PRIMATES,13(3): 276-296, September 1972
277
Tool Use in Captive Hamadryas Baboons BENJAMIN B. BECK
Chicago ZoologicalPark ABSTRACT. A subadult male of a captive harem of hamadryas baboons learned, by instrumental trial-and-error learning, to use a tool to supply food for the harem. The behavior did not spread through the harem by imitation. The subadult male's rank in the harem's dominance hierarchy did not change as a result of his tool use.
INTRODUCTION The use and manufacture of tools were of critical importance in hominid evolution. The study of extant nonhuman primates is considered to have special relevance to the understanding of the behavior of early hominids. Therefore, HALL (1963a) and KORTLANDT and KooIJ (1963) emphasized tool behavior in nonhuman primates in their reviews. Observations by VA~ LAWICK-GOODALL(1968) have revealed that chimpanzees (Pa~ troglodytes) engage in a greater variety of tool behaviors than that known for any wild nonhuman primate. In the field, experimental analysis is difficult although KORTLANDT(1967) has had some success in this regard. Tool use in captive primates has been studied experimentally by BIRCH (1945), KrmOUSTOV (in TOBIAS, 1965), KLf]VER(1933), K6HLER(1927), MENZELet al. (1970), PARKER(1969), SCHILLER (1957) and YERKES(e.g. YERKES& YERKES, 1929). Most of these workers, however, used pongids as subjects and none included detailed consideration of social dynamics involved in the behavior. Accordingly, I will present here results of an experiment designed to examine the acquisition of tool use by monkeys in light of modern principles of behavior analysis and to examine concomitant and consequent social dynamics. Specifically, the study was designed to answer four questions: (1) Will hamadryas baboons, which are reported not to use tools in the wild, do so when environmental conditions make tool use possible and advantageous ? (2) Is tool use initially acquired by instrumental, trialand-error learning or by insight? (3) Does tool use spread by imitation through the user's social group? (4) Does the tool user's position in the dominance hierarchy of his group change as a result of his tool use?
SUBJECTS AND METHODS A harem of hamadryas baboons (Papio hamadryas) housed in the primate house at the Chicago Zoological Park, Brookfield, Illinois, U.S.A. was used as subjects. At the
278
B.B. BECK
time of the research, the harem consisted of six animals, none of whom had been used previously for experimentation: MI: A fully mature silver-mantled male, estimated to be over 40 years old at the time of the experiment. He suffers from muscular degeneration in the hindlimbs, chronic hyperextension of several digits of each manus and extensive dental wear. All appear to be concomitants of his exceptional age. While his locomotor and manipulative abilities are somewhat impaired, he is socially and sexually active, being dominant in the group and having sired five offspring in the last four years (19671971). FI: A mature female, estimated to be between seven and eight years old at the time of the experiment. She has given birth to the aforementioned five infants. One (female) was born about two months after the present experiment was completed; the other four (M2, F2, F3, M3) formed the balance of the harem. M2: A subadult male born 20 April 1967. F2: A juvenile female born 11 August 1968. She had shown no sign of perineal estrous swelling as of the end of the experiment. F3: A juvenile female born 21 August 1969. M3: An infant male born 9 May 1970. He was weaned during the study and his natal coat began to change. This group was chosen because it is established, healthy and reproductively active. It approximates the basic social unit (harem) of P. hamadryas in the wild (KUMMER, 1968). In Ethiopia, harems commonly include one adult male, one adult female, several immature animals (sometimes offspring of the harem adults and sometimes, in the case of juvenile females, following the adult male as an adult female does) and young adult "follower" males. The present group fits this description closely although it is not clear whether F2, in terms of behavioral dynamics, is best considered as offspring or as a following female and whether M2 is best considered as offspring or as a "follower" male. While approximating the basic social unit of hamadryas, my experimental group did not have the opportunity to associate with others in "bands" or "troops" as occurs in the wild and in larger zoo colonies (KUMMER,1957; ZUCI~ERMAN,1932). An additional reason for using this group was that whereas P. hamadryas apparently do not use tools in the wild (KuMMER, 1968), individuals of other species of Papio do have the capacity to do so. BOLWIG(1961, 1964) reported tool use by a captive female Papio anubis. My experiment was conducted between 12 September and 11 December, 1970. Prior to beginning the experiment, I observed the group for 40 hours over 20 days to become familiar with behavior patterns and social interactions. During the experiment, the group was housed in a cage 356 cm wide x 183 cm deep × 239 cm high. A horizontal shelf, 28 cm wide, ran the full length of the rear wall, 114 cm above the floor. A similar shelf ran diagonally from the floor to the horizontal shelf and a rope hung freely from the top of the cage. A short, stout unattached branch and a short section of plastic pipe were present as objects for play and manipulation. Illumination was provided mainly by daylight through a skylight above the cage. The group's basic daily diet consisted of about 2 kg prepared monkey chow (Purina), two oranges cut
Tool Use in Captive Baboons
279
into eighths and two quartered"Gainesburgers" (Purina) plus various tidbits provided by zoo visitors. Water was freely available at all times. I began the experiment by observing the group's social behavior for periods of 60 to 120 minutes between 0800 and 1030 on 21 days between 12 September and 12 October, 1970. Total observation time was 241 t minutes. This time of the morning was chosen because few or no visitors are present and because KUMMERand KURT (1965) found that this is the period of maximal social activity for both wild and zoohoused hamadryas. I observed from behind the visitors' guardrail, about 205 cm from the front of the cage. The animals had already become accustomed to my presence during the preliminary observation period and I did not feed, speak to or otherwise interact with them during observation. Observation was discontinued during the few cases of distraction by zoo visitors or keepers; the beginning or termination of an observation period was never dependent upon the group's activity as this can bias results (ALTMANN& ALTMANN, 1970). The dyadic frequencies of nine behavior patterns were recorded: Threat: Brow raising, protruding or bobbing head, lunging, slapping substrate. M1 and (rarely) M2 also used the "cheek-pumping" threat characteristic of hamadryas males. Fight: Snapping, biting, slapping (recipient), chasing. Displacement (Supplantation) : Gaining possession of or occupying a spatial feature, area or manipulable object from another animal. This frequently occurred in play when one animal pushed another from the shelf or rope. Flatten (Crouch, Press): Flexing hindlimbs and, at high intensity, the forelimbs so that the ventrum touches the substrate. This was frequently accompanied by retracting lips ("grinning"), squealing, screaming or cackling. Present: Lifting or everting tail and presenting posterior to another's face. The legs are frequently flexed slightly and the presenting animal often looks back toward recipient who sometimes touches or mouthes perineum of presenting animal. M2 sometimes presented to M1 in a different fashion: if M2 were standing quadrupedally, he either lifted a hindleg and exposed his genitalia or rose bipedally and exposed his genitalia; if M2 were sitting, he oriented toward M1 and spread his legs, thereby exposing his genitalia. Because the context of this "inguinal presenting" was similar to that in which the usual posterior presenting occurred, both were scored as one type of event. F2, F3 and M3 used the inguinal present a few times. Mount: Standing with hindfeet on substrate and placing hands on the dorsum of recipient. This was frequently accompanied by pelvic thrusting. Mounting with intromission (two cases) was not scored in this category. Allogroom: Manual parting and/or picking through the pelage of another. Frequently accompanied by removal of material from pelage with hands or mouth but mere licking or mouthing of pelage was not scored. Play: Elements of many of the above patterns occurred during play but were scored as play only if accompanied by distinctive "chuttering" vocalization which is characteristic of play in this species. Embrace: Two animals sitting facing one another with each surrounding the trunk of the other with his arms. M3 clinging to F1 was not scored in this category.
