Introduction

All social insect colonies share a fundamental division of labour between one or more reproductive queens and non-reproductive workers, which in the Hymenoptera are always female (Andersson 1984). Depending on the species, the number of workers in a single nest may vary from less than 10 to several million (Wilson and Holldobler 2005). In highly populous colonies, workers tend to become subdivided into groups specialising on particular tasks (Bourke 1999; Ferguson-Gow et al. 2014). Many eusocial species accomplish this through separation of the workers into alternate, highly dimorphic specialised developmental castes such as soldiers, major workers and minor workers, and this strategy is common in ants and termites (Wheeler 1991; Miura et al. 1998; Bourke 1999).

In species without worker dimorphism, each worker must instead have the flexibility to carry out most or all of the necessary tasks for colony function. A common mechanism is age polyethism, in which colony tasks are performed in sequence according to worker age, usually beginning with brood nursing behaviours and ending with foraging (von Frisch 1954, Robinson 1987, O’Donnell and Jeanne 1993, Johnson 2010). Whilst age polyethism is not mutually exclusive with caste based division of labour (e.g. Gruter et al. 2012; Yanagihara et al. 2018), it plays a prominent role in eusocial species with workers morphologically similar to one another, such as honeybees, stingless bees, and some species of wasps and ants (von Frisch 1954, Robinson 1987, O’Donnell and Jeanne 1993, Bernadou et al. 2015, Mateus et al. 2019). A third proposed mechanism of task allocation is called ‘foraging for work’, which posits that young workers roam the nest until they come across a task that needs doing, and then usually continue performing that task, although nest disturbance may trigger a shift in tasks performed (Sendova-Franks and Franks 1993). Although this latter mechanism was defined from the results of mathematical modelling than on empirical data, it has been observed as the primary means of task allocation in an ant species lacking worker sub-castes with only a very weak association of tasks with worker age (Sendova-Franks and Franks 1993). Foraging for work may also co-occur to some extent with age polyethism in other species (Tripet and Nonacs 2004). Given it involves chance encounters by individuals looking for work in different parts of the nest, it may explain idiosyncrasies in behaviour amongst workers, even those emerging on the same day (Nakata 1995), and how middle-aged honeybees individually specialise as undertakers, guards or food storers before eventually moving on to foraging (Trumbo et al. 1997).

Mechanistically, age polyethism appears to be a developmental process mediated by an increase in the level of juvenile hormone with age, and this has been shown experimentally in worker honeybees and wasps (Rutz et al. 1976, Robinson 1987, O’Donnell and Jeanne 1993). The rate at which hormone levels increase is flexible and responds to colony stress, which allows workers to progress to foraging more quickly in response to colony needs such as loss of foragers (Robinson 1987, Giray and Robinson 1994). Expression of juvenile hormone in honeybees is also affected by queen and brood pheromones, allowing a regulation mechanism where at low density of nurse bees, juvenile hormone is suppressed by these pheromones whereas at high densities some nurse bees are crowded away from the queen and brood and thus allowed to continue the transition into middle-aged worker bee tasks (Kaatz et al. 1992; Le Conte et al. 2001; Johnson 2010). There are also genetic differences among honeybee colonies (Calderone and Page 1988, Giray and Robinson 1994). Some colonies are more prone to precocious foragers or overage nurses and are respectively more easily able to compensate for a deficit or surplus of foragers (Giray and Robinson 1994). In an ant species showing age polyethism, a loss of nurse workers led to some of the foragers quickly reverting to nursing behaviour, whereas nurse bees could not immediately skip ahead to foraging (Tripet and Nonacs 2004). This seems to support the theory of hormone development being necessary for emergence of later-stage behaviours. Nonetheless, emergence of end-stage behaviours such as guarding and foraging is not inevitable. For example some individual workers within a species of stingless bee and ant continued or returned to performing within-nest tasks throughout their lives whilst others in their age cohort foraged (Bernadou et al. 2015; Mateus et al. 2019).

