S1 Supplemental Data Single Units in the Pigeon Brain Integrate Reward Amount and Time-to-Reward in an Impulsive Choice Task ceivable reasons why a pigeon occasionally chose the nonpreferred Tobias Kalenscher, Sabine Windmann, Bettina Diekamp, Jonas Rose, Onur Gu ¨ ntu ¨ rku ¨ n, and Michael Colombo Supplemental Results and Discussion small reward although it generally chose the large reward: First, the animal might have chosen the small reward by accident. In this case, Response Categories We recorded single-unit activity from 159 NCL neurons. Of these, it would have anticipated a large reward, and hence the reward- correlated activity should reflect the subjective value of the large 79 units (49.7%; all p  0.05; t test) showed significantly increased activity in relation to baseline activity to the sound cue indicating reward. Second, it might have chosen the small reward because it truly preferred the small reward on this particular trial. Again, in this trial onset, and 62 units (39%) showed further increased activity during the waiting period between sound offset and motor onset. case one would expect that the neural activity reflected the relatively higher preference for the small reward compared to the large reward. One hundred twenty-eight units (80.5%) had significantly elevated activity during the fixed-ratio motor response. One hundred fourteen The only way to overcome this problem is by comparing trials in which pigeons show a clear preference for the large reward—that cells (71.7%) showed significantly increased activity during the delay period. Ninety-eight of these delay units (86.0%) exhibited a short is, large-reward choices before the PS—with trials in which pigeons show a clear preference for the small reward—that is, small-reward phasic response at the beginning of the delay with a latency of approximately 80–100 ms and an approximate duration of 100 ms choices following the PS. Hence, the lack of statistical power and the ambiguity in interpreting the odd choices made it necessary to and a sustained increased response with varying latency and a duration of at least 1000 ms. Sixteen delay units (14.0%) exhibited perform between-block comparisons. only the sustained response. One hundred thirty-five units showed increased activity (84.9%) during the reward period. Eighty-one units Supplemental Experimental Procedures (50.9%) showed sustained firing of varying duration (500–2000 ms) after reward offset. None of the neurons showed evidence of reward- Animals amount, delay-, or pecking-site-related activity, except neurons with All subjects were kept and treated according to the University of sustained delay activity. Otago Code of Ethical Conduct for the Manipulation of Animals, and the research was approved by the University of Otago Animal Ethics Committee. Eight pigeons (Columba livia) were used in this experi- Properties of the 14 Neurons of Interest ment. During training, all animals were kept on a food-deprivation The main text reported the neural activity pattern in the first 1500 schedule at 80% of their free-feeding body weight. ms, but we also looked at the following activity in longer delays. Not all of the 14 selected neurons of interest maintained their dis- charge level during the entire delay. We compared the discharge Apparatus Pigeons were trained and tested in a standard shielded and sound- rate in the late-delay phase (delay after 1500 ms until reward delivery) with the discharge rate in the early phase (delay 0–1500 ms) and attenuated operant chamber (34  70  34 cm) equipped with two round pecking keys, two shutters to prevent access to the keys, the baseline discharge rate (significance was assumed when p  0.05; t test for paired samples). In eight neurons, late-delay activity one centrally located feeder, and one white houselight. All items were symmetrically arranged at the front wall of the box. Pecking was always significantly above baseline level and either did increase or was not significantly different in comparison to early-delay activ- keys had a diameter of 2.5 cm and were positioned 22 cm above the floor and 10 cm apart from each other. Throughout the entire ity, suggesting that these neurons showed climbing or maintained activity throughout the entire delay. In four neurons, the late-delay training and testing sessions, the houselight was turned on, the left key was always illuminated in white, and the right key was always activity was not significantly different from baseline activity but was significantly below early-delay activity, and in two neurons, the late- illuminated in green. delay activity dropped below baseline activity. We have no specific interpretation for these diverse discharge Group Assignment and Pretraining patterns other than that the timing of the long delays was not com- Pigeons were randomly assigned to two groups. Four pigeons re- pletely learned for some of the pigeons/neurons. Alternative expla- ceived the large reward after pecking the left key, and the remaining nations, such as coding of a temporal difference error, are also four pigeons received the large reward after pecking the right key. conceivable, but future work needs to investigate the functional After an autoshaping procedure, during which pigeons acquired the significance of the stability or instability of late-delay activity. association between responding to a pecking key and subsequent food reward, pigeons were trained with only one active key (the large-reward key). Pecking this key led to 4 s of access to food, Notes on Our Conclusions consisting of a mixture of seeds, after a constant delay of 500 Our finding that the difference in large- and small-reward-related ms. Pigeons were trained until they succesfully completed 50 trials activity correlated with the animals’ current and future preference within 1 hr in each of three consecutive sessions. Subsequently, bias is of particular interest and requires some elaboration. In our pigeons were trained with the other key alone (the small-reward analysis, we considered only choices that were consistent with the key), which was associated with 2 s of access to food after a con- overall preference bias (and eliminated odd choices) for the follow- stant delay of 500 ms. Pigeons were again trained until they success- ing reasons: Because pigeons mainly chose only one of the two fully reached the criterion. The amount of food consumed during response alternatives within one block, the number of small-reward the pretraining, training, and testing sessions was insufficient to choices before the PS was too small to allow for a within-block raise the body weight above 80% normal body weight. Moreover, statistical analysis, as was the number of large reward choices the grains used as rewards were low-calorie grains causing very following the PS. little satiation. Moreover, occasional choices of the nonpreferred response alter- native are hard to interpret. If delay activity indeed correlates with the subjective value of the anticipated reward, as proposed in this Experimental Design After satisfactorily meeting the pretraining requirements, pigeons paper, then one would expect equally high activity on preferred and odd choice trials. For instance, consider pre-PS trials in which the were subjected to the main task (Figure 1), which is a modified version of the adjusting delay procedure developed by Evenden pigeon showed preference for the large reward. There are two con-