The Psychological Record, 1975,25, 573-581.
MEASURING DOMINANCE IN RATS DAVID R. DREWS and FRED H. WULCZYN' Juniata College
The social behavior of a colony of 9 male albino rats was recorded over a 5-weekperiod. These observations were followed by a series of paired food competitions which yielded a standard laboratory measure of dominance. Each of these measures was correlated with every other measure. Factor analysis of the resulting correlation matrix revealed a cluster of closely related behaviors, including fighting, boxing, grooming, mounting, and total interactions. Food competition was not significantly related to any of the behavioral indices. These findings do provide a basis for a unidimensional dominance concept of some descriptive breadth and also raise questions concerning the relation between measures obtained via seminaturalistic observation vs those typically used in laboratory studies of dominance.
The present study was performed primarily in an attempt to help clarify some of the increasingly problematic aspects of the dominance concept. Although the notion of dominance has been used by a broad spectrum of students of animal behavior as both a descriptive and an explanatory tool, recent writers have begun to question its utility (Gartlan, 1968; Van Krevald. 1970; CrOOk, 1970). Much of this criticism has been broadly directed at definitional problems, most often focusing on the difficulties involved in presenting a precise definition. These problems have been exacerbated by the wide variety of measurement techniques used to assess dominance. For example, laboratory psychologists have emphasized priority of access measures, most often using pairs of animals in situation where they are required to compete for a limited amount of food (e.g., Uyeno, 1960), for the opportunity to perform a previously rewarded operant response (e.g., Monroe, 1970), for safety from a strong electric shock (e.g., Hamilton, 1960), or for escape from an underwater tube (e.g., Howells & Kise, 1974). On the other hand, those working in naturalistic or seminaturalistic settings have often de-emphasized these kinds of priority of access measures in favor of observational measures of agonistic, sexual, grooming, or locomotor behaviors (e.g., Bernstein, 1969; Barnett, 1963). A broad variety of approaches to the measurement of dominance is valuable in the event that different measures show some relation to one another. If this is true, then multiple measures add, at the very least, considerable descriptive breadth. However, if different , We would like to thank Paul W. Sheppard and Pauline Masters for their help with the factor analysis. We also wish to thank Michael Harrison for his helpful criticisms of an earlier draft. Reprint requests should be addressed to David Drews, Psychology Department, Juniata College, Huntingdon, Pennsylvania 16652.
574
DREWS AND WULCZYN
measures fail to correlate, questions may be raised concerning either the utility of particular measures or, more importantly, whether one ought to forego altogether the use of dominance as a descriptive tool. Concern with this issue has grown recently, but a clear picture is yet to emerge from existing research. Thus while food and water competition measures may correlate (Baenninger, 1970), Syme, Pollard, Syme, & Reid (1974) have suggested that these correlations depend heavily on the details of measurement techniques. Correlations between various kinds of tube tests and food competition generally appear to be minimal (e.g., Lindzey, Manosevitz, & Winston, 1966; Howells & Kise, 1974), possibly because of topographic differences in the responses required by each situation. A more detailed account of the similar problems associated with competitive measures of dominance may be found in Syme (1975). In addition, correlations between limited access or competitive measures and agonistic behavior have been found by Lindzey et al. (1966) and by Ruskin and Corman (1971), but not by Baenninger (1970). In general, then, strong evidence favoring a unidimensional dominance concept has not been forthcoming, at least not from the laboratory situations most often studied by psychologists. While ethologists are able to provide some evidence for such a concept, very little ofit is concerned with rats. And although detailed descriptions of various aspects of the rat's social behavior are available and generally encouraging (e.g., Barnett, 1963; Seward, 1945; Grant, 1963), the primary intent of this research has not been to present evidence for a cluster of related behaviors which might profitably be given the dominance label. Of particular importance is some attempt to assess the relationships between laboratory measures of dominance and measures taken from naturalistic and seminaturalistic observations. Without some such tangible interface, laboratory and field students of animal behavior might very well be talking about different processes when they refer to dominance. Yet, except for those studies using some index of agonistic behavior (e.g., Baenninger, 1970), little effort has been devoted to assessing this particular set of potential relationships. The present study, therefore, represents an attempt to gather observational data which might provide evidence for a unidimensional dominance concept in the case of rats. It also represents an attempt to assess the relationships between such measures and a standard laboratory measure of dominance, food competition. Subjects
METHOD
Nine male, albino rats of the Wistar strain, 16-17 weeks old on arrival, were used. After being housed individually in standard wire cages for 2 weeks, each animal was weighed and marked for individual identification with gentian violet stain before being placed in the group environment.
