The Psychological Record, 1997,47,145-166 THE EFFECTS OF TERMINAL-LINK STIMULUS ARRANGEMENTS ON PREFERENCE IN CONCURRENT CHAINS LAUREL COLTON and JAY MOORE University of Wisconsin-Milwaukee Pigeons served as subjects in two concurrent-chains experiments in which responding on one or the other of two side keys occasionally produced either a short- ( 10-s) or a long- ( 20-s) duration terminal-link schedule of food reinforcement according to equal variable-interval 30-s initial-link schedules. In Experiment 1, the terminal link was in effect randomly on either key, or always on the opposite key, in relation to the initial link from which it was entered. In addition, terminal-link stimuli were differential keylights, nondifferential keylights, or differential houselights. Preference tended toward indifference with a random location of a terminal link and nondifferential keylights. However, the birds preferred the shorter terminal link when the terminal link was always in effect on the same key, or always on the opposite key, with either differential or nondifferential keylights. The birds also preferred the shorter terminal link with a random location, nondifferential keylights, and differential houselights. In Experiment 2, the terminal link was always in effect on the center key. As before, terminal-link stimuli were differential keylights, nondifferential keylights, or differential houselights. Preference tended toward indifference with nondifferential keylights. However, the birds preferred the shorter terminal link with either differential keylights or differential houselights. Taken together, the data indicate that terminal-link stimuli in concurrent chains function as conditioned reinforcers, rather than simple discriminative stimuli, although other features of the terminal links such as stimulus location act in concert with the visual stimuli to influence preference. The concurrent-chains procedure has long been used in studies of choice (e.g., Autor, 1960,1969; Herrnstein, 1964). Typically, pigeons are presented with two concurrently available response keys, each illuminated by a stimulus correlated with the initial link of a chain schedule of reinforcement. Occasionally, a response on either of the keys produces a terminal-link stimulus in whose presence primary Reprint requests may be sent to Laurel Colton, Department of Psychology, University of Wisconsin-Milwaukee, Garland Hall, Box 413, Milwaukee, WI 53201.
146 COLTON AND MOORE reinforcement (e.g., food) occurs according to a designated schedule. Following the delivery of reinforcement, the initial links are reinstated for the next period of choice, and the cycle begins again. An experimental session consists of a series of such exposures, typically until a specified number of reinforcers is obtained. The major dependent variable in the concurrent-chains procedure is the relative response rate in the initial links, or choice proportion, calculated by dividing the rate of initial-link responding on one key by the rate on both keys. A long-standing experimental question concerns the contribution of terminal-link stimuli to responding in concurrent chains. Two possibilities have emerged in the literature. The first is that the stimuli are merely discriminative for a given rate of primary reinforcement and without any conditioned reinforcing strength (Baum, 1973, p. 144). The second is that they function as conditioned reinforcers and actually mediate initiallink choice responding (Williams & Fantino, 1978, p. 86). A number of studies have reported data on this question over the years. For example, Autor (1960, 1969) presented pigeons with a choice between various aperiodic terminal-link schedules (e.g., variable-interval, VI, and variable-time). A terminal link was always located on the same key from which the terminal link was entered. Autor found a higher rate of initiallink responding on the key leading to the higher terminal-link rate of reinforcement. He concluded that responding in concurrent chains was mediated through the conditioned reinforcement provided by the terminallink stimuli, a conclusion that is consistent with a more molecular view of the variables controlling preference in concurrent chains. Herrnstein (1964) also presented pigeons with a choice between aperiodic terminal-link schedules (e.g., VI and variable-ratio), in a procedure similar to Autor's where the terminal link was located on the same key from which the terminal link was entered. He found that the pigeons' initial-link responding was distributed in proportion to the relative rate of terminal-link reinforcement. Herrnstein's results are often said to support a molar view of behavioral processes, according to which (a) the overall correlation between responding and reinforcement is primary, and (b) putative conditioned reinforcers function merely as discriminative stimuli without conditioned reinforcing value of their own (e.g., Baum, 1973, p. 144). Later studies continued the attempt to dissociate the possible roles of terminal-link stimuli in concurrent chains. Williams and Fantino (1978) gave pigeons a choice between terminal links consisting of different fixed-interval () schedules. The terminal-link stimuli could appear randomly on left or right keys. In addition, the terminal-link stimuli appearing on each key were either correlated or uncorrelated with the schedule in effect. The use of the randomly alternating position of the terminal-link key and the use of correlated or uncorrelated terminal-link stimuli allowed Williams and Fantino to examine the effects of overall time to reinforcement independent of any immediate changes in stimuli that could be construed as differential conditioned reinforcement. In
EFFECTS OF TERMINAL-LINK STIMULUS 147 some conditions, because the terminal-link stimulus was uncorrelated with left- and right-key schedules, any difference in initial-link responding could be considered a result of the overall delay to reinforcement. Williams and Fantino (1978) found higher preferences for the shorter terminal link when terminal-link stimuli were correlated rather than uncorrelated with schedule duration. They concluded that their findings "support the role of conditioned reinforcement as a determinant of choice in the concurrent-chains procedure" (p. 86). Moreover, their findings were inconsistent with the molar notions that choice behavior is a function only of the primary reinforcement presented at the end of each cycle and that terminal-link stimuli are solely discriminative. In related studies, Omino (1993) and Omino and Ito (1993) examined preference in concurrent chains using a procedure similar to Williams and Fantino (1978). In both Omino (1993) and Omino and Ito (1993), the terminal links were fixed-time (FT) schedules. Entrance into one of the terminal links changed the keylight from white to an "uncued blackout" condition or remained white ("uncued white" condition) while the other key went dark. In both cases, preference for the FT terminallink schedule with the shorter delay to reinforcement increased with increases in absolute values of paired delays in both uncued blackout and uncued white conditions. Furthermore, preference for the shorter delay was higher in the uncued