Behavioural Processes,31 (1994) 29-38 0 1994 Elsevier Science B.V. All rights reserved 0376-6357/94/$07.00 BEPROC

29

00511

Methodological improvements

for the study

of reconci I iation H.C. Veenema Department

*, M. Das and F. Aureli

of Ethology and Socio-ecology University of Utrecht, P. 0. Box 80086, 3508 TB Utrecht, the Netherlands

(Accepted 13 August 1993)

Abstract The present paper suggests methodological improvements for the study of reconciliation, i.e. affiliative interactions between former opponents shortly after agonistic conflicts. Three methods have been suggested to determine whether post-conflict affiliation between former opponents is higher than what would be expected by chance. Two of these methods may fail to find this higher level when the analyses are based on long-lasting observations. The third method, however, solves this potential shortcoming by identifying the ‘relevant’ duration of the observations to be considered. We also emphasize the importance of distinguishing post-conflict affiliative interactions on the basis of their timing following a conflict in order to examine their conciliatory functions. Finally we suggest a correction of the conciliatory tendency, a measure used to compare the frequency of reconciliation between dyads of individuals that may have different baseline levels of affiliation. A comparison between the original measure and the corrected one shows that only the latter is independent of the baseline level of affiliation and is, therefore, more suitable for the study of intra- and inter-specific differences in the frequency of reconciliation.

Key words: Aggression; Conciliatory tendency; Method; Reconciliation

Introduction In 1979 de Waal and van Roosmalen introduced the term ‘reconciliation’ for affiliative interactions between former opponents shortly after agonistic conflicts. Since then three

* Corresponding author. SSDl 0376.6357(93)E0041

-Y

30

aspects of these post-conflict interactions have been especially studied: 1) their occurrence relative to what would be expected by chance; 2) their function in reconciling former opponents; and 3) the factors affecting their occurrence. This paper reviews the methods used to examine these three aspects of reconciliation, points out problems of data analysis and suggests methodological improvements.

The occurrence

of post-conflict

interactions

Three methods have been used to demonstrate that post-conflict situations are characterized

by a high affiliative tendency between former opponents. The first

method (the

‘PC-MC method’) was developed by de Waal and Yoshihara (1983). Two types of observations are made. Post-conflict observations (PC) consist of a focal sample of one of the two opponents starting right after the end of an agonistic conflict. A day later a control observation of the same individual is made at the same time as the corresponding PC using the same method (matched control observation or MC). In this way multiple PC-MC pairs are collected. To determine whether conflicts lead to a change in the subsequent interactions between former opponents and whether this change involves attraction or dispersion, the timing of the first affiliative interaction between former opponents during one PC is compared with that during the corresponding MC. The analysis is done per each PC-MC pair. In polyadic conflicts, when more than one aggressor-victim dyad is present, each dyad is considered separately. This means that for a single polyadic conflict several PC-MC pairs are analyzed, one for each aggressor-victim dyad. If the first affiliative interaction between former opponents occurs only in the PC or earlier in the PC than in the MC, the PC-MC pair is said to be ‘attracted’; if the interaction takes place earlier (or only) in the MC, the PC-MC pair is said to be ‘dispersed’. When there is no interaction in either the PC or the MC, or when the interaction occurs at the same time in both, the PC-MC pair is considered ‘neutral’ (de Waal and Ren 1988). If the timing of the first affiliative interaction in the PCs is similar to that in the MCs, the number of attracted and dispersed pairs should not differ from the 1 : 1 ratio expected by chance. If the number of attracted pairs is higher than the number of dispersed pairs, we may conclude that the former opponents show higher affiliative tendency in post-conflict situations. A second method (the ‘rate method’) compares the rate of affiliative interactions between former opponents in the PCs with that during the MCs (Judge, 1991) or with that during baseline observations (de Waal, 1987). If the rate is higher in the PCs, the former opponents can be said to show higher affiliative tendency in post-conflict situations. A third method (the ‘time rule’> for demonstrating such a tendency was introduced by Aureli et al. (1989). This method compares the frequency of the first affiliative interaction between former opponents as a function of time during the PCs with the equivalent distribution during the MCs. If the frequency of such interactions is higher in the PCs the method allows one to define a time window after a conflict in which any affiliative interaction between former opponents can be defined operationally as reconciliation (after checking that the pattern is representative of the majority of individuals: Aureli and van Schai k, 1991 a>. In spite of differences in the characteristics of the control observations, the affiliative interactions to be considered as potentially conciliatory, and conflicts have been analyzed (see Kappeler and van Schaik, 1992, and discussion and suggested standardized criteria), all recent studies of Old