280
B.B. BECK
For the most part, these categories and criteria were derived from KUMMER'Swork. The initiator and receiver of all of these behaviors were noted with the exception of play and embrace where such a distinction could not be made confidently. In this paper I shall deal only with the dyadic frequencies of threat, fight, displacement, flatten, present, and groom. I have included descriptions of mount, play, and embrace because their occurrence affected the frequencies of the other events due to the scoring method used. An event was scored at its onset. If 30 seconds or more elapsed between the termination of an event and the onset of the same event between the same animals in the same roles, the event was scored again. For example, ifFl groomed M1, stopped for 35 seconds and groomed M1 again, "F1 groom M I " would be scored twice but if she stopped for only 25 seconds, it would be scored only once. However, if the initiator of an event initiated another scored event (with the same or another recipient) or the same event with another recipient and then resumed the first, the first and third events were scored separately regardless of the intervening interval. For example, if F2 groomed 3/11, stopped and played for 10 seconds with M3 (or groomed M3) and then groomed M1 again, "F2 groom M I " would be scored twice. Thus while mount, play and embrace were judged, a posteriorL not to be meaningful in terms of this report, their occurrence did affect the absolute frequencies of the events included herein. All event frequencies were recorded using paper, pencil, and stopwatch. The events were later recorded on six by six matrices and, additionally, a cumulative record was maintained. Events which could not be categorized unambiguously were not scored. The period of preliminary observation had convinced me that 40 hours of observation were sufficient to gain representative samples of the scored behaviors for this group. Accordingly, after 2411 minutes of the morning social observation, I began the next phase of the experiment on 13 October. This consisted of discontinuing the normal feeding routine in which keepers provided the food noted above at 1430. Instead, at about the same time, I set up a black wooden platform, 183 cm wide × 152 cm deep, extending out from and perpendicular to the front of the cage at the level of the cage floor. An L-shaped "rake" tool fashioned of 1/4 inch (.635 cm) threaded steel rod was then attached by a chain to a bar of the cage front. The chain was 81 cm long, the shaft of the tool was 84 cm long and the hook of the tool was 11.5 cm long. The tool was positioned on the board so that the shaft formed about a 30 ° angle in regard to the cage front and the hook pointed back toward the cage. The farthest extent of reach through the bars of any animal in the group was 45.7 cm and the tool was positioned within this limit. It was accessible to all animals either by reaching it directly or pulling it in by the chain. An aluminum pan, 22.9 × 12.7 × 6.4 cm was filled with food and placed on the board with the 22.9 cm side parallel to the cage front, 81.3 cm away. The pan was filled with items from the basic daily diet (monkey chow, "Gainesburgers" and oranges) and with such novel and highly preferred items as lettuce, spinach, carrots, apples, bananas, and sunflower seeds. The group was then allowed to secure the pan with the tool for a total of 602 minutes over seven afternoons from 13 October to 23 October. As the food was not secured during this time, the pan was moved in so that its edge was only 50.8 cm from the cage front. After a further opportunity of 74 minutes and 35 seconds, one of the group secured the food with the
Tool Use in Captive Baboons
281
tool on 28 October. Before this solution, the animals were fed only the basic diet after the afternoon session. When solution did occur, the animals were allowed to eat all of the food in the pan and then another trial was begun. Once the behavior was mastered, the animals provided themselves with all of their food through use of the tool on four trials each day with the exception of a few days on which I could not be present to present the problem. A small "breakfast" portion of monkey chow was given at 1000 and an undetermined but small amount of food was provided by visitors. Signs were posted informing visitors that the group was being used in an experiment and requesting them not to feed or distract the animals. For the most part, visitors complied with these requests. Once solution occurred, I began the "post-tool mastery" phase of the experiment. On 25 afternoons from 29 October to 2 December, food was made available to the group contingent upon their use of the tool. Additionally, I again conducted a total of 2411 minutes of social observation on 25 mornings during this period. The procedure was identical to that used in the first phase with the exception that no cumulative record of dyadic event frequencies was maintained in order to avoid biasing the results. Each tool trial was timed, to the nearest second, from the time of presentation to the time the pan was secured. Qualitative aspects of manipulation and use of the tool were noted. Still and cine photographs were taken periodically throughout the experiment. In summary, this design allowed me to ascertain the dynamics of the group's learning to use a tool to get the major portion of their food (including highly preferred but otherwise unavailable food) and to compare aspects of the group's social behavior before and after their acquisition of the tool behavior. RESULTS USE OF THE TOOL
