Regulation of nest construction behaviour inPolybia occidentalis

Anim. Behav., 1996, 52, 473–488

Regulation of nest construction behaviour in Polybia occidentalis ROBERT L. JEANNE Departments of Entomology and Zoology, University of Wisconsin at Madison (Received 5 July 1995; initial acceptance 3 October 1995; final acceptance 7 December 1995; MS. number: 7357)

Abstract. Nest construction by the wasp Polybia occidentalis involves the coordination of three groups of specialized workers. Wood-pulp foragers deliver their loads to builders, who use it to construct the nest. Water foragers transfer water to both pulp foragers and builders, who use it to keep the pulp moist during handling. During a round of nest construction or repair, the levels of activity in the three groups are kept in balance, minimizing queuing delays during transfer of these materials. If the operation is perturbed, balance between the groups is restored, indicating that the system self-regulates. The aim of this study was to determine the sources of information used by each of the three task groups to adjust its level of activity to those of the others. Colonies’ responses to experimentally-induced increases and decreases in the flow of each material during nest repair were videotaped in the field in Costa Rica. The results showed that only the builders received information about the nest damage directly from the nest. Feedback between builders determines the level of building activity, which in turn sets the size of the entire repair operation. The level of pulp-foraging activity was determined by information about demand for pulp received by pulp foragers from builders. Water foragers similarly adjusted their group’s activity rate in response to feedback received from pulp foragers and builders about the demand for water. Such a chained flow of information from one task group to the next, in reverse order of material flow, allows each task group to fine-tune the size of its effort to that of the group it serves, enabling the whole system to respond efficiently to perturbations. ?

Nesting behaviour in the social insects is one of the most spectacular manifestations of emergent behaviour in the biological world. The nests themselves are often remarkably complex and intricately adapted to provide housing for the colony, protection against predators and parasites, and facilitate homeostatic control of physical conditions for the adults and brood. The nest may be immensely larger than a single individual could construct, and the fabrication of the whole structure may take longer than the lifetime of an individual worker, yet a worker can contribute labour to the effort in ways that are appropriate to the stage of construction at the time (Wilson 1971). This coordination is achieved without any evidence of central control; no one individual or group has a ‘blueprint’ for the nest from which Correspondence: R. L. Jeanne, Department of Entomology, University of Wisconsin, 1630 Linden Drive, Madison, WI 53706, U.S.A. (email: jeanne@ entomology.wisc.edu). 0003–3472/96/090473+16 $18.00/0

?

1996 The Association for the Study of Animal Behaviour

it directs others. Instead, the organization of workers into a nest-constructing group is characteristic of a distributed process, from which the nest structure, the product of colony-level behaviour, emerges (Deneubourg et al. 1990; Camazine 1991, 1993). One of the challenges of insect sociobiology is to explain how such colony-level behaviour emerges from the individual decisions of members of the colony. Some of the more elaborate nests constructed by social insects are found among the wasps. In the complexity and regularity of their nests and the diversity of their construction techniques, wasps equal or surpass many of the ants and bees (Jeanne 1975; Wenzel 1991). Polybia occidentalis, a common wasp of the neotropics, is an excellent species in which to analyse how nesting behaviour is organized in an advanced eusocial insect (Jeanne 1986, 1987; O’Donnell & Jeanne 1990). Colonies of this wasp are founded by a swarm of workers and queens. Within a few days of settling on a site, the workers in the swarm complete 1996 The Association for the Study of Animal Behaviour

473

474

Animal Behaviour, 52, 3

construction of a nest consisting of several stacked combs. Each comb is covered by an envelope, which serves as the substrate for the next comb. After the nest structure is complete, the colony spends the next several months rearing worker offspring, while enlarging the nest from time to time in bouts of rapid construction (Forsyth 1978). Nest construction in this species represents a social operation that is all the more remarkable in that it involves the coordination of three groups of specialized workers. Two kinds of materials, wood pulp and water, are collected by two groups of foragers to be used by the third group, the builders, to construct the nest (Jeanne 1986). Water foragers regurgitate their loads to both builders and pulp foragers, who use it to moisten wood pulp. Wood-pulp foragers distribute their loads to builders, who add the material to the appropriate places on the nest. Thus each of the three task groups interacts with the other two while transferring materials. Although a worker that is engaged in one of the nest construction tasks may switch to one of the other tasks, this normally occurs at low rates (Jeanne 1986). During a round of nest construction or repair, the number of workers allocated to each of the three groups is such that the three tasks are kept in balance and queuing delays are kept to a minimum (Jeanne 1986). On average, pulp foragers are able to find recipients for their loads within 7 s of landing on the nest, water foragers transfer their entire croploads of water within 35 s of arriving on the nest, and builders find both water and pulp for their next bout of building within about 38 s of using up the previous load (Jeanne 1986). If the operation is perturbed, for example by removing some of the workers in one of the groups, the balance can be restored (Jeanne 1987). This ability implies some manner of selfregulation of the whole system. Results of an earlier study suggested that water foragers set the pace for the operation (Jeanne 1987). The effects of manipulations done in that study, however, were not studied in detail. The purpose of this study was to investigate how this regulation is achieved by analysing the sources of information used by each of the three worker groups. Three possibilities for how control is exercised are put forward here as working hypotheses. First and most simply, individual workers may decide whether to become active by responding directly