DOMINANCE IN RATS
575
Apparatus
During the observational phase of the study, all subjects were h'lused in a circular enclosure constructed of corrugated aluminum Wl.1Ch was placed on a hardware cloth floor. This enclosure was approximately 173 cm in diameter, with 92 cm high walls. Its interior was divided into quadrants by means of a 7.5 cm high, unpainted wooden partition. An additional 7.5 cm partition was added to this to isolate the single quadrant in which food was located. This partition extended to within 10 cm of the walls of the enclosure. Both food and water were available on an ad lib basis. Water could be obtained in three quadrants from bottles whose tubes protruded through the walls of the enclosure. In one of these quadrants a food bin, constructed of hardware cloth and measuring approximately 30 cm x 15 cm x 10 cm was attached to the wall. Standard food pellets could be obtained either by gnawing through the sides of the food bin or by removing them from its open top. A standard 24-hour timer was used to control a 12/12 light dark cycle. During the light part of that cycle (0700-1900 hours), illumination was provided by a single 100-watt light bulb. Observations, which were done solely during the dark part of the cycle, were aided by two 25-watt red lights. For the food competition tests, a 46 cm x 46 cm x 20 cm arena, constructed of plywood on three sides with corrugated aluminum for the fourth side, was used. A 3.8 cm diameter hole, located 3.8 cm from the floor in the center of the long plywood wall, provided access to a food trough containing a sweetened, wet mash. Activities Recorded
Based on the available literature and earlier lab work, several behaviors were recorded. Fightingwas recorded whenever two rats engaged in what might best be described as "vigorous wrestling." This typically involved rolling and tumbling around at speeds too rapid to permit identification of more discrete behaviors without the aid of high-speed cameras. Animals may hold onto one another, and bites may be delivered. Typically, fights ended when one animal rolled onto his back and assumed a submissive position, although this was not always the case. While this category approximates the descriptions of Barnett (1963) and Calhoun (1963), it is probably broader. No attempt was made to differentiate fighting from what Calhoun (1963) calls "mock-combat" and forms of threat behavior described by Seward (1945) and Barnett (1963) were not observed systematically. Winners and losers were distinguished on the basis of the submissive posture noted above or on the basis of escape behavior. In several instances where winners could not be determined, the behavior was not recorded. Boxingwas recorded whenever two animals stood on their hind legs and slapped one another or "boxed" with their paws. This activity might last several seconds and could lead to pinning,
576
DREWS AND WULCZYN
grooming, mounting, or behaviors not recorded in the present study. Often the winner and loser of such an encounter could be determined on the basis of escape beha vior or by noting the first animal to break off contact. In anum ber of cases, however, boxing led directly to pinning (see below), and this measure had to be used to determine the winner and loser. Pinning refers to those situations in which one rat rolled onto his back, generally remaining motionless, while the other placed his forepaws and often his head on the submitting rat's stomach. This behavior pattern appears to be identical to what Baenninger (1970) has called "spontaneous competition." For analysis purposes, the animal who was on its back was defined as the loser. Grooming was recorded whenever one rats was observed to engage in nibbling or licking the fur of another's head, neck, or back. The identity of both the groomer and the rat being groomed was noted. Self-grooming was not recorded. Mounting was recorded when one animal approached another from behind and grasped him with forepaws placed behind his hind legs. In each case the identity of both the mounter and the animal mounted were noted. While Grant (1963) regards this as homosexual behavior, his data also suggest that mounting might be related to agonistic behavior. In addition to the measures already noted, an index of total interaction was calculated for each animal by cumulating all of the instances in which it was observed to participate in any of the forms of interaction noted above. This was done in an attempt to create a measure related to Calhoun's (1967) concept of "velocity." For purposes of analysis, two weight measures, weight at the end of the experiment and percentage of weight gain from the beginning to the end, were also noted. Procedure
