Observations on larval cannibalism and other behaviors in a captive
colony of Amblyopone oregonensis

Amblyopone workers with a centipede they have paralyzed

Alex Wild
Department of Entomology
University of California, Davis
Davis, CA 95616
e-mail: alwild [at] ucdavis.edu

Amblyopone is perhaps the most paradoxical element of the North American myrmecofauna. On one hand Amblyopone ants are regarded as being our most primitive formicids, an idea not without merit-  they are an ancient lineage of simple-looking, sluggish ants with an exceptionally basic social repetoire (Holldobler and Wilson 1990, Traniello 1982).  At the same time, we would be hard-pressed to find an ant with a more apparently derived ecological niche.  Most Amblyopone species, to the extent they are known, are specialist predators. A. oregonensis, the western North American species of the present study, appears to prey exclusively on geophilomorph centipedes (Ward 1988).

The oddness of the feeding habits of Amblyopone extends well beyond their prey specialization and has led to the use of the evocative name "dracula ants" (e.g., Saux et al 2004).  Masuko (1986) noted that the Japanese species A. silvestrii practice a form of non-lethal cannibalism where the queens obtain nearly all of their nutrients from the hemolymph of larvae, whose integument they unceremoniously puncture prior to feeding.  This behavior, called larval hemolymph feeding (LHF), has also been seen in the Amblyoponine ant Adetomyrma (B. Fisher, unpublished) and the proceratiine ant Proceratium (Masuko 1986), with a similar behavior expressed in Leptanilla mediated through larval glands (Masuko 1989). 

On a recent expedition to California's Sierra Nevada I had the good fortune to find and capture a colony of A. oregonensis.  I took the discovery as an opportunity to observe Amblyopone's distinctive predatory behavior in captivity and to find out if A. oregonensis engages in larval hemolymph feeding.  I report here my notes and photographs on the colony describing centipede predation, confirm that this species also practices LHF, and make several additional observations.


I collected a colony fragment of A. oregonensis on 25 June 2004 at Fern Glen, 8 km northwest of Quincy, Plumas Co., California (40º00'N 120º59'W, 1030m).  The colony was nesting in a large rotting Pseudotsuga log in an old growth Douglas fir forest (Fig.1).  This is a known population of A. oregonensis, Ward (1988) conducted prey choice experiments with ants collected at the same site. At the time of collection, the colony fragment contained 11 queens, 56 workers, and hundreds of larvae of varying instars. No cocoons were seen in the nest at the time of collection.  A small number of geophilomorph centipedes were also collected from the same log and maintained alive for later feeding to the colony. The nest was then transported to the Sagehen Creek field station near Truckee, California for observation.

The colony was kept alive for 18 days in a petri dish nest lined with moist plaster of paris (Fig. 2).  A  plastic snap-cap vial was attached with tubing to the side of the nest as a foraging area.  This configuration was easy to observe under a microscope, and the ants could be fed without disturbing the brood nest.  I observed the nest for approximately one hour a day at sporadic intervals using a Wild M5 stereomicroscope at 6x - 50x magnification.  I took general notes on the behavior of the ants and photographed some of the behaviors using a Canon digital SLR camera. After 18 days, the colony was killed by freezing and the ants were preserved in 95% ethanol. Voucher specimens have been deposited at the U. C. Davis Bohart Museum collection and at the Museum of Natural History in Geneva, Switzerland.

Figure 1. The collection location of the study colony, an old growth Douglas Fir forest near Quincy, California.

Figure 2. The petri-dish nest that housed the study colony.

(click on images to enlarge)

Bringing down the beast: 
Predatory behavior of A. oregonensis on geophilomorph centipedes

I provided the colony with 10 geophilomorph centipedes (Fig. 3), one per day for ten days.  Four centipedes were captured at Fern Glen when the colony was collected, and six were found under stones at the Sagehen Creek field station.  Geophilomorphs were difficult to locate at Sagehen Creek, and after 10 days I was unsuccessful in finding any additional prey items for the colony.  The ants went without food for the remainder of the observation period. 