white than in the uncued blackout conditions. Thus, Omino (1993) concluded that "the terminal-link keylight stimulus condition differentially exerts its effect on choice as a conditioned reinforcer in the concurrent-chains procedure" (p. 591). More recently, Alsop, Stewart, and Honig (1994) examined preference using a three-key procedure, where initial links were in effect on the side keys and terminal links on a center key. Fixed-interval terminal-link schedules were varied across conditions. In one case, the center key was always illuminated with white light during the terminal links, regardless of which initial-link schedule had arranged access to the terminal link. In another case, the center key was illuminated with red light if the left initial-link key had arranged access to the terminal link, or blue light if the right initial-link key had arranged access to the terminal link. Preferences for the shorter terminal-link schedule were greater when different keylights were used, a finding consistent with Omino (1993), Omino and Ito (1993), and Williams and Fantino (1978). Based on the procedures in the studies cited above, the relevant research can be organized into a 3 x 4 table in terms of (a) the location of a terminal link relative to its associated initial link and (b) whether an unique exteroceptive stimulus is present during the terminal link. Table 1 presents a summary of terminal-link stimulus arrangements employed in the relevant research on choice in concurrent chains. The horizontal axis of Table 1 indicates the location of the terminal link relative to the location of the initial link from which the terminal link was entered. The first three columns pertain to terminal-link stimulus arrangements in which two keys are used. In the same arrangement, the
148 COLTON AND MOORE Table 1 Studies Concerning Conditioned Reinforcement and Preference in Concurrent Chains and Possible Terminal-Link Stimulus Conditions Location of terminal-link stimulus Nature of terminal- Same Random Opposite Center link stimulus (two-key) (two-key) (two-key (three-key) Cued Autor (1960,1969) Williams & Alsop et al. Herrnstein (1964) Fantino (1978) (1994) Uncued Omino (1993) Williams & Alsop et al. Omino & Ito (1993) Fantino (1978) (1994) Houselight >< >< terminal link was always located on the same key from which the terminal link was entered. This procedure is the conventional procedure used in the vast majority of concurrent chains studies. In the random arrangement, the terminal link appears randomly on the same key or the opposite key from which the terminal link was entered. In the opposite arrangement, the terminal link always appears on the opposite key from which the terminal link was entered. The fourth column pertains to arrangements in which three keys are used. In the center arrangement, the initial links appear on the two side keys, and a terminal link is always in effect on the center key. The vertical axis of Table 1 indicates the nature of the terminallink stimuli. Here, we will attempt to maintain consistency by using the terminology of the previous studies. In cued conditions, the terminallink stimuli are differential keylights (i.e., different keylight colors associated with each terminal link). In uncued conditions, the terminal-link stimuli are nondifferential keylights. In houselight conditions, the terminal-link stimuli are differential houselights. Table 1 presents the six relevant studies noted above according to the terminal-link stimulus arrangements used: same-cued (Autor, 1960, 1969; Herrnstein, 1964), same-uncued (Omino, 1993; Omino & Ito, 1993), random-cued and -uncued (Williams & Fantino, 1978), and center-cued and -uncued (Alsop et ai., 1994). The present research sought to determine the generality of the prior results and also fill in the unexamined cells of the matrix in Table 1 in order to clarify the role of terminal-link stimuli. Experiment 1 employed a modified two-key version of the concurrent-chains procedure. Pigeons chose between terminal links consisting of schedules of different durations. In some conditions, terminal links appeared randomly on either the left or the right key. In other conditions, they always appeared on the key opposite that on which initial-link responding had accessed the terminal link. Behavior was examined in cued conditions, uncued conditions, and with differential houselights. Differential houselights were
EFFECTS OF TERMINAL-LINK STIMULUS 149 used in recognition of prior research indicating that such stimuli influence choice responding (Brownstein & Pliskoff, 1968). Thus, Experiment 1 permitted the investigation of two previously unexamined terminal-link stimulus conditions and the effects of terminal-link keylight stimuli on stimulus control of initial-link responding. It also examined the effect of differential houselights on terminal-link stimulus control in a manner not previously used. Experiment 2 employed a three-key version of the concurrent-chains procedure much like that of Alsop et al. (1994). Pigeons chose between terminal links consisting of schedules of different durations. Terminallink keylight stimuli and schedules were in effect on a center key. As in Experiment 1, behavior was examined in cued conditions, uncued conditions, and with differential houselights. Thus, in addition to examining the generality of the findings of Alsop et al. (1994) using a three-key arrangement, Experiment 2 determined whether preference develops in this center-key arrangement when nondifferential keylights and differential houselights are employed. Data suggesting that patterns of preference and other characteristics of initial-link responding remain relatively constant across all manipulations would indicate that overall rate of primary reinforcement in the terminal links is central, independent of any contribution of conditioned reinforcement. These data would suggest that the primary role of terminal-link stimuli is simply to enhance discrimination of terminal-link events (Baum, 1973; Schneider, 1972; Schuster, 1969). However, data suggesting that preferences are higher in cued conditions than in uncued conditions would indicate that conditioned reinforcement does contribute to responding, as reported by Williams and Fantino (1978) and others (see Williams, 1994, for a review). General Method Subjects Four adult male pigeons of mixed breeds and different experimental histories (R-222, R-213, R-1580, R-187) served as subjects. The birds were maintained at approximately 80% of their free-feeding weights. Sessions were conducted at approximately the same time of day, 6 days per week, under 23-h food deprivation. Apparatus Three-key experimental chambers for pigeons were used. Only the two side keys were used in Experiment 1, and all three keys were used in Experiment 2. Conventional electromechanical equipment located in an adjacent room controlled events and recorded the data. Houselights were used only on specific occasions for differential cues, rather than for general illumination. These houselights were either red or green 11 OV AC, 6-W light bulbs, illuminated only during the terminal links. The bulbs were located behind a plastic diffusing panel at the top rear of the experimental chambers.