The first 2 weeks following introduction of the animals into the circular enclosure were used to train the primary observer. A second observer, used solely to assess the reliability of the scoring system, was also trained at this time. Periodic reliability checks, conducted at a number of points during the study, yielded satisfactory measures. Data were gathered 5 nights a week for the next 5 weeks. Each night two I-hour observation periods were run. All observations were conducted between the hours of 1900-0700. Specific hours were chosen at random with the restriction that observation periods could not run consecutively. A second restriction equalized the number of times a particular hour was observed over the course of the study. During each hour the observer recorded the frequency of each of the activities noted above, including the identity of the participants and the outcome of each encounter. Because one night was missed, the present study represents a total of 48 hours of observation. At the conclusion of the observational phase of the study, food competition tests were run. Prior to the actual competition sessions,
DOMINANCE IN RATS
577
all animals were deprived offood for 24 hours. Each subject was then individually placed in the arena until he was observed to feed for 60 seconds. Following an additional 12-hour deprivation period, this procedure was repeated. For the actual tests nine of the 36 possible dyads were run each night, with competitions beginning at 2200 hours. To minimize the possibility of satiation, no animal was run twice in successioll. These tests continued for a period of 8 days until each dyad had been run twice. During each 3-minute test, the amount of time each animal controlled access to the food trough was measured. Individual times were then cumulated to provide a composite measure. RESULTS For each behavioral category total scores for individual animals were used to calculate Pearson Product Moment correlations, each measure being correlated with every other measure. The resulting correlation matrix appears in Table 1. Each correlation was then subjected to a test of significance following Kurtz (1965). Reliability coefficients for each measure were calculated by regressing data from the first half of the study on second-half data. These coefficients appear in parentheses. Finally, total observed frequencies for each measure appear in the bottom row of Table 1. Examination of the reliability coefficients indicates that most of the measures were satisfactorily reliable from the first to the second half of the study. The most notable exceptions are the measures of fighting wins and losses. This lack of reliability may be attributed to the fact that most of the fights (14 of 17) occurred during the first half of the study. Boxing losses and grooming also fail to reach satisfactory levels of reliability. Inasmuch as boxing wins and being groomed are highly reliable, it appears that the same animals are winning boxing encounters and being groomed in each half of the study but that these animals are interacting with different animals over time. With regard to the correlations, it is particularly notable that the one standard laboratory measure of dominance, food competition, is not significantly related to any of the observational measures. The only measure that does correlate significantly with food competition is weight at the end of the experiment. But while food competition is not related to any of the observational measures, it is clear that several of these measures are closely related to one another. These relations were pursued by means of factor analyses, using a principal-factors solution with orthogonal rotations. Several analyses were done, using between two and five factors. These analyses always revealed one stable factor, which accounted for 38.6% of the variance, and a second moderately stable factor. Taken together, these accounted for 61 % of the variance. The two-factor solution is presented in Table 2. Following Kerlinger's (1973)
2 .186 (.229)
17
17
r....:l
:§~
1 (.055)
··p<.05 a p between .10 and .05
1 2 3 4 5 6 7 8 9 10 11 12 13 14 Total frequency
I"
!~
I::
61
3 .774* .309 (.915)
~~
~"
61
4 .755* .518 .705* (.516)
I! ~.3
408
5 .401 -.226 .292 .273 (.696)
0
..:l
408
6 -.226 -.448 .114 -.367 .326 (.798)
•
f~ f~ 8 .637a -.017 .751.425 .008 .228 -.292 (.770)
604
604
"
~
E
'g
.068 -.442 ··.282 -.084 .713* .244 (4.75)
"7
...~
...
..
TABLE 1 Overall Correlations Between Measures
210
II
.801* .329 .958* .762* .324 -.038 -.180 .663 a (.925)
0
~
::"!!
...
g
210
10 .005 .336 -.067 .277 -.719* -.390 -.398 .186 -.053 (.604)
::"!!
.B
."
11 .417 -.339 .161 .002 -.182 -.293 .038 .201 .297 .173
r..u
8 o! f ..