I observed the behavior of Amblyopone when introducing 8 of the 10 centipedes to the colony.  The ants acted similarly in all cases.  On sensing the presence of a centipede, the ants turn to face it and show no hesitation in grabbing the centipede with their mandibles and stinging it.  To sting, an ant grasps the prey in her mandibles, bends her metasoma forward under her body and thrusts it, sting extruded, into the prey (Figs. 4-6).  This behavior appears to be the same regardless of the number of ants attacking the centipede.  The centipedes react by pulling away, and a large centipede can drag an attacking ant across the substrate.  The paralyzing effect of the venom is not immediately apparent, although after 10 minutes the animal becomes effectively immotile, the legs still moving slightly in an uncoordinated manner.

Figure 3. A geophilomorph centipede, the prey of A. oregonesis, found guarding her egg cluster in a rotting log at Fern Glen, the collection site of the study colony.

Figure 4. A worker ant sinks her sting deep into the head capsule of a centipede.

Figure 5. Two Amblyopone workers attempt to sting the front end of a centipede while the back end pulls them across the substrate.

Figure 6. Typical stinging behavior in Amblyopone, metasoma projected forward under the body.

The ants can sense when a centipede has been paralyzed.  Once the animal is subdued, the ants cease their stinging behavior and instead masticate the centipede with their mandibles (Fig. 7).  To determine whether the ants’ change in behavior was a reaction to a cue from the centipede or a simple matter of elapsed time, I removed one centipede after it had been stung by a single Amblyopone worker and let it sit in a separate vial overnight before reintroduction to the ants.  The centipede remained paralyzed during this time.  When the ants were allowed to approach this centipede, they did not attempt to sting it again but rather bit and manipulated it with their mandibles, as they do for a more recently-stricken centipede.  The cessation of stinging and initiation of mastication, then, seems to be due to a cue from the centipede.

Figure 7. The ants have paralyzed the centipede and are beginning to work it over with their mandibles.

Figure 8. A worker ant brings a small centipede she has paralyzed back to the nest.

In contrast to reports that Amblyopone sometimes move their larvae to prey (Holldobler and Wilson 1990, pg. 168), I found that A. oregonensis always moves the prey, even the largest centipedes, directly into the brood nest (Fig. 8) and only carry larvae to the centipede once the prey is within the brood nest.  In general, the process from the initial stinging of the centipede to the arrival of the paralyzed centipede in the brood nest takes about an hour.

Once in the brood nest, a centipede is not immediately fed to the larvae.  Rather, the adult ants spend one to several hours masticating the centipede and licking the juices from the centipede’s wounds (Fig. 9).  Ants masticate the centipede by grasping softer places on the centipede integument, often at the base of legs between the sclerites, with the tips of their mandibles and “chewing” by alternately crossing the left then the right mandible over each other several times in a scissoring-motion (Fig. 10).  This process often releases droplets of hemolymph from the centipede, which the adult ants readily imbibe. 

Figure 9. Amblyopone workers masticating a paralyzed centipede.

Figure 10. Mandibular movements used by Amblyopone to masticate the integument of prey centipedes.

Larval numbers on a centipede increase over time as the centipede is ‘softened up’ by adult mastication.  Within a few hours a centipede is covered in larvae, the larval heads and necks often buried deep within the body of the centipede (Fig 11).  In all 10 centipede feedings, the prey was entirely consumed, save the head capsule and other hard sclerotized parts, within 24 hours.

Figure 11. Amblyopone larvae and adult workers feeding on the carcass of a centipede.

Caught in the act of cannibalism: 
larval hemolymph feeding in Amblyopone oregonensis

I was fortunate to observe several instances of larval hemolymph feeding in A. oregonensis during the short period I kept the colony in captivity.  Four observations were from queens and two were from workers.  The adult ants engaging in LHF appear somewhat agitated, they quickly grab and handle larvae in a haphazard manner, squeezing the larvae with their mandibles with more force than is ordinarily used in larval care and transport, and without apparent regard to where on the larva they were squeezing (Fig 12).  An individual larva is usually repositioned and squeezed several times by an adult before either being successfully punctured or passed over by the adult for another larva.  Interestingly, most of the squeezing fails to puncture the larval integument.  However, in all 6 instances where an adult ant succeeded in piercing the larval integument, the ant had managed to sink the tips of its mandibles in between the dorsal sclerites of larval abdominal segments 1 and 2, or between segments 2 and 3 (Fig 13), as though those areas were more vulnerable than elsewhere on the body.  The behavior of the ants did not indicate that the ants were aiming for these puncture points; rather, it appeared to be more a matter of persistence and luck. 

Figure 12. A queen manhandles a larva with her mandibles.

Figure 13. A worker squeezes a larva at a scarring site.

Punctures resulted in a droplet of hemolymph, which the adult ant would rush forward to lap up with her mouthparts.  The extent of the wound is variable.  In one instance, a queen was rewarded not just with hemolymph but with a stream of fat bodies. 

Most of the larger larvae in the colony bore dark dorsal scars, between abdominal sclerites 2-3, 3-4, and sometimes 4-5, the very locations where successful LHF was observed (Fig. 14).  These scars in some cases must pre-date the capture of the colony, as photographs of the brood nest before the colony was removed from the nest log at Fern Glen reveal scars on many of the larvae (Fig. 15).  Not all large larvae have such marks, suggesting that these are indeed scars and not glands or other structures derived through development.

Figure 14. A larva bearing the scars of LHF, the dark spots in between sclerites towards the head. Notice that one of the larvae in the background also displays scars.

Figure 15. The study colony in the nest log before capture.  Some of the larvae show LHF scars.

Workers in the brood nest spent considerable time licking the larvae (Fig 16), especially after carrying larvae in their mandibles.  Such grooming behavior may be related to LHF, as the ants tend to concentrate on the part of the larval body touched by the mandibles even though the amount of force used by the ants is noticeably less than in instances where ants succeeding in producing visible drops of hemolymph from puncture points.

In one instance, a queen behaving in the agitated manner typical for LHF squeezed just posteriad of the larval head capsule and the larva produced a droplet from its mouth that the queen then consumed (as in Fig 17).  It is worth noting that Traniello (1982) reported several observations of this behavior from the closely related A. pallipes. Given the single observation during this study, however, it is probable that this sort of larval trophallaxis is rarer than LHF in A. oregonensis (Traniello 1982). 

Figure 16. A worker grooming a larva.

Figure 17. A queen licks the mouthparts of a larva.

The prevalence of LHF may be related to the availability of prey centipedes in the nest.  LHF was not observed in workers until after I ran out of prey centipedes to feed the colony, although admittedly I only recorded two observations of worker LHF.  In other Amblyopone species, LHF is only known from queens (Masuko 1986).

In most cases larvae did not appear to be seriously harmed by LHF, and given the prevalence of large larvae in the nest bearing scars (> 90%) it is likely that there can be a low direct fitness cost to most LHF.  However, on two occassions larvae were observed being consumed by other larvae and by adult workers from wounds that appears to have been the result of LHF punctures, although I did not witness the LHF directly.  Incidence of larval cannibalism increased as I neared and passed the end of the prey centipede supply, but it is not clear if these were all initiated by LHF events.  Figure 18 shows a larva being consumed by a nestmate larva. 

Figure 18. Larva-on-larva cannibalism.

Differential vulnerability of larval integuments to puncture

Intrigued by the observation that LHF did not appear successful until the feeding ant focused on particular parts of the larval body, I decided to explore the relative susceptibility of different parts of the larval integument to puncture.

Once the colony had been killed at the end of the observation period, I took 20 larvae and subjected them to punctures on three different parts of the body.  First, the middle of the sclerites on the abdomen; second, the sutures between the abdominal sclerites exclusive of the anterior scarring locations; and third, on the scarring locations themselves.  I have observed adult ants attempting to puncture the larval integument in each of these areas.  Holding a larva firmly with forceps to create hydrostatic pressure in the body cavity, I poked the integument lightly with the sharp end of a #2 stainless steel insect pin, attempting to keep the amount of pressure applied by the pin roughly constant.  For each larva I applied this procedure to 6-8 sclerite sites, 6-8 inter-sclerite sites, and 1-3 scarring sites.  There are few scarring sites on each larva relative to the total size of the larva, so there are fewer overall trials for this part of the larval body.

I found that scarring sites were much more easily punctured than either the sclerites or the non-scarring intersclerites (Fig 19).  The pin produced hemolymph in 58% of the scarring site trials, compared to 6% of the non-scarring intersclerite sites and 1% of the sclerite sites.  The scarring sites are indeed more vulnerable to puncture.  However, it isn't clear if this is due to the fact that most of those sites were already scarred.  There were relatively few non-scarred larvae in the colony, and although several were included in the puncture trials it would be informative to replicate this experiment with all unscarred larvae.