150 COLTON AND MOORE General Procedure The initial links in both Experiments 1 and 2 were equal, interdependent, variable-interval (VI) 30-s schedules (Stubbs & Pliskoff, 1969). The schedules were interdependent in the sense that whenever one terminal link was assigned, the other initial-link schedule became inoperative. This control procedure ensured that the pigeons entered each terminal link equally often. During the initial links, the keylights were white. With the exceptions noted below, an 10-s schedule was in effect during one terminal link and an 20-s schedule was in effect during the other. During a terminal link, responding on the operative key produced a primary reinforcer (3-s access to mixed grain) according to the schedule in effect. The initial links were reinstated after each primary reinforcer. As in Williams and Fantino (1978), but unlike Alsop et al. (1994), no changeover delay was used. Each session terminated after the delivery of 40 reinforcers. Each condition was in effect until a stability criterion was satisfied. This criterion required that the birds perform in at least 15 sessions. The final nine sessions were divided into three blocks of three sessions each. A standard choice proportion (Chung & Herrnstein, 1967; Herrnstein, 1961) was calculated for each session and a mean choice proportion obtained for each block of sessions. Responding was considered stable when choice proportions did not differ by more than.05 and showed no increasing or decreasing trend across blocks. The data of primary interest were the asymtotic preference levels for each condition, expressed as the average choice proportion for the nine stable sessions. All four birds served in both experiments. Pigeons R-222 and R-187 served first in Experiment 1 and then in Experiment 2. Pigeons R-213 and R-1580 served first in Experiment 2 and then in Experiment 1. Experiment 1 Procedure In Experiment 1, all key arrangements involved only the two side keys. Behavior was examined in seven conditions appearing in Table 2: Same-cued, Same-uncued, Random-cued, Random-uncued, Oppositecued, Opposite-uncued, and Houselight. With the exception of the houselight condition, these conditions are presented schematically in Figure 1. In cued conditions, the terminal-link stimulus in effect when a terminal link was entered from the left initial link was a green light, and from the right initial link, a red light. In uncued conditions, a circle was in effect during the terminal links, regardless of whether the terminal link was entered from the left or right initial link. In the opposite arrangement (Figure 1, third row), the terminal link always appeared on the opposite key from which the terminal link was entered. The effects of this manipulation have never been examined before. Thus, these conditions attempted to extend the study of Williams and Fantino (1978) and to determine whether the location of the terminal link influenced terminal-link stimulus control of initial-link responding.
EFFECTS OF TERMINAL-LINK STIMULUS 151 Table 2 Experiment 1: Sequence of Conditions and Terminal-Link Schedules for Each Subject Terminal-link schedule after entry from: Sequence for each subject -- Condition Left key Right key R-222 R-213 R-1580 R-187 Same-cued 10 s 20 s 1, 13 2,13 1, 13 2, 13 20 s 10 s 2 1 2 1 Same-uncued 10 s 20 s 11 7 12 12 20 s 10 s 12 8 11 11 Opposite-cued 10 s 20 s 4 3 6 10 20 s 10 s 3 4 5 9 Opposite-u ncued 10 s 20 s 5 12 10 7 20 s 10 s 6 11 9 8 Random-cued 10 s 20 s 8 5 7 3 20 s 10 s 7 6 8 4 Random-uncued 10 s 20 s 9 10 4 6 20 s 10 s 10 9 3 5 Houselight 10 s 20 s 15 15 20 s 10 s 14 14 Note. Birds R-222 and R-187 were not exposed to the houselight condition in Experiment 1. In the houselight condition, the location of a terminal link was random, and the keylights were nondifferential (i.e., a circle) as in the random-uncued condition. However, a red houselight was in effect during a terminal link entered from the right initial-link key, and a green houselight was in effect during a terminal link entered from the left initiallink key. The effects of this manipulation also have never been examined before. The manipulation sought to determine whether differential houselights would control initial-link responding, in the absence of differential keylights and reliable location of terminal-link keylight stimuli. As indicated in Table 2, an initial determination of preference was made between 10-s and 20-s schedules. Then, a reversal was conducted in which the schedule that was formerly 10 s became 20 s, and vice versa. A reversal served to demonstrate experimental control by the conditions in effect. In Figure 1, these initial and reversed assignments of schedules to the terminal links are respectively indicated by two numbers for each (i.e., 10 and 20). Each subject was exposed to its conditions in a different order, and reversals were performed immediately after initial determinations of preference. All subjects began with the same-cued and same-uncued conditions. However, they were exposed to these conditions only once, because these conditions were common to both Experiments 1 and 2. For ease of comparison, all four birds are listed in Table 2. Upon completing all experimental conditions in Experiments 1 and 2, all subjects were exposed to a final same-cued condition in which left and right terminal links were 10 sand 20 s, respectively. This final determination confirmed performance in the same-cued condition.
152 COLTON AND MOORE Same-cued VI 30 VI 30 Same-uncued VI 30 VI30 W 10/20 20/10 10/20 20/10 Random-cued VI 30 VI 30 W Random-uncued VI 30 VI 30 W 10/20 (p-.s) 20/10 (p=-.5) 10/20 (p=.5) 20/10 (p=.5) Opposite-cued VI 30 VI 30 Opposite-uncued VI 30 VI 30 10/20 20/10 10/20 20/10 Figure 1. Key arrangements used in Experiment 1. After the behavior of Pigeons R-213 and R-1580 was examined under the same, random, and opposite arrangements, these two birds were exposed to the houselight condition. The keylight stimuli in this condition were the same as those in the random-uncued condition, except that a green housejight was in effect when a terminal link was entered from the left key, and a red housejight was in effect when a terminal link was entered from the right key. Results Figure 2 presents the average asymptotic preferences in terms of the shorter terminal-link schedule ( 10 s) for all four birds. The preferences in each condition are averaged across the last nine stable
EFFECTS OF TERMINAL-LINK STIMULUS 153 Cued o Uncued 1.0 R-222 R-213 0.5 (]) (.) 55 0.0 ~ (]) 051.0 ~ a.. R-187 R-1580 0.5 0.0 Same Opposite Random Same Opposite Random Houselight TERMINAL-LINK ARRANGMENT Figure 2. Experiment 1: Average asymptotic preference (choice proportions for 10-s over 20-s terminal-link schedule) in cued and uncued key arrangements for each subject. sessions for initial determinations and reversals. The only exception was the same-cued condition for which preferences in initial, reversed, and final determinations were averaged. Shaded bars represent cued conditions and unshaded bars represent uncued conditions. The same-cued and sameuncued conditions are shown for all four birds. The houselight condition for Pigeons R-213 and R-1S80 is represented by a shaded bar. Figure 2 shows that with the same arrangement, all four birds preferred the 10-s schedule in both the same-cued and sameuncued conditions. The finding of equivalent preference in these two conditions contrasts with that of Omino (1993), who found somewhat lower preferences in the same-uncued condition (his "uncued white" condition).