6
'l:I 'l:I
2600
12 .852* .069 .869* .657 a .545 .239 .183 .682.866-.129 .296
fo<,s
30:1e
u
6 'l:I
13 .359 .437 .468 .466 .235 -.207 -.372 .252 .381 -.280 -.471 .168
&:~
...~j..
j~ C ...
'0
.451 -.142 .517 .102 -.358 -.003 -.491 .712.465 .285 .673li .410 -.108
14-
."j &i;
21
N
n
~t"'
I:'
~ ~
Ii!tIj
~ at
DOMINANCE IN RATS
579
tentative suggestion that loadings comparable to significant correlations be regarded as significant, all measures with loadings over .665 are noted with an asterisk. TABLE 2 Factor Loadings on All Measures Measure Fighting wins Fighting losses Boxing wins Boxing losses Pinning wins Pinning losses Groomer Groomed Mounter Mounted Food Competition Time Total interaction Percentage of weight gain End weight
Rotated 1
.899* .235 .959* .778* .398 .021 -.109 .780* .954* -.059 .240 .894* .448 .480
Factors 2
.028 .498 .067 .179 -.880* -.591 -.796* .198 .075 .792* .312 -.327 .007 .501
* Significant loadings>. 665
Factor 1 has high loadings on fighting wins, boxing wins and losses, being groomed, mounting, and total interactions. Since the dominance concept originated in reference to agonistic behavior (Schjelderup- Ebbe, 1922), it is reasonable to call Factor 1 dominance. It is interesting that pinning has only a moderate loading on this factor. Since pinning was used occasionally to identify winners of fights and boxing encounters, a higher, and possibly artifactual, loading might have been expected. Factor 2 is bipolar with strong negative loadings on pinning wins and grooming and a high positive loading on being mounted. Interpretation of this factor is less clear, especially since solutions with more than two factors reduce considerably the loadings on being mounted. One therefore hesitates to pursue this factor. The very low reliability of fighting is somewhat difficult to reconcile with its high loading on the dominance factor. For this reason, the first and second halves of the study were factor-analyzed separately. These analyses revealed that both fighting wins and being groomed loaded heavily on Factor 1 in the first but not the second half of the study. The other measures with high loadings on Factor 1 in Table 2 had high loadings in both halves of the experiment. DISCUSSION In general, the results of the factor analysis do provide support for the notion of a unidimensional dominance concept. In laboratory rats dominant animals win fights, mount other rats, are groomed, and engage in a good deal of boxing, although the outcome of boxing encounters appears to be relatively unimportant. These animals are
580
DREWS AND WULCZYN
also more likely to engage in all forms of social behavior. This particular finding is of some interest, since Baenninger (1960) has indicated that total interaction is positively related to dominance, and Calhoun (1967) has made the same point even though his measure of "velocity" is operationally different from the present measure of total interaction. Confidence that this measure may provide a useful, easily obtainable index of dominance is bolstered by the fact that total interactions are not significantly correlated with any of the behaviors outside of the dominance factor. The above description of dominant behaviors requires some qualification, since that cluster of behaviors undergoes some change over time. In particular, a large majority of the fights observed occurred during the first half of the study. This finding is consistent with observations that fighting is more probable in the earlier stages of group formation and that, as dominant-subordinate relationships are formed, fighting declines (Scott & Frederickson, 1951). Subsequently, these relationships may be maintained with "lower intensity" behaviors, such as boxing and mounting. The high reliabilities of boxing wins and mounting are compatible with this account. As in the case of fighting wins, being groomed has a high loading on the dominance factor only during the first half of the study. Why this should happen is unclear, especially since this behavior is reliable and shows only a slight decline in frequency over time. The reduced loadings of fighting wins and being groomed in the second half of the study should not be allowed to obscure the fact that a cluster of closely related behaviors exists that may provide the basis for a dominance concept of some descriptive breadth. Of equal importance is the fact that none of the behaviors in this cluster are related to the authors' measure of food competition. This finding replicates and extends Baenninger's (1970) report that pinning and success in food competition are unrelated. While Ruskin and Corman (1971) have found such a relationship, it should be noted that their measures of both food competition and agonistic behavior were taken atthe same time. Using mice, however, Lindzey et al. (1966) found agonistic behavior to be related to food competition when these measures were taken independently. Both species differences and less obvious procedural variations may account for this discrepancy. It is possible to offer a number of explanations for the lack of a relationship between the observational indices of dominance and food competition. Syme (1974), in an excellent review of competitive dominance measures, indicates that they often fail to correlate with one another. This raises the possibility that the present measure of food competition was simply an unfortunate choice and that some of the other commonly used priority of access measures might be highly related to our dominance factor. Such an outcome would provide a laboratory measure with considerably greater predictive power than the present measure offood competition. More importantly, it would provide the kind of linkage which appears crucial to any attempt to
DOMINANCE IN RATS
581
combine the findings and principles drawn from field and laboratory studies of dominance. Alternatively, it is possible that such correlations would not obtain and that field and laboratory studies of dominance are examllllllg different and unrelated processes. Although unfortunate, this knowledge would obviously be important. Clearly, more effort toward assessing these relationships and finding laboratory measures with wider generality is needed.
