Introduction

Many of the complex behaviours displayed by insects are facilitated by learning (Dukas 2008). For example, insects may exhibit extraordinary behavioural plasticity via the associative learning of sensory cues that lead to a reward or a punishment (e.g. Daly and Smith 2000; Matsumoto and Mizunami 2000; Vergoz et al. 2007). In the case of nectar-collecting insects, like bees, this includes learning to associate between a positive food reward (e.g. nectar) and floral cues (such as flower scent). In learning this association, individuals ensure they will return to a profitable food source and be able to find other similar food sources in the environment.

Learning in nectar-feeding insects can be studied under laboratory conditions by observing a proboscis extension response (PER) (Bitterman et al. 1983). This response is a reflex of the insect and is used as an unconditioned response to a conditioned stimulus, in much the same way as salivation was an unconditioned response of dogs in Pavlov’s famous tests (Takeda 1961). That is, the insect is repeatedly offered a food reward following a particular stimulus and, over time (if learning has occurred), will extend its proboscis in response to that stimulus alone. PER tests were originally established via studies on the European honey bee (Apis mellifera) and have been widely used to investigate the learning of olfactory stimuli in particular (Giurfa and Sandoz 2012). PER tests of associative learning have since been adapted for use in other bees (Laloi et al. 1999; McCabe et al 2007; Henske et al 2015), as well as moths (Daly and Smith 2000) and fruit flies (Chabaud et al. 2006).

Stingless bees are the most speciose of all the social bee clades, with over 600 species in tropical ecosystems across the world (Engel et al. 2023). The foraging ecology of stingless bees is relatively poorly known compared to that of honey bees. As stingless bees are also generalists, however, using an exceptionally wide range of floral resources (Bueno et al. 2023), they are presumably capable of learning floral odours and recalling that a particular odour is associated with a reward in subsequent foraging trips. To date, PER has been used to investigate olfactory learning in a handful of stingless bee species from the Neotropics (McCabe et al. 2007; McCabe and Farina 2010; Nocelli et al. 2017) and from Africa (Henske et al 2015), with varying results. For example, Melipona scutellaris demonstrated no odour conditioning of the proboscis extension reflex (Abramon et al. 1999), while differential odour conditioning of the PER was found for naïve workers of Melipona quadrifasciata (McCabe et al 2007) and Meliponula ferruginea and M. bocandei (Henske et al 2015). In Tetragonisca angustula, workers demonstrated learning of a positively conditioned odour via PER only if they had previous in-hive experience with that odour (McCabe and Farina 2010).

Stingless bees might also learn odours other than those of flowers when foraging, such as the odours left behind in the environment by other bees. Many bees will passively deposit chemical “footprints” at flowers in the act of feeding, and social bees may additionally deposit pheromones at flowers to recruit nestmates (e.g. Stout and Goulson 2001; Nieh et al 2004). Other bees may then incorporate such odours into their foraging decisions, being either attracted to, or repelled by, them (Stout and Goulson 2001; Nieh et al. 2004; Lichtenberg et al. 2011; Gloag et al. 2021). The mechanisms that underlie this behaviour remain poorly understood. However, where olfactory associative learning of floral odours is already part of a bee’s foraging ecology, it seems likely that this also contributes to forager’s responses to bee odours at flowers. Indeed, a range of bee pheromones, including some associated with food-marking recruitment in social species, comprise volatiles also produced by plants, such as esters and terpenes (Jarau et al. 2009; Bortolotti and Costa 2014; Beran et al. 2019).

A recent study (Gloag et al. 2021) revealed that foragers of the Australian stingless bee Tetragonula carbonaria and two other Australian stingless bees (T. clypearis and Austroplebeia australis) are all capable of responding to honey bee odours at food sources when making foraging decisions, with foragers attracted to artificial feeders recently used by honey bees. Honey bees are relatively recent arrivals in Australia, having been introduced to the continent around 200 years ago for commercial purposes (Paton 1993). They now exist as a large feral population across the continent in addition to an extensive managed population. Since stingless bees and A. mellifera have come into sympatry in Australia only recently, and exhibit high niche overlap in diets (Elliot et.al 2021), learning is presumably the mechanism by which T. carbonaria can form associations between honey bee odours and food rewards.

In this study, we use a PER protocol to investigate the olfactory learning ability of the Australian stingless bee Tetragonula carbonaria. Specifically, we aimed to (i) confirm that PER can be used to investigate odour learning in this species, and (ii) confirm that foragers are capable of olfactory associative learning across a range of odours (including a heterospecific bee odour: the Nasonov pheromone of A. mellifera) and thus that this form of learning is part of their foraging behaviour.

Methods

We performed a differential conditioning experiment using PER on T. carbonaria workers collected from colonies at The University of Sydney, Australia, during spring–summer 2018–2019. The experimental design was adapted from a protocol used for honey bees (Bitterman et al. 1983). In differential conditioning tests, each bee must learn to respond to a rewarded odour but not an unrewarded one, such that the test provides a within-group control of odour-association (Matsumoto et al. 2012).

