Division of labor among individuals in a colony is paramount to the study of social insect societies. Although division of labor in the social Hymenoptera has been well studied, research on the other order of social insects, Isoptera, has been lacking historically. However, enough current information on termite division of labor now exists to make synthesis possible. Interestingly, patterns of social organization in the lower termites appear to be fundamentally different from patterns in the social Hymenoptera, while there is considerable convergence in social organization between higher termites and the social Hymenoptera. This paper reviews our current understanding of division of labor in lower and higher termites, with a brief discussion of characteristics that promote division of labor in insect societies.

Termite colonies, like colonies of social Hymenopterans, are composed of individuals representing discrete castes that differ in morphology and/or behavior. Division of labor among these individuals leads to greater colony efficiency and ultimately maintains colony integrity (Oster and Wilson, 1978). Thus, understanding the division of labor in social insect societies is fundamental to a basic understanding of the evolution of sociality.

Although division of labor has been well studied in the social Hymenoptera, such studies on termites have been neglected historically. The relative lack of information on termite division of labor is probably a reflection of the difficulties associated with studying termite behavior in natural environments (Miura and Matsumoto, 1997, 1998a). Most termites are subterranean, living cryptically in nests and galleries, thus making it impossible to observe colonies without disturbing nest sites. However, though difficult to obtain, there is currently enough information on termite division of labor to make synthesis possible. This paper will synthesize our current understanding of division of labor in termite colonies. First, a general overview of the importance of caste developmental schemes to patterns of social organization will be presented, followed by an evaluation of the evidence for division of labor in lower and higher termites. In conclusion, emerging themes will be discussed and areas for future research will be highlighted.

The social organization of termites and social Hymenopterans is fundamentally different, largely due to differences in patterns of caste development (Noirot, 1989). First, termites are hemimetabolous, meaning that development is gradual with individuals passing through several instars, each showing progressive growth but with little morphological differentiation until the final moult to a winged imago (the final developmental stage). The termite developmental pathway is subject to many deviations on route to the terminal imago, and these deviations yield individuals that participate in social tasks. In contrast, social Hymenopterans are holometabolous, meaning that larval forms are morphologically dissimilar to the adult form and are typically inactive. Thus, while termite colonies include active immature individuals that continually grow, only adults (imagos) are active in Hymenopteran colonies (Roisin, 2000). The other major difference between caste patterns in termites and social Hymenoptera is that castes in termites are usually comprised of both males and females, while exclusively females in the Hymenoptera perform social tasks.

Division of labor in Hymenopteran colonies is typically based on differences in morphology or age of adult individuals (Oster and Wilson, 1978). Age-based polyethism, often referred to as temporal polyethism, is well studied in many Hymenopterans. The general pattern that emerges from these studies is that younger individuals tend to remain inside the nest where they perform duties of nest maintenance and brood care, while older individuals engage in activities outside of the nest (Holldobler and Wilson, 1990). Because temporal polyethism describes changes in the behavior of adults, these changes are independent of any changes in morphology. However, the applicability of this term to describe division of labor in termites is questionable because changes in age of active individuals are inextricably linked with morphological changes. Thus, division of labor in termite colonies may be age-based or morphologically-based but it is not possible to separate the two. Nevertheless, because of their hemimetabolous development, it has been argued that termites should have an almost fully discretized caste system such that individuals of different developmental stages perform certain tasks exclusively (Oster and Wilson, 1978).

Patterns of caste development vary among termite families, and these differences may have consequences for division of labor within colonies (Wilson, 1985). Termites can generally be divided into two groups: the "lower" termites, which includes all families except for Termitidae, and the "higher" termites (family Termitidae) (Pearce and Waite, 1994). Evidence for division of labor in termite colonies will be considered separately for lower and higher termites.

In general, lower termites inhabit nest sites that also serve as feeding sites (not always the case), and there is no terminal worker caste. Workers retain the ability to reproduce throughout all life stages, and thus there is no truly sterile worker caste. The developmental pathway is typically linear, leading to a final imago (alate stage), but deviations to other terminal castes such as replacement reproductive or soldier are possible. A generalized development scheme for the lower family Kalotermitidae is a representative example of caste developmental patterns of lower termites (Figure 1; from Roisin, 2000.)

There is little evidence that tasks are allocated discretely among workers of different age or morphology in lower termite societies, though only a few species have been examined. Rosengaus and Traniello (1993) studied division of labor in colonies of the dampwood termite, Zootermopsis angusticollis. Behaviors performed by marked individuals representing all VII instars were placed into one of six categories: (1) individual maintenance, (2) brood care, (3) care of reproductives, (4) miscellaneous social interactions, (5) feeding behavior, and (6) nest construction and repair. These behavioral categories were subdivided further into specific acts. Behaviors falling in categories two through six were considered to be task-related because they achieved a purpose for the colony. The researchers calculated the total act repertoire size for each instar, and determined whether different instars tended to specialize in certain tasks. Moreover, they tested whether the caste organization was continuous or discrete by calculating the relative performance measure (RPM), which is the probability that a particular instar will perform a given task divided by the highest such probability for that behavioral act.

Rosengaus and Traniello (1993) found that instars I and II performed only two acts (allogrooming nestmates and reproductives), and typically remained inactive. Thus, they argue that these instars are functionally equivalent to dependent larval forms of the social Hymenoptera and should not be considered part of the worker caste. Instars III—VII performed 64-100% of all colony tasks, and there was no correlation between age and task performed, suggesting a lack of age-based division of labor in these colonies. Instead, they argue that there is evidence for one functional worker caste that spans instars III—VII (Rosengaus and Traniello, 1993).

Crosland et al. (1997) and Crosland et al. (1998) reported similar findings from studies of temporal polyethism in the lower termite, Reticulitermes fukienensis. In two studies they examined task performance by three different size classes of workers, small, medium and large. Since the developmental pathway is linear, size is correlated with age, as the smallest workers are the youngest and the largest workers are the oldest. Crosland et al. (1997) compared the three age classes to determine the degree to which each would carry out specific tasks grouped into three categories: (1) foraging-related, (2) care of eggs, larvae and queen, and (3) other important tasks such as corpse burying, alarm calling and time spent stationary. Crosland et al. (1998) compared the age classes for their performance of tasks partitioned into four categories: (1) tunnel construction, (2) covered gallery construction, (3) gallery repair, and (4) feeding. Thus, taken together, these two studies are an extensive examination of the potential for age-based division of labor in Reticulitermes fukienensis.

Crosland et al. (1997) found no evidence for temporal polyethism in the performance of tasks divided into the three behavioral categories. In most cases, all three worker groups performed all tasks and the oldest workers performed all tasks at a higher frequency than did small or medium workers. Thus, unlike the social Hymenoptera where younger workers tend to specialize in brood and queen care and older workers tend to shift to work outside the nest, there is no age-based specialization of workers in this termite species, as older workers simply perform all tasks at a higher frequency.

In the latter study, Crosland et al. (1998) found that large workers also carried out all tasks more efficiently than did small or medium workers. There was some evidence for temporal polyethism as smaller workers were unable to participate in the building of covered galleries or tunnels. However, the distribution of behavioral performance is not discrete, as medium and large workers performed all behaviors examined and younger workers were not more efficient at the performance of any task.

Several hypotheses have been proposed to explain the apparent lack of age-based (size based) division of labor in colonies of lower termites. Rosengaus and Traniello (1993) argued that physiological, developmental, and ecological factors interact to determine the degree to which social insects display age-based division of labor. Lower termites such as Zootermopsis angusticollis have small colony sizes, slow growth rates, low queen oviposition rates, and most live in simple nests within food sources, making extranidal foraging unnecessary. These characteristics may not favor the spatial segregation of tasks and reduce the necessity for specialized brood care in these species. Crosland et al. (1997) noted that lower activity levels of young workers might simply be explained by constraints imposed by the fragility of body parts which remains until the cuticle hardens. The risk of injury would be high and costly to the colony, so younger individuals remain, for the most part, inactive.

Finally, young workers in colonies of lower termites might act selfishly by performing fewer altruistic acts. Workers of lower termites retain the ability to gain reproductive status in the parent nest, and it has been suggested that the switch is made more easily by younger workers (Noirot, 1985). Therefore, young workers may be less altruistic and do less work for the colony because of the potential to gain reproductive status (Crosland et al., 1997; Traniello and Rosengaus, 1997; Rosengaus and Traniello, 1993).