154 COLTON AND MOORE In the opposite arrangement, all four birds again preferred the 10- s schedule in both the opposite-cued and opposite-uncued conditions. Note that the average preference exhibited by R-1580 in the oppositecued condition is slightly less than that exhibited in the opposite-uncued condition (see Figure 2). No other birds exhibited this pattern. In the random arrangement, all four birds preferred the 10-s schedule in the random-cued condition. However, three of the four birds tended toward indifference in the random-uncued condition. The only exception was the slight preference exhibited by R-187. Inspection of the data in Table 3 indicates that any preferences that appeared during initial determinations in this condition were unreliable in that they never fully reversed for Pigeons R-222, R-213, and R-1580. Table 3 Experiment 1: Absolute and Relative Response-Rate and Reinforcement Data T-L I-L T-L T-L schedule (s) resp/min resp/min IRI (s) Bird Condition L R L R L R L R P (SO) Sessions R-222 Same-cued 10 20 125 5 102 35 11 22.96 (.02) 24 20 10 11 59 58 52 20 13.83 (.03) 30 Same-uncued 10 20 31 8 108 100 11 20.85 (.06) 25 20 10 7 65 62 183 20 11.90 (.04) 25 Random-cued 10 20 59 18 100 77 11 21.76 (.04) 16 20 10 12 51 77 102 20 11.81 (.04) 18 Random-uncued 10 20 42 21 101 123 11 20.67 (.04) 24 20 10 51 21 142 111 20 11.30 (.04) 15 Opposite-cued 10 20 73 8 110 64 11 20.90 (.04) 20 20 10 10 38 54 58 22 11.79 (.08) 16 Opposite-uncued 10 20 92 6 61 70 12 21.94 (.01) 15 20 10 17 36 55 62 21 11.70 (.14) 30* Final Same-cued 10 20 83 11 144 109 11 20.89 (.03) 16 R-187 Same-cued 10 20 30 11 169 37 11 21.73 (.03) 34 20 10 4107 169 102 20 11.96 (.01) 15 Same-uncued 10 20 93 13 71 45 11 21.88 (03) 19 20 10 13 82 60 51 20 12.86 (.05) 30* Random-cued 10 20 53 3 210 167 10 20.94 (.02) 16 20 10 5 53 161 192 20 10.91 (.04) 19 Random-uncued 10 20 30 9 53 89 11 20.77 (.04) 19 20 10 16 18 126 99 20 11.55 (.10) 30* Opposite-cued 10 20 33 7 112 99 10 20.80 (.09) 17 20 10 9 60 94 154 20 10.87 (.05) 24 Opposite-uncued 10 20 36 7 106 131 10 20.89 (.04) 16 20 10 9 43 112 142 20 10.83 (.05) 24 Final Same-cued 10 20 117 7 253 54 10 20.94 (.03) 23 R-1580 Same-cued 10 20 71 16 201 126 11 20.82 (. 02) 16 20 10 7 89 140 203 20 11.93 (.01) 23 Same-uncued 10 20 96 12 245 165 10 20.89 (.02) 20 20 10 14116 120 278 20 11.89 (.04) 28 Random-cued 10 20 44 9 141 86 11 20.83 (.01) 30 20 10 5 76 84 175 20 10.92 (.03) 23 Random-uncued 10 20 26 33 102 110 11 20.44 (.04) 17 20 10 30 37 96 93 20 11.56 (.03) 18
EFFECTS OF TERMINAL-LINK STIMULUS 155 R-1S80 Opposite-cued 10 20 34 21 96 78 10 21.61 (.06) 20 20 10 17 40 58 150 20 10.69 (.07) 15 Opposite-uncued 10 20 49 14 109 95 11 20.77 (.03) 18 20 10 5 71 87 163 20 10.94 (.01) 15 Final Same-cued 10 20 92 18 230 184 10 20.84 (.02) 15 Houselight 10 20 65 15 125 86 11 20.81 (.03) 19 20 10 5 57 75 108 20 10.08 (.02) 17 R-213 Same-cued 10 20 97 14 145 83 11 20.88 (.03) 22 20 10 7 96 90 101 21 11.93 (.02) 21 Same-uncued 10 20 110 33 116 90 10 21.78 (.05) 30* 20 10 12151 112151 20 11.93 (.03) 18 Random-cued 10 20 104 31 91 74 10 20.77 (.06) 16 20 10 8 84 94 139 20 10.91 (.03) 25 Random-uncued 10 20 16 18 92 123 10 20.47 (.03) 18 20 10 24 32 131 87 20 11.58 (.06) 30* Opposite-cued 10 20 111 12 91 106 10 20.90 (.03) 17 20 10 10 82 57 167 20 10.90 (.03) 23 Opposite-uncued 10 20 144 21 113115 10 20.87 (.03) 16 20 10 18 73 73 136 20 10.80 (.03) 20 Final Same-cued 10 20 128 25 116 156 10 20.84 (.00) 27 Houselight 10 20 141 9 4 91 10 20.94 (.03) 15 20 10 40 71 116 147 20 10.35 (.06) 19 Note. Data presented for each pigeon and condition are: IRI = interreinforcement interval; P = asymptotic level of preference; SO = choice proportion standard deviation; where T-L = terminal-link and I-L = initial-link. The choice proportion (P) is expressed in terms of the 10-s terminal link. Choice proportions for the shorter terminal-link schedule ( 10 s) were calculated, e.g., L / (L + R), where Land R represent the number of responses during left and right initial links, respectively. Absolute response rates during the initial links were calculated, e.g., L / total time in initial links. Absolute response rates during the terminal links were calculated, e.g., rate during 10 / (rate during 10 + rate during 20). Interreinforcement intervals were calculated, e.g., total time in left terminal link / number of reinforcements obtained in left terminal link. * Preference failed to stabilize within 30 sessions. In the houselight condition, both birds reliably preferred the 10- s schedule. The preferences here were exhibited even though the birds were responding on a key which appeared in a random location with a nondifferential keylight. Table 3 displays additional data from Experiment 1. Absolute response rates during the initial links were again averaged across the final nine sessions of the initial determinations and reversals. These absolute response rates were higher on the key leading to the 10-s terminal-link schedule in all conditions except the random-uncued condition. Rather, in this condition, response rates were often higher during the initial link leading to the longer terminal-link schedule ( 20 s). For three of the four birds, initial-link response rates did not reverse with reversals in terminal links for this random-uncued condition. In this condition, only R-187 responded with higher rates during whichever initial link led to the shorter terminal-link schedule. Recall that this bird also slightly preferred whichever key was correlated with the shorter terminal-link schedule in initial and reversed random-uncued conditions.
