Behavioral Ecology and Sociobiology

, Volume 61, Issue 9, pp 1449–1457

“Selfish worker policing” controls reproduction in a Temnothorax ant


    • Lehrstuhl Biologie IUniversität Regensburg
    • Ecole Normale Supérieure
  • Elisabeth Brunner
    • Lehrstuhl Biologie IUniversität Regensburg
  • Jürgen Heinze
    • Lehrstuhl Biologie IUniversität Regensburg
Original Paper

DOI: 10.1007/s00265-007-0377-3

Cite this article as:
Stroeymeyt, N., Brunner, E. & Heinze, J. Behav Ecol Sociobiol (2007) 61: 1449. doi:10.1007/s00265-007-0377-3


Animal societies, including those of humans, are under constant threat by selfish individuals, who attempt to enforce their own interests at the cost of the group. In the societies of bees, wasps, and ants, such individual selfishness can be prevented by “policing,” whereby workers or queens impede the reproduction of other individuals by aggression, immobilization, or egg eating. In this study, we report on a particular kind of reproduction control in the ant Temnothorax unifasciatus, which can be considered as a selfish act itself. We experimentally induced workers to lay eggs by dividing several colonies into two halves, one with and one without a queen. In queenless colonies, workers established rank orders by aggression and several top-ranking workers started to reproduce. Upon reunification, egg-laying workers mostly stopped behaving aggressively. They were neither attacked by the queen nor by random workers, but instead received infrequent, nondestructive, targeted aggression from a few workers, most of which became fertile when the queen was later removed. The introduction of differentially stained worker-laid and queen-laid eggs in queenright fragments did not lead to a selective removal of worker-laid eggs. Hence, there appears to be no collective worker policing in T. unifasciatus. Instead, reproduction appears to be controlled mostly through a few attacks from high-ranking workers, which, in this way, might attempt to selfishly increase their chances of future reproduction.


Kin conflictWorker policingDominanceTemnothorax unifasciatus


Insect societies have long fascinated both researchers and amateurs because all individuals seem to cooperate harmoniously to increase the reproductive output of the society as a whole. However, as group members are usually not genetically identical, their reproductive interests may differ and intra-group conflicts arise at many levels (Heinze 2004; Bourke 2005; Ratnieks et al. 2006).

A prominent conflict in insect societies concerns the origin of males. Workers of most social Hymenoptera (bees, ants, and wasps) have functional ovaries and are able to produce males from viable, haploid eggs. Individuals can increase their direct fitness by selfishly laying eggs but, in this way, might disturb group-level performance and decrease the average inclusive fitness of their nestmates. Social insects have evolved several methods of “policing” to prevent such exploitation by egoistic group members (Ratnieks 1988; Ratnieks and Reeve 1992; Bourke and Franks 1995; Monnin and Ratnieks 2001; Ratnieks and Wenseleers 2005; Ratnieks et al. 2006).

Relatedness theory provides a theoretical framework to predict the occurrence of policing in insect societies. For example, workers are more closely related to their own sons (relatedness coefficient: r = 0.5) than to the males produced by the queen (r = 0.25) or other workers (r ≤ 0.375), and conflict over male parentage among workers and between queen and workers is, thus, expected (Ratnieks 1988; Ratnieks and Reeve 1992; Bourke and Franks 1995; Monnin and Ratnieks 2001; Heinze 2004; Ratnieks et al. 2006; Wenseleers and Ratnieks 2006). When the queen mates multiply (polyandry) or when a colony contains multiple related queens (polygyny), workers are more closely related to queen-produced males than to average worker-produced males. They would, therefore, benefit from taking sides with the queen and jointly preventing other workers from selfish egg-laying by the means of “worker policing” (Ratnieks 1988; Pamilo 1991). By contrast, in a colony with a single, singly-mated queen (monogyny and monandry), workers are more closely related to worker-produced males than to queen-produced males (Ratnieks 1988). In such a situation, worker policing is not expected, unless it benefits individuals in other ways, such as increasing colony efficiency (Ratnieks 1988) or biasing the colony sex ratio in favor of females, close to the workers’ optimum (Foster and Ratnieks 2001).