REFERENCES BAENNINGER, L. P. 1966. The reliability of dominance orders in rats. Animal Beha vior, 14, 367 -371. BAENNINGER, L. P. 1970. Social dominance orders in the rat: "Spontaneous," food, and water competition. Journal of Comparative and Physiological Psychology, 71, 202-209. BARNETT, S. A. 1963. The rat: A study in behavior. Chicago: Aldine. BERNSTEIN, I. S. 1969. Stability of the status hierarchy in a pigtail monkey group (Maca nemistrina). Animal Behavior, 17, 452-458. CALHOUN, J. B. 1963. The ecology and sociology of the Norway rat. Bethesda: U.S. Department of Health, Education & Welfare. CALHOUN, J. B.1967. Ecologicaifactors in the development of behavioral anomalies. In J. Zubin (Ed.), Comparative Psychopathology. New York: Grune & Stratton. CROOK, J. H. 1970. Social organization and the environment: Aspects of contemporary social ethology. Animal Behavior, 18, 197-209. GARTLAN, J. S. 1968. Structure and function in primate society. Folia Primatologica, 8, 89120. GRANT, E. C.1963. An analysis of the social behavior of the male laboratory rat. Behavior, 21, 260-281. HAMILTON, C. L. 1960. Comparison of two methods of dominance testing in the monkey. Psychological Reports, 6, 247-250. HOWELLS, G. N., & KISE, G. 1974. The measurement of social dominance in rats. The Psychological Record, 24, 101-108. KERLINGER, F. N. 1973. Foundations of Behavioral Research (2nd ed.) New York: Holt, Rinehart & Winston. KURTZ, K. H. 1965. Foundations of Psychological Research. Boston: Allyn & Bacon. LINDZEY, G., MANOSEVITZ, M., & WINSTON, H. 1966. Social dominance in the mouse. Psychonomic Science, 5, 451-452. MONROE, B. D. 1970. Influence of successful and unsuccessful social experiences on dominant-subordinate behavior of the rat. Developmental Psychology, 2, 359-363. RUSKIN, R. S., & CORMAN, C. D.1971. A multivariate study of competition in a free-operant situation. Psychonomic Science, 23, 251-252. SCHJELDERUP-EBBE, T. 1922. Beitrage zur Biologie und Social-und Individual-psychologie bei Gallus domesticus. Griefswald, Germany: Adler. SCOTT, J. P. & FREDERICKSON, E. 1951. The causes of fighting in mice and rats. PhysiolOgical Zoology, 24, 273-309. SEWARD, J. P. 1945. Aggressive behavior in the rat. I. General characteristics: Age and sex differences. Journal of Comparative Psychology, 38, 175-197. SYME, G. J. 1974. Competitive orders as measures of social dominance. Animal Behavior, 22, 931-940. SYME, G. J., POLLARD, J. S., SYME, L. A., & REID, R. M. 1974. An analysis of the limited access measure of social dominance in rats. Animal Behavior, 22, 486-500. UYENO, E. T. 1960. Hereditary and environmental aspects of dominance behavior in the albino rat. Journal of Comparative and Physiological Psychology, 53, 138-141. VAN KREVALD, D. 1970. A selective review of dominance-subordination relations in animals. Genetic Psychology Monographs, 81, 143-173.