As was noted above, the pan was initially placed 81.3 cm from the cage and, at this distance, no animal succeeded in getting the food with the tool in 602 minutes of exposure. The pan was then placed 50.8 cm from the cage and, after a total of 74 minutes and 35 seconds, M 2 succeeded in getting the food by using the tool on 28 October. Since 60 minutes of this time had elapsed in one unsuccessful trial on a previous day, the actual elapsed time in the first successful trial was 14 minutes and 35 seconds. For the first five minutes of this trial, only M 2 manipulated the tool. This consisted of underhanded pushing, sliding and throwing the tool toward the pan with one or both hands, each time retrieving the tool by pulling it in directly or by pulling on the chain. Additionally, he several times stood the tool on end with the hook end up and dropped or pushed it out and down. He had exhibited all of these behaviors before but there was a noticeable increase in the persistence with which they were oriented toward the pan. Five minutes into the trial he hit the pan, actually moving it away from the cage. He then looked intently at the pan, pushed the tool out but pulled it back without contacting the pan again. He resumed the underhanded pushing,
282
B.B. BECK
sliding and throwing actions and shortly after nine minutes, the tool fell behind the pan; in retrieving the tool he moved the pan in toward the cage. He continued the same actions without hitting the pan again but at 14 minutes and 35 seconds he reached out with his hand and secured the pan. The movement at nine minutes actually had gotten the pan within reach although he did not successfully reach it until more than five minutes later. As he pulled the pan in, he picked out several bits of food and began to eat them. F1, with M3 riding on her back, ran suddenly and quickly to the food and filled her mouth and cheek pouches literally to overflowing. With the exception of the few pieces eaten by M2 and one piece by 343, F1 ate all of the food. As a result of studies on problem solving of a different type, I have concluded previously that "insight is uniquely characterized by the sudden appearance of a complete adaptive ("correct") complex response sequence following a period of nonproblem-directed responding" (BECK, 1967, p. 106). The behavior used by M2 to solve the tool problem did not appear suddenly but rather was a repetition of behaviors which he had shown frequently in the test situation. Nor was it complete: the pan was moved only a short distance so that it was barely within reach and, even then, more than five minutes elapsed before he actually took it. Finally, there was no period of non-problem-directed activity just prior to solution. In fact, M2 had manipulated the tool and worked at the problem for all but a few seconds of the first successful trial. The solution to this problem, then, was the result of instrumental, trial-anderror learning; there was no evidence for insightful solution. The opportunity to manipulate and investigate novel objects is reinforcing and, especially in primates, high rates of manipulation can be maintained with no other reinforcement (BUTLER,1965; HARLOW,1950; HARLOWet al., 1950). 342 manipulated the tool persistently throughout the experiment. On the fourth through the ninth day of presentation of the tool problem (before solution had occurred), I took a five second time sample every three minutes noting simply which animal(s) was touching (in physical contact with) the tool. 342 was touching the tool in 56.3 ~ of these
MI
FI
M2
F2
F3
M3
NONE
GROUP MEMBER
Fig. 1. Percent of time samples during which group members were touching the tool.
Tool Use in Captive Baboons
283
8oo1 70O
300
1
I
.'2
b
3
I
4
[
S
I
6
BLOCKS OF
/
7
I
8
I
9
J
10
I
11
t
12
[
13
EIGHT TRIALS
Fig. 2. M2's average time to solution in consecutive blocks of eight tool trials. samples, more than two and one-half times as frequently as any other group member (see Fig. 1). In the course of this manipulation, one animal was by chance bound to move the pan within reach and the most frequent manipulator had the highest probability of doing so. Once the food is secured, it will act as a reinforcer which would support higher rates of manipulation, lead to alterations in the topography of the manipulative behavior so that it is more accurately directed toward securing the pan and, consequently, cause decreased time to solution in subsequent trials until an asymptote is reached. Figure 2 plots average time to solution for consecutive blocks of eight trials each. The first block represents trials from three different days but, as solution time became shorter, it was possible to present four trials per day and the second through the thirteenth blocks each represents performance from two consecutive days. Trials occasionally had to be discontinued because the pan was upset or pushed off the board. The durations of such trials are not included in Figure 2. Likewise, all of the unsuccessful pre-solution trials (totalling 662 minutes) are omitted with the duration of the first successful trial being computed as 14 minutes and 35 seconds. Some of the variation in solution time is due to lack of precise control of food deprivation (the times of the afternoon tool sessions differed slightly and there was some feeding by visitors), to social interactions in the group during presentation of the problem and also to transient distractions resulting from conducting the tool sessions in public space. Nonetheless, average solution time decreased sharply in the third block and reached an asymptote of about 40 seconds by the eighth block. Variability also decreased at asymptote with solution time ranging from 7 to 143 seconds in the eighth through the thirteenth block. M 2 ' s solution behavior at asymptote consisted of sitting in front of the display and "throwing" the tool: he would hold the shaft in both hands and markedly flex his arms at the elbow so that his forearms were perpendicular to the board and his hands near his shoulders. Then he would extend his arms rapidly and open his hands so that the tool was propelled outward (see Fig. 3). If the entire tool landed behind the pan, the pan would be moved toward the cage when the tool was retrieved with the chain (see Figs. 4, 5, & 6). If only the hook landed behind the pan or the shaft landed on the pan, 11/12 frequently pulled in the tool slowly and deliberately which usually resulted
284
B.B. BECK
in the pan's being brought within reach. If the tool landed to one side of the pan, M 2 often swept the pan into reach by moving the tool shaft sidewise in an arc. Thus the initial component of asymptotic solution behavior was picking up and "throwing" the tool until all or a part of it landed behind or to one side of the pan. The second component, that by which the tray was actually retrieved, showed three general variants, each most adaptive to a different variation in the tool-pan configuration which resulted after the tool was thrown. IMITATION
On the third successful trial (28 October), M 2 got the pan very nearly within reach at 25 minutes and 15 seconds with the throwing-retrieving sequence. F1 rushed over quickly and M 2 moved away leaving the tool lying between the pan and the cage. F1 reached for the pan but could not get it, picked up the tool and flipped it to one side of the pan so that the hook was behind it. She pulled in the shaft a bit moving the tray just into her reach and secured the food. This seems to have been a clear case of imitation but in the total of 104 successful trials, this was the only time that any member of the group other than M 2 secured the food with the tool. On one day after the "post-tool mastery" phase, I twice separated M 2 from the group into an adjacent cage. The first separation lasted 60 minutes and then M 2 was allowed to return and use the tool to get the food. Several hours later, I again allowed M 2 to use the tool to get the food and then separated him for another 60 minutes. All group members manipulated the tool duringM2's separations but none got the food with it. It should be noted that the group seemed agitated during the separation, emitting many "alarm barks" and trying to open the shift door between the two cages. Nevertheless, with the one exception noted above, no group member other than M 2 used the tool successfully in 10a trials over 26 days or in the two hours of M2's separation. Tool use did not spread through the group by imitation. Although no imitation occurred, M2's continued use of the tool caused and was dependent upon an alteration in Fl's behavior. As was noted above, M 2 ate only a few bits of the food he secured on the first successful trial. Because F1 was dominant over M 2 (vide infra), he was displaced from the food at her approach and she ate most of it. For the first 24 minutes of the second successful trial, F1 sat directly in front of the display thus displacing M 2 and preventing him from using the tool. She then moved away and M 2 began to manipulate the tool. He secured the food at 44 minutes and 35 seconds at which time F1 again rushed over and ate all of the food. This pattern was firmly established by the second day of successful trials. When F1 sat near the display, M 2 would not work. Her dominance precluded his occupying the area from which the tool could be used successfully. When she left, M 2 immediately began to work. If she returned before he secured the food, he stopped and moved away. This would be repeated until he got the pan at which point she would rush over and eat most of the food (see Figs. 7, 8, 9, & 10). Except for M1 who usually ate beside F1, the other group members could not get more than an occasional piece of food until the ttiird or fc,urth trial of each daily session. At asymptote (Blocks 8-13), FI sat at the far end of the cage watching M 2 attentively while he worked. M 2 glanced toward her intermittently. Both engaged frequently in "nose-wiping" which seems to
Tool Use in Captive Baboons
Fig. 3. M 2 throws the tool toward the pan. F2 sits to the (reader's) left of M 2 and F3 stands to the right. F1 is barely visible at left of photograph watching M 2 working.
Fig. 4. The tool lands behind the pan.
Fig. 5. M2 pulls on the chain and begins to move the pan toward the cage.
Figs. 6 & 7. As M 2 moves the pan toward the cage, F1 begins to move toward the food.
285
286
B.B. BEcK
Fig. 7. (See Fig. 6)
Fig. 8. F1 gets the pan which is now within reach. M1 is Visible at rear of cage.
Fig. 9. F1 eats. 341 moves toward food. M3 is visible behind F1.
Fig. 10. F1 and M1 eat. M 2 leaves without having eaten.
Tool Use in Captive Baboons
287
indicate excitement or agitation (KUMMER, 1957). When M 2 got the food within reach, she rushed over and ate. Thus not only had M 2 learned to use the tool but F 1 had learned not to approach the display until M 2 had moved the pan within reach. F l ' s behavior represents a simple operant discrimination in which she was not reinforced for staying near the display (because she displaced M 2 thus preventing him from getting the food) and was reinforced for staying at the far end of the cage (because only then would M 2 work and get the food). As a result, she was conditioned to stay at the far end of the cage, thus allowing relatively quick and efficient procurement of the food. CHANGE IN DOMINANCE Based on the 2411 minutes of morning social observation before the tool was introduced, I was able to estimate the dominance hierarchy of the harem. Table 1 is a matrix of pooled dyadic frequencies for threat, fight, and displacement for this "pretool mastery" phase. These events were combined because each is thought to reflect the dominance of the initiator over the receiver and the ranks derived from each measure in this experiment were very similar. The matrix format was arranged so as to maximize the number of interactions above the diagonal and thus be consistent with dominance. Of the total of 2185 such interactions, 2024 (92.6 ~ ) fell above the diagonal and there were no cases where an animal ranked as less dominant initiated more interactions than he received with an animal who is more dominant. That is, the hierarchy was strong and linear. I also used the "submissive" patterns of present and flatten to estimate the dominance hierarchy. Table 2 presents this matrix. In this case, the matrix format was chosen to maximize the number of interactions below the diagonal and thus be consistent with dominance. Of the total of 305 such interactions, 286 (93.8 ~o) fell below the line and the hierarchy was again linear.