We collected exiting foragers from the hive entrance and placed them into a 200 µl pipette tip that had the end cut off to provide an opening of approx. 2 mm diameter. Workers were then harnessed into the tip with masking tape, allowing for their head, antennae and mandibles to move freely. We harnessed 10 to 15 bees at one time. We left the bees inside a box for 2–3 h at 25 °C to become hungry and motivated to learn.

We then subjected the bees to a differential PER conditioning protocol, in which they were presented with two odours alternatively, one rewarded (rewarded conditional stimulus, CS +) and one not (unrewarded stimulus, CS-). The rewarded odour was followed with exposure to a 50% sucrose solution (the unconditioned stimulus, US). We used two sets of odour pairs: (1) lavender (Lavender Oil Spike, David Craig Galenicals) vs. vanilla (Queen Natural Vanilla Extract) and (2) linalool (Sigma-Aldrich) vs. Synthetic Nasonov pheromone. Lavender, vanilla and linalool are all plant odours associated with flowers or fruits (Knudsen et al 1993). Nasonov is a honey bee pheromone used to signal to workers the location of the swarm or nest (Morse and Boch 1971) or, more rarely, food or water sources (Free and Williams 1970; Williams et al. 1981; Fernandez et al. 2003). It is comprised of seven terpenoids: geraniol, nerolic acid, geranic acid, (E)-citral, (Z)-citral, (E-E)-farnesol, and nerol (Pickett et al 1980), but synthetic versions comprising only the citrals, geraniol and acids elicit honey bee responses most similar to natural Nasonov (Free et al. 1981; Schmidt 1994). Here we used a Synthetic Nasonov of equal parts citral (E- and Z-citral; Sigma-Aldrich), geraniol (Sigma-Aldrich), and geranic and nerolic acids (Sigma-Aldrich) (Schmidt 1994). We included Nasonov among our test odours because it is an example of a heterospecific bee odour, unlikely that T. carbonaria foragers were pre-trained to associate it with food (given that honey bees rarely use it at food sources) and the synthetic form can be readily made in a lab.

Each odour pair was tested twice, reversing the rewarded odour (N = 28 bees per test; 4 tests total). Four T. carbonaria colonies provided the foragers for this experiment, with 5–8 workers used per colony per test. Per test, each bee was exposed to 10 trials, where each trial consisted of: (1) exposure to an odour, (2) either presentation of the US (for CS+) or not (for CS−), and (3) scoring for PER. Each odour was presented to the bees five times in a pseudo-random odour (CS−, CS+ , CS+ , CS−, CS−, CS+ , CS−, CS+ , CS+ , CS−). We administrated odours to harnessed bees using an Aqua One Precision Air Pump 2500, which delivered a continuous airflow for the length of the trial. The airflow was directed so it would pass through a 3 mL syringe with a piece of filter paper that had been soaked in 4 µL of the odour. A trial lasted 1 min and consisted of 20 s airflow, 12 s odour (CS) and 28 s airflow, with an inter-trial time ranging from 10 to 15 min depending on the number of bees being trained. For trials of the CS + , the US was presented during the last 3 s of odour presentation after touching the bee’s antennae with a toothpick dipped in US. A positive PER score was only considered if the bee extended its proboscis in the first 9 s of odour onset (i.e. before antennal contact), with the proboscis extending past the end of the mandibles. Individuals that presented a response to the first CS were excluded from the rest of the experiment, as they may have had previous associations with the test odours.

After the 10 trials, to determine if the bees had learnt the odour, we left the harnessed bees in the box for 15 min and then presented them with a non-rewarded presentation of both odours: the “test trial”. This final trial followed the same protocol as previous learning trials. A positive response was recorded if the proboscis of the bee extended past the end of the mandibles before the completion of the odour. A PER during the test trial showed that they had learnt that a reward was associated with that odour. We then compared for each test trial the proportion of responses to the rewarded odour vs. the non-rewarded one using a Fisher’s exact test (Minitab 18 2019). To determine if bees responded equally to all four odours as the conditioned stimulus, we used a Chi-square test of equality of proportions.

Results

T. carbonaria workers were significantly more likely to respond via proboscis extension (PER) to the rewarded odour (CS +) than the unrewarded odour (CS-) during the test trials for all four odours we tested (Fisher’s Exact Tests: vanilla CS+ , N = 9 of 28 vs N = 0 of 28, p = 0.001; lavender CS+ , N = 5 of 28 vs N = 0 of 28, p = 0.05; linalool CS+ , N = 9 of 28 vs N = 1 of 28, p = 0.011; Nasonov CS+ , N = 9 of 28 vs N = 1 of 28, p = 0.011; Fig. 1). Thus, T. carbonaria demonstrated associative learning of all odours.