Workers of many eusocial insect species have been shown to prevent other workers from producing males in the presence of a fertile queen by two main mechanisms (Ratnieks et al. 2006; Wenseleers and Ratnieks 2006): eating worker-laid eggs (ants: Kikuta and Tsuji 1999; D’Ettorre et al. 2004; Endler et al. 2004; Helanterä and Sundström 2005; bees: Ratnieks and Visscher 1989; Visscher 1996; wasps: Foster et al. 2002; Saigo and Tsuchida 2004; Wenseleers et al. 2005) and attacking ovary-developed workers (ants: Gobin et al. 1999; Kikuta and Tsuji 1999; Liebig et al. 1999; Hartmann et al. 2003; Iwanishi et al. 2003; bees: Visscher and Dukas 1995). Such “worker policing” is traditionally thought to increase the policing workers’ indirect fitness by favoring the rearing of more closely related males or by increasing colony efficiency and sexual output (Monnin and Ratnieks 2001). However, it was rarely studied whether their direct fitness might be affected as well.

Worker policing is generally thought to benefit the whole colony. All workers are, therefore, expected to police selfish individuals. However, the identity and social status of policing workers have rarely been studied, and it is unclear whether all or only particular subsets of workers really take part in the control of reproduction.

In an attempt to clarify these points, we studied the model ant species Temnothorax unifasciatus, a monogynous and monandrous ant with small colonies (Heinze et al. 1997). In contradiction with relatedness predictions, workers in natural queenright colonies have a very low share in male production (less than 3%) and produce mostly trophic eggs (Heinze et al. 1997; Hammond and Keller 2004). This low level of worker reproduction could be due to worker policing. In this study, we investigate the occurrence of worker policing in T. unifasciatus, using individually marked workers to determine the identity and status of policing individuals. We first show that neither queens nor workers preferentially destroy worker-laid eggs. We then show that only a few workers attack egg-layers, and that these workers become fertile themselves when the queen is removed. Worker attacks in T. unifasciatus, therefore, appear in part to increase an attacker’s potential direct fitness and can be considered as selfish acts.

Materials and methods

Collection and housing of colonies

Colonies of T. unifasciatus (Latreille 1798) were collected in sun-exposed rock crevices in different sites in Central and Southern Europe (see below). During the time of the experiments, all colonies were kept in small plastic boxes (10 × 10 × 3 cm3) with a regularly moistened plaster floor in incubators under artificial summer conditions (12 h light 28°C; 12 h dark 17°C; Buschinger 1974; Heinze and Ortius 1991) and were provided three times a week with water, honey, and pieces of cockroaches or tuna.

Policing by egg eating

Worker-produced males are normally extremely rare in colonies of T. unifasciatus with a fertile queen (Heinze et al. 1997; Hammond and Keller 2004). To determine whether worker reproduction in queenright colonies is prevented by the selective removal of worker-laid eggs by workers or by the queen, we followed an experimental setup previously used in honeybees (Ratnieks and Visscher 1989), wasps (Foster et al. 2002), and ants (D’Ettorre et al. 2004; Endler et al. 2004) by introducing non-nestmate queen-laid and worker-laid eggs into queenright and queenless discriminator colonies.

