference for the smaller, more immediate volume in all groups. There were, however, differences in the rate of preference change. Most noticeably, adult B6 mice were more tolerant of the increasing delay than the remaining groups, which were more similar to each other. A mixed-model ANOVA was used to analyze the data. The model included two between-subject factors (Age and Strain) and one within-subject factor (Delay). The ANOVA revealed a significant main effect of Strain (F(1,43) = 8.1, p < .01) and Delay (F(6,258) = 26.3, p < .0001). Importantly, the analysis also identified a significant Age × Strain interaction (F(1,43) = 11.9, p < .003) and a Age × Strain × Delay interaction (6, 258) = 3.8, p < .001). The significant three-way interaction reflects the increased tolerance to delay observed in B6 adult mice when compared to the other three strains.
Figure 1 Age and strain differences in tolerance to delayed to reward. The y-axis shows the percentage of large-reward choices as a function of delay. Data from one-second delay test has been separated because it also constitutes a replication of the training (more ...)
One consequence of the fixed session length is that the number of choice opportunities necessarily decreases with increasing delays to the large volume, and the number of choices could have fallen below a value that permits meaningful comparisons. shows the average number of choices made for each session during the testing phase. Although the number of choices declined for the adult B6 mice, the data show that on average approximately 25 choices were still made, which approximates or exceeds the number of free choices commonly measured in other tests of delay discounting (e.g., Green, Myerson, & Calvert, 2010
; Madden et al. 2010
). So, the number of choices remained adequate enough for meaningful analysis of decision making, even under the longest delay.
The relative insensitivity of adult B6 mice to increasing reward delay could have been due to perseverative response patterns. That is, choice may have been insensitive to the reinforcement contingency, and instead was under the control of some other stimulus feature of the environment. To address such possibilities, adult mice were examined for perseverative behavior in a resistance to extinction task (see Methods). As the data from the extinction sessions were not normally distributed, we analyzed the results in terms of group medians, rather than means. Prior to extinction, median responses per minute were 15.3 (adult D2) and 14.6 (adult B6) responses per minute; rates of responding were not different between strains (Wilcoxon test, p > .05). The results of the extinction test are shown in . The upper left graph shows median responses per minute for each minute of extinction. Response rate was increased early in the extinction component, consistent with extinction “bursting” commonly observed upon reinforcement withdrawal (Minor, 1987
; Weissman, 1959
). Rate of responding declined across the remainder of the extinction test. Over the last 5 minutes of extinction, median rates were 4.2 (adult D2) and 3.4 (adult B6) responses per minute; these values were significantly less than values observed prior to extinction for both the D2 (Wilcoxon test, Z = -2.134, p < .033) and B6 (Wilcoxon test, Z = -2.432, p < .015) mice. The upper right graph shows box plots summarizing total response output across the 30-min extinction session. Total responses emitted during extinction did not differ between adult D2 and B6 mice (Mann-Whitney U
, p > .05).
Figure 2 Effects of extinction on fixed-ratio maintained behavior of adult mice. The top set of graphs show the effects of the removal of the reinforcement contingency on nose poking. The top left figure shows the median number of responses per minute of a 30-min (more ...)
Extinction of one operant response can result in recovery or “resurgence” of previously reinforced responses (e.g., Lieving & Lattal, 2003
). So, it was of interest to examine responding on the inactive hole during extinction. On the session prior to extinction, median rates were 0.8 (D2) and 0.4 (B6) responses per minute. During extinction, responding on the active hole increased somewhat, reaching 1.6 (D2) and 1.5 (B6) responses per minute over the last 5 minutes of extinction, but these increases were not significantly different from the prior session when response rates were calculated over the last 5 minutes or over the entire session (Wilcoxon test, p > .05 in all cases). Total number of inactive hole responses are shown in box plots in the bottom right of . Total number of responses did not differ under the extinction test (Mann-Whitney U
, p > .05).
Collectively, the data show that removal of reinforcement resulted in increases in response rate early in the extinction test followed by decreasing rates over the remainder of the session. By the end of the extinction test, rates of responding were reduced compared to level under the FR 1 contingency. There was some increase in responding on the inactive hole, but those effects were not significant. Most importantly, B6 evinced no greater propensity than D2 mice to persist responding during the extinction test. Responding by both strains fell below levels maintained by the contingency once it was removed. Thus, it seems unlikely that the apparent insensitivity to delayed reinforcement observed in B6 mice in the prior condition was due to perseverative responding.