156 COLTON AND MOORE Table 3 also presents absolute response rates during the terminal links, averaged across the final nine sessions of the initial determinations and reversals. Figure 3 presents relative response rates during the terminal links, expressed in terms of the shorter 10-s terminal link, for each condition. All relative response-rate data were averaged over initial determinations and reversals. Absolute response rates during the terminal links showed no systematic trends across conditions. Relative response rates during the 10-s terminal link were often higher in cued conditions (>.50, as shown in Figure 3) than in uncued conditions (approximately.50, as shown in Figure 3). However, some conditions produced the opposite result (for example, compare same-cued and -uncued conditions for Pigeons R-222, R-213, and R-1580). C/) ~ c: :::i 1.0 cts c: EO.5 ~ ~ 0> c: "C ::J 0.0 Cl... cts a: 1.0 C/) c: o Q. C/) a: 0.5 > -..::; ~ a: 0.0 R-222 R-187 Cued o Uncued R-213 R-1580 Same Opposite Random Same Opposite Random Houselight TERMINAL-LINK ARRANGEMENT Figure 3. Experiment 1: Relative terminal-link responses/min expressed in terms of the shorter ( 10-s) terminal-link schedule and averaged over initial determinations and reversals.
EFFECTS OF TERMINAL-LINK STIMULUS 157 Interreinforcement intervals (IRis) were calculated (in seconds) from the final nine sessions of terminal-link data for each condition. Table 3 shows that the obtained I Rls were approximately 10 sand 20 s, corresponding to terminal-link schedules of 10 sand 20 s, respectively. After subjects had completed all conditions, a final same-cued ( 10 vs. 20) assessment reconfirmed preferences for the shorter terminallink schedule as observed previously. Table 3 includes choice-proportion standard deviations and the number of sessions required for preferences to stabilize in each condition of Experiment 1. No systematic differences in either of these measures emerged between conditions. Cases in which preferences failed to stabilize within 30 sessions occurred in uncued conditions only and are indicated by asterisks in Table 3. Experiment 2 Procedure The same four birds served in Experiment 2. Behavior was examined in the four comparisons listed in Table 4: Same-cued, Centercued, Center-uncued, and Houselight. With the exception of the houselight condition, these conditions appear in Figure 4. The same key arrangement was described in Experiment 1 and employed two keys (Figure 4, first row). Once again, in this arrangement, terminal-link stimuli appeared on the same key on which entry had been gained. The second key arrangement involved three keys, and was called the center arrangement (Figure 4, second row). In this arrangement, terminal links were in effect on the center key rather than on the two side keys. Behavior was examined in cued conditions (Figure 4, left column) and uncued conditions (Figure 4, right column). In the center-cued condition, a green light was in effect on the center key when a terminal link was entered from the left initial-link key, and a red light was in effect Table 4 Experiment 2: Sequence of Conditions and Terminal-Link Schedules for Each Subject Terminal-link schedule after entry from: Sequence for each subject Condition Left key Right key R-222 R-213 R-1580 R-187 --------------------------------------- Same-cued 10 s 20 s 1 2 1 2 20 s 10 s 2 1 2 1 Center-cued 10 s 20 s 3 5 4 6 20 s 10 s 4 6 3 5 Center-uncued 10 s 20 s 6 3 5 4 20 s 10 s 5 4 6 3 Houselight 10 s 20 s 8 8 20 s 10 s 7 7 Note. Birds R-213 and R-1580 were not exposed to the houselight condition in Experiment 2.
158 COLTON AND MOORE 10/20 20/10 Center-cued VI 30 VI 30 Center-uncued VI 30 VI 30 10/20 1 0/20 Figure 4. Key arrangements used in Experiment 2. when a terminal link was entered from the right initial-link key. In the center-uncued condition, a circle was in effect on the center key, regardless of the initial link from which the terminal link was entered. These conditions examined the generality of Alsop et al. (1994). Pigeons R-222 and R-187 were also exposed to the center arrangement with nondifferential keylights (i.e., the center-uncued condition) and with differential houselights. In this condition, a circle was in effect on the center key, regardless of which initial link had resulted in terminal-link access. However, if the terminal link was entered from the left key, a green houselight was in effect, and if the terminal link was entered from the right key, a red houselight was in effect. For each of the four combinations of location and terminal-link stimuli, an initial determination of preference was made between 10-s and 20-s schedules. Then, a reversal of these terminal-link schedules was performed, in which the schedule that was formerly 10 s became 20 s, and vice versa. A reversal served to demonstrate experimental control by the conditions in effect. In Figure 4, these initial and reversed assignments of schedules to the terminal links are respectively indicated by two numbers for each : 10 and 20. Again, the birds were exposed to these conditions in a different order (see Table 4), with the exception that
EFFECTS OF TERMINAL-LINK STIMULUS 159 reversals were performed immediately after initial determinations of preference. For ease of comparison, all four birds are listed in Table 4. In summary, Experiment 2 attempted to examine the generality of Alsop et al. (1994), and to determine whether preference would develop in the center-key arrangement with nondifferential keylights and differential houselights. Results Figure 5 presents the average asymptotic preferences in terms of the shorter terminal-link schedule ( 10 s) for all four birds. The preferences in each condition are averaged across the last nine stable 1.0 R-222 Cued o Uncued R-213 0.5 U ~ 0.0 '- '+- CD CD 1.0 '- CL R-187 R-1580 0.5 0.0 Same Center Houselight Same Center TERMINAL-LINK ARRANGMENT Figure 5. Experiment 2: Average asymptotic preference (choice proportion for 10-s over 20-s terminal-link schedule) in cued and uncued key arrangements for each subject.
160 COLTON AND MOORE sessions for initial determinations and reversals. The only exception was the same-cued condition for which preferences in initial, reversed, and final determinations were averaged. The shaded bars represent cued conditions, and the unshaded bars represent uncued conditions. The houselight condition for R-222 and R-187 is represented by a shaded bar. Figure 5 shows that all four birds preferred the 10-s terminal-link schedule in the same-cued condition. Figure 5 also shows that all four subjects preferred the 10-s terminal-link schedule in the center-cued condition. Their preferences here were comparable to those in the same arrangement. However, three of the four birds tended toward indifference in the center-uncued condition, a result similar to that obtained with the random-uncued condition in Experiment 1. The slight preferences that did appear in the center-uncued condition were unreliable in that they never fully reversed for R-213, R-1580, and R-187 (see Table 5). A fourth bird (R-222), however, did exhibit a slight preference for whichever key was correlated with the shorter terminal-link schedule during both initial determinations and reversals. These data are consistent with those of Alsop et al. (1994). Table 5 Experiment 2: Absolute and Relative Response-Rate and Reinforcement Data T-L I-L T-L T-L schedule (s) resp/min resp/min IRI (s) Bird Condition L R L R L R L R P (s) Sessions R-222 Same-cued 10 20 125 5 102 35 11 22.96 (.02) 24 20 10 11 59 58 52 20 13.83 (.03) 30 Center-cued 10 20 86 13 266 145 10 20.87 (.03) 18 20 10 16 84 135221 20 11.84 (.01) 19 Center-uncued 10 20 50 27 181 222 10 20.65 (.03) 30 20 10 51 90 99247 20 11.64 (.02) 28 Final Same-cued 10 20 83 11 144 109 11 20.89 (.03) 16 Houselight 10 20 42 6 209 105 10 21.87 (.03) 24 20 10 8 61 107 49 20 11.13 (.09) 30* R-187 Same-cued 10 20 30 11 169 37 11 21.73 (.03) 34 20 10 4107 169 102 20 11.96 (.01) 15 Center-cued 10 20 59 11 219 88 10 20.85 (.04) 17 20 10 13121 103 246 20 10.90 (.02) 20 Center-uncued 10 20 72 29 184 154 11 20.72 (.03) 22 20 10 43 21 150 136 20 11.32 (.03) 23 Final Same-cued 10 20 117 7 253 54 10 20.94 (.03) 23 Houselight 10 20 73 12 27 39 12 22.86 (.03) 24 20 10 15 38 43 26 20 11.28 (.01) 30* R-1580 Same-cued 10 20 71 16 201 126 11 20.82 (.02) 16 20 10 7 89 140 203 20 11.93 (.01) 23 Center-cued 10 20 95 24 194 184 10 20.80 (.06) 20 20 10 3103 179 219 20 11.97 (.02) 20 Center-uncued 10 20 45 56 85 122 10 21.45 (.07) 30* 20 10 43 53 120 99 20 11.54 (.04) 16 Final Same-cued 10 20 92 18 230 184 10 20.84 (.02) 15
EFFECTS OF TERMINAL-LINK STIMULUS 161 R-213 Same-cued 10 20 97 14 145 83 11 20.88 (.03) 22 20 10 7 96 90 101 21 11.93 (.02) 21 Center-cued 10 20 109 30 98 78 10 21.79 (.03) 16 20 10 13124 75 110 20 11.90 (.02) 15 Center-uncued 10 20 19 40 82 87 10 20.32 (.05) 22 20 10 15 67 85 80 20 11.82 (.09) 24 Final Same-cued 10 20 128 25 116 156 10 20.84 (.00) 27 Note. Data presented for each pigeon and condition are: IRI = interreinforcement interval; P = asymptotic level of preference; SO = choice proportion standard deviation; where T-L = terminal-link and I-L = initial-link. The choice proportion (P) is expressed in terms of the 10-s terminal link. Choice proportions for the shorter terminal-link schedule ( 10 s) were calculated, e.g., L / (L + R), where Land R represent the number of responses during left and right initial links, respectively. Absolute response rates during the initial links were calculated, e.g., L / total time in initial links. Absolute response rates during the terminal links were calculated, e.g., rate during 10 / (rate during 10 + rate during 20). Interreinforcement intervals were calculated, e.g., total time in left terminal link / number of reinforcements obtained in left terminal link. * Preference failed to stabilize within 30 sessions. Finally, Figure 5 shows that R-222 and R-187 preferred the 10-s schedule in the houselight condition of the center-key arrangement employing nondifferential keylights. The preferences in this condition were just as robust as in the center-cued condition. The outcome of this manipulation has not been reported before. Table 5 presents further data from Experiment 2. Absolute response rates during the initial links were calculated using data from the final nine sessions of each condition. Table 5 presents initial-link absolute response rates across conditions. Higher response rates on the key leading to the 10-s terminal-link schedule occurred in same- and center-cued conditions. However, no such trend occurred in the center-uncued condition. Rather, in this condition, response rates were often higher during the initial link leading to the longer terminal-link schedule ( 20 s). For three of the four birds, initial-link response rates did not reverse when terminal-link conditions were reversed. In the center-uncued condition, only R-222 responded with higher rates during whichever initial link led to the shorter terminal-link schedule. Recall that this bird also slightly preferred whichever key was correlated with the shorter terminal-link schedule in initial and reversed center-uncued conditions. Absolute response rates during the terminal links were also calculated using data from the final nine sessions of each condition. Terminal-link absolute response rates for each condition appear in Table 5. Figure 6 presents relative response rates, expressed in terms of the shorter 10-s terminal link, for each condition. All relative response rate data were averaged over initial determinations and reversals. Relative response rates during the 10-s terminal link were higher in cued conditions (>.50, as shown in Figure 6) than in uncued conditions (approximately.50, as shown in Figure 6). IRis were calculated (in seconds) from the final nine sessions of terminal-link data for each condition. As in Experiment 1, obtained IRis were approximately 10 sand 20 s, corresponding to terminal-link schedules of 10 sand 20 s, respectively.