In stingless bees, workers must perform a range of different tasks, including the construction, provisioning and cleaning of brood cells, building and maintenance of food storage pots and connective structures, removal of waste, nest temperature regulation, as well as foraging for nectar, pollen and resin (Bassindale and Matthews 1955; Sommeijer 1984; Roubik 2012; Gruter 2020). Colony tasks appear to be strongly influenced by worker age (Bassindale and Matthews 1955; Sommeijer 1984; Hammel et al. 2016; Mateus et al. 2019; Lau et al. 2022). For example, workers of the Neotropical stingless bee Melipona favosa typically focus on brood-related tasks (building cells, discharging food into cells and capping them) during the first 20 days of life, whereas guarding the entrance and dumping waste is performed by slightly older workers and foraging is the primary task of workers more than 60 days old (Sommeijer 1984). A similar series of tasks were observed in a West African stingless bee, Hypotrigona gribodoi (Bassindale and Matthews 1955), but there were some subtle differences. For one, nest entrance guarding and maintenance tasks co-occurred with foraging in the older H. gribodoi bees. There was also a transition within brood care behaviours from brood cell building and filling to helping new bees emerge and breaking down old cells (Bassindale and Matthews 1955). In some species, the job of guarding the nest is performed by a cryptic guard caste (Gruter et al. 2012, 2017; Wittwer and Elgar 2018), or a rare ‘elite caste’ that perform their own slightly different series of behaviours, with guarding predominating over foraging as the terminal specialisation (Hammel et al. 2016).

Despite these examples, age-related task behaviour has not been investigated in the vast majority of the world’s ~ 58 genera and ~ 500 species of stingless bee (Gruter 2020). One such genus lacking study on timing of behaviours is the species rich Tetragonula of Asia and Australia. Although there is evidence guarding behaviour in T. carbonaria is caste based (Wittwer and Elgar 2018), other stingless bees with guard castes still show temporal polyethism (e.g. Hammel et al. 2016). Published results on age progression of colony tasks for this stingless bee genus, however, include only a preliminary study with a single colony of T. carbonaria (Lau et al. 2022).

In the present study we aimed to describe the natural history of age-related task behaviour in two Australian Tetragonula bee species: T. carbonaria and T. hockingsi (Hymenoptera: Apidae: Meliponini). These are two of three stingless bee species found in the study area (Brisbane) and the most common species kept in hive boxes across Australia (Heard 2016). Workers of the two species are 4 mm and 4.5 mm long, respectively, and are polylectic pollinators (Heard 1999; Wilson et al. 2021). Both species are found on Australia’s east coast. Tetragonula hockingsi (Cockerell) is distributed along the mostly tropical Queensland coast, with subtropical Brisbane (27.5°S) marking the southern end of its range. Tetragonula carbonaria is mostly distributed in subtropical areas, and can tolerate coastal temperate climates as far as 36°S (Heard 2016). Both species are cavity nesters with a single queen and up to 10 000 workers (Heard 1999). Brood development occurs within cells arranged in layered sheets, which vary in architecture across species (Brito et al. 2012). In contrast to honeybees, colony reproduction in stingless bees usually occurs gradually over a few weeks by scout bees finding a new nest site, workers flying materials over from the mother nest and building a nest entrance and food pots, and then a virgin queen from the mother nest entering the new nest following a mating flight (Inoue et al. 1984). Both our Tetragonula species, particularly T. hockingsi, can also reproduce by usurping another stingless bee nest. This involves workers sacrificing themselves in a series of attacks, often over several months, to eliminate the weaker colony’s population – thereby allowing a daughter queen to inherit a pre-built nest already stocked with resources (Cunningham et al. 2014; Lau et al. 2022).

Preliminary results with one colony of Tetragonula carbonaria have shown an age progression of colony tasks. We intended to substantiate this by repeating with more colonies of T. carbonaria as well as the related species T. hockingsi, for which no information on timing of worker behaviour has been published. We expected that brood care tasks would be performed first and foraging last, but this needs to be quantified, as does as the sequence of tasks in the intermediate range of worker ages and how discrete they are relative to one another.

Materials and methods

Marking methods

We marked newly emerged adult bees (callows) with light grey colouration rather than the black they darken to after ~ 2–4 days. We extracted groups of between 25 and 100 callows (depending on availability and the size of the colony into which they were to be introduced) by either opening hives and directly aspirating callows from the brood comb using a pooter, or by extracting pupal brood that was kept separately and checked regularly for emergence of callows.

As T. carbonaria and T. hockingsi will accept callows from other nests of the same species (Lau et al. 2022) we pooled callows from both the destination colony and the other two conspecific colonies when necessary to increase the number of marked bees. Each callow was marked by gently pinning its legs between thumb and forefinger and painting a coloured dot on its thorax using a Posca PC-1MR 0.7 mm marker pen (Mitsubishi Pencil, Tokyo, Japan). All bees within a cohort were marked in the same colour. These marked bees were then released into a single colony for observation.

Observations — preliminary study with T. carbonaria micro-colonies

Preliminary observations were made with T. carbonaria micro-colonies in Annerley (Brisbane, Queensland, Australia) in 2016. We used two single layer wooden boxes (20 cm wide, 28 cm long and 9 cm deep) fitted with observation windows. These two boxes housed micro-colonies. The micro-colonies were queen-less colonies containing a small amount of brood and around 400 workers, and were established by transferring nest material and bees from a strong donor colony. As the micro-colonies had little stored food sources, we provided small amounts of 50% sugar solution periodically, when large numbers of bees gathered at the entrance, presumably because they needed sustenance.