Fig. 1
figure 1

PER results showing the percentage of T. carbonaria workers (N = 28 in all cases) that responded during the trials of the training phase (left) and the testing phase (right), which was 15 min after training for each of four tests: a, b Lavender CS + vs. Vanilla CS−, c, dVanilla CS + vs Lavender CS−, e, f Nasonov CS + v Linalool CS−, g, h Linalool CS + vs Nasonov CS−

Full size image

A maximum of 32% of bees (9 of 28) responded positively to any CS + odour during either trial or test trials (Fig. 1), and PER rates during test trials for the CS + odour were similar across all four odours (Chi-square test of homogeneity: χ2 = 2.1, p = 0.55).

Discussion

The success of social bees when foraging, like other nectar-foraging insects, is greatly aided by an ability to learn and differentiate between the olfactory cues of different flowers. They must learn to quickly associate an odour with a reward, or lack of a reward, then retain memories of these associations in at least the short term (Dukas 2008). Here, we confirm that T. carbonaria can rapidly learn to associate an odour with a food reward, discriminate between two odours where only one is rewarded, and maintain the memory of these odour-food associations for at least 15 min.

The PER protocol is based on the reflex of the insect proboscis in the anticipation of food, and was developed to test the associative learning of honey bees. Perhaps not surprisingly, therefore, there is a higher rate of responses with this protocol in honey bees (70–100%, Bitterman et al. 1983) compared with most other nectar-feeding insects in which it has been employed to date (e.g. stingless bees: 0–60%, McCabe et al. 2007, Henske et al. 2015; bumble bees: 40%, Laloi, Sandoz et al. 1999). It seems likely that these differences reflect the suitability of the protocol for use in different taxa, rather than differences in true learning ability. T. carbonaria’s response rates of 32% are within the range for other meliponines tested using similar PER protocols (McCabe et al. 2007, Henske et al. 2015). These response rates are sufficient to assess learning with reasonable sample sizes and thus show that PER conditioning is a suitable technique for studying learning in this species. However, further modifications to the protocol we adopted may well improve further on T. carbonaria response rates. For example, “free movement PER” (in which bees are free to walk inside a plastic tube) was recently used in some Neotropical stingless bees to elicit rates of response of more than 50%, compared to 12.5% or less for the same species with a traditional harnessed-bee method (Nocelli et al. 2017).

T. carbonaria was similarly efficient at learning all the four odours we tested. Vanilla, lavender and linalool are all-natural occurring odours or odour blends in plants (Knudsen et al. 1993). Geraniol, citral, geranic and nerolic acids can also be components of floral scents, but when admixed simulate the Nasonov pheromone of honey bee workers (Schmidt 1994). In some contexts, honey bees may use Nasonov when foraging, such as to mark water sources (Free and Williams 1970) or other resources that lack their own odour (Fernandez et al. 2003), but Nasonov is principally used to communicate nest locations rather than food. Consistent with this, none of the T. carbonaria in our trials spontaneously extended their proboscis in response to Nasonov prior to training, suggesting that they had not learnt to associate this particular honey bee odour with food during their own foraging experiences, despite the high density of honey bees at our study site. Honey bees also leave behind odour deposits at their food sources from their tarsal gland, a type of chemical ‘footprint’ (Stout and Goulson 2001). We consider that previous evidence showing T. carbonaria’s preference for feeders recently visited by honey bees (Gloag et al. 2021) are best explained by the learning of honey bee footprint odours, rather than Nasonov. Indeed, testing T. carbonaria’s response to honey bee footprints via either PER or other means would be an interesting next step in understanding the foraging interactions of the two species. Nevertheless, the ability of T. carbonaria workers in this study to learn to associate a honey bee pheromone with a reward supports the view that T. carbonaria’s capacity for olfactory learning extends to odours deposited by other bees on flowers (both pheromones and footprint cues). That is, individual T. carbonaria could learn over the course of repeated foraging bouts that certain interspecific odours are regularly associated with food (or the absence of food) and respond accordingly.

The idea that bees’ responses to interspecific pheromones and other odours at flowers is based on short-term associative learning, rather than fixed behaviour patterns or predispositions for certain odours, is consistent with current understanding of the critical importance of odour learning for bee foraging (Giurfa and Sandoz 2012). It also has two important implications for understanding this foraging strategy. First, it indicates that the use of interspecific odours at flowers must be a highly flexible strategy, with foragers capable of changing their responses day to day depending on their foraging experiences. Responses to heterospecific odours at food sources will, therefore, be highly context-dependent. Second, it indicates that the use of heterospecific odours when foraging could readily occur between native and introduced species; that is, there is no requirement for an extensive period of sympatry before these interactions arise. As such, the indirect responses of bees to the odours of heterospecifics at flowers could be an overlooked type of interaction between native and introduced bees (Gloag et al. 2021). The extent to which learning between a given species pair is also facilitated by similarity in the odours of each species (e.g. shared chemistry of glandular compounds), however, remains to be investigated.

Many aspects of the behaviour and cognition of T. carbonaria, and of stingless bees more generally, remain to be investigated. This study confirms that the PER protocol can be applied to T. carbonaria, and so opens up the possibility of using this protocol in the future to further explore the cognitive ability of this species, and in turn improve our understanding of their foraging behaviour and ecology.