Table 1. Pooled dominance behavior (threat, fight, displacement) ;"pre-tool mastery" phase. Initiator Receiver M1 F1 M2 F2 F3 M3
Z
341
F1
M2
F2
F3
M3
-6 100 8 0
11 -1 0 0
800 16 -6 3
330 30 167 -29
35 15 195 125 --
91 21 145 33 10
0
114
0
12
1
826
2
558
5
375
--
300
Z
1267 88 608 172 42 8
2185
Table 2. Pooled submissive behavior (flatten, present); "pre-tool mastery" phase. Initiator Receiver M1 F1 M2 172 F3 M3
Z
M1 --
66 82 15 7 3 173
F1 0
-9 15 14 4 42
M2
F2
F3
M3
13 2 -13 41 2 71
0 3 0 -14 1 18
0 0 0 1 -0 1
0 0 0 0 0 -0
Z
13 71 91 44 76 10 305
288
B.B. BECK
The dominance hierarchy determined for this group using either dominant or submissive behaviors yields identical rankings. M 2 ranks third. Like mounting, the direction of dyadic grooming interactions correlates poorly with agonistic measures of dominance (BERNSTnN, 1970). However, since allogrooming can be considered as a measure of cohesiveness between members of a primate social group, I thought it might reflect social consequences of M 2 ' s tool use. Table 3 presents the matrix for allogrooming in the "pre-tool mastery" phase. The format chosen is derived from the dominance rankings established by dominant and submissive behaviors. Of a total of 629 grooming interactions, 361 (57.4 ~ ) fell below the diagonal (consistent with dominance if subordinates groom dominants) which confirms the poor correlation between grooming and other measures of dominance. It should be noted that M 2 was a very infrequent groomer and was groomed less than any other group member. After the tool task had been solved, I again conducted 2411 minutes of morning social observation. During these days of observation M 2 continued to provide food for the group by his use of the tool in the afternoon. It should be stressed that the data on social behavior in the "post-tool mastery" phase were gathered in the morning and not during the afternoon tool sessions. Table 4 is the matrix of pooled dyadic frequencies for threat, fight and displacement in this phase and can be compared directly with Table 1 from the "pre-tool mastery" phase. The overall frequency of such interactions increased to 2403 with most (2260; 94 700)again falling above the diagonal. There was no change in M 2 ' s rank. The only change in the hierarchy as determined by these measures was that M 3 ranked fifth and F 3 sixth. M 3 was weaned, growing rapidly and widening his sphere of social interaction during this period. His dominance over F 3 results in his more frequently displacing her in rough-and-tumble play.
Table 3. Allogrooming; "pre-tool mastery" phase.
Initiator
Receiver M1
M1 F1 M2 F2 F3 M3
X
-113 11 51 54 4 233
F1
43 -1 47 23 5 119
M2
3 17 -15 16 1 52
1:2
F3
M3
X
0 32 6 -15 2 55
2 27 6 23 -3 61
0 52 1 43 13 -109
48 241 25 179 121 15 629
Table 4. Pooled dominance behavior (threat, fight, displacement); "post-tool mastery" phase. Initiator Receiver M1 M1 FI M2 1:2 M3 F3
2
-8 73 1 0 0 82
F1
M2
F2
M3
F3
X
30 -2 1 0 0 33
789 31 -5 2 2 829
192 36 242 -12 23 505
102 15 252 51 -14 434
25 17 280 167 31 -520
1138 107 849 225 45 39 2403
289
Tool Use in Captive Baboons Table 5. Pooled submissive behavior (flatten, present); "post-tool mastery" phase. Initiator Receiver M1 F1 M2 F2 F3 M3
2
M1 --
F1 0
55 35 8 I 5 104
-5 17 10 7 39
M2
16 1 -20 26 13 76
F2
0 4 2 ~ 7 2 15
F3
M3
0 0 1 0 -1 2
1 0 0 0 0 -1
17 60 43 45 44 28 237
Table 6. Allogrooming; "post-tool mastery" phase. Initiator Receiver M1 F1 M2 F2 M3 F3
X
M1
F1
M2
F2
M3
F3
-I08 42 99 l0 53 312
128 -1 41 5 15 190
7 16 -29 3 30 85
0 46 3 -14 17 80
15 47 3 40 -10 115
1 32 13 17 3 -66
151 249 62 226 35 125 848
Table 5 presents the matrix for submissive behaviors in the "post-tool mastery" phase and can be compared to Table 2. The overall frequency of such interactions decreased to 237 but a sizeable portion (212; 89.5 ~ ) again fell below the diagonal. Again, M 2 ranked third. There was no change in M 3 ' s and F3's rank according to this measure but the paucity of submissive interactions between them makes equivocal their ranking on this basis. Table 6 presents the allogrooming matrix from the "post-tool mastery" phase. While the number of allogrooming events increased substantially to 848, only slightly more than half (477; 56.3 ~o) again fell below the diagonal (if M 3 is considered dominant to F3). M 2 was the recipient of grooming more often than he was before his tool use and he also initiated more grooming events. However, the same trend was shown by every group member. M 2 groomed M 1 almost four times as frequently in the "post-tool mastery" phase and M 1 groomed F1 almost three times as often. These increases were disproportionately large. Several events occurred in the morning social observation shortly after M 2 began to use the tool which deserve mention but, in retrospect, appear to have been coincidental. On the morning of 29 October (the day after the first successful tool trials) M 1 was lethargic, moved laboriously and showed a marked tremor in the right rear leg and foot. N o cause for these symptoms could be determined and they disappeared without treatment by 30 October. On the morning of 31 October, M 2 mounted F2. There was pelvic thrusting but his feet were on the floor of the cage. It appeared that intromission had occurred and when M 2 dismounted, he began to touch and lick his penis, eating the ejaculate which was clearly visible. This was the first time M 2 was ever observed to ejaculate and he did so again on 3 November, this time mounting F3. He has not been observed to ejaculate since. The relationship of the onset of ejaculation to his use of the tool is unclear but, in the absence of other social consequences of his tool use, I suspect that it was coincidental.
290
B.B. BECK
DISCUSSION USE OF THE TOOL
M2 became a facile and effective tool-user. He acquired this behavior through instrumental trial-and-error learning; there was no evidence of insight. It should be noted that I did not "shape" the behavior by reinforcing successive approximations of the response, an operant conditioning technique often used to train animals to emit complex responses. Eventual acquisition of the food resulted from the increasing probability of success with continued self-reinforcing manipulation of a novel object, the tool. Once the "correct" response was emitted, the food reinforcement acted to increase the probability of subsequent "correct" responses and therefore to decrease the time required for solution to an asymptotic level in subsequent trials. Because use of the tool in this situation is a fortuitous result of exploratory and manipulative responses, the animal which most frequently manipulated the tool was most likely to use the tool successfully. This was the case with M2. In general, exploratory and manipulative responses, like play, decrease in frequency upon sexual maturity (MENzEL, 1966). F1 and All rarely touched the tool. One would expect, then, that most innovative behaviors would be initiated by young animals. Such has been the case with most innovations among Japanese macaque (Macaca fuscata) troops (KAWAMURA,1959). To the degree that characteristics of the environment and dominance factors differentially affect the opportunity for pre-solution manipulation by young animals, they also will influence which becomes an innovator. In a situation like the present one, where the young animals are ordered in a dominance hierarchy and there is only one manipulandum, the most dominant immature animal is most likely to be an innovator because he gets most access to the manipulandum. Because M2 was subordinate to F1 (who ate most of the food), he was not reinforced on every trial. However, intermittent reinforcement on a ratio schedule (one reinforcement every one + n correct responses where n can be fixed or variable) produces higher rates of responding and greater resistance to extinction than does reinforcement of every correct response (e.g. REYNOLDS, 1968). FI's domination of the food may therefore have actually supported M2's tool use. It should be emphasized that in setting up the tool problem, the tool was positioned between the cage and the pan; no portion of the tool was behind the pan which would allow the animal to get the pan simply by pulling the tool in. Rather, the animal was responsible for mechanically connecting the tool and pan in a way necessary to secure the pan. Caged primates (and I suspect wild primates also) immediately pull in toward themselves any attractive or non-fear-producing object. Doing so allows consumption or close examination of the object and doing otherwise often results in its loss to a captor or conspecific. This tendency can be considered a response set and the "pull in" set dominated the group's pre-solution manipulation of the tool. This, of course, would not allow acquisition of the pan and the animals had to reverse this set to get the pan, i.e., they had to push the tool away from themselves so as to get it behind the pan. M2 was the first to exhibit such behavior consistently and he did so on the day preceding solution. The "throwing" component of his asymptotic solution behavior
Tool Use in Captive Baboons
291
was essentially pushing the tool away from himself. Successful use of the tool, then, involved a reversal of response set, from "pull-in" to "push-out." RUMBAUGH(1969, 1970), using a two-choice discrimination which does not involve tools, has found that some primate taxa are distinguishable on the basis of the ability to reverse a learned response set. Differences in the ability to use tools of the sort employed here may correlate with RUMBAVGH'S"transfer indices" which express the ability to reverse response set. Reversal learning may in fact be involved in many other types of complex learning tasks, e.g. detour problems, in that many involve a reversal or change in response set. KUMMER (1968) did not observe any tool use by wild hamadryas baboons in his extensive study and reports that they rarely even manipulate inedible objects. The present experiment, however, demonstrates that at least one hamadryas has the genetically-based capacity to use tools. There is no reason to suppose that M 2 is unique in this regard; the capacity is undoubtedly widely distributed in the species. Hamadryas baboons have the sensory, motor and cognitive attributes sufficient to use tools but do not do so until their environment contains those structural elements and reinforcement contingencies which make such behavior possible and adaptive. Although they did not address themselves specifically to hamadryas, such a result was perceptively anticipated by HALL (1963a) and MENZEL(1966). KOgTLANOTand KooIJ (1963) and KORTLANDT(1966, 1967) noted differences between forest- and savannahdwelling chimpanzees in the intensity and effectiveness of agonistic tool use. The observed differences in these two populations probably do not represent differences in their genetically-based capacity to use tools but rather in the opportunities for adaptive tool use offered by their different environments. Such a consideration has led to a revision of the "dehumanization" hypothesis (KORTLANDT & VAN ZON, 1969). CROOK (1970) has reviewed the effects of variation in environmental parameters on many dimensions of primate social life. As nonhuman primates continue to be studied in a variety of ecological conditions, I suspect that more cases of tool use in a greater variety of primate taxa will also be catalogued. These general considerations and the results of my experiment lend strength to the hypothesis that tool use by hominid ancestors first occurred upon a change in their ecology which presented them for the first time with the environmental conditions conducive and/or necessary for capitalizing on a hitherto unused genetically-based capacity to use tools. Such a preadaptation could have resulted from selective pressures favoring certain locomotor adaptions and learning skills such as those involved in elaborate social systems. Tool behavior was undoubtedly a critically important factor in hominid evolution and radiation (e.g. BARTHOLOMEW& BmDSELL,1953; DART, 1960; WAS~URN, 1960). LANCASTEg(1968), however, argues that tool use was present even before the hominid grade was attained. The present experiment supports this conclusion since tool use is possible with sensory, motor and cognitive capacities no more specialized than that of extant baboons. IMITATION Excluding Fl's use of the tool on the third successful trial, no group member other
292
B.B. BECK