Although lower termites do not appear to have highly discretized division of labor, individuals in colonies of Reticulitermes fukieninsis do show behavioral plasticity for task performance in response to the removal of other castes (Crosland and Traniello, 1997). Crosland and Traniello (1997) found that in the absence of large workers, medium workers performed significantly more behaviors typically associated with large workers, including exploratory behavior, alarm signaling, and corpse burying. In honey bees one age group of workers is often capable of compensating for the loss of another age group by patterns of behavioral flexibility (Giray et al., 2000). The same is true, at least for this species of lower termite, but the compensation occurs among individuals of different instars rather than diffferent aged adults and is therefore more analogous to intercaste flexibility seen in some ants (Crosland and Traniello, 1997).

The generalized caste developmental pathway of the higher termites is fundamentally quite different and more complex than that of the lower termites. The most important demarcation is that the first moult for higher termites is characterized by an irreversible divergence between two possible developmental pathways. There is a lack of consistency in the terminology used to describe these two developmental lines, but Roisin (2000) argues that "nymphal" and "apterous" are the most appropriate terms. The nymphal line produces alates, which are sexual and can disperse, and can also lead to the production of nymphoid replacement reproductives, which remain in the nest to reproduce. The apterous line is characterized by an inability to grow wings and fly. This line leads to development of sexually dimorphic workers who typically go through several instars and become permanently sterile, but are capable of becoming ergatoid replacement reproductives early in development. Typically, the soldier line is derived early in development from only one sex of the worker lines. The caste developmental pattern for the genus Microcerotermes is provided as a general example of caste patterns in higher termites (Figure 2; from Roisin, 2000).

Unlike lower termites, higher termites exhibit characteristics that are likely to promote age-based division of labor. Colonies are typically large, oviposition rates are high, nest architecture is complex, and individuals must forage outside the nest, often above ground. Moreover, higher termites have a true sterile worker caste in addition to reproductive and soldier castes (Rosengaus and Traniello, 1993). Because colonies of higher termites often nest above ground and forage in the open rather than in tunnels or gallery systems, it is much easier to perform behavioral analyses on these species. Hence, division of labor in higher termites is well studied in comparison to lower termites.

In an early study on higher termites, Gerber et al. (1998) studied division of labor and temporal polyethism in Macrotermes bellicosus. M. bellicosus feeds on plant litter that is brought back to the nest and stored for later consumption. After the food passes through the gut, the feaces are used for construction of a comb, which serves as a growth site for the fungus, Termitomyces (Gerber et al. 1988). Using sampling methods, Gerber et al. (1988) estimated that the foraging force is composed predominantly of major workers (70%), while minor workers perform the majority of spontaneous and emergency construction (97% and 85% of samples respectively) and queen care (85%). Soldiers also engaged in emergency construction (17%). Evidence for age-related polyethism was only indirect as the authors could only assume that the presence of earth particles in the gut was indicative of older individuals. Individuals were analyzed for gut contents, which indicated that older workers participate in foraging and spontaneous building, whereas the age of workers caring for the fungus comb and queen were mixed.

Temporal polyethism in M. bellicosus was confirmed in a recent study that utilized newly developed marking methods in combination with metal and photo detectors to determine if the transition between work inside the nest to work outside the nest was correlated with age (Hinze and Leuthold, 1999). Metal detectors and photo detectors were placed at the entrance to the queen chamber and the entrance to the nest in order to monitor activity within the nest and outside the nest. Individual major and minor workers of known age were marked with metal wires, and thus could be detected when passing in and out of the queen chamber or nest. Activity of marked workers showed a gradual change from work inside the nest to work outside the nest. The change occurred in major workers 13-25 days after adult moult, and in minor workers 9-32 days after adult moult, thus indicating a temporal component to the observed polyethism (Hinze and Leuthold, 1999).

There is also temporal division of labor among soldiers in Macrotermes as Konate et al. (2000) demonstrated that only older soldiers accompany workers on foraging trips while younger soldiers remain in the nest. Moreover, younger individuals show flexibility by replacing older workers when a significant number of older individuals are lost. Division of labor in Macrotermes appears to be highly convergent with some Hymenopterans in that younger individuals perform intranidal tasks while older workers perform predominantly extranidal tasks. However, the system of polyethism is not rigid, as there is considerable overlap and task sharing by individuals of different instars (Traniello and Leuthold, 2000).