Ten queenright colonies (E1-E5, collected in September 2004 in Waldenhausen, Baden-Württemberg, Germany; 12 to 81 workers, mean 49.6 ± SD 25.1; E6-E10, collected in April 2005 in Schönhofen, Bavaria, Germany; 150 to 230 workers; mean 195 ± SD 33.2) were each divided into two fragments, one with the queen (queenright half) and one without (queenless half). Workers and brood were equally distributed among both fragments. As it has been previously shown in many Temnothorax ants (Heinze et al. 1997), workers in queenless colony halves started to lay eggs about two weeks after colony splitting. To be able to distinguish worker- from queen-laid eggs, we stained some of the eggs by smearing tuna oil either neat or mixed with red or black fat-soluble dyes (Sudan IV and Sudan black B; Heinze et al. 1995) once onto the heads of workers and repeatedly onto the heads of queens and by, thereafter, providing individuals only with tuna impregnated with unstained or stained oil as protein source. We, therefore, obtained three kinds of eggs: white, red, and blue eggs (see Fig. 1). Two different treatments were assigned to each mother colony—one color for all individuals of the queenless colony part and another one for the queen—and the color combinations were varied as much as possible. We simultaneously introduced equal numbers of queen-laid and worker-laid eggs from the queenright and queenless halves of the same source colony into a queenright or queenless discriminator colony. Colonies were chosen in a way that the three kinds of eggs (introduced queen-laid eggs, introduced worker-laid eggs, and eggs produced by the discriminator colony) had different colors. Egg survival was then monitored daily for four days. Food shortage might lead to increased egg eating. To determine whether worker- and queen-laid eggs are more strongly discriminated under such conditions, we starved some discriminator colonies for five days before introducing the eggs. Normally fed discriminator colonies received eight eggs of each kind in each trial, whereas starved discriminator colonies only received six eggs of each kind in each trial because of a shortage in available worker-laid eggs. We applied Gehan’s Wilcoxon test to every discriminator colony to compare the survival of queen-laid and worker-laid eggs in each trial, and we used a generalized linear model with binomial error structure applied to the pooled data to determine whether egg survival on the fourth day after introduction was influenced by the following factors: identity of egg-layers (queens or workers), nature of discriminator colonies (queenright or queenless), eggs’ color, starvation, and identity of source and discriminator colonies.
Fig. 1

Blue, red, and unstained worker-laid eggs obtained two weeks after smearing individuals’ head and feeding them with fat soluble dyes (respectively, Sudan black B, Sudan IV, and no dye) dissolved in tuna oil

Policing by aggression

To determine whether worker reproduction is prevented by worker or queen policing through aggressive behavior, we applied a standard experimental design (Kikuta and Tsuji 1999; Liebig et al. 1999; Hartmann et al. 2003; Iwanishi et al. 2003) by splitting colonies into queenright and queenless fragments until workers in queenless fragments started laying eggs, and by thereafter reuniting both fragments and monitoring aggressive interactions. One might argue that this procedure is highly artificial. However, colonies of Temnothorax ants often occupy fragile nest sites and are known to move frequently (Möglich 1978; Foitzik and Heinze 1998). During relocations, colonies may temporarily split and later fuse again (Mallon et al. 2001; Pratt et al. 2002; Franks et al. 2003; Pratt 2005; Pratt et al. 2005).

For this experiment, we collected five colonies near Mont Ventoux, Provence, France, in July 2003 (A1–A5, 43 to 61 workers; mean 53 ± SD 7.2). We individually marked all workers in these colonies by tying 19- to 24-μm thin tungsten or copper wires (courtesy of Elektrisola, Eckenhagen) between their segments and, thereafter, divided each colony into queenright and queenless halves. Workers in queenless colony fragments are known to establish near-linear dominance hierarchies by violent antennal boxing and biting, and only the top-ranking workers usually start to lay haploid eggs within a few weeks after colony splitting (Heinze et al. 1997).

We identified top-ranking workers by observing worker behavior in queenless colony fragments. Each colony fragment was repeatedly observed for 30-min periods resulting in a total observation time of 10 h per colony spread over several weeks. We focused on the nest area where the brood and most workers were gathered, and we recorded every of the following actions occurring in this area: aggressive interactions (violent antennal boxing, biting, and immobilization), trophallaxis, brood care, and self- and allo-grooming. The more aggressive individuals were, the higher their assumed rank in the colony hierarchy. At the end of the experiment, we checked whether workers’ rank in the colony hierarchy was correlated with their reproductive status by dissecting the ovaries of all workers (Buschinger and Alloway 1978), measuring ovariole length and counting the total number of mature and developing oocytes. For statistical analyses, we focused on those workers that were involved each in more than 1% of the total aggressive acts (“dominant workers”; A1, A2, and A4: four workers each, A3: seven workers; A5: five workers). To reach the sample size required for Wilcoxon matched-pairs tests, we included the next most aggressive worker from colonies A1, A2, and A4 in the analysis.

After two (colony A1) or four weeks (colonies A2–A5) of observation, i.e., six or fourteen to twenty-one days after first worker-laid eggs had appeared in the queenless halves, we reunited the colony fragments and observed each colony for a total of 8 h spread over nine days, recording the same behaviors as mentioned above. Once the workers that behaved aggressively towards high-ranking workers were identified, we removed the queens from the reunited colonies and again observed hierarchy formation and other behaviors in each “orphaned colony” for 5 h spread over five days, now involving all the colony’s workers. The total observation period of orphaned colonies was limited to five days because hierarchies in queenless fragments usually reach a stable state within a few days. In one colony (A1), several workers, including the dominants, left the colony a few days after reunion and shortly thereafter died of a fungal disease that had infected most of the colony. This colony was excluded from the subsequent experiments and analyses.

We analyzed the data by comparing the behavior of different categories of workers in each colony (e.g., dominant workers vs other workers) using chi-square or Fisher’s exact tests computed by Statistica 5.0 and two-sample permutations test computed by the program Rundom Projects 2.01 Lite (Jadwiszczak 2003). Whenever it was necessary, we combined the independent p values obtained from all colonies using Stouffer’s method (Whitlock 2005) to obtain a single level of significance for our whole data-set.


Policing by egg eating

Egg survival was very variable among discriminator colonies (see Fig. 2). In particular, eggs generally disappeared faster in queenless than in queenright colony fragments (generalized linear model with binomial error structure [GLZ], Wald Statistics (W.s.) = 16.59, p < 0.00005).
Fig. 2

Survival rate of queen-laid (black bars) and worker-laid (white bars) eggs four days after their introduction in discriminator colonies. The discriminator colony’s name and status (plus sign queenright; minus sign queenless) and the source colony’s name are given for each trial. Normal feeding treatment: eight queen-laid and eight worker-laid eggs were introduced in each discriminator colony. Starvation treatment: six queen-laid and six worker-laid eggs were introduced in each discriminator colony. p values were obtained with Gehan’s Wilcoxon test

The survival of introduced queen-laid and worker-laid eggs did not differ (GLZ, W.s. = 0.21; p = 0.65), whether the discriminator colony was with or without a queen (GLZ, interaction between identity of the egg-layer and presence/absence of a queen in the discriminator colony: W.s. = 0.005, p = 0.94). The survival rates of simultaneously introduced queen-laid and worker-laid eggs were indeed similar in 18 of 20 trials and differed in only two starved, queenless discriminator colonies (see Fig. 2; Gehan-Wilcoxon test: discriminator colony E3-, source colony E1: p = 0.029; discriminator colony E6-, source colony E7: p = 0.030; p > 0.1 in all other trials). However, in one of these colonies, queen-laid eggs disappeared faster, whereas in the other one, worker-laid eggs disappeared faster, so that no consistent conclusion could be drawn from these two colonies. Moreover, after Bonferroni correction, the corresponding p values were not below the new significance level of 0.0083. Hence, worker-laid eggs did not appear to be differentially removed either in queenright or queenless discriminator colonies.

The dye used to stain the eggs and the feeding status of the discriminator colony (starved or unstarved) did not influence egg survival (Dye: GLZ, W.s. = 0.55, p = 0.76; Gehan Wilcoxon tests, p > 0.05 in all trials; feeding status: GLZ, W.s. = 0.41, p = 0.52). Furthermore, we could not detect an interaction between the origin of an egg (queen-laid or worker-laid) and these two factors (GLZ, interaction origin/dye: W.s. = 3.27, p = 0.195; interaction origin/feeding status: W.s. = 0.04, p = 0.85). Hence, policing did not appear to be affected by the dye used or by discriminator colony starvation.

Policing by aggression-split colonies

Workers almost never interact aggressively in the presence of a fertile queen (Heinze et al. 1997). In contrast, the level of antagonistic interactions among workers (antennal boxing, biting, etc.) increased strongly in all queenless colony parts within one or two days after colony splitting (see Fig. 3). In each colony, only a few workers were consistently aggressive during this first observation period (A1, A2, and A4: four workers each, A3: seven workers; A5: five workers; see “Materials and methods”). These are referred to as “dominant” workers. First worker-laid eggs appeared around 10 days after colony splitting, and their number increased gradually until the reunion of colony halves. Ovary dissection confirmed later that these eggs were mostly laid by dominant workers (see below).
Fig. 3

