Skip to main content

Thank you for visiting You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

How aggressive ant-guards assist seed-set in Acacia flowers


The phenomenon of ant-guarding on Acacia trees is probably the best known case of a mutualism between plants and animals, the ants conferring biotic defence against herbivores and perhaps against encroaching vegetation1,2,3. However, as with many defence mutualisms, sometimes the interests of the plant and its defender conflict: for example, when they are in flower the Acacia trees require the presence and service of other insects to effect cross-pollination. How is pollinator access achieved in the face of aggressive ant-guards? Here we report that ants are deterred from young flowers at the crucial stage of dehiscence, allowing bees and other pollinators to visit and transfer pollen. This deterrence appears to be a response to a volatile chemical signal from young flowers, perhaps from the pollen itself. Ants patrol the young (undehisced) buds, and also return to the flowers after dehiscence, protecting the fertilized ovules and developing seeds. The outcome is a directly improved seed-set in the presence of ants (rather than an indirect extra reproductive resource allocation due to decreased defoliation4,5,6).


Acacia zanzibarica and A. drepanolobium are classic ant-guarded trees of eastern Africa7; ants (such as Crematogaster8) reside on the trees inside modified thorns (pseudogalls) and patrol the branches, attacking any insect or vertebrate herbivores encountered. As in other guarded Acacias1,2,3, herbivory is thereby much reduced. Cross-pollination could be allowed through reduced ant activity during flowering but, as flowers appear soon after seasonal rains and accompany the flush of vegetative growth, this would leave the tree foliage and the flowers themselves vulnerable to damage.

A. zanzibarica and A. drepanolobium flowers occur in spherical inflorescences of 20–100 florets bearing conspicuous exposed anthers. Flowering occurs mainly in November–January in Mkomazi. The flowers open at around 05:00 and start to dehisce at 08:00–09:00 (ref. 9). They appear pristine through day 1, persisting through day 2 in a tattered non-functional state before being shed, leaving a small proportion of ovules effectively fertilized. Seed-set is apparent within 2–3 days, although seeds take about a month to mature.

Flower visitors operate on clear diurnal patterns9. As in other Acacias10, most pollen collection is carried out by solitary bees (Megachilidae and Halictidae), making 98% of visits to A. drepanolobium and 40% of visits to A. zanzibarica. Bee activity begins from first dehiscence and peaks at 11:00–14:00 (ref. 9); around 75–80% of available pollen is removed within these 3 hours. The flowers contain little or no nectar and, unlike some sympatric Acacia species, are therefore relatively unattractive to nectarivorous pollinators such as butterflies9. However, flower predation by beetles (melioids and scarabaeids) is a significant problem, with anthers, stigmas and ovaries being chewed in about 10% of flowers.

Ants also show a daily activity pattern (Table 1), patrolling twigs adjacent to their pseudogall in the early morning, declining in numbers from 09:00–10:00, only rarely in evidence until around 14:00, then increasing during most of the afternoon. Therefore, there are strictly defined times when ants are not present on flowers, corresponding to peak dehiscence and visitation of flowers. Because of the temporal patterning of their behaviours, pollinators meet relatively few ants on the flowers. Nevertheless, ants and pollinators such as bees do overlap in their flower-visiting activities, especially at the onset of dehiscence. How then is conflict avoided?

Table 1 Numbers of patrolling ants

Ant activities in relation to flower age (Fig. 1) show that ant–flower interactions are more subtle than indicated by an assessment of mere presence or absence. In the early morning, the ants visit more of, and make longer visits to, older (day-2) flowers and the unopened buds (not shown), but significantly avoid younger flowers and stay only briefly (often for 1–5 seconds) when they do encounter a young flower. During such visits, they also differ in their behaviour, with their antennae and mandibles being unusually active and the abdomen often being cocked upwards. Ants could be reacting to flower occupancy by pollinators, but ‘occupied’ flowers are rare at these times (07:00–10:00) and ants make equally short visits to both occupied (6.47 ± 7.20 s; n = 11) and unoccupied (7.15 ± 5.45 s; n = 75) young flowers. In contrast, when ant activity increases again in the afternoon, younger flowers are intensively visited and for significantly longer than during morning visits (Table 2: age, not significant; time, P = 0.000; time/age interaction, P = 0.005). These patterns could be explained by young flowers producing a deterrent to ants when they first dehisce.