Two groups of callows (one per colony) were investigated (released on 9th April 2016 and 5th May 2016). The day on which marked callows were released was considered day 1. Following their introduction, visual observations of the behaviours they performed were recorded 3 days a week until no marked bees were observed for three continuous observation days. Data were collected three times for each colony within an observation day to determine any variation in their behaviour within a day. The duration of each observation period was roughly 20 min, with 10 min at the observation window to observe bees within the nest and 10 min at the entrance to observe bees flying in and out. We did not make observations when it was raining or temperatures were below 18 °C, as T. carbonaria is not capable of flight under these conditions (Heard and Hendrikz 1993).

Behavioural data from the two micro-colonies were recorded for only 25 days, because after this time all the bees had either gradually died or, for one of the two micro-colonies, appeared to have gone missing following a storm. Most of the marked bees cared for brood and maintained nest structures such as propolis connecting scaffolds and food pots during the observation period. None of the marked workers was ever observed foraging or engaging in waste removal (Table 1) but instead appeared to be starving, as indicated by the crowds of bees gathering around the internal hive entrance (Figure S1). Both foraging and waste removal were performed only by some of the original unmarked bees introduced during micro-colony creation and these activities ceased as the colony population started to decline. Waste materials and bodies of dead bees therefore accumulated inside the nest after the first week (Figure S1). In addition, few changes in nest structure were detected during the observation period despite the availability of propolis building material, which was provided at micro-colony establishment. Empty brood cells following bee emergence were left in place without being broken down as they would be in typical, strong colonies.

Table 1 Median age (with 80% range) and total number of times each behavioural category was observed across all marked cohorts in the two micro-colonies of Tetragonula carbonaria
Full size table

Only limited changes in behaviour were detected in marked workers of the two micro-colonies. All behavioural categories were observed evenly across the observation period. The exceptions were brood care, which decreased as most of the pre-existing brood from the parental colony emerged, and food collection (collecting the provided sugar solution and transferring to food storage pots) which increased when sugar solution was supplied a week after micro-colony establishment (Table 1).

Given that micro-colonies have smaller worker populations than typical, strong colonies, they were chosen for the experiment because the marked bees would be more likely to be visible. However, the bees in the micro-colonies appeared too stressed for the normal development of their behaviour (see Results). This highlighted the need for developing observational means in regular-sized colonies, but they allowed enough observation for development of a categorisation of behaviours to underpin the structure of subsequent observations.

Observations on larger colonies with a queen

The behaviour of bees in larger colonies of T. carbonaria and T. hockingsi was studied through 2020–2022 at the University of Queensland, St Lucia campus in Brisbane. Nests of T. carbonaria and T. hockingsi were kept independently of one another in wooden observation hive boxes (3 of each species) on a covered outdoor rooftop area which shaded the boxes and plastic tubes from direct sunlight during the warmer months. The observation hives had internal dimensions of 24 cm wide, 35 cm long and 4 cm high (volume 3.36L). Although this is still smaller than many natural nests, which may have volumes of around 6–8 L in T. carbonaria and up to 10L in T. hockingsi (Heard 2016 p. 120), the colonies were nonetheless large enough to survive for a year or more without supplemental feeding. A clear plastic observation window formed the top of the hive space, and a removable wooden cover was placed over it (Fig. 1). The boxes had a hole in the front fitted with a clear plastic tube (~ 50 cm long) along which the bees entered or left the hive (Fig. 1). Because the hive entrances were only about 1 m apart from one another, plastic coloured shapes were used to provide unique nest entrancees to aid in nest recognition and limit worker drift amongst neighbouring colonies.

Fig. 1
figure 1

top left: Extracting bees from a colony for marking. The hive box is propped atop its wooden cover that usually covers the observation window. Top right: newly released marked callows of T. carbonaria on the brood comb. Bottom: bees entering or leaving the hive through the entrance tube. The two bees in the blue ellipse are carrying out waste pellets held with their mandibles, whilst the bee in the red ellipse is carrying in pollen on its hind legs

Full size image

Into each of the three colonies of our two species, we released new marked cohorts of callows. Callows were marked and released opportunistically at various times in spring, summer or autumn, with observations of marked callows within the colony made during all seasons. The number of cohorts released into an individual colony varied from one to three. Each marked cohort was labelled by its species (TC = Tetragonula carbonaria and TH = Tetragonula hockingsi). As we used three colonies for each species, we gave each cohort a number 1, 2 or 3 corresponding to which colony it was introduced into. As some of the colonies had two or three cohorts introduced into them over the course of the experiment, we used the letters A = first, B = second and C = third to distinguish them. For example TC1A was the first cohort to be released into T. carbonaria colony 1, whilst TH3C was the third cohort to be released into T. hockingsi colony 3. We never released more than one cohort into the same colony simultaneously. We used different colours for different colonies in case workers of an older or younger cohort drifted into a different colony. All of the bees observed during the study were of the ‘right’ colour for that colony, which suggests most of the bees could recognise their nest entrance.

Following the introduction of marked callows, we observed the colony once in a day on 2 days each week until no marked bees were seen on five consecutive observations. Observations were made in the early morning in summer (typically around 0700–0900 h), later morning (~ 0800–1000 h) during spring and autumn and early afternoon (~ 1200–1400 h) in winter. Although we continued observations into winter for longer lived cohorts marked in autumn, we did not start new trials in winter. This is because the bees extended the propolis involucrum surrounding the brood comb to cover the brood almost completely during the coldest months, and this hid most of the bees working on the brood comb from view.

During an observation, we were careful to observe all parts of the colony, including the central brood comb, the food pots and propolis structures clustered around the comb, and the bees walking along the inside edges and floor of the hive box. The latter were visible through gaps between the food pots and the web of propolis connectives. Each observation consisted of three searches of the nest through the observation window, during which one observer (usually LCJ, occasionally JPH) recorded behaviours (e.g. inspecting brood cells, visiting food pots, constructing propolis structures) of all marked bees visible. The three searches typically lasted five minutes. In most cases, the number of marked bees simultaneously visible was five or fewer, low enough to be able to remember individuals that had already been counted. Nonetheless, we cannot discount the possibility that a marked bee disappearing into the lower brood cells or beneath a food pot and reappearing again somewhere else may have been counted twice. We allowed deliberate recording of the same bee more than once if it was performing more than one category of behaviour. In cases where an exceptionally large number of bees (> 15) were visible at once (often appearing and disappearing at high frequency) we used a shorter ‘snapshot’ of 2 min to reduce the risk of double counting bees within an observation frame.

Each of the three observations of behaviour inside the nest were spaced with a roughly five minute break in between, during which time the lid was replaced we instead scanned for marked bees entering and leaving the hive through the clear tubing at the hive entrance. We usually scanned a total of between 100 and 600 bees in the entrance tube per observation days in search of marked bees. The exact number depended on the colony’s population and activity, although sometimes rainy or cold weather meant there were few or no bees in the tube. For each marked bee we observed in the entrance tube, we recorded its direction of movement (in, out, or lingering near entrance), and whether it was carrying pollen or resin on its hind legs or a waste pellet in its mandibles. When we returned to scanning bees inside the colony for the second and third observation frame, we discounted a bee if we noticed one in the exact same position as we remembered from before. In most cases, however, the bees had shifted in that time.

Statistics

We used a generalised linear mixed model with binomial errors for each of the six main behaviours observed in the full sized hives (brood care, propolis, pot construction, visiting food pots, walking and foraging) to test the effects of bee age on the proportion of marked bees observed performing this behaviour. We used the proportion of bees performing a given behaviour on each of the 144 observation days across all cohorts (65 for T. carbonaria, 79 for T. hockingsi) as our raw data. The two bee species were analysed separately with regard to correlating age and proportion of bees performing each of the main behaviours. We also compared the two species to test whether they differed in how commonly each behavioural category was performed. Because the relationship of many behavioural categories with age was unlikely to be linear, we initially ran all models including a quadratic and cubic term for age. This allowed relationships such as where frequency of a task rose to a peak and then declined (quadratic) or rose to a peak, declined, then plateaued (cubic) to be adequately modelled. As we were combining multiple cohorts, which varied substantially in the pace of progression through colony tasks, we included cohort as random factor. We ran simplified models removing one or more higher order terms and retained a simplified model if it had a lower BIC value or a value of AIC no more than two units higher than a more complex model (see Supplementary material). We also used binomial generalised linear mixed models to compare the most common behaviours across species, with species as the fixed factor and cohort as the random factor. Statistics were run in R version 4.3.0 (R Core Team 2023) using the lme4 package (Bates et al. 2015).

We did not test the numerous rare behaviours (see Results) against age as the number of observations for each was low, reducing statistical power to the extent that any positive results would have a high chance of being type-I errors and negative results type II errors.

Additional entrance tube observations

We made 15 observations (6 in T. hockingsi, 9 in T. carbonaria) on sets of 100 consecutive bees (marked or not) that were entering through the tube where we counted how many were bringing in pollen. We were unable to tell whether these bees in the entrance tube were carrying nectar, as they usually moved too quickly for close examination of their abdominal size. We also observed 16 sets (7 in T. hockingsi, 9 in T. carbonaria) of 100 consecutive bees that were leaving the colony, and recorded how many of these were carrying out waste balls.