Average frequency of aggressive interactions (aggressive acts per hour) among T. unifasciatus workers during different periods of the experiment. For each colony from left to right: split colony, queenless half (white); reunited colony (vertical hatching); orphaned colonies (black). No aggression was ever observed in queenright colony halves, and therefore, the corresponding bars are not shown in the graph. For colony A1, data are missing for the last step of the experiment (orphaned colony)

Policing by aggression-reunited colonies

Reunification of the two colony halves led to striking changes in the behavior of formerly dominant workers: They immediately stopped behaving aggressively (Wilcoxon matched-pairs tests, t = 0, p < 0.05 in all five colonies) and instead engaged in social activities at a very high rate. For example, many former dominants groomed the queen more frequently than other nestmate workers (two-sample permutation tests, p < 0.1 in three of five colonies, p > 0.2 in A3 and A4) and/or engaged more in brood care (p < 0.05 in three of five colonies, p = 0.102 in A2, p = 0.086 in A3). In return, they appeared to be groomed more often (p < 0.05 in four of five colonies, p = 0.126 in A4).

Aggression in reunited colonies was lower than in queenless colony halves, but still higher than in natural queenright colonies (see Fig. 3). Antagonistic interactions were qualitatively similar to those observed in queenless colonies (mostly antennal boxing) and never resulted in injuries. In total, 17 out of 24 former dominant individuals were attacked by other workers. In four of five colonies, former dominant workers received much more attacks than expected from their proportion in the colony (Yates’ corrected χ2 test, A1: χ2 = 18.74, p < 0.0001; A2: χ2 = 32.57, p < 0.0001; A3: χ2 = 10.2, p < 0.002; A4, χ2 = 9.74, p < 0.002; A5: χ2 = 2.17, p = 0.14) and also received individually significantly more attacks than other workers (see Fig. 4; two-sample permutation tests, A1 and A2: p < 0.005; A3: p < 0.0005; A4: p = 0.106; A5: p < 0.05). The consensus p value obtained from the meta-analysis confirmed that, overall, aggression was higher towards former dominant workers than towards other workers (Stouffer’s method; S = −5.53, p < 0.0001). Therefore, worker aggression did not result from a change in colony odor due to colony splitting but was clearly associated with reproductive status.
Fig. 4

Number of attacks received per individual in the presence of the queen in colonies A1 to A5. Minimum and maximum (horizontal lines), first and third quartiles (rectangle), and the median (point) are represented for each colony for dominant workers from the queenless part on the left and for other workers on the right. The number of workers in each category is given in parentheses. *p < 0.05; ***p < 0.005; *****p < 0.0005; n.s. nonsignificant (two-sample permutation tests)

Policing by aggression-orphaned colonies

The removal of the queens from the reunited colonies resulted in a new, at least tenfold increase in the frequency of antagonistic interactions among workers (see Fig. 3). The resulting new hierarchies differed from the hierarchies in the queenless colony halves: Approximately half of the former dominant workers were no longer involved in the hierarchy (median 50%, range 40–66.7%), whereas most of the workers that had attacked the former dominants in the presence of the queen were now involved in the new hierarchies (range 50–100% of aggressive individuals in a colony, median 65%).

Our results suggest on the one hand that new dominant workers had been more involved in aggression against former dominants in reunited colonies than other workers: in all colonies except A4, they had had a significantly larger share in the aggression against dominants than expected from their proportion in the colony (A2, χ2 = 75.77, p < 0.0001; A3, χ2 = 32.95, p < 0.0001; A5, χ2 = 7.76, p < 0.01). They had also been individually significantly more aggressive after reunification than other workers (see Fig. 5; two-sample permutation tests, A2: p < 0.001; A3: p < 0.0005; A5: p = 0.062; A4: p = 0.259). The consensus p value obtained from the meta-analysis confirmed that, overall, future dominant workers were more aggressive (Stouffer’s method: S = −4.28, p < 0.0001). On the other hand, it appears that aggressive workers in reunited colonies had a significantly greater chance of becoming dominant than non-aggressive workers (Fisher’s exact test, A2: p = 0.002; A3: p = 0.002; A5: p = 0.02; A4: p = 0.26). In A4, only two workers had been aggressive in reunited colonies; one of them established itself as a new dominant after queen removal, while the other was attacked by both a former and a new dominant and finally left the nest. Hence, there appears to be a strong link between future dominance rank and aggressiveness in the presence of the queen.
Fig. 5