than M2 used the tool so as to get the food. In view of the rapid spread of innovative behaviors through troops of Japanese macaques (KAWAMURA,1959), this result was unexpected. It may be that these animals simply do not have the ability to learn tool use by imitation. While primates seem to be capable of accelerating their mastery of a variety of learning tasks by observing a conspecific who has already mastered the task (DAR'BY & RIOPELLE, 1959; HALL, 1963b; HALL & GOSWELL, 1964; STEPHENSON, 1967), there is no documented, unequivocal case of imitation or observational learning of tool use in nonhuman primates. While young wild chimpanzees gain some information about using stems to fish termites from subterranean nests by watching skilled adults, each individual actually acquires the behavior through instrumental, trial-and-error learning (VAN LAWlCK-GOODALL,1968). GOODALL(1963) noted that baboons (P. anubis) also eat termites eagerly but wait until the insects emerge from the nests and catch them in flight. The baboons have not learned to use stems to get the termites despite ample opportunity to observe chimpanzees doing so. A journalistic report (Anon., 1970) describes MENZEL'S observations of captive chimpanzees using poles as ladders. One animal began the behavior spontaneously (through trialand-error learning) and other group members are said to have learned to prop the poles by imitation. No mention is made, however, of the criteria for excluding the possibility that each animal learned the behavior individually. At the time of this writing, MENZEL'Sown report is not yet available. It may be that nonhuman primates, at least monkeys, are not able to learn to use tools by imitation, although learning such behavior may be accelerated by observing a skilled tool user. Alternatively, it may be that other group members can and did learn the behavior by imitation but simply did not perform it. Once M2 had mastered the behavior, he dominated the tool and secured the food quickly. Because F2, 173, and 313 were subordinate to him, they rarely got access to the tool after mastery and thus may have had no opportunity to make imitative responses. Of course, when M2 was separated, the younger animals should have been free to imitate but the relative brevity of the separation and the consequent disturbance may have interfered. If M2's separation had been extended or more than one manipulandum had been provided in his presence, the younger animals may have shown imitation. Ml's locomotor impairment may have precluded his imitating 312. F1 did solve the problem once but then learned a behavior which allowed her to get the food with a minimum of effort. In terms of conservation of energy, it was to her advantage to allow 312 to get the pan, from which she got whatever type and amount of food she wanted. As noted above, this behavior was also the result of instrumental learning; there is no implication that F1 purposively exploited M2. MEIERet al. (1969) found that young animals in a group of rhesus macaques (Macaca rnulatta) were the first to learn an operant response (not involving a tool) to get food and older, dominant animals often simply took the food from the young ones rather than learning or emitting the response themselves. Whatever the cause for the lack of imitation in the present experiment, it seems clear that if food had been limited and the experiment terminal, some animals would have died if imitation or cooperative sharing of the food had not occurred. I am currently conducting another experiment where the group members will be
Tool Use in Captive Baboons
293
able to observe the tool user on a greater number of trials (in the unlikely event that 104 were not enough to support imitation) and where the tool user will be separated for a longer period. CHANGEIN DOMINANCE There was no change in M2's rank in the group's dominance hierarchy despite his using the tool to provide the major portion of the food supply for over a month. This may have been due to the limited upward mobility of young hamadryas males; they rarely compete for dominance with the lead male of the harem with which they are associated (KUMMER,1968). In a setting with more animals, M2 may have been at an advantage, as a result of his tool use, in competing with other young males for harem females. However, the question of the effect of tool use on dominance would be more appropriately examined in those cercopithecines, e.g. Papio cynocephalus, where the basic social structure is a troop with a number of adult males ordered in a dominance hierarchy. While there were no dominance rank changes as a result of tool use, there were differences in social behavior between the "pre- and post-tool mastery" phases. There was a disproportionately large increase in Ml's grooming F1 and in M2"s grooming M1 in the latter phase. KUMMER(1968) hypothesizes that "it often appears as if grooming were used to strengthen a social bond when it is in danger of breaking apart" (p. 45), Perhaps M2's use of the tool was placing a strain on the dyadic bonds between M1 and F1 and between M1 and M2. The overall increase in grooming may even indicate a weakening in the cohesiveness of the harem. The overall increase in dominance behaviors in the "post-tool mastery" phase can be attributed largely to M3's weaning. He was more accessible to M2, F2, and F3 for play. Consequently, M2 displaced M3 more frequently in play and M2 displaced F2 and F3 more frequently as they played with 3"13.However, part of the overall increase in dominance behavior is due to Ml's initiating nearly three times as many dominance events (mainly threats) toward F1 in the second phase, possibly indicating a weakening of this bond. The overall decrease in submissive behaviors in the "post-tool mastery" phase can be attributed almost entirely to fewer submissive events initiated by both M2 and F1 toward M1 which again suggests a weakening of these dyadic relationships. While not conclusive, these data suggest that M2's tool use was straining the social bonds between him and M1 and between M1 and F1. However, there was no change in dominance rank; the advantages accruing to the tool user were limited to the specific resource gained from the use of the tool and did not include increased status and the consequent increased access to other incentives. SUMMARY 1. A captive subadult male hamadryas baboon learned, without training, to use a tool to secure food. 2. The behavior was acquired through instrumental, trial-and-error learning with no evidence of insight.
294
B.B. BECK
3. Since wild hamadryas apparently do not use tools, it is critical to consider those environmental factors that make tool use possible and adaptive as well as the genetically-based capacity for tool use in discussions of tool behavior in both extant and extinct primates. 4. Tool use did not spread through the user's social group. Baboons may not have the ability to learn to use tools by observation or imitation. 5. The tool user's rank in the harem's dominance hierarchy did not change but there were indications that some dyadic bonds were strained as a result of his tool use.
Acknowledgements. J. ROWELL,W. HAHN, J. O'CONNOR, R. McDONALD, and J. GRANT, primate keepers, offered invaluable assistance and patience throughout the experiment. Drs. R. HORWICH, G. RnaB, and R. TUTTLE made helpful suggestions and critically read the manuscript. L. LAFRANCEphotographed the work, D. FISERAtyped the manuscript and L. BECK prepared the figures. I am sincerely grateful to all.