162 COLTON AND MOORE Cued o Uncued C/) ~ c :.:J 1.0 al c: E 0.5 ~ ~ 0) C c ~ 0.0 o..- cts a: C/) c o a. w 1.0 0:: 0.5 > ~ as c:: 0.0 Same R-222 R-213 R-187 R-1580 Center Houselight Same TERMINAL-LINK ARRANGEMENT Center Figure 6. Experiment 2: Relative terminal-link responses/min expressed in terms of the shorter ( 10-s) terminal-link schedule and averaged over initial determinations and reversals. After subjects had completed all conditions, a final same-cued ( 10 vs. 20) assessment reconfirmed clear preferences for the shorter terminal-link schedule as observed previously. Table 5 includes the standard deviations for the choice proportions and the number of sessions required in each condition of Experiment 2. No systematic differences in either of these measures emerged between conditions. Cases in which preferences failed to stabilize within 30 sessions occurred in uncued conditions only and are indicated by asterisks in Table 5.
EFFECTS OF TERMINAL-LINK STIMULUS 163 General Discussion The present two experiments examined the effects of terminal-link stimulus arrangements on preference in concurrent chains. In Experiment 1, preferences were low or nonexistent in the randomuncued condition. In contrast, preferences were strong and reliable in the same-cued, same-uncued, opposite-cued, opposite-uncued, and random-cued conditions. Strong and reliable preferences were also observed in the houselight condition. In Experiment 2, preferences were low or nonexistent in the center-uncued condition. In contrast, preferences were strong and reliable in the same-cued and center-cued conditions. Strong and reliable preferences were also observed under the houselight condition. These findings were observed when there was little variation in the obtained terminal-link IRis. Primary reinforcement was reliably obtained after about 10 s in the 10-s terminal link, and after about 20 s in the 20-s terminal link. Taken together, these results confirm the earlier reports of Williams and Fantino (1978), Omino (1993), Omino and Ito (1993), and Alsop et al. (1994). In addition, they extend the findings to procedures involving an opposite arrangement, and to those involving differential houselights. The results indicate that terminal-link stimuli were more than just discriminative about the IRI. If a characteristic stimulus, such as a differential keylight or houselight, was correlated with terminal-link IRI, the birds' initial-link responding came under its control. These characteristic terminal-link stimuli may be regarded as conditioned reinforcers, a view consistent with that of Williams and Fantino (1978). Williams and Fantino (1978) reasoned that when terminal-link schedules were associated with different stimuli (cued conditions), differences in both delayed reinforcement and immediate conditioned reinforcement should influence preference in the initial links. However, when terminal-link schedules were associated with the same stimuli (uncued conditions), differences in immediate conditioned reinforcement should be eliminated and initial-link responding should differ only with respect to overall delay of reinforcement. The findings of the present study are in agreement with this reasoning. In all cued conditions, when the nature and/or location of the terminal-link keylight stimuli were consistent, preferences for the shorter terminal-link schedule were reliably high. In uncued conditions, when terminal-link keylight stimuli were nondifferential in nature but were consistent and differential in their location, (i.e., same- and opposite-uncued conditions), preferences for the shorter terminal-link schedule were also reliably high. However, when terminal-link keylight stimuli were both uncued and nondifferential in location (i.e., center- and random-uncued conditions), preferences tended toward indifference. Thus, the present research supports the majority of studies concerning the effect of terminal-link stimuli on preference and shows that such stimuli may serve as conditioned reinforcers, rather than just discriminative stimuli for terminal-link IRis.
164 COLTON AND MOORE A variety of findings in the stimulus control literature also testify to the importance of consistency between stimulus and response. For example, in the area of animal discrimination learning, Bitterman (1966) noted that Yerkes' (1907) original apparatus required the rat merely to run past the discriminative stimulus to reach a goal box containing food. Unfortunately, this "Yerkes box" resulted in very poor discrimination learning. However, a jumping apparatus developed by Lashley (1930) allowed the rat to make contact with the stimuli to be discriminated; the discriminative stimuli were also the locations of the required response. The rat jumped directly at one of two movable stimulus windows in order to access food located on a platform behind the window. This refinement yielded better discrimination learning. Indeed, Bitterman, Tyler, and Elam (1955) showed that rats learn a discrimination (e.g., vertical versus horizontal stripes, large circle versus small circle) with fewer errors in a Lashley jumping stand if the stimuli are located on the stimulus-windows rather than adjacent to the windows. Discrimination learning studies using auditory stimuli also support the influence of stimulus location on stimulus control under certain conditions. For example, Harrison, Iversen, and Pratt (1977) found that stimulus control in monkeys developed better when the auditory stimulus was closer to, rather than separate from, the manipulandum (see also Dobrezeska, Szweijkowska, and Konorski's study using dogs as subjects, cited in Davey, 1986, p. 213). Thus, when a noise burst was presented through the same key on which a response was reinforced, stimulus control was achieved in 1 or 2 sessions. However, 10 sessions were required to achieve stimulus control in two other conditions in which (a) the noise burst was presented through one key and responding on the other key was reinforced and (b) reinforcement was contingent upon responding on a key nearer the sound source. Indeed, the connection between stimulus presentation and response is said to account, in part, for the effect of spatial arrangement of cues, response, and reinforcement on the learning of discriminations (Hearst & Jenkins, 1974). In particular, research on sign-tracking has examined how behavior directed toward (or away from) stimuli as a result of the relation between the stimulus and reinforcement (or the absence of reinforcement) brings an organism into contact with an informative stimulus. Sign-tracking also involves an overt response directed toward that stimulus. In the case of a feature-positive effect, behavior seems to come under the control of a localized stimulus more readily when the appearance or presentation of the stimulus is correlated with reinforcement than when the disappearance or absence of the stimulus is correlated with reinforcement. Like the earlier studies, the present experiments varied the relation between the nature and location of terminal-link stimuli and the location of the preceding initial-link response. Holding the location of visual stimuli constant from initial to terminal links clearly facilitated the development of systematic responding in the present research.
EFFECTS OF TERMINAL-LINK STIMULUS 165 In conclusion, the findings of this study indicate that initial-link responding may be influenced by more than just terminal-link primary reinforcement. Other terminal-link variables (Le., the nature and location of terminal-link stimuli) playa role as well. A conditioned-reinforcing value may be conferred upon terminal-link stimuli by various parameters of the link such that initial-link responding is reinforced to a greater or lesser degree by the stimuli (Fantino, 1977). Thus, the present study joins Williams and Fantino (1978), Alsop, Stewart, and Honig (1994), and amino (1993) in showing that conditioned reinforcement contributes significantly to choice in concurrent chains. References ALSOP, B., STEWART, K. E., & HONIG, K. (1994). Cued and uncued terminal links in concurrent-chains schedules. Journal of the Experimental Analysis of Behavior, 62, 385-397. AUTOR, S. M. (1960). The strength of conditioned reinforcers as a function of frequency and probability of reinforcement. Unpublished doctoral dissertation, Harvard University. AUTOR, S. M. (1969). The strength of conditioned reinforcers as a function of frequency and probability of reinforcement. In D. P. Hendry (Ed.), Conditioned reinforcement (pp. 127-162}. Homewood, IL: Dorsey Press. BAUM, W. M. (1973). The correlation-based law of effect. Journal of the Experimental Analysis of Behavior, 20, 137-153. BITTERMAN, M. E. (1966). Animal learning. In J. B. Sidowski (Ed.), Experimental methods and instrumentation in psychology (pp. 451-484). New York: McGraw-Hili, Inc. BITTERMAN, M. E., TYLER, D. W., & ELAM, C. B. (1955). Simultaneous and successive discrimination under identical stimulating conditions. American Journal of Psychology, 68, 237-248. BROWNSTEIN, A. J., & PLiSKOFF, S. S. (1968). Some effects of relative reinforcement rate and changeover delay in response-independent concurrent schedules of reinforcement. Journal of the Experimental Analysis of Behavior, 11,683-688. CHUNG, S., & HERRNSTEIN, R. J. (1967). Choice and delay of reinforcement. Journal of the Experimental Analysis of Behavior, 10,67-74. DAVEY, G. (Ed.). (1966). Animal learning and conditioning (pp. 212-213). Baltimore: University Park Press. FANTINO, E. (1977). Conditioned reinforcement: Choice and information. In W. K. Honig & J. E. R. Staddon (Eds.), Handbook of operant behavior (pp. 313-331). Englewood Cliffs, NJ: Prentice-Hall. HARRISON, J. M., IVERSEN, S. D., & PRATT, S. R. (1977). Control of responding by location of auditory stimuli: Adjacency of sound and response. Journal of the Experimental Analysis of Behavior, 28, 243-251. HEARST, E., & JENKINS, H. M. (1974). Sign tracking: The stimulus-reinforcer relation and direct action. Austin, TX: Psychonomic Society.
166 COLTON AND MOORE HERRNSTEIN, R. J. (1961). Relative and absolute strength of response as a function of frequency of reinforcement. Journal of the Experimental Analysis of Behavior, 4, 267-272. HERRNSTEIN, R. J. (1964). Secondary reinforcement and the rate of primary reinforcement. Journal of the Experimental Analysis of Behavior, 7, 27-36. LASHLEY, K. S. (1930). The mechanics of vision. Part I. A method for rapid analysis of pattern vision in the rat. Journal of Genetic Psychology, 37, 453-460. OMINO, T. (1993). A quantitative analysis of sensitivity to the conditioned reinforcing value of terminal-link stimuli in a concurrent-chains schedule. Journal of the Experimental Analysis of Behavior, 60, 587-594. OMINO, T., & ITO, M. (1993). Choice and delay of reinforcement: Effects of terminal-link stimuli and response conditions. Journal of the Experimental Analysis of Behavior, 59, 361-371. SCHNEIDER, J. W. (1972). Choice between two-component chained and tandem schedules. Journal of the Experimental Analysis of Behavior, 18, 45-60. SCHUSTER, R. H. (1969). A functional analysis of conditioned reinforcement. In D. P Hendry (Ed.), Conditioned reinforcement (pp. 192-234). Homewood, IL: Dorsey Press. STUBBS, D. A., & PLiSKOFF, S. S. (1969). Concurrent responding with a fixed relative rate of reinforcement. Journal of the Experimental Analysis of Behavior, 12,887-895. WILLIAMS, B. A., & FANTINO, E. (1978). Effects on choice of reinforcement delay and conditioned reinforcement. Journal of the Experimental Analysis of Behavior, 29, 77-86. WILLIAMS, B. A. (1994). Conditioned reinforcement: Experimental and theoretical issues. The Behavior Analyst, 17, 261-285. YERKES, R. W. (1907). The dancing mouse. New York: MacMillan.