Results

Observations on larger colonies with a queen

In the colonies with queens observed in 2020–2022 the lifespan of an introduced marked cohort varied from 28 to 92 days across the five T. carbonaria cohorts (median: 40 days) and from 21 to 114 days across the six T. hockingsi cohorts (median: 50 days) (Fig. 2). Oddly, the two cohorts lasting 21 days and 114 days were in the same colony, with the latter cohort introduced shortly after the 21 day cohort had died off. The 21 day cohort was not observed to forage (Fig. 2), suggesting the bees died prematurely. Other short-lived cohorts in other colonies, however, lasting 35–55 days, progressed through the full series of tasks from brood care to foraging despite living less than half as long as the longest lived cohort of its respective species (Fig. 2). The transitions between tasks were not absolute. Some marked bees typically still performed inside-nest behaviours such as maintaining propolis structures during the peak of a marked cohort’s foraging activity. Even caring for brood, which is mostly performed by the youngest bees (Figs. 2, 3), was once observed being undertaken by a 104-day old bee.

Fig. 2
figure 2

Median and 80% range of ages in days of bees in each marked cohort that were observed performing each category of task. The length of the grey bars indicates the lifespan of each cohort

Full size image
Fig. 3
figure 3

Percentage of bees on each observation day that performed each of six common behaviours: Tetragonula carbonaria (left column), T. hockingsi (right column). Age of the observed bees is scaled by percentage of their respective cohort lifespans that had elapsed at the time of observation. Shading is underneath line of best fit through the plotted points

Full size image

The most commonly observed brood care behaviour was inspecting or cleaning cells that had already been filled and capped (i.e. bees moved slowly over the brood cells, stopping frequently and often making contact with the cells with their legs or mouthparts). This is partly because only the top of several layers of brood comb could be seen. In Tetragonula nests, brood is built in horizontal layers stacked on top of one another within the brood chamber. When bees emerge from brood cells ~ 50 days after construction, those empty cells are removed leaving a layer of empty space into which new brood cells are constructed. In this way brood dynamics follows a 50 day cycle, with layers fixed in place for 50 days until emergence and deconstruction. Most of the time the visible top brood layers consisted of capped cells that contained larvae or pupae, so the worker bees visible on these cells were presumably cleaning them, removing outer propolis of pupal cells, and/or perhaps checking for signs of emergence or problems with the developing bees. When the visible top layer of brood cells was new brood being constructed and provisioned, we observed that a bee filling a brood cell lowers her head deep inside the brood cell so that her abdomen (usually more swollen than that of other workers) pointed upwards (Table 2). If a bee was, instead, manipulating the outer edge of an unfinished brood cell with its mouthparts, we assumed she was building the cell (Table 2). Likewise, bees working on propolis structures or food pots appeared to be using their mouthparts to add propolis material to the structure (Table 2). Bees constructing food pots sometimes formed small groups of 2–4 individuals working inside a propolis pot. On the other hand, a bee that simply dipped her head inside a filled food pot was recorded as visiting food pots (Table 2), most likely either taking food from it or regurgitating nectar transferred to it from a forager.

Table 2 Ethogram of behavioural categories recorded for Tetragonula carbonaria and T. hockingsi
Full size table

Overall relative frequencies of behaviours

Brood care and constructing connecting structures from propolis were the two most commonly observed behaviours amongst individuals of both species (Table 3), followed by walking, building food pots and foraging (observed as entering and leaving the nest). When accounting for cohort differences, the frequency of brood care behaviour did not differ across the two species (z =  – 0.56, P = 0.58), and nor did building infrastructure (z =  – 1.2, P = 0.25), building food pots (z =  – 1.7, P = 0.082), visiting food pots (z = 0.85, P = 0.4), walking (z = 1.25, P = 0.21) or entering and leaving the nest (z = 0.83, P = 0.40) (Table 3).

Table 3 Total number of times each behavioural category was observed across all marked cohorts of each species, ranked from overall most to least common
Full size table

Age separation in the six most common behaviours

Brood care

Brood-care behaviour for both species peaked amongst young bees and sharply declined with age (Figs. 2, 3). With T. carbonaria, the best model included only the linear term (see Supplementary material) and revealed a highly significant effect of age on frequency of brood caring behaviour (z =  – 11.5, P = 2 × 10–16). With T. hockingsi brood care behaviour was best described with a parabolic model, in which both the linear (z =  – 5.3, P = 9.48 × 10–9) and quadratic (z = 6.0, P = 1.61 × 10–9) terms were highly significant.

Visiting food pots

Bees visiting the nectar or pollen pots were typically young to middle-aged bees (Figs. 2, 3). With T. carbonaria, proportion of bees visiting food pots peaked early (Fig. 3) and had a quadratic relationship with age (z =  – 2.31, P = 0.021). The linear term in the model was also significant (z = -2.35, P = 0.019). T. hockingsi also showed a somewhat significant quadratic relationship between age and visiting food pots (z =  – 2.17, P = 0.030), and this peaked for middle-aged bees (Fig. 3,) but there was no significant linear component to this relationship (z =  – 1.57, P = 0.12).

Build/maintain propolis infrastructure

Contributing to propolis infrastructure was performed by bees of a wide range of ages but was most common amongst slightly older bees than those exhibiting brood care (Fig. 3). With T. carbonaria, the model we retained was the parabolic model, with a significant quadratic term (z =  – 3.22, P = 0.0013). The linear term was not significant (z = 0.266, P = 0.82). With T. hockingsi, however, there was a strong cubic relationship between the proportion of bees building infrastructure and their age (z = 4.3, P = 0.000015). The quadratic term was more weakly significant (z = 2.2, P = 0.025) and the linear term non-significant (z =  – 1.5, P = 0.125).

Food-pot construction

Food-pot construction or maintenance was most common amongst middle-aged workers of both species (Figs. 2, 3). With T. carbonaria, the best model showed a strong parabolic relationship across food pot construction and age (z =  – 5.9, P = 4 × 10–9) and the linear term was also significant (z =  – 3.6, P = 0.00037). By contrast food pot construction by T. hockingsi had a strong cubic relationship with age (z = 4.0, P = 0.000057). The linear (z =  – 0.91, P = 0.36) and quadratic (z =  – 0.80, P = 0.43) terms were not significant.

Walking

The proportion of T. carbonaria bees walking did not vary significantly with age, with none of the terms being significant (all P > 0.4) and the model over-fitted. This did not change with removal of the higher order terms, and the lowest AIC and BIC values were found with the null model with only the cohort random factor included (see Supplementary material). By contrast, with T. hockingsi the cubic term was somewhat significant (z =  – 2.1, P = 0.036) and both the quadratic (z =  – 2.6, P = 0.0086) and linear (z = 3.4, P = 0.00068) terms were more significant. Walking appears to be common among older T. hockingsi bees, especially in cohort TC3C (Fig. 2), but not discernably so for T. carbonaria (Fig. 3).

Foraging (entering or leaving nest)

Foraging occupied the oldest bees of both species (Figs. 2, 3). With T. carbonaria, the proportion of bees foraging increased in a linear correlation with age (z = 2.6, P = 0.0099), however the quadratic (z = 0.21, P = 0.83) and cubic (z = 1.8, P = 0.072) terms were not significant. With T. hockingsi, the cubic term was also not significant (z = 0.93, P = 0.35), and removing it from the model resulted in lower AIC and BIC values. In the simplified model, the relationship of bees foraging with age had highly significant quadratic (z =  – 4.5, P = 7.9 × 10–6) and linear (z = 5.4, P = 8.3 × 10–8) terms.

We noticed that bees carrying pollen into the nest entrance tended to be older than those entering without pollen. For example, in colony TH2A the pollen foragers had a median age of 50 days (N = 4) compared to 26 days for bees returning to the hive without pollen (N = 8) and in TC1A the pollen foragers had a median age of 77 days (N = 3) compared to 54 (N = 10) for other returning bees. This could mean that foragers tend to gather nectar initially and pollen at an older age. However, we only observed 14 pollen foragers across all the hives of both species, too few to assert this confidently or to warrant a separate analysis.

Rare behaviours

A further eight distinct behaviours were observed less commonly (Table 3). The most common among these were resting, trophallaxis (bees passing regurgitated food mouth to mouth) and guarding. Resting occurred across a range of ages from day 1 to 83, whereas visiting food pots seemed to occupy relatively young bees, although not as young on average as brood carers (Fig. 2). Trophallaxis also occurred across a range of ages. We observed it between two callows and between a callow and the queen, showing that bees may give or receive food by trophallaxis from an early age. We were unable to determine, however, whether the older bees engaged in trophallaxis were giving or receiving food. Guarding the nest entrance occurred in bees shortly before they began foraging, but with some overlap (Fig. 2). Only seven marked bees across the two species were seen carrying waste, either in the tube or inside the hive. The remaining behaviours, namely fanning wings, antennal contact with another bee, depositing wax and spreading resin on the ceiling (plastic observation window), were also observed only on average once per cohort or fewer (Table 3). We only once observed a bee depositing pollen from its legs into a pollen pot, and we counted this as pollen foraging.