in the choice of in how polyadic Cords, 1993, for World monkeys

31

and apes have demonstrated

the occurrence of post-conflict

affiliation

between former

opponents using one of the three methods (for reviews see de Waal, 1989, 1993; Kappeler and van Schaik, 1992). In addition, most of these studies reported that post-conflict affiliation is especially high between former opponents who are, thus, selectively attracted to each other. The ‘PC-MC method’ and the ‘rate method’ might fail to demonstrate higher post-conflict affiliation if they are used to analyze data from long-lasting observations. The duration of the observations influences the frequency of affiliative interactions between former opponents and, therefore, the number of attracted and dispersed pairs (‘PC-MC method’) in a given situation: the longer the observations, the more attracted and dispersed pairs will be found as a result of the baseline level of interactions. In most of the studies the difference in the level of affiliation between PCs and MCs was limited to the first minutes after a conflict (reviewed in Kappeler and van Schaik, 1992). This indicates that the effect of the previous conflict on the affiliation between former opponents is limited to a short period of time. The increased number of attracted and dispersed pairs due to long-lasting observations might, therefore, result in failing to find any difference from the 1 : 1 ratio. In these cases, however, the difference between the number of attracted and dispersed pairs calculated by using only the first minutes of the observations might be large enough to demonstrate higher affiliative tendency after a conflict. A similar outcome may be obtained when the rate of affiliative interactions (‘rate method’) is calculated over a prolonged period due to long-lasting PCs and MCs. One way to avoid this problem is to check the time distribution of the first affiliative interactions in the PCs and MCs in order to identify the ‘relevant’ duration of the observations to be considered (cf. Aureli et al., 1989; Aureli and van Schaik, 1991a). Kappeler’s (1993) study of the redfronted lemurs provides an example of discrepant results using different methods. Using the ‘PC-MC method’, he failed to find higher affiliative tendency between former opponents following a conflict, even though, according to the ‘time rule’, affiliative interactions occurred significantly earlier in the PCs than in the MCs. The latter method clearly showed that the higher frequency of affiliative interactions in the PCs was limited to the first two minutes. If the ‘PC-MC method’ were applied only during the ‘relevant’ two minute period, a significantly greater number of attracted pairs would probably have been found.

The function The term ‘reconciliation’ suggests that post-conflict affiliative interactions restore the former opponents’ relationship which was disturbed by the previous conflict. The term is justified only if it can be demonstrated that the occurrence of such interactions changes the subsequent behavior of former opponents towards each other (e.g. by restoring tolerance around resources and reducing the probability of renewed attacks against the victim). To determine whether post-conflict affiliative interactions achieve these effects, it is necessary to define these interactions in such a way that allows an examination of their consequences. The ‘PC-MC method’ is not suited to investigating functional aspects (and was not introduced for this purpose). Whether or not a post-conflict affiliative interaction between former opponents will result in an attracted pair depends entirely on the timing of occurrence of an affiliative interaction in the corresponding MC, the latter being due to

32

chance. Therefore, post-conflict interactions that result in attracted pairs cannot be directly considered as potential cases of reconciliation. The ‘time rule’ is more suited to investigating functional aspects. This method allows post-conflict affiliative interactions to be defined as potentially conciliatory on the basis of their timing after a conflict. Such a definition, though rather conservative (Cords, 1993), is particularly useful when examining the function of post-conflict interactions in coping with the consequences of agonistic conflicts (e.g. tension and renewed attacks) that may depend on the time elapsed after the conflict (Aureli and van Schaik, 1991 b). A different approach was used by Cords (1992) to test the hypothesis that post-conflict affiliative interactions restore tolerance between former opponents around resources. She did so in an experimental setting where conflicts were provoked between temporarily isolated dyads of individuals. The time elapsed between the conflict and the tolerance test and between the affiliative contact (that could function as reconciliation) and the tolerance test could be controlled by the experimenter. The experimental approach has clear advantages over naturalistic observations because several variables can be controlled but manipulations may also create disturbances and, therefore, recommended (de Waal, 1992; Cords, in press).