Figure 19. Results of the larval puncture assay demonstrating that scarring sites are more vulnerable to puncture than elsewhere on the larva.

That some parts of the larvae are much more vulnerable than others raises the question of why adult ants do not appear to be able to target those sites for LHF, instead of hitting them by what appears to be an inefficient process of trial and error.  It may be that the larvae are adapted to avoid being fed upon too readily by making it difficult to find the sites.  But this possibility raises the issue of why the larvae are vulnerable at all, if there is pressure on them to evolve defensive mechanisms.  There may well be an evolutionary tension between selection on individuals and between selection on colonies.

Miscellaneous observations

Although the study colony contained 11 queens, only three queens had enlarged abdomens indicating ovary development (visible distension between abdominal segments 4 and 5).  Two of these were observed laying eggs in the nest.  Interestingly, these three queens also showed signs of damage beyond anything seen in workers or non-laying queens.  All three were missing parts of their legs in a manner that suggests past fighting (Fig. 20), even though no aggression was observed during the course of the study.  It may be that dominance-related conflict and fighting only takes places at particular times of year, or at particular points in the brood cycle.

One queen was observed several times engaging in what appears to be mate-calling behavior.  She would stand in various places in the nest, raising her abdomen with her genital opening exposed and her sting extruded (Fig. 21). 

Figure 20. One of the laying queens, showing battle scars. She is missing her left hind leg below the coxa.

Figure 21. A queen in mate-calling position.

Cocoon-spinning behavior is essentially as has been reported in other species with enclosed pupae (Holldobler and Wilson 1990).  Adult ants place debris in a bank around larvae of the right age, and the larvae use the debris as scaffolding to produce a network of silk (Fig 22).  Once the cocoon is completed, worker ants remove the debris from the silken casing.

Figure 22. Cocoon-spinning in Amblyopone oregonensis, with a debris scaffolding visable around the larva.


A recent paper by Saux et al (2004) detailing a molecular phylogeny of Amblyoponine ants could have some rather bizarre implications for the ancestral state of all ants, if the specialized predation and larval cannibalism habits reported here and elsewhere in the ants are mapped to Saux et al's inferred maximum parsimony and maximum likelihood phylogenies.  The taxon sampling in that study is complete enough to suggest that the assemblage of taxa that engage in LHF- Leptanilla, Proceratium, and various Amblyoponines- spans the root node of all extant ants.  These taxa are all specialized predators on other arthropods, and interestingly, Leptanilla and many Amblyoponines are restricted to geophilomorph centipedes.  The most parsimonius reconstruction of the common ant ancestor, if Saux et al's phylogeny is correct, shows an ant with behavior much like that of Amblyopone.  Having a centipede predator that feeds from larval hemolymph as the archetypal ant is highly speculative, of course, but still food for a myrmecologist's thoughts.

Literature Cited:

Holldobler, B. and E. O.Wilson. 1990. The Ants. Harvard University Press, Cambridge (USA).

Masuko, K. 1986. Larval hemolymph feeding: a nondestructive parental cannibalism in the primitive ant Amblyopone silvestrii Wheeler (Hymenoptera: Formicidae). Behav. Ecol. Sociobiol. 19: 249-255.

Masuko, K. 1989. Larval hemolymph feeding in the ant Leptanilla japonica by use of a specialized duct organ, the "larval hemolymph tap" (Hymenoptera: Formicidae). Behav. Ecol. Sociobiol. 24: 127-132.

Saux, C., B. L. Fisher and G. L. Spicer. 2004. Dracula ant phylogeny as inferred by nuclear 28s rDNA sequences and implications for ant systematics (Hymenoptera: Formicidae: Amblyoponinae). Molecular Phylogenetics and Evolution 33: 457-468.

Traniello, J. F. A. 1982. Population structure and social organization in the primitive ant Amblyopone pallipes (Hymenoptera: Formicidae). Psyche 89: 65-80.

Ward, P. S. 1988. Mesic elements in the western Nearctic ant fauna: taxonomic and biological notes on Amblyopone, Proceratium and Smithistruma (Hymenoptera: Formicidae). J. Kans. Entomol. Soc. 61: 102-124.

Related Links:
Amblyopone at antweb.org
Amblyopone at myrmecos.net
Amblyopone at Japanese ant image database

 Special thanks to Jo-Anne Holley and Phil Ward for commenting on the manuscript.