to the nest damage. If individuals differ in their probability of working in response to a given degree of nest damage, then, through positive feedback, the size of the damage will directly determine the number of active workers in the task group. I refer to this as the ‘nest feedback’ hypothesis. A second possibility is that there may be some form of self-regulation of the size of task groups via negative feedback from active members of the same group. A low level of activity by a task group may feed back positively to idle members of the group, stimulating them to become active. Conversely, a high level of activity being performed by the task group may, through negative feedback, inhibit the recruitment of additional workers to the group. I call this the ‘task-mates feedback’ hypothesis. Finally, there may be feedback across task groups. Because workers in each task group interact with workers in both of the other groups while transferring materials, workers in a task group may decide to become active not in response to the level of activity of their own group, but in accordance with their perception of the supply or demand for a material, as imposed by members of other task groups. An excess of activity by pulp foragers, for example, could feed back positively to members of the builder group, stimulating more of them to become active. Conversely, a low level of activity of one group relative to the others could feed back negatively to active members of these other groups, causing some of them to become inactive. I refer to this as the ‘non-taskmates feedback’ hypothesis. I tested these three hypotheses for each of the three task groups by analysing their responses to perturbations I imposed as colonies were engaged in nest repair. As an ancillary result, I estimated the speed with which the system self-corrects following perturbation. The focus is on group, rather than individual, responses. The question of the cues by which information is transmitted from one worker to another was not addressed in this study. METHODS Study Site and Population I conducted field studies on P. occidentalis at Centro Ecológico ‘La Pacífica’, near Can˜as,

Jeanne: Regulation of nest construction Guanacaste, Costa Rica (10)25*N, 85)7*W). Can˜as is within the tropical dry forest (moist province transition) life zone (Tosi 1969), which is marked by strong wet–dry seasonality, the wet season extending from May through October. Mean annual rainfall (1978–1987) for Can˜as is 1369 mm. The native tropical dry forest has been broken up at La Pacífica by pasture and cropland. Colonies were located in shrubs and trees in pastures, hedgerows, forest edges, and on and around buildings. I chose moderate-sized colonies (containing several hundred adults) for study on the basis of ease of access for observation. Data were collected on two colonies during June–August 1989 and on two colonies during June–July 1990. Induction of Nest Construction Behaviour Between bouts of nest expansion, construction behaviour occurs frequently but at a low level in the form of the lengthening of brood cells to accommodate the growth of larvae and the thickening of the upper walls of the nest to strengthen the supporting structure. To induce more extensive nest construction behaviour, I removed up to one-half of the lower envelope of the nest in the late afternoon. The construction behaviour of the wasps participating in the resulting envelope repair resembled in every discernible way behaviour observed during normal construction, except that repair was not completed in one continuous effort. Work did not begin until dawn of the day following the damage and was terminated by late morning. If the repair was not complete, the workers resumed construction behaviour the following morning. Marking Individual Workers I marked wasps for individual recognition by using Decocolor> paint pens to paint unique colour codes on the thorax. Pulp and water foragers were captured for marking just after their return to the nest by grasping them gently around the waist with spring forceps and lifting them from the nest. With rare exceptions, this could be accomplished without disturbance to other resident workers. Wasps were held without anaesthesia in reverse-spring forceps for marking and immediately released after marking. Because active foragers could be readily identified as they landed and because their numbers typically did

475

not exceed 20 at one time, it was easy to mark all that were active at a given time. Builders were much more numerous (60+) than foragers and could be assigned unambiguously to the builder task group only when they were adding pulp to the edge of the envelope. Because of the firm grip they obtained by straddling the edge, the mechanical disturbance of pulling them free with forceps frequently caused an alarm reaction in other workers, resulting in a disruption of building activity. Each successive disturbance led to heightened alarm in the colony and a longer delay before building returned to normal and the next builder could be marked. This, combined with the large supply of replacement builders, made it unfeasible to attempt to mark all the builders individually. It was necessary to mark new workers every 2–3 days because of the continual turnover of individuals in the three task groups resulting from recruitment of young workers and the aging and death of older ones. Videotaping Nest Repair When marking was complete, I again removed a portion of the envelope at the end of the afternoon. The next morning I videotaped repair behaviour at the nest, using a Panasonic model AG-450 S-VHS camcorder set approximately 1.3 m from the nest and zoomed so the nest filled the screen. This magnification allowed the behaviour of individuals to be resolved, but was insufficient to allow the colour of the paint spots to be discerned. To circumvent this shortcoming, I narrated observations onto the tape, identifying marked individuals orally while using a grass straw to indicate them visually without touching or disturbing them. Because foragers on the nest confined their movements to the front of the envelope over the entrance and because the edge of the envelope being repaired faced front, virtually all repair-related behaviour was visible on the tape. The only exception was when builders temporarily disappeared from view on the back of the lower comb as they walked from the upper part of the nest to a construction site on the envelope’s edge. Manipulations Videotaping a morning’s repair behaviour began with 18–80 minutes of undisturbed activity