The conciliatory

external validation is always

tendency

Even though all the post-conflict affiliative interactions between former opponents probably function as reconciliation, a measure that takes into account the baseline level of affiliation is needed for comparisons among different dyads of individuals. To compare the frequency of reconciliation between subgroups of individuals (e.g. kin or non-kin) or between species, at least four measures have been used (Kappeler and van Schaik, 1992). In this paper we focus on the measure that has been most commonly used since its introduction by de Waal and Yoshihara (1983). They defined the ‘conciliatory tendency’ between former opponents as the proportion of PC-MC pairs (attracted, dispersed and neutral pairs) that were attracted; we will refer to this measure as ‘dW&Y’s measure’. One disadvantage of this measure is that it is not independent of the duration of observations. An increase in the duration of observations will increase the chance that at least one affiliative interaction between former opponents will take place in the PC and in the MC (see above). Thus, the number of both attracted pairs and dispersed pairs will increase with the duration of observations. Since dW&Y’s measure is based only on the relative number of attracted pairs, the conciliatory tendency will be higher simply because observations last longer. A second major disadvantage of dW&Y’s measure is that it is affected by baseline levels of affiliative interactions. De Waal and Yoshihara (1983) argued that their measure had a built-in correction for baseline levels of interactions because attracted pairs are the result of a comparison between the timing of interactions in the PCs and MCs. This is only partly true, however. For example, in a hypothetical case in which the ratio of attracted to dispersed pairs does not differ from 1 : 1, the number of attracted pairs is determined only by chance and, therefore, is (statistically) equal to the number of dispersed pairs. In such a situation, however, dyads of individuals with a high baseline level of affiliation will have more attracted (and dispersed) pairs than dyads with lower baseline levels, and, thus, apparently higher conciliatory tendency. This result is clearly incorrect, since in this case the conciliatory tendency for both types of dyads should be zero. In fact, since reconcilia-

33 TABLE

1

Number

of attracted,

a given situation in a situation

dispersed

and total

(I case), in a situation

with

a lower

in the same number

post-conflict

of attracted

Number

PC-MC

pairs and two measures of conciliatory

with a higher affiliative

baseline

tendency

level of affiliative between

former

contacts

opponents

Conciliatory

of pairs

tendency

total

dW&Y

corrected

I case II case

a a+b

d d+b

t

t

a/t (a + b)/t

(a - d)/t (a - d)/t

III case

a

d+b

t

a/t

(a-d-b)/t

value.

= the measure

corrected

introduced

= the corrected

See text for definitions

tion

but resulting

measure

dispersed

b = a constant

in

pairs (III case)

attracted

dW&Y

tendency

(II case) and

measure

by de Waal and Yoshihara suggested

in the present

(1983). paper.

of the measures.

does not occur, the number of attracted pairs is just an expression of the baseline level

of affiliative interactions. By contrast, when the attracted pairs outnumber the dispersed pairs, the former can be considered to consist of two components. One component is due to chance and its magnitude depends on the baseline level of interactions and, therefore, equals (statistically) the number of dispersed pairs. The second component reflects the increase in affiliative tendency between former opponents due to the previous conflict. A measure of conciliatory tendency should express only the second component: thus, we should not take into account the total number of attracted pairs but should remove the component due to chance. This can be done by subtracting the number of dispersed pairs from the number of attracted pairs since the number of dispersed pairs estimates the number of attracted pairs that is due to chance. Thus, the conciliatory tendency should be defined as: (number