REFERENCES ALTMANN,S. A. & J. ALTMANN,1970. Baboon Ecology. University of Chicago Press, Chicago. Anon., 1970. MENZELreporting on spontaneous use of poles as ladders. Delta Primate Rep. BARTHOLOMEW, G. A. ~¢ J. B. BIRDSELL, 1953. Ecology and the protohominids. Amer. Anthrop., 55: 481-498. BECK, B. B., 1967. A study of problem solving by gibbons. Behav., 28 : 95-109. BERNSTEIN, I. S., 1970. Primate status hierarchies. In: Primate Behavior (Vol. 1), L. A. ROSENBLUM(ed.), Academic Press, New York, pp. 71-109. BIRCH, H. G., 1945. The relation of previous experience to insightful problem solving. J. comp. Psychol., 38: 367-383. BOLWIG, N., 1961. An intelligent tool-using baboon. S. Afr. J. Sci., 57: 147-152. - - , 1964. Observations on the mental and manipulative abilities of a captive baboon (Papio doguera). Behav., 22: 24-40. BUTLER, R. A., 1965. Investigative behavior. In: Behavior of Nonhuman Primates (Vol. II), A, M. SCHRIER,H. F. HARLOW, ~%F. STOLLNITZ(eds.), Academic Press, New York, pp. 463-493. CROOK,J. H., 1970. The socio-ecology of primates. In: SocialBehavior in Birds andMarnmals, J. H. CROOK(ed.), Academic Press, New York, pp. 103-166. DARBY, C. L. 8¢ A. J. RIOPELLE, 1959. Observational learning in the rhesus monkey. J. comp. physiol. Psychol., 52: 94-98. DART, R. A., 1960. The bone tool-manufacturing ability of Australopithecus prometheus. Amer. Anthrop., 62: 134-143. GOODALL, J., 1963. My life among wild chimpanzees. Nat. Geog., 124: 272-308. HALL, K. R. L., 1963a. Tool-using performances as indicators of behavioral adaptability. Curt. Anthrop., 4: 479-494. - - , 1963b. Observational learning in monkeys and apes. Brit. J. Psyehol., 54: 201-226. t~ M. J. GOSWELL, 1964. Aspects of social learning in captive patas monkeys. Primates, 5: 59-70. HARLOW, H. F., 1950. Learning and satiation of response in intrinsically motivated complex puzzle performance by monkeys. J. comp. physiol. Psychol., 43 : 289-294. HARLOW, H. F., M. K. HARLOW,~¢ D. R. MEYER, 1950. Learning motivated by a manipulation drive. J. exp. Psychol., 40: 228-234. KAWAMURA, S., 1959. The process of sub-culture propagation among Japanese macaques. Primates, 2: 43-60. -
-
Tool Use in Captive Baboons
295
KLUVER,H., 1933. Behavior Mechanisms in Monkeys. University of Chicago Press, Chicago. K~3HLER,W., 1927. The Mentality of Apes (2nd ed.). Routledge & Kegan Paul, London. KORTLANDT,A., 1966. On tool-use among primates. Curr. Anthrop., 7: 215-216. - - . , 1967. Experimentation with chimpanzees in the wild. In: Neue Ergebnisse der Primatologie, D. STARCK, R. SCHNEIDER,& H. J. KtJHN (eds.), Gustav Fischer Verlag, Stuttgart, pp. 208-224. - & M. KooIJ, 1963. Protohominid behaviour in primates. Syrup. zooI. Soe. Lond., 10: 61-88. & J. C. J. VAN ZON, 1969. The present state of research on the dehumanization hypothesis of African ape evolution. In: Proceedings of the Second International Congress ofPrimatology (Vol. 3), H. O. HOrER (ed.), S. Karger, Basel, pp. 10-13. KUMMER, H., 1957. Soziales Verhalten einer Mantelpavianengruppe. Beiheft 33 Schweiz. Z. Psychol., Verlag Hans Huber, Bern/Stuttgart. - - , 1968. Social Organization of Hamadryas Baboons. University of Chicago Press, Chicago. & F. KURT, 1965. A comparison of social behavior in captive and wild hamadryas baboons. In: The Baboon in Medical Research, H. VAGTBORG(ed.), University of Texas Press, Austin, pp. 65-80. LANCASTER,J. B., 1968. On the evolution of tool-using behavior. Amer. Anthrop., 70: 56-66. LAWICK-GOODALL,J. VAN, 1968. The behaviour of free-living chimpanzees in the Combe Stream Reserve. An. Behav. Mono., 1 : 161-311. MEIER, G. W., H. C. WILCOXIN,R. ORLANDO,& D. C. PAULSON, 1969. The development of food-reinforced behaviors in a social situation. In: Proceedings of the Second International Congress of Primatology (Vol. 1), C. R. CARVENTER (ed.), S. Karger, Basel, pp. 143-148. MENZEL, E. W., 1966. Responsiveness to objects in free-ranging Japanese monkeys. Behav., 26: 130-150. - - , J. K. DAVENVORT,& C. M. ROGERS, 1970. The development of tool-using in wildborn and restriction reared chimpanzees. Folia primat., 12: 273-283. PAR~ER, C. E., 1969. Responsiveness, manipulation and implementation behavior in chimpanzees, gorillas and orangutans. In: Proceedings of the Second International Congress of Primatology (Vol. 1), C. R. CARVENTER(ed.), S. Karger, Basel, pp. 160-166. REYNOLDS, G. S., 1968. A Primer of Operant Conditioning. Scott, Foresman, Glenview, Ill. R~MBAUGH, D. M., 1969, The transfer index. An alternative measure of learning set. In: Proceedings of the Second International Congress of Primatology (Vol. 1), C. R. CARVENTER(ed.), S. Karger, Basel, pp. 267-273. - - , 1970. Learning skills of anthropoids. In : Primate Behavior (Vol. 1), L. A. ROSENBLt~M(ed.), Academic Press, New York, pp. 1-70. SCmLLER, P. H., 1957. Innate motor action as a basis for learning. In: Instinctive Behavior, C. H. SCHILLER(ed.), International Universities Press, New York, pp. 264-287. STEVHENSON, G. R., 1967. Cultural acquisition of a specific learned response among rhesus monkeys. In: Neue Ergebnisse der Primatologie, D. STARCK, R. SCHNEIDER, & H. J. KUnN (eds.), Gustav Fischer Verlag, Stuttgart, pp. 279-288. TOBIAS, P. V., 1965. Australopithecus, Homo habilis, tool-using and tool-making. S. Afr. Archeo. Bull., 20: 167-192. WASHBLrRN,S. L., 1960. Tools and human evolution. Sci. Amer., 203 : 62-75. YERKES, R. M. & A. W. YERKES, 1929. The Great Apes. Yale University Press, New Haven. ZtJCKERMAN, S., 1932. The Social Life of Monkeys and Apes. Routledge & Kegan Paul, London. -
-
-
-
- - R e c e i v e d November 2, 1971; Accepted February 5, 1972. Author's Address: BENJAMINB. BECK, Chicago ZoologicalPark, Brookfield, Illinois 60513, U.S.A.