476

Animal Behaviour, 52, 3

to record a baseline level of activity. This was followed by one of the following manipulations. I attempted at least two repetitions of each. Water forager removal. Removing the majority of active water foragers decreased the rate of inflow of water. Water supplementation. Adding water to the nest increased the availability of water. Water was pipetted onto the upper surface of the lower envelope, just behind the edge being repaired, and spritzed onto the entire upper surface of the nest. Water was supplemented ad libitum for 25–30 min, then allowed to run out. Pulp forager removal. Removing the majority of active pulp foragers decreased the rate of inflow of pulp. Pulp supplementation. Providing wood fibre increased the availability of nest material. Wads of pulp were scraped with a knife from moistened sources used by pulp foragers and put onto the tips of spatulas anchored so as to touch the nest in the region where builders searched for newly arrived pulp foragers. Supplemental pulp was made available ad libitum for 35–50 min, then removed. From time to time, small amounts of water were pipetted onto the wads to keep them moist. Builder removal. Removing the majority of active builders decreased the rate of building. In all cases, I placed removed workers in screened containers in an ice chest, held them until the end of the morning’s experiment, then released them below the nest. Data Collection I extracted the following data from each day’s experiment directly from the videotapes into a computer spreadsheet: identification number of each forager; type of load (water or pulp); time of return to the nest; time of start of transfer of load to a receiver; time of completion of transfer; and time of take-off for the next foraging trip. In each case, the times entered were the elapsed time from the start of observations, to the nearest second, as measured with a stopwatch recorded on the videotape. For each pulp forager’s return, I tallied the number of times its offer of pulp to a nestmate was rejected (loaded forager turns towards nestmate and pauses, nestmate turns away or contacts load but takes none). The rate of building activity was measured every 5 min as the instantaneous number of build-

ers actively adding pulp to the edge of the envelope. For each measure, two consecutive counts were made and averaged. Data Analysis I calculated the durations (in seconds) of the following intervals for each water and pulp forager: from landing on the nest to the start of load transfer (unloading delay); from the start to the end of load transfer (unloading duration). I also computed the interval between successive arrivals of foragers of each load type (inter-trip interval). Behavioural events during the course of each experiment were grouped into 5-min intervals for analysis and graphing. I assessed the effect of each manipulation by comparing behaviour after the manipulation began to behaviour before it started. The ‘before’ period extended from the start of observations through the 5-min interval prior to the one in which the manipulation was begun. If I began the manipulation in the second half of its 5-min interval, then the before period included that interval as well. The effects of the manipulation were observed as modified behaviour consisting of a two-part response. The ‘initial response’ is the disruption in work rates caused by the manipulation. The ‘corrective response’ is the workers’ adjustment to the disruption, which returns the rates to pre-manipulation values. The onset and duration of these responses varied from one manipulation to another. For example, one source of variability was the time it took to complete the removal of workers. Another source may have been variation in the readiness of replacements to begin work. Because of this variability, arbitrarily assigning rigid times to onset and termination of the response periods would often have meant missing the manipulations’ effects. Therefore, in an effort to capture and analyse the actual initial and corrective responses of each manipulation, I adopted the following operationally defined conventions. The ‘initial response’ period began with the first 5-min interval in which the rate of performance of the behaviour in question dropped to or below 0.67 (or rose to 1.5) of its mean rate during the before period. It continued until, but did not include, the interval in which the rate rose (fell) to the mean of the before rate, or through to the end of observations. The ‘corrective response’ period began with the first 5-min interval whose

Jeanne: Regulation of nest construction

RESULTS Not all experimental manipulations yielded usable data. In some cases too many foragers were unmarked. In others, bad weather intervened, or smoke from burning residue on nearby rice fields disrupted the wasps’ behaviour patterns. In still others, the nest repair effort was weak, or the manipulation was followed by a slowing or stopping of repair activity. If the building effort did not remain strong through the manipulation and with a large fraction of the foragers marked, results are not reported. This was the case for 11 of the 21 experiments attempted. The results of the remaining 10 experiments are reported. Water Forager Removal According to the nest feedback hypothesis, workers in the water forager pool respond directly

4

(a) ***

Water foragers

3

2

1 * 0 16

(b)

14 12 Water arrivals

value dropped (rose) to the mean of the before period, even if this occurred before the end of the manipulation, and continued until the beginning of the next manipulation or until the end of observations. If the mean value of the before period was not reached, no corrective response period was defined. The number of individual pulp foragers active in a 5-min interval included all individuals arriving at the nest during that interval, plus those arriving during the next interval (because pulp foraging trips last approximately 5 min, a forager landing in one 5-min interval would have to have started its trip during the previous interval). If a given pulp forager waited more than 30 min between successive arrivals (measured from return to return), she was arbitrarily assumed to have stopped foraging in that interval and was not counted. Water foragers were tallied in the same manner except that, because round trip time was shorter than for pulp arrivals (approximately 1 min), they were considered to have become inactive when the gap between successive trips exceeded 15 min. I tested comparisons between the ‘before’ and ‘initial response’ periods and between the ‘initial response’ and ‘corrective response’ periods using the Mann–Whitney U-test on the 5-min interval data (number of forager arrivals, foragers, building acts and rejected pulp offers). For continuous data (unloading delay, unloading duration), raw times were used.

477

***

10 8 6 4 2 0

* 0

20

40 60 Time (min)

80

Figure 1. Effects of the removal of water foragers, colony 90, 5 August 1989. (a) Number of individual water foragers active. (b) Number of water forager arrivals. Black rectangle indicates the period during which the three water foragers were removed. Bold horizontal lines indicate the ranges and mean values for the ‘before’ (pre-manipulation) period (left), the ‘initial response’ period (middle) and the ‘corrective response’ period (right). Asterisks indicate the level of confidence that the difference between the denoted period and the previous period is not due to chance: *, P<0.05; **, P<0.01; ***, P<0.001.

to the damage and decide whether to become active according to the size of the stimulus. This hypothesis predicts that if active water foragers are removed, they will not be replaced because the manipulation did not affect the size of the stimulus. The results (Fig. 1) fail to support this hypothesis. The three removed water foragers were

Animal Behaviour, 52, 3

Water Supplementation Water foragers ignored the droplets of water on the nest, and pulp foragers and builders used them. On one occasion, at the same time builders and pulp foragers used the water on the envelope, workers inside the nest treated the water inside the envelope as waste, sucking it up and regurgitating it out of the entrance. This observation indicates that demand for a resource is strictly localized within the nest. The task-mates feedback hypothesis predicts that the availability of supplemental water at the nest would cause no change in rate of water foraging, because adding excess water to the nest does not alter the ongoing rate of activity in the group. The results fail to support this hypothesis. Adding water resulted in a reduction in the number of active water foragers and a significant drop in overall water foraging rate within 20 min (Figs 2, 3). As did the water forager removal manipulation, this result also fails to support the nest feedback hypothesis, which predicts no change in water foraging rate in response to supplemental water because there has been no change in the size of the nest damage. Together, the results of this and the previous manipulations support the feedback from nontask-mates hypothesis for water foragers and suggest that water foragers decide whether to be active or idle in response to the level of demand

7 (a) 6 Water foragers

replaced within approximately 30 min, returning the foraging rate to pre-removal levels. Of the two marked replacement water foragers, one switched from pulp foraging and the other had been idle but had been a water forager on previous days. The observation that water foragers did not inspect the damage area between trips also fails to support the nest feedback hypothesis. In fact, water foragers typically landed on the envelope above the entrance, regurgitated to one or more nestmates, then took off again without venturing beyond the upper part of the nest. Therefore, either water foragers respond to feedback within their own group to maintain a certain level of activity (task-mates feedback hypothesis), or their activity level is a direct response to demand for water from builders and pulp foragers (feedback from non-task-mates hypothesis). The next manipulation tested between these two alternatives.

5 ***

4 3 2 1 0 14 (b) 12

Water arrivals

478

10 8 **

6 4 2 0

0

20

40 60 80 Time (min)

100

120

Figure 2. Effects of the addition of water to the nest, colony 89, 13 July 1989. (a) Number of individual water foragers active. (b) Number of water forager arrivals. Black rectangle indicates the period during which excess water was available on the nest (see Fig. 1).

for water exerted by the other two groups, a conclusion further supported by the results of the next manipulation. Pulp Forager Removal The removal of pulp foragers on all three dates had a depressive effect on the rate of water foraging, which tracked pulp foraging rate (Figs 4, 5, 6c), although the changes were significant only on one date. This effect supports the non-taskmate hypothesis, that water foraging rate is responsive to the demand exerted for water by pulp foragers and builders.

Jeanne: Regulation of nest construction 3

8

(a)

479

(a)

Pulp foragers

Water foragers

6 2

*

1

4

***

2

0 10

0 (b)

10

(b)

8

6

Pulp arrivals

Water arrivals

8

4

6

4

**

2

2 0

0

20 Time (min)

40

0 14

Figure 3. Effects of the addition of water to the nest, colony 90, 1 August 1989 (see Fig. 2).

(c)

12 10 Water arrivals

The nest feedback hypothesis with respect to pulp foragers predicts that removed pulp foragers should not be replaced, because each individual in the pool is making a decision to become active or not according to its direct perception of the level of damage to the nest. The results show that on two of three dates new pulp foragers were recruited within 20–45 min following rounds of removal and that rates of foraging recovered (Figs 4, 5a, b). One recruit had been marked as a pulp forager two days earlier; the other nine were all unmarked. On the third date, pulp foragers were not replaced (Fig. 6a). The preponderance of the evidence leads to the rejection of the nest feedback hypothesis. As was the case for water foragers, this conclusion is supported by the observation that pulp foragers did not inspect the damage area between trips. They landed on

***

NS

8 6 4

NS

**

2 0

0

20

40 60 80 Time (min)

100

120

Figure 4. Effects of the removal of pulp foragers, colony 87, 21 July 1989. (a) Number of individual pulp foragers active. (b) Number of pulp forager arrivals. (c) Number of water forager arrivals. Black rectangles indicate the periods during which, respectively, 6 and 3 pulp foragers were removed (see Fig. 1).