Number of aggressions initiated per individual in the presence of the queen. Minimum and maximum (horizontal lines), first and third quartiles (rectangle), and the median (point) are represented for each colony for future dominant workers on the left and for other workers on the right. The number of workers in each category is given in parentheses. ****p < 0.001; *****p < 0.0005; n.s. nonsignificant (two-sample permutation tests)

Ovary dissections

Ovary dissections showed that all queens were mated and fully fertile. Although the ovaries of several non-aggressive workers contained one, occasionally two trophic eggs (see also Heinze et al. 1997), those of dominant workers contained significantly more developing and mature oocytes than those of subordinate workers (see Fig. 6).
Fig. 6

Number of developing and mature oocytes observed after ovary dissection per individual. Minimum and maximum (horizontal lines), first and third quartiles (rectangle), and the median (point) are represented for each colony for dominant workers on the left and for subordinate workers on the right. ****p < 0.001; *****p < 0.0005 (two-sample permutation tests)


In the monogynous and monandrous ant T. unifasciatus, workers in queenright colonies have a very low share in male production and mostly produce trophic eggs (Heinze et al. 1997; Hammond and Keller 2004), whereas they lay viable haploid eggs developing into males, when the queen is removed (Heinze et al. 1997). We investigated the proximate mechanisms involved in the prevention of worker reproduction in T. unifasciatus and, in particular, whether mechanisms described from other ant species, such as self-restraint, queen inhibition, and queen or worker policing, do occur.

Our results did not give any evidence for policing by egg eating by workers or the queen. Introduced worker-laid eggs were not removed faster than queen-laid eggs in either queenless or queenright discriminator colonies. Our experimental design allowed us to minimize any influence of colony odor on policing, as introduced worker-laid and queen-laid eggs came from the same source colony in each trial. Moreover, we were careful to avoid any contact between introduced and own eggs to minimize the risk of odor scrambling that may affect egg discrimination and policing intensity (Ratnieks 1992; Peeters and Tsuji 1993; Nakata and Tsuji 1996). It is, thus, safe to conclude that policing by egg eating does not commonly occur in T. unifasciatus, in contrast with other ant species (e.g., Kikuta and Tsuji 1999; Monnin and Peeters 1997; D’Ettorre et al. 2004; Endler et al. 2004). Our experimental design did not enable us to exclude that worker-laid eggs might be eaten directly after oviposition, as observed in the queenless ant Diacamma (Kikuta and Tsuji 1999). However, as there is no overt aggression in queenright colonies, workers could probably evade such egg robbing.

Queens were never seen involved in aggressive interactions, suggesting that they do not police reproductive workers. Instead, queens appeared to be indifferent to worker reproduction, in contrast to other eusocial Hymenoptera species with small colonies where active queen policing was described (e.g., bees: Michener and Brothers 1974; Koedam et al. 2001; Cnaani et al. 2002; wasps: Saigo and Tsuchida 2004; Wenseleers et al. 2005; ants: Bourke 1988, 1991; Nakata and Tsuji 1996; Monnin and Peeters 1997; Kikuta and Tsuji 1999; reviewed in Wenseleers et al. 2006). Conversely, the presence of egg-laying workers in reunited colonies led to an increased level of aggression among workers compared with natural colonies. Aggression was not randomly distributed but selectively directed against those individuals that had become fertile when separated from the queen. It can, thus, be considered as worker policing by aggression. In our trials, policed workers were never killed or harmed, which again contrasts with other eusocial Hymenoptera species (Gobin et al. 1999; Monnin and Ratnieks 2001; Hartmann et al. 2003), where policing by aggression can lead to the death of egg-laying workers.