of attracted pairs - number of dispersed pairs)/total

number of pairs

Table 1 shows two examples of how our correction improves the measure of the conciliatory tendency. A comparison of the first and second cases illustrates how dW&Y’s measure and our corrected measure respond to an equal increase in the number of attracted and dispersed pairs. Case I represents the findings for a hypothetical group under certain conditions. Case II could represent 1) the same group with a higher baseline level of interactions due to different conditions, 2) a different group with the same frequency of post-conflict affiliative contacts between former opponents but a higher baseline level of interactions, or 3) the same group sampled with PCs and MCs of longer duration (see above). Since there is no difference in the actual post-conflict affiliative tendency between cases I and II, the conciliatory tendency should be the same. Our corrected measure gives equal values in the two cases, whereas dW&Y’s measure results in a higher conciliatory tendency in case II. This example shows that only the corrected measure is independent of the baseline level of interactions and solely reflects the post-conflict affiliative tendency between former opponents. This corrected measure also eliminates the problem of how the duration of the observation affects the value of the measure. In fact, subtraction of the

34

number of dispersed pairs from the number of attracted pairs removes the number of attracted pairs due to interactions occurring after the ‘relevant’ period (see above). Case III reflects a situation with a higher baseline level of interactions than in case I (as indicated by the different number of dispersed pairs). Case III does not differ from case I in the number of attracted pairs, however. This means the component that reflects post-conflict affiliation between former opponents is higher in case I because the component that depends on the baseline level of interactions (i.e. that equals (statistically) the number of dispersed pairs) is lower in case I. We expect, therefore, to find a higher conciliatory tendency in case I. The corrected measure performs as expected, whereas dW&Y’s measure does not distinguish between the two components and results in equal values in both cases. Generally, then, the corrected measure does discriminate between the two components of attracted pairs. This corrected measure allows us to compare the conciliatory tendency of different types of dyads independently of the baseline level of affiliation. Of special interest is a re-examination of published results on the effect of kinship on reconciliation since dyads of relatives usually have higher baseline level of affiliation. We were able to re-calculate the conciliatory tendency using the corrected measure for four studies. Using dW&Y’s measure, de Waal and co-workers found that kin reconciled more often than non-kin among both rhesus macaques (lumped data: 30% vs. 19%; de Waal and Yoshihara, 1983) and stumptail macaques (lumped data: 55% vs. 47%; de Waal and Ren, 1988). We could not confirm these results using the corrected measure. Among rhesus macaques only 11% of conflicts among kin were reconciled compared to 10% among non-kin (lumped data; sample size too small for statistical analysis at the individual level). Among stumptail macaques kin did not reconcile more often than non-kin (mean values: 38% vs. 37%; Wilcoxon matched pairs test: n = 8, t = 16, NS; only individuals with 5 or more PC-MC pairs were considered as in the original paper). In two other studies, the original results based on dW&Y’s measure were confirmed by our findings. York and Rowell (1988) reported a conciliatory tendency of 48% for related patas monkeys and 26% for unrelated ones (lumped data); using the corrected measure we found 29% and 18% respectively. Thus, a substantial difference between the two classes of individuals remained but we were unable to use statistical tests because of a lack of individual scores in the original paper. Aureli et al. (1989) used a different method derived from the ‘time rule’ to analyze differences among different classes of long-tailed macaques. They found that kin reconciled more often than non-kin. They also reported that dW&Y’s measure gave a very similar outcome. By using our corrected measure we could confirm the results (mean values: 40% for kin vs. 14% for non-kin; Wilcoxon matched pairs test: n = 9, t = 5, P < 0.05; only individuals with 3 or more PC-MC pairs were considered). This re-examination of published results does not give a final answer on the effect of kinship on reconciliation frequency. We need more data on species such as rhesus macaques and patas monkeys for proper statistical analysis. We also need data from more than one group per species before claiming inter-specific differences in the role of kinship. The investigation of the factors that can mediate the effects of kinship on reconciliation frequency (e.g. the value and/or the security of the social relationship and the partner accessibility) might be necessary to fully understand the phenomenon (Cords and Aureli, in press). The corrected measure is also important to explore inter-specific differences in the conciliatory tendency since baseline levels of affiliative contacts differ among species. A