Additional observations at entrance tube

Observations on sets of 100 bees consecutively entering the nest through the clear tube showed wide variation in the proportion carrying pollen. A median of 10% of T. carbonaria bees were carrying pollen into the nest, and this ranged from 1 to 36% (n = 9). Tetragonula hockingsi bees were similar in this respect with a median of 11.5% carrying in pollen, ranging from 2 to 31% (n = 6).

Our observations on sets of 100 consecutive bees that were instead leaving their nest showed extreme variation in the proportion carrying out waste pellets. With T. carbonaria, the median was 8%, however this ranged from 0 to 73% (n = 9). For T. hockingsi bees, the median was 4% and ranged from 1 to 30% (n = 7). Levels of waste carrying was generally low, and the extremely high level of 73% in one set of 100 T. carbonaria bees occurred on the first sunny day after an extended period of wet weather.

Discussion

Our results demonstrate age progression (polyethism) of colony tasks in Tetragonula carbonaria and T. hockingsi. Cleaning, filling and building of brood cells formed the primary occupation of workers during the first quarter to a third of their cohort’s maximum lifespan (Fig. 3), which is a similar nurse bee task duration to honeybees (Seeley 1982) and other stingless bees (e.g. Sommeijer 1984; Mateus et al. 2019). Building and maintaining infrastructure (including the involucrum protecting the brood cells and the filamentous support structures throughout the nest) was a common task for bees at any age, whilst building food pots rose from near zero frequency for the youngest bees to peak in bees at about 30–40% of their cohort lifespan (Fig. 3). Visiting food pots to add or remove nectar peaked in young bees for T. carbonaria and mid aged bees of T. hockingsi. The proportion of bees foraging in both species peaked during the final 80–100% of a cohort’s lifespan (Fig. 3). The performance of brood care first and foraging last further supports a unifying pattern in worker age polyethism across a variety of taxa including ants, honeybees, stingless bees and wasps (von Frish 1954; Bassindale and Matthews 1955; Seeley 1982; Sommeijer 1984; O’Donnell and Jeanne 1993; Bernadou et al. 2015; Mateus et al. 2019). This is despite independent origins of eusociality in Vespid wasps and Apid bees (Hines et al. 2007; Payne 2014) and of complex eusociality from primitively eusocial ancestors in honeybees and stingless bees (Cardinal and Danforth 2011; Payne 2014), indicating an evolutionary convergence. This makes sense given all bees begin their adult lives on the brood comb and most likely shared a natural selection pressure to postpone foraging until an older age given it is the most exhausting and dangerous of tasks (see Neukirch 1982; Visscher and Dukas 1997; Ippolito et al. 2020). Although the sequence of behaviours was mostly consistent across Tetragonula cohorts and species (Fig. 3), there was clear flexibility in how quickly bees could move through their sequence of tasks, with the median age of a forager varying from 24 to 87 days across cohorts (Fig. 2). This suggests that T. carbonaria and T. hockingsi colonies, like honeybees (Robinson 1987, Giray and Robinson 1994), can vary the pace of worker polyethism in response to stress or colony needs.

In comparison to other stingless bees, the two Tetragonula species show many similarities in task progression sequence, but also some subtle differences. In Tetragonula, brood care behaviour was observed in newly emerged bees, which is consistent with observations in Neotropical species such as Tetragonisca angustula (Grosso and Bego 2002) but contrasts with the African species Hypotrigona gribodoi, in which workers were not reported to do any work in the first 3 days after emergence (Bassindale and Matthews 1955). We did not notice any worker oviposition in our Tetragonula species, and it has not been reported elsewhere for this genus. By contrast, workers of the Neotropical Melipona favosa and M. marginata routinely lay eggs that are consumed by the queen (Sommeijer 1984). Trophic egg laying occurs less commonly in the Australian species Austroplebia australis and A. symei (= cassiae) (Drummond et al. 2001), whilst in H. gribodoi, oviposition occurred only in a queen’s absence, and the eggs developed into males (Bassindale and Matthews 1955). In H. gribodoi and M. marginata, depositing wax was observed only in young workers (Bassindale and Matthews 1955; Mateus et al. 2019), whereas the few bees we observed depositing wax ranged from 11 to 49 days in age. This suggests wax glands may continue functioning for longer in Tetragonula compared to other stingless bee species (e.g. Justino et al. 2018). There was also a difference in timing of building food storage pots. Tetragonula carbonaria and T. hockingsi workers building pots were older than those inspecting brood cells (Figs. 2 and 3), whereas the peak frequency of these two jobs in Tetragonisca angustula were at almost exactly the same age in Hammel et al. (2016). On the other hand, guarding seemed to be consistently a behaviour taking place just before or concurrently with foraging (Fig. 2; Bassindale and Matthews 1955; Sommeijer 1984; Hammel et al. 2016).

