Reward magnitude and instrumental responses: Consistent and partial reward

Reward magnitude and instrumental responses: Consistent and partial reward 1 GAR VIN McCAIN, University of Texas at Arlington, Arlington, Tex. 76010 Two studies using partial or consistent large (500 mg) and small (45 mg) reward are presented. 1n both studies, after extended re ward acquisition, differences are negligible or nonexistent. Results from the partial re ward groups indicate extinction differences also disappear after extended training. These results do not seem to be in line with usual assumptions regarding the effects of re ward magnitude. Over the past 2 years aseries of studies involving the effects of reward magnitude have been run in this laboratory. This is a preliminary re port giving only two of these studies. Several others have been summarized elsewhere (McCain, 1969). The problem of the effects of reward magnitude is critical to a number of interpretations of learning, and the empirical relations are important to alm ost any learning analysis. The two studies presented followed several studies in which the usual assumption that larger rewards produce more vigorous responses during acquisition was questioned. EXPERIMENT 1 This study, as well as a number of other studies in the series, was run as a reward-shift study. The reward-shift data will be presented in a different context. The foeus of this study is on comparison of aequisition effects of large (500 mg) and small (45 mg) consistent re ward. Subjects and Apparatus The Ss were 24 rats of the Wistar strain from the colony of the University of Texas at Arlington. All Ss were approximately 90 days old at the beginning of training. Approximately equal numbers of eaeh sex were used. The straight alley was approximately 6 ft long. Four successive time measures were taken in the alley; the first was for a 12-in. seetion beginning about 12 in. from the startbox door, the second was about 18 in., and the goal measure was approximately 10 in., terminating 8 in. from the goal cup. The fourth measure incJuded the time from the startbox door to 8 in. from the goal cup. Procedure Ss were plaeed on 23-h deprivation on Day 1. On Days 2-6, Ss were handled in Fig. 1. Running times for groups on 4S-mg and SOO-mg consistent reward schedule. Psychon. Sei., 1970, Vol. 19 (3) groups for 1 h daily. Food was available on the handling table during this hour. On Day 7, Ss were assigned to groups on a random basis. Each S explored the runway for 5 min daily on Days 7-9. After exploration, Ss were returned to their horne cages and, about 15 min later, each S received his appropriate training reward in a goal cup. A few minutes later, an appropriate amount oflab blocks was given to total 15 g when added to the reward. Ss were maintained on 15-g total daily food for the remainder of the study. On Days 10 and 11, each S received two trials per day and four trials on Days 12 and 13. Six trials per day were given throughout the remainder of training. Ss were brought into the running room in squads of six in a carrying cage with individual compartments. Each S was given a single trial in rotation. This procedure gives an intertrial interval of approximately 6 min. Equal numbers of Ss from eaeh group were arranged in a random order in each squad. The two groups were designated on the basis of their reward schedule, 1-500 (a single 500-mg Noyes pellet on each trial) and 1-45 (one 45-mg Noyes pellet per trial). All Ss received 54 acquisition trials. Results . Figure 1 shows the acquisition data for the full runway measure. As may be noted, Group 1-500 has substantially shorter running times over the early stages of training. Analysis of the first 4 days of acquisition gives a significant difference (F = 6.94, df= 1/22, p< .02). Later in training, there is very little apparent difference in performance of Groups 1-500 and 1-45. An analysis of the data confrrmed the impression (F = 1.01). EXPERIMENT 2 The series of studies involving consistent reward with different magnitudes suggests that the partial reinforcement situation should also be investigated. Wagner (1961) found that larger rewards gave more vigorous performance and greater resistance to extinction when given on a random partial reinforcement schedule. Wagner ran two levels of training, 16 and 60 trials. The present study included a group given extended training, since extended training appears to change the consistent reward situation. Subjects and Apparatus The Ss were 44 rats of the Long-Evans strain from the colony of the University of Texas at Arlington. All Ss were approximately 90 days old at the beginning of training. Approximately equal numbers of each sex were used. The apparatus was the same as that used in Experiment 1. Procedure Ss were placed on 23-h deprivation on Day 1 and handled and fed on Days 2-6. Ss explored the test alley on Days 7-10 for approximately 5 min daily. On Day 9, Ss were divided into two groups, on a random basis. Group 45 was to receive one 45-mg Noyes pellet as reward and Group 500 was to receive one 500-mg pellet. On Days 9 and 10, Ss received one goalbox (GB) placement per day with the appropriate reward. On Day 11, each S received two running trials, four trials on Day 12, and six trials per day thereafter. A schedule of 50% randorn reward was used. Ss were 20 uw !!2. w ~ D-----{J 145 • 1-500 • 15 z « o UJ ~ 5 Z « w ~ 2 3 4 5 6 7 8 9 10 DAYS 139 G u.J 25 20 45 Short V) W ~ 500 Short 20 \~ the small reward group declines slightly but not significantly more than the large-reward group. Z z z 15 ::J a:: Z Fig. 2. a and b. Running tim es for groups given partial 45- or 500-mg reward at two levels of aequisition training. DISCUSSION Taken alone, these two studies are not u.J sufficient to convince anyone that the ::E effects of reward magnitude are sharply EXTINCTlON z 5 decreased or disappear after extended <t: u.J training. Aseries of nine other consisten t ::E reward studies has been run in our laboratory. Six of these studies have been 1-2 4 3 5 6 7 8 9 10 11 12 put into one paper that is now under editorial consideration. In addition, Black (in press) and Bloom & Milstead (1969) DAYS have substantial evidence that, during acquisition, magnitude differences 15 disappear under some conditions. A 0 - - 0 45 Lang 2b reasonable conc1usion is that after about 500 lang 60 consistently reinforced acquisition trials G in a straight alley, the acquisition effects of u.J !Odifferent reward magnitudes are either u.J minimal or absent. Our other studies ~ 10 indicate that extinction differences I-ACQUISITION disappear or become minimal after about o 100+ consistently reinforced acquisition z trials. The situation as regards random z Z partial reinforcement is not so c1ear. The ::J a:: study presented here is our only complete EXTINCTION study of its kind. Single studies must, of z 5 <t: necessity, be received with substantial Ci skepticism. Further work is under way. u.J ::E A number of possible c (...truncated)