Worker policing has been defined as “coercive actions that reduce direct reproduction by other individuals” (Monnin and Ratnieks 2001). In our study, we could not prove that fertile workers effectively stopped laying eggs after being attacked, as their ovaries were dissected only after the queen was removed again. However, we observed that egg-laying workers switched from an aggressive dominance behavior to a social behavior involving frequent brood care and queen grooming shortly after colony reunion. This behavior and studies on queenright colonies suggest that fertile workers indeed stopped pursuing egg-laying in reunited colonies. One might argue that these workers were not actually forced to stop reproducing because of worker policing, but “voluntarily” refrained from doing so in reaction to the presence of the queen—in other words, that they showed self-restraint (Keller and Nonacs 1993). Worker policing and self-restraint are not necessarily incompatible: Efficient policing makes attempts of reproduction by selfish workers unprofitable and, thus, favors the emergence and maintenance of self-restraint in a society (Wenseleers et al. 2004a,b). The presence of the queen—signaled, for example, by queen pheromones or by other workers’ attacks—might, in this context, constitute an honest signal for workers to determine whether they risk being policed or not. An alternative explanation would be that queen pheromones inhibit workers’ ovarian development. Our experimental setup did not allow us to investigate this hypothesis. However, although such an action of queen pheromones has often been proposed (Wilson 1971; Michener 1974; Watson et al. 1985; Hölldobler and Wilson 1990; Ross and Matthews 1991), it has never been demonstrated. It might also be evolutionarily unstable because workers should be selected to escape such a queen control (Keller and Nonacs 1993). Future studies might help to clarify this point.

Policing individuals in insect societies are generally thought to increase their inclusive fitness by preventing selfish workers from producing less closely related offspring, inhibiting them from disrupting an efficient division of labor, and/or creating a more female-biased sex ratio (Ratnieks 1988; Ratnieks and Reeve 1992; Bourke and Franks 1995; Monnin and Ratnieks 2001; Ratnieks and Wenseleers 2005; Hartmann et al. 2003; Foster and Ratnieks 2001). Hence, worker policing benefits all workers, and one might expect all workers to police, irrespective of their social status and task in the colony. In contrast, Frank (1996, 2003) suggested that only resource-rich, strong individuals should engage in policing because group efficiency would be decreased if weaker individuals wasted their resources on policing instead of cooperating. Indeed, we found that policing workers in T. unifasciatus were not random individuals but mostly workers that became dominant and fertile themselves after the queen was removed. This pattern was not significant in colony A4, but this is probably due to the very low rate of aggression in that colony. Similar investigations with higher sample size may be required to confirm our results, which suggest a strong link between policing and dominance. Worker aggression might be simultaneously aimed at decreasing egg-laying by other workers and increasing the aggressor’s chances of future reproduction when the queen dies through achieving a higher rank in the colony hierarchy. Worker policing, thus, enables policing workers to increase both their indirect and direct fitness. Such “selfish policing” has recently been observed in the wasps Polistes chinensis antennalis and Dolichovespula sylvestris, where some worker-laid eggs were eaten by other egg-laying workers (Saigo and Tsuchida 2004; Wenseleers et al. 2005; Wenseleers and Ratnieks 2006).

Selfish policing implies that workers form a hierarchy in queenright colonies of T. unifasciatus. The existence of dominance hierarchies in the presence of a fertile queen has been reported in ponerine ants (Oliveira and Hölldobler 1990), slave-making ants (Franks and Scovell 1983), and the related species Temnothorax allardycei (Cole 1981). However, in those species, top-ranking workers contribute significantly to reproduction in queenright colonies, which is not the case in T. unifasciatus (Hammond and Keller 2004; Heinze et al. 1997). Our observations, therefore, rather suggest the existence of a hierarchy of “hopeful reproductive” workers in a species with caste dimorphism, as commonly observed in queenless species (Monnin and Peeters 1999). The existence of such a hierarchy may be surprising, since the lifespan of T. unifasciatus queens is much higher than that of workers. A worker’s opportunity to reproduce, therefore, appears to be low, and it may not be advantageous to spend much energy on dominance interactions. However, field work revealed that a high proportion of natural colonies are queenless (35% of all collected colonies in a German population; J. Heinze, unpublished data). The chance of replacing a queen might, thus, be higher than suggested by the lifespans of queens and workers.


This work was supported by the Deutsche Forschungsgemeinschaft (He 1623/17-1) to J. H. The experiments performed in this study comply with Germany laws. We would like to thank two anonymous referees for their very helpful comments on the manuscript and Tom Wenseleers for his help on statistics.

Copyright information

© Springer-Verlag 2007