Animal Behaviour, 52, 3

480 12

7

(a)

(a) 6

10

5 Pulp foragers

Pulp foragers

** 8 6 4

**

3 2

2

***

1

0 16

4

0 6

(b)

(b)

14

5

* Pulp arrivals

Pulp arrivals

12 10 8 6 4

4 3 2

** 1

2 0

***

0

14 (c)

(c) 6

**

10

Water arrivals

Water arrivals

12

8 6

*

4

4 NS NS

2

2 0

0

20

40 60 Time (min)

80

100

Figure 5. Effects of the removal of pulp foragers, colony 90, 7 August 1989. Six foragers removed in each period (see Fig. 4).

0

0

20

40 60 Time (min)

80

100

Figure 6. Effects of the removal of pulp foragers, colony 90, 29 July 1989. Four foragers were removed (see Fig. 4).

Jeanne: Regulation of nest construction the envelope above the entrance, unloaded their pulp to one or more nestmates there, then took off again after obtaining water from water foragers on the upper part of the nest. The results instead support the remaining two hypotheses: that pulp foragers either respond to feedback within their own group to maintain a certain group level of activity (task-mates feedback hypothesis), or their activity level is a direct response to demand for pulp from builders (non-task-mates feedback hypothesis). The next manipulation tested between these two alternatives. Pulp Supplementation Builders readily discovered and used the pulp wads placed in contact with the nest, carving off large lumps which they then subdivided with other builders, just as they did with loads gathered by foragers. Pulp foragers ignored these resources. According to the task-mates feedback hypothesis, the overall pulp foraging level is maintained by feedback within the group. This hypothesis predicts that adding supplemental pulp to the nest should cause no change in the rate of pulp foraging, because adding excess pulp does not alter the rate of activity in the group. The outcome failed to support this hypothesis (Figs 7, 8a, b). On both dates, adding pulp resulted in a significant drop in the pulp foraging rate and a reduction in the number of active pulp foragers within 20 min of the beginning of the manipulation. Both the number of foragers and the pulp foraging rate rebounded following the removal of supplemental pulp from the nest, but did not reach premanipulation levels. These results also do not support the nest feedback hypothesis, which predicts no change in pulp foraging rate in response to supplemental material, because there has been no change in the size of the stimulus. The results of the pulp forager removal and pulp supplementation manipulations, taken together, support the non-task-mates feedback hypothesis for pulp foragers, and suggest that the level of group activity by pulp foragers is sensitive to the level of demand for pulp imposed by the users of pulp, the builders. This demand was reduced when supplemental pulp was available to the builders. Adding excess pulp to the nest also caused an increase in the building rate (Figs 7, 8c). This

481

response, however, was not primarily due to an increase in the number of active builders. If more builders had been recruited to handle the increased flow of pulp, the number of building acts would have remained high and the level of pulp foraging would have dipped temporarily but then rebounded during the period of excess pulp availability. Instead, pulp foragers almost immediately experienced increased unloading delays, unloading duration, and number of rejected offers of pulp (Figs 7, 8d–f), effects that would be expected if the builders were using the supplemental pulp supplies and neglecting pulp foragers. Pulp foragers responded by reducing the overall rate of foraging, as was shown above, which reduced their unloading difficulties, in one case even while the supplemental pulp was still available (Fig. 8d–f). Following the removal of supplemental pulp, the pulp foraging rate rebounded. On 15 July 1990 (Fig. 8), when pulp was available for an unusually long period, the number of building acts peaked and then declined while supplemental pulp was still available, further supporting the conclusion that additional builders were not recruited. These results lead to the rejection of the hypothesis that builders adjust their numbers and rates in response to the amount of pulp coming into the nest (non-task-mates feedback hypothesis). On the other hand, the nest feedback hypothesis is supported for builders. If members of the builder task group were making individual decisions to be active or idle according to the level of the nest damage, making extra pulp available would not cause a change in work rates because the level of nest damage has not been changed. Although the number of building acts temporarily increased, the argument made above suggests that this was due to a faster cycling time by already active builders rather than to the recruitment of more builders. In addition to supporting the nest feedback hypothesis, the results also support the hypothesis that builders are responding to feedback from within their own group to keep the group rate of building more or less constant (task-mates feedback hypothesis). The final manipulation tested between these two alternatives. Builder Removal In one instance (third round of builder removal on 13 July 1990), grasping a worker caused local alarm due to jarring of the nest or to the release of

Animal Behaviour, 52, 3

482 10

60 (a)

(d)

56 Unloading delay (s)

Pulp foragers

8

6 ***

4

2

24 20 16 12 ***

8

* 4

0

0 24

10 (b) Unloading duration (s)

8 Pulp arrivals

(e)

22

6

4

**

2

12 10 8

NS

6 NS

4 2

0

0

18

8 (c)

16

(f)

Building acts

Rejected pulp offers

*

14 12 10 8 6

*

4

6

4 *** 2 **

2 0

0

20

40 60 Time (min)

80

100

0

0

20

40 60 Time (min)

80

100

Figure 7. Effects of the addition of supplemental pulp to the nest, colony 87, 17 July 1989. (a) Number of individual pulp foragers active. (b) Number of pulp forager arrivals. (c) Number of building acts being performed simultaneously. (d) Unloading delay for pulp foragers (time from landing at the nest until the start of unloading pulp). (e) Unloading duration (time from start to end of unloading pulp). (f) Rejected pulp offers (number of offers of pulp made by pulp forager that were rejected). (d–f) Bars=mean, error bars=. Black rectangles indicate the periods during which moist pulp wads were available at the nest (see Fig. 1).