One problem with trying to compare behavioural sequences across studies is that there are inevitably one or more behaviours missing (not observed or observed rarely) from each study that nonetheless must have taken place at some point for the colony to function. For example Mateus et al. (2019) observed pot construction too rarely to include in the analysis and Sommeijer (1984) did not observe it at all, whereas we seldom observed wax deposition or waste removal which was observed very frequently in other studies (e.g. Bassindale and Matthews 1955; Sommeijer 1984), and this probably highlights the need for repeat observations on multiple colonies at different times of year to develop a more complete picture. Nonetheless, the inconsistencies in timing across species hint that whilst brood care as a job for young workers and guarding and foraging as jobs for old workers seems evolutionarily conserved, the relative sequence of the intermediate jobs may be more labile. Whether this variation involves adaptation to different habitats or is merely a consequence of relaxed selection pressure on the relative sequence of within-nest behaviours could be an avenue for future investigation.

Given the evidence that T. carbonaria guards and foragers differ in size, antennal morphology and aggressiveness suggestive of ‘cryptic castes’ (Wittwer and Elgar 2018), it may be that workers develop into foragers or guards as alternative terminal specialisations after performing the same or a similar sequence of tasks initially (see Fig. 2). For example the Neotropical stingless bee Tetragonisca angustula has a physically larger ‘rare elite’ caste that spend more time guarding the entrance rather than foraging as older bees, but still perform a series of within nest tasks as younger bees (Hammel et al. 2016). However, in our study the propolis ring that forms the entrance to the hive box itself was occluded by the wooden box, meaning guards could only be seen when they ventured forwards into the clear tube. If younger guards tend to have safer positions, as observed in soldier termite castes (Yanagihara et al. 2018), the older marked guards we observed may have been guards all along, but merely hidden from view at younger ages. A focussed study on guards, perhaps involving regular sampling inside the propolis-rimmed hive entrance for any bees marked whilst doing other colony tasks would be needed to answer this question.

Amongst the bees entering and leaving the nest, ~ 80% were neither carrying out waste nor carrying pollen into the nest. Whilst some foragers may be unsuccessful, is likely that most returning foragers had collected nectar rather than pollen. If foragers tend to collect either nectar or pollen rather than both together on a foraging trip, this may be related to forager specialisation on resource types, or individual floral constancy (see Grant 1950; Wilson and Stine 1996), both of which have been recorded in stingless bees (Pangestika et al. 2017; Gruter 2020 p. 214). Our findings of 10% and 11.5% of returning foragers carrying pollen are comparable to similar observations on honeybees with 11–13% seen carrying pollen by Ovige and Hoover (2018). Likewise, a study on bumblebees found 10% of returning foragers carried pollen in the early morning, though increasing to 50% later in the day (Peat and Goulson 2005).

In summary, our results provide an outline of colony behaviour in two species of Australian stingless bees for the first time. These findings contribute to a picture of polyethism as a flexible process with a mixture of evolutionarily shared and unique features. Our findings raise a number of questions for future research on stingless bee task allocation. Are bees returning with pollen actually older than nectar foragers, and if so is this a form of age-dependent resource specialisation or merely a result of experience? And are bees building or filling brood cells older than those simply cleaning or inspecting the cells, as is the case with honeybees (Seeley 1982)? In the latter case, observing these types of brood care behaviour is constrained by the observation hives we used, in which brood filling and building behaviours will be seen only when the layer of youngest brood, called the advancing front, happens to be at the top layer (Heard 2016), however targeted experiments with different parts of the brood comb experimentally exposed or timed to a transition point where both open and capped brood cells are visible at the top layer could allow more detailed study of brood care sub-behaviours.

Beyond behavioural observations, improving our understanding of the hormone basis of polyethism in stingless bees would be a valuable step forward. Results to date looking at juvenile hormone in worker task transitions have been inconsistent. Data from some species suggest a similar mechanism to honeybees and wasps with an increase in hormone levels (or up-regulation of genes in the synthesis pathway) driving behavioural progression from nurse to forager, whilst other species show a corresponding decrease (Cardoso-Junior et al. 2017; Brito et al. 2021; de Souza and Hartfelder 2023). In the latter cases, the authors hypothesised that juvenile hormone had not been adapted from its ancestral state and polyethism was regulated by a different process (Cardoso-Junior et al. 2017; de Souza and Hartfelder 2023). An improved understanding of how polyethism works in stingless bees and its variability amongst species has the potential to help us understand the level of predictability in evolutionary processes, and possibly how these behavioural adaptations may be influenced in an increasingly human-modified environment.