35 re-analysis of the only published inter-specific

comparison that used control observations

(de Waal and Ren, 1988) confirmed the previous findings. Stumptail macaques reconciled more frequently than rhesus macaques using both dW&Y’s measure (mean values: 51% vs. 18%) and the corrected measure (mean values: 41% vs. 9%; Mann-Whitney test: nl = 19, n2 = 37, Z= 5.13, P < 0.001; only individuals with 5 or more PC-MC pairs were considered). Finally, another advantage of the corrected measure is that it enables us to compare the results of studies that used different duration of observations. We note, however, that both dW&Y’s measure and the corrected measure are sensitive to inaccuracy in the classification of neutral pairs. Two different types of neutral pairs can be distinguished: one when no interaction takes place in either the PC or the MC, another when interactions take place equally soon in the PC and the MC. The second type of neutral pairs occurs mostly because of imprecise measurement: interactions are considered to be simultaneous when they occur during the same sampling interval (usually one minute). If a shorter interval were chosen, some of these pairs would be classified as either attracted or dispersed. If the timing of interactions were recorded more precisely, we would be able to classify all the PC-MC pairs with affiliative interactions as either attracted or dispersed, leaving as neutral pairs only those with no interactions in either PC or MC. Moreover, since the number of attracted pairs is by definition higher than the number of dispersed pairs in groups that reconcile, the pairs misclassified as neutral should follow the same pattern. This means that by using a one-minute sampling interval we misclassify as neutral pairs a larger proportion of actual attracted pairs than actual dispersed pairs: therefore, we underestimate the conciliatory tendency. Fortunately, in practice this type of error is rather small. In fact, in three studies using one-minute sampling intervals the proportion of attracted and dispersed pairs misclassified as neutral pairs out of the total number of pairs was very low: 4/527 (0.8%) in long-tailed macaques (Aureli et al., 1989); II/457 (2.4%) in Barbary macaques (Aureli et al., in press); l/640 (0.2%) in Japanese macaques (Aureli et al., 1993). This shows that misclassifications of PC-MC pairs as neutral usually have a minor impact on the accuracy of the corrected measure but this effect can be minimized by choosing a shorter sampling interval.

Conclusions One of the most promising areas of the study of reconciliation is the investigation of intra- and inter-specific variation, including non-primate social species. The comparison between different studies becomes a fundamental tool for this type of research and, therefore, requires the use of similar methodology. Recent articles addressed some of the differences in the methods for the study of reconciliation (Kappeler and van Schaik, 1992; Cords, 1993). They discussed and suggested criteria for the choice of affiliative interactions to be considered as potentially conciliatory, for treating polyadic conflicts, and for collecting control observations. The present paper stresses additional aspects. First, all studies should initially demonstrate that the affiliative tendency between former opponents after a conflict is higher than what would be expected by chance. The three methods discussed in this paper are equally convenient for this purpose. A recommendation is to check the ‘relevant’ duration of observations by comparing the time distribution of the frequency of the first affiliative interactions between the PCs and MCs (cf. Aureli et al., 1989).

36

Second, there is clearly no need to investigate the conciliatory function of post-conflict affiliation in every study because such a function can be assumed on the basis of previous findings. It should be emphasized, however, that when functional aspects are investigated, the combination of an experimental approach with naturalistic observations is highly recommended. In naturalistic studies it is important to distinguish post-conflict affiliative interactions on the basis of their timing following a conflict in order to examine their conciliatory functions. Finally, the comparison of the frequency of reconciliation between dyads of individuals within and between groups and species is an area of growing interest. This paper suggests a correction for the measure of conciliatory tendency introduced by de Waal and Yoshihara (1983). The corrected measure of conciliatory tendency permits a more accurate assessment of the post-conflict affiliation solely due to the previous conflict because it is independent

of the baseline

reconciliation affiliation.

frequency

level of affiliation.

between

Therefore,

it allows

proper

dyads, groups, or species with different

comparisons

of

levels of

baseline

Acknowledgements We are deeply grateful to Marina Cords for helpful discussion manuscript at different stages. We thank Care1 van Eck and

and careful reading of the Desiree Verleur for early

discussion and Jennifer Ashley, Josep Call, Charles Chaffin, Jan van Hooff, Peter Judge, Peter Kappeler, Dario Maestripieri, Care1 van Schaik, Peter Verbeek and Frans de Waal for comments on earlier versions of the manuscript. We also thank Frans de Waal for access to his data for new calculations.

References Aureli,

F. and van Schaik,

Aureli,

F. and van Schaik, C.P.,

IVVla.

1991 b. Post-conflict

F., Das, M., Verleur,

sylvanus).

F., van Schaik, C.P. and van Hooff,

behaviour in long-tailed macaques (Macaca

J.A.R.A.M., 1989.

fascicularis).

Post-conflict

Functional

aspects of reconciliation

Am. J. Primatol.,

F., Veenema, H.C., van Panthaleon van Eck, C. and van Hooff, J.A.R.A.M.,

Cords, M., 1992.

Post-conflict

reunions

and reconciliation

social interactions

in press.

tion, consolation and redirection in Japanese macaques (Macaca 44:

14.

1993.

Int. J. Primatol.,

J.A.R.A.M.,

among captive long-tailed macaques (Macaca Aureli,

behaviour in long-tailed macaques (Macaca

Ethology, 89: 101-l

D. and van Hooff,

among Barbary macaques (Macaca Aureli,

Post-conflict

II Coping with the uncertainty.

fascicularis): Aureli,

C.P.,

I The social events. Ethology, 89: 89-100.

fascicdaris):

fuscata).

19: 39-51. 1993.

Behaviour,

Reconcilia124:

in long-tailed macaques. Anim.

I-21. Behav.,

57-61.

Cords, M., 1993.

On operationally

Cords, M., 1993.

Experimental

Primatol.,

Pereira F.B.M.,

F., 1993. Patterns of reconciliation

and L.A.

Behavior, Oxford de Waal,

Am. J. PrimatoJ., 29: 255-267.

in press.

Cords, M. and Aureli, M.E.

defining reconciliation.

approaches to the study of primate conflict resolution (abstract). Int. J.

Fairbanks

University

1987.

Press, New York,

Tension

bonobos (Pan paniscus).

(Editors),

Juvenile

among juvenile Primates:

long-tailed macaques. In:

Life History,

Development,

and

in press.

regulation and nonreproductive

Nat. Geogr. Res., 3: 318-335.

functions

of sex among captive

37 de Waal,

F.B.M.,

1989.

Peacemaking among Primates.

Harvard

University

Press, Cambridge, MA,

294 pp. de Waal, F.B.M.,

1992.

Complementary

social cognition. Behaviour, de Waal,

F.B.M.,

unresolved

1993.

issues.

methods and convergent evidence in the study of primate

297-320.

Reconciliation

In: W.A.

Press, New York, pp. 11 I-l de Waal, F.B.M.

118:

among primates:

Mason and S.P.

and Ren, R.M.,

Behaviour,

Kappeler,

SUNY

1988.

Comparison

of the reconciliation

behavior of stumptail

and

42.

5: 55-66.

and Yoshihara, D., 1983.

Reconciliation and redirected affection in rhesus monkeys.

85: 224-241.

Judge, P.J., 1991. Primatol.,

evidence and

and van Roosmalen, A., 1979. Reconciliation and consolation among chimpanzees.

Behav. Ecol. Sociobiol., de Waal, F.B.M.

of empirical

Primate Social Conflict,

44.

rhesus macaques. Ethology, 78: 129-l de Waal, F.B.M.

a review

Mendoza (Editors),

Dyadic and triadic reconciliation

in pigtail macaques (Macaca nemestrina).

Am. J.

23: 225-237.

P.M,

1993.

redfronted (Eulemur

Reconciliation fulvus rufus)

and post-conflict lemurs. Anim.

Kappeler, P.M. and van Schaik C.P., 1992.

behaviour

in ringtailed (Lemur

catta) and

Behav., 45: 901-915.

Methodological and evolutionary

aspects of reconciliation

in primates. Ethology, 92: 51-69. York, A.D. and Rowell, T.E., bus patas. Anim.

1988.

Reconciliation following

Behav., 36: 502-509.

aggression in patas monkeys,

Erythroce-

Methodological improvements for the study of reconciliation.

The present paper suggests methodological improvements for the study of reconciliation, i.e. affiliative interactions between former opponents shortly...
722KB Sizes 1 Downloads 3 Views