Figure 1

Duration of visits of young and old flowers of Acacia zanzibarica through the diurnal cycle, as mean values (± s.e.m.) from 4 days of observation.

Table 2 ANOVA for duration of ant visits

The changing response of ants to young flowers suggests that there is a progressive decay of ant-repellant stimulus after dehiscence. This could result from a short-lived volatile chemical signal from the young flowers. Some flowers do contain ant-repellant nectar11,12, but this is an implausible source of repellant in Acacia zanzibarica, where nectar volumes are negligible. The repellant must be present either on the corolla surfaces or on the pollen itself. It might therefore be transferable by contact between flowers, either by direct wiping of chemical(s) or by transfer of whole pollen grains/polyads. To test this, we watched ant behaviour on selected A. zanzibarica trees, treated as described under Methods. Figure 2 shows that old flowers receive longer morning visits than young ones, as expected (Fig. 1); wiping old flowers with other old flowers has no effect, whereas wiping with young flowers produces avoidance and shorter visits, as for untreated young flowers (ANOVA, F = 17.63, P = 0.000), and also elicits an exaggerated behavioural response. Some component transferable from the young dehiscing flowers (therefore a chemical rather than a tactile cue) is acting as an ant deterrent. If freshly dehiscing flowers have a volatile ant-deterrent, this also explains why just one day of observation (27 November 1995) gave ant-visit durations that were invariant between young (40 ± 11; n = 12) and old (34 ± 6 s; n = 9) flowers; rain on that morning had presumably washed away the deterrent cue.

Figure 2

Duration of visits (mean ± s.e.m.; n above bars) between 07:00–09:30 from two days of observation with four age/treatment groups of A. zanzibarica flowers (see Methods).

In this model, ants could aid Acacia pollination by keeping the very young flower buds pristine and uneaten, then by their absence allowing access to pollinators at peak flower fertility (with the small cost of losing some young flowers), and finally by returning to deter flower-feeding beetles and seed predators. Therefore, if ant abundance per branch is consistent over the flowering and fruiting period, it should be positively related to eventual seed-set on a given branch. Counts of ants per branch on 10 branches from a single tree over 4 days show a highly significant rank effect of branch (Friedman, P = 0.006), indicating that each branch has a rather fixed number of patrollers over time (as expected, given the presence of pseudogalls of fixed size and number). To show the effects of ant activities on overall pollination and seed-set, we therefore compared five branches from each of eleven A. zanzibarica trees (Fig. 3). The mean ant abundance per branch is positively correlated with the mean fruit-set per inflorescence for each of the 11 trees. Using a repeated measure ANCOVA (Table 3) with ‘trees’ as units and ‘branches’ as repeats, the relationship between ant activity and fruit-set is highly significant (P = 0.000). None of the other ‘non-ant’ factors measured (see Methods) is a significant predictor of fruit-set. Nor are there likely to be environmental factors differing between the 11 adjacent trees that could independently affect both ant abundance and fruit-set5,6. Therefore the presence of guarding ants per se can be taken to assist fruit-set in this Acacia species.

Figure 3: Effects of ant-guard abundance on seed-set.

Ant abundance is the number per branch for 5 branches on each of 11 trees; seed-set was scored on the same branch lengths. The line shown is the best-fit linear regression.

Table 3 Repeated-measures ANCOVA for fruitset


All studies were carried out on Acacia trees in the western end of Mkomazi Game Reserve, Northern Tanzania. Flower visiting was scored by quiet continuous observation of tagged 0.6-m branch lengths, with individual flowers noted and aged. To assess the consistency of ant activity patterns, patrolling ants were scored at different times of day over 4 days (on 22, 24, 27 and 28 November 1995) on ten 0.5-m tagged lengths of branch on one tree.