Soldiers of some species of higher termites not only specialize in colony defense, but also play an important role in the location of new food sources and the organization of foraging behavior of the colony (Traniello, 1981). Soldiers of Nasutitermes costalis initially act as scouts by exploring the environment for new food items. Traniello and Busher (1985) described three phases of foraging in N. costalis. In the first phase, soldiers encountering food items primarily recruit other soldiers to the feeding site. In the next phase, workers begin to arrive at the feeding site in large numbers. The final phase is characterized by additional heavy recruitment of workers and a decline in soldier recruitment. As the number of workers continues to increase, soldiers line the periphery of the foraging column until a gallery that encloses the foraging trail is constructed.

Traniello and Busher (1985) found that the organization of the foraging phases is mediated by chemical communication within and between castes. Soldiers and workers can secrete pheromone trails using the sternal gland. Pheromone cues are not caste-specific, but rather soldiers and workers have different behavioral response thresholds to the pheromone. Soldiers are more sensitive to the presence of trail pheromones, and have lower sternal gland volumes than workers. Hence, soldiers respond to the initial low volume secretions of other soldiers, while workers remain insensitive. Workers respond only when the pheromone is present in a quantity at or above their behavioral response threshold, which delays the onset of worker recruitment. Thus, the division of labor among soldiers and workers is regulated by differences in pheromone sensitivities and the response thresholds of these individuals.

Much of the information available on division of labor during foraging activity comes from studies of species that forage for lichens or litter in large open-air processionals. Three genera, Hospitalitermes, Lacessititermes, and Longipeditermes, that exhibit such conspicuous foraging behavior are distributed throughout southeastern Asia, and the foraging behavior of two (Hospitalitermes and Longipditermes) has been fairly well studied. Hospitalitermes forages in the open for lichens whereas Longipeditermes forages for leaf litter. After food is located by soldiers that act as scouts, colonies forage in large columns containing thousands of individuals; foraging columns were comprised of between 300,000-500,000 for Hospitalitermes colonies (Miura and Mastumoto, 1998a). Soldiers typically flank the periphery of the foraging column, while workers move back and forth between the soldiers. Workers make food balls at foraging sites that are transferred to another worker and subsequently carried back to the nest for later consumption.

Colonies of the termites, H. medioflavus and H. rufus exhibit division of labor among trimorphic workers during foraging activity (Miura and Matsumoto, 1998a; Miura and Matsumoto, 1995). Workers can be grouped into minor, medium, and major groups based on head capsule width, and this polymorphism is associated with the task each performs. Workers termed, "gnawers," gnaw food material and pass small pieces of food to "carriers," who carry food balls back to the nest once they have attained an adequate size. Minor workers act almost exclusively as "gnawers," whereas major workers serve almost exclusively as "carriers." Medium workers perform both tasks at high frequencies. Thus, Hospitalitermes colonies exhibit division of labor between workers and soldiers, and labor is subdivided further among trimorphic workers, resulting in a complex, but efficient foraging system.

A similar pattern of division of labor between soldiers and workers is present in the litter foraging termite, Longipeditermes longipes, as soldiers act as scouts and guard the foraging column while workers compile and carry food balls. However, in contrast to Hospitalitermes workers, L. longipes workers are monomorphic. Nevertheless, there is a division of labor between individual workers, as some act as "gnawers" and some act as "carriers." The only obvious difference between division of labor among Hospitalitermes foragers and those of L. longipes is that task performance is not associated with worker polymorphism (Hoare and Jones, 1998; Miura and Matsumoto, 1998b). Moreover, soldiers of this species are dimorphic, but as of yet there are no quantitative data that indicate a division of labor between soldiers of different sizes (Miura and Matsumoto, 1998b).

Oster and Wilson (1978) argued that termites should have an almost fully discretized caste system. However, this is certainly not the case, at least for those species studied to date. In fact, individuals in termite colonies show considerable overlap for the ability to perform particular tasks and the allocation of jobs is not discrete. Perhaps the most insightful information gleaned from studies of termite division of labor is that patterns of social organization observed in the social Hymenoptera are not necessarily the rule for all social insects. Ecological parameters, such as nesting habitats and foraging behaviors, coupled with developmental constraints can lead to patterns of social organization in the lower termites that are fundametally different from patterns observed in the Hymenoptera. Nevertheless, there is considerable convergence in social organization between the higher termites and the social Hymenoptera. Hence, fruitful insight can be gained by an elucidation of the conditions that promote division of labor, and comparative studies between Isoptera and the social Hymenoptera are essential for further progress on this matter.