Jeanne: Regulation of nest construction 12

483

36 (a)

(d)

32

10 Unloading delay (s)

Pulp foragers

28 8 *** 6 4

24 20 16 12 *

8 2

***

4 0

0

14

75

(b) Unloading duration (s)

12

Pulp arrivals

10 8 *

6 *

4

(e)

60

2

28 24 20 16 12

*

8 ***

4 0

0 14

16

(c)

(f)

12 8

* Rejected pulp offers

Building acts

12 10 8

***

6 4

4

***

***

2 0

0

20

40 60 80 Time (min)

100

120

0

0

20

40 60 80 Time (min)

100

120

Figure 8. Effects of the addition of supplemental pulp to the nest, colony 134, 15 July 1990 (see Fig. 7).

venom by the captured worker (Jeanne 1981). In this case, I waited until the colony calmed before removing additional builders.

Results of the builder removal experiments distinguish between the nest feedback and taskmates feedback hypotheses for builders. If the nest

Animal Behaviour, 52, 3 16 (a) 14

Building acts

12 10 8 6 4

***

2 0 14 (b) 12 10 Pulp foragers

feedback hypothesis were true and each builder responded primarily to the severity of nest damage, then removing a portion of the builders would not lead to their replacement by previously inactive builders, because the level of the stimulus is unchanged. Thus the level of building activity would remain depressed. On the other hand, if the task-mates feedback hypothesis were true and the size of the force is somehow maintained at a given level via feedback among members of the pool of builders, then the removed builders should eventually be replaced. The results lead to the rejection of the former hypothesis and support of the latter (Figs 9, 10a). Following each round of removal, the level of building activity, measured as number of building acts, rebounded, although the increase was not significant in every case. The rebound is best explained as the recruitment of newly active replacement builders and suggests that somehow workers in the builder pool, both active and inactive, seek to maintain a set global level of repair activity. When this level is artificially reduced, some of the previously idle builders become active. The results of this manipulation also show that the rate of pulp foraging and number of active pulp foragers track the rate of building activity. On 27 June 1990, the numbers of pulp foragers and arrivals dropped precipitously, recovering slightly but temporarily, just as did builder activity (Fig. 9b, c). On 13 July the rate of pulp forager arrival again mirrored the drop in building rates, but the strong corrective response of builder activity following the first two rounds of removals damped this effect on pulp forager numbers until after the third round of removals (Fig. 10b, c). These results provide additional support for the non-task-mate feedback hypothesis that the pulp forager group adjusts its rate of activity to the demand for pulp as set by builders.

8 6 ***

4 2 0 14 (c) 12 10 Pulp arrivals

484

8 6 4 *** 2

DISCUSSION 0

Ultimately, the information keeping all three worker groups active at nest repair must come from the nest damage itself. Of the three groups, however, the builder task group appears to be the only one responding to this information directly. First, only the builders regularly come into contact with the damage; pulp and water foragers do not. Second, the builders do not respond to

0

20

40 60 Time (min)

80

100

Figure 9. Effects of the removal of builders, colony 33, 27 June 1990. (a) Number of building acts being performed simultaneously. (b) Number of individual pulp foragers active. (c) Number of pulp forager arrivals. Black rectangles indicate the periods during which, respectively, 28 and 24 builders were removed (see Fig. 1).

Jeanne: Regulation of nest construction 10

(a)

8 Building acts

NS

*

**

6

***

4 * 2

0 14

(b)

12

Pulp foragers

10 8 *** 6 4 2 0 14

(c)

12

Pulp arrivals

10 8 *** 6 4 2 0

0

60

120 180 Time (min)

240

Figure 10. Effects of the removal of builders, colony 134, 13 July 1990. Black rectangles indicate the periods during which, respectively, 27, 32, 1 (caused alarm) and 33 builders were removed (see Fig. 9).

485

feedback from other groups, but foragers do. The fact that the builder group is able to return to a constant level of activity in response to perturbations indicates some mechanism of homeostatic control of global work effort. Thus, although builders respond to the nest damage, this is not their only source of information. Results reported here support the hypothesis that additional feedback between active and inactive builders regulates the number of builders that are active. One mechanism by which these two sources of information could interact to produce the observed result is the following. When the nest is damaged, each potential builder makes a decision whether to become active according to the balance between the positive feedback from the damage to the nest (stimulus) and the negative feedback from those already working (inhibition). A builder with a high sensitivity to the stimulus of nest damage may become active, but one with a lower sensitivity may be inhibited from becoming active by the level of activity already ongoing. The mechanism of the negative feedback from the active subgroup is not known, but it could be as simple as the number of builders on the repair site at any one time. As long as there is room for more builders along the damaged envelope edge, additional members of the builder pool will become active, but an edge that is filled with building workers inhibits the recruitment of additional active builders. Further work will be required to test this hypothesis. Pulp foragers are one step removed from the repair work. They neither inspect the damage site nor do they add material to it. Yet, like builders, if some of the active foragers are removed the global level of pulp foraging rebounds to close to the original level through the recruitment of newly active replacements. Unlike builders, however, pulp foragers do not self-regulate their group size. Instead, their activity level appears to be set by the level of demand for pulp as imposed by the builders. Although it is true that a forager gets feedback from her active task-mates in the form of a reduction of demand on her, this feedback is indirect in that it comes by way of the builders. The effect of demand is seen if extra pulp is made available at the nest or if some of the builders are removed. The resulting decrease in the demand for foraged pulp causes a decrease in both the number of active pulp foragers and the global rate of pulp foraging. The downward adjustment of group

486

Animal Behaviour, 52, 3

foraging rate in response to decreased demand is achieved in two ways. The immediate response is for the active foragers to reduce their individual foraging rates. This was seen on 13 July 1990, where the arrival rate (Fig. 10c) closely mirrored the building rate (Fig. 10a). This form of slowdown could be entirely due to the increased unloading difficulty pulp foragers experience at the nest, causing them to spend more time on the nest before making the next trip. The second mechanism is through a reduction in the number of active foragers. This response is less sensitive to short-term changes in demand, only coming into play following a sustained drop in demand (Fig. 10b). Water foragers are two steps removed from the ultimate source of information, the damaged nest. That is, nest repair requires pulp, but it is pulp that requires water, so water foragers could not assess the need for water even if they did visit the site of the nest damage. Like pulp foragers, water foragers appear instead to obtain information about the need for their material through feedback from water users, namely the pulp foragers and the builders. Reducing the need for water by spraying it on the nest caused a decrease in the overall rate of water foraging. Removal of pulp foragers also reduced demand for water by reducing the rates of pulp foraging and building. The results of these experiments offer little insight into how idle pulp and water foragers become active. One possibility is that idle foragers monitor demand levels simply by being on the envelope where they are contacted both by workers offering materials and workers soliciting materials. An increase in the frequency of solicitations from builders seeking pulp, for example, may induce an idle pulp forager to begin working. Determination of the cues prompting idle foragers to become active awaits further investigation. In sum, none of the workers respond solely to the nest damage; all respond to feedback either from other members of their own group or to members of one or both of the other groups. The ultimate source of information to which the entire work force is responding is indeed the nest, but only the builders appear to be perceiving it directly. The foragers do not have direct access to this information because they do not come in contact with that part of the nest while they are active. Instead, the information they use to set levels of activity comes directly from the builders,

Nest

+ Builders

– + Pulp foragers

+ Water foragers

+ Pulp sources

Water sources

Figure 11. Effects of materials and information between task groups during nest construction. Thin arrows indicate direction of flow of nesting materials (water and wood pulp). Thick arrows indicate the major pathways of information flow among the task groups. The major form of feedback along each information pathway is indicated with a (+) (positive, or stimulatory effect) or a (") (negative, or inhibitory effect).

and in the case of water foragers, from builders and pulp foragers. Thus the builders set the pace for the entire operation, with the pulp and water foragers adjusting their rates in response to the need for materials set by the builders. In other words, in this demand-driven system, the pathway of information flow (Fig. 11) is the opposite of the direction of material flow. Results from a preliminary study (Jeanne 1987) suggested that the system was driven by the supply of water brought in by water foragers. In that study, removed water foragers were not replaced within an hour. The present study, however, shows that water foragers can be replaced. Whether removed foragers are replaced within a short time may depend on the availability of experienced foragers of the same type. Of the three tasks, water-foraging rates are especially dependent upon the presence of one or two ‘elite’, or hard-working, water foragers, which bring in most of the water. What makes a water forager an ‘elite’ is not known, but evidently not every worker in the water forager pool is able to step in and become an elite within an hour. The failure of the colony to replace its water foragers in the earlier study may simply signify that the availability of elite water foragers is limited and may vary from colony to colony or from day to

Jeanne: Regulation of nest construction day. The present study clearly supports the hypothesis that the system is driven by demand, not supply. The organization of information flow in this manner has advantages. If instead each group were to respond independently and directly to the nest damage and ignore feedback from its own or other task groups, the number of active individuals in each group would be solely a function of individual tendencies to work and would bear no relation to the level of demand. This arrangement could lead to excessive queuing delays as workers waited for donors or receivers. Chaining information flow hierarchically from one group to the next in reverse order of material flow keeps the levels of activity of the three task groups in balance with one another. Chaining incorporates information about the level of demand from each user one step up the line and thus allows finetuning of the size of each group to decrease queuing delays and increase productivity. Each of the three groups has a pool of ‘reserves’ from which replacements are drawn if some of the active workers are removed. The character of this reserve pool is poorly understood. On one hand, reserves could be workers that have previously performed the task but happen to be inactive on a given day. Alternatively, they may consist largely of inexperienced, younger workers who have not yet performed the task in question. Circumstantial evidence suggests both possibilities occur. The rapid activation (30 min) of replacements following removals suggests that these are experienced workers; indeed, some of the replacement pulp and water foragers were marked and had performed their tasks on previous days. That the supply of these is limited, however, is suggested by the low level of recruitment of builders following the third round of removals on 13 July 1990 (Fig. 10). Seeley (1989), in his experimental studies of integration of nectar foraging and storage in honey bees, found that after the supply of ready food-storer bees was exhausted by repeated removal, subsequent replacements took longer to appear. Gordon (1989, 1995) has found that auxiliary workers in seed harvester ants are recruited largely from lower age groups. The lag time of 20–35 min before a response to a manipulation suggests a damped response to changes in feedback, which may be adaptive. Given the stochasticity that inevitably exists in the arrival intervals of workers to each queue, too

487

tight a response to a heightened unloading difficulty could lead to excessive oscillations in the numbers in each group. It is worth noting that the lag time for foragers to slow and drop out in response to provisioning of excess water and pulp appeared to be less (on the order of 20 min) than the 30–35 min it took replacements to be recruited following the removal of foragers and builders. This difference suggests that the cues that keep workers active give more direct and immediate information about the level of demand than do the cues idle workers respond to by becoming active. Finally, a considerable amount of variability existed in the data collected, not only in the overall size of the nest repair effort and relative numbers of active workers in each task group from one experiment to the next, but also in the strength of the response to a particular manipulation. Some of this variability is evident in the data presented: compare, for example, the relative lack of corrective response in pulp forager numbers following removal on 29 July 1989 (Fig. 6) with the other two replicates (Figs 4, 5). Variability also existed in the experiments that were not reported. In some cases, a weak nest repair response was shown by colonies preparing for seasonal evacuation of the nest (Jeanne et al. 1988). In other cases, dramatically different levels of repair activity were witnessed on successive days on the same colony under seemingly similar weather conditions. This variability only serves to indicate that other factors in addition to the feedback mechanism postulated here influence the system. Such factors may include colony size, colony stage, state of the brood population and weather (including relative humidity). Further investigation will be required to determine the causes of this unexplained variation.

ACKNOWLEDGMENTS Erik Nordheim and Eric Espe are gratefully acknowledged for advice and assistance with statistical analyses. Sean O’Donnell assisted in the field. I thank Sean O’Donnell, Gene Robinson, Cristie Hurd and Stephanie Overmyer for helpful discussion and for constructive criticism of early drafts. Research supported by NSF # 8517519 and by the College of Agricultural and Life Sciences, University of Wisconsin, Madison.

Animal Behaviour, 52, 3

488 REFERENCES

Camazine, S. 1991. Self-organizing pattern formation on the combs of honey bee colonies. Behav. Ecol. Sociobiol., 28, 61–76. Camazine, S. 1993. The regulation of pollen foraging by honey bees—how foragers assess the colony’s need for pollen. Behav. Ecol. Sociobiol., 32, 265–272. Deneubourg, J. L., Aron, S., Goss, S. & Pasteels, J. M. 1990. The self-organizing exploratory patterns of the argentine ant. J. Ins. Behav., 3, 159–168. Forsyth, A. B. 1978. Studies on the behavioral ecology of polygynous social wasps. Ph.D. dissertation, Harvard University. Gordon, D. M. 1989. Dynamics of task switching in harvester ants. Anim. Behav., 38, 194–204. Gordon, D. M. 1995. The development of organization in an ant colony. Am. Scient., 83, 50–57. Jeanne, R. L. 1975. The adaptiveness of social wasp nest architecture. Q. Rev. Biol., 50, 267–287. Jeanne, R. L. 1981. Alarm recruitment, attack behavior, and the role of the alarm pheromone in Polybia occidentalis. Behav. Ecol. Sociobiol., 9, 143–148. Jeanne, R. L. 1986. The organization of work in Polybia occidentalis: costs and benefits of specialization in a social wasp. Behav. Ecol. Sociobiol., 19, 333–341.

Jeanne, R. L. 1987. Do water foragers pace nest construction activity in Polybia occidentalis? Experientia, 54 (supplement), 241–251. Jeanne, R. L., Downing, H. A. & Post, D. C. 1988. Age polyethism and individual variation in Polybia occidentalis, an advanced eusocial wasp. In: Interindividual Behavioral Variability in Social Insects (Ed. by R. L. Jeanne), pp. 323–357. Boulder, Colorado: Westview Press. O’Donnell, S. & Jeanne, R. L. 1990. Forager specialization and the control of nest repair in Polybia occidentalis Olivier (Hymenoptera: Vespidae). Behav. Ecol. Sociobiol., 27, 359–364. Seeley, T. D. 1989. Social foraging in honey bees: how nectar foragers assess their colony’s nutritional status. Behav. Ecol. Sociobiol., 24, 181–199. Tosi, J. A. Jr. 1969. Ecological map of Costa Rica. San José, Costa Rica: Tropical Science Center. Wenzel, J. W. 1991. Evolution of nest architecture. In: The Social Biology of Wasps (Ed. by Kenneth G. Ross & Robert W. Matthews), pp. 480–519. Ithaca, New York: Cornell University Press. Wilson, E. O. 1971. The Insect Societies. Cambridge, Massachusetts: Harvard University Press.