At about 07:00 on two days, we set up and tagged flowers, treated in the following categories: untouched young flowers; untouched old flowers; young flowers ‘wiped’ onto old flowers; old flowers ‘wiped’ onto old flowers, as a control for handling effects. Wiping involved smearing the recipient flower's surfaces with either a young or an old flower from the same tree; the picked ‘donor’ flower, held in forceps, was moved three times over the recipient flower to give contact with most of the spherical exposed surface of massed anthers and stigmas.

To assess ant effects on fruit-set, we selected eleven A. zanzibarica trees, all growing close together in a small stand on flat ground to reduce edaphic variations, and tagged equal 0.5-m lengths of five branches on each tree. We then scored ant abundance on each branch totalled across 12 sampling times (07:30–14:30) on 2 December 1995. Numbers and sizes of pseudogalls on that branch were also scored, together with numbers of flowers of each age, the height, aspect and shading of branch, and the tree size (height and trunk girth), to rule out other possible causes of variation in seed-set. Finally, we counted all seeds on each former flower-head site separately (range, 0–11 per flower site, with 0–39 sites per branch).


  1. 1

    Janzen, D. H. Coevolution of mutualism between ants and Acacia in Central America. Evolution 20, 249–275 (1966).

    Article  Google Scholar 

  2. 2

    Hocking, B. Insect associations with the swollen-thorn acacias. Trans. R. Ent. Soc. Lond. 122, 211–255.

  3. 3

    Knox, R. B., Marginson, R., Kenrick, J. & Beattie, A. J. in Insects and the Plant Surface (eds Juniper, B. E. & Southwood, T. R. E.) 295–307 (Arnold, London, (1986)).

    Google Scholar 

  4. 4

    Del-Claro, K., Berto, V. & Reu, W. Effect of herbivore deterrence by ants on the fruit set of an extrafloral nectary plant Qualea multiflora (Vochysiaceae). J. Trop. Ecol. 12, 887–792 (1996).

    Article  Google Scholar 

  5. 5

    Niesenbaum, R. A. Linking herbivory and pollination: defoliation and selective fruit abortion in Lindera benzoin. Ecology 77, 2324–2331 (1996).

    Article  Google Scholar 

  6. 6

    Koptur, S. in Insect–Plant Interactions (ed. Bernays, E.) 81–129 (CRC Press, Boca Raton, (1992)).

    Google Scholar 

  7. 7

    Ross, J. H. An analysis of the African Acacia species: their distribution, possible origins and relationships. Bothalia 13, 389–413 (1981).

    Article  Google Scholar 

  8. 8

    Young, T. P., Stubblefield, C. H. & Isbell, L. A. Ants on swollen thorn Acacias: species coexistence in a simple system. Oecologia 109, 98–107 (1997).

    ADS  Article  Google Scholar 

  9. 9

    Stone, G. N., Willmer, P. G. & Nee, S. Daily partitioning of pollinators in an African Acacia community. Proc. R. Soc. Lond. B 263, 1389–1393 (1996).

    ADS  Article  Google Scholar 

  10. 10

    Tybirk, K. Pollination, breeding system and seed abortion in some African acacias. Bot. J. Linn. Soc. 112, 107–137 (1993).

    Article  Google Scholar 

  11. 11

    Feinsinger, P. & Swarm, L. A. How common are ant-repellent nectars? Biotropica 10, 238–139 (1978).

    Article  Google Scholar 

  12. 12

    Prys-Jones, O. E. & Willmer, P. G. The biology of alkaline nectar in the Purple Toothwort (Lathrea clandestina): ground level defence. Biol. J. Linn. Soc. 45, 373–388 (1992).

    Article  Google Scholar 

Download references


We thank D. Mafunde at Mkomazi Research Station for expert field assistance; T. Morgan and N. McWilliam for invaluable logistic back-up; and M. Ritchie and J. Graves for statistical advice. We thank the Royal Geographical Society, the Darwin Initiative and the British Council for financial support, as part of the Mkomazi Ecological Research Programme, and the Department of Wildlife of the Tanzanian Government.

Author information



Corresponding author

Correspondence to P. G. Willmer.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Willmer, P., Stone, G. How aggressive ant-guards assist seed-set in Acacia flowers. Nature 388, 165–167 (1997).

Download citation

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.


Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing