ANIMAL BEHAVIOUR, 2000, 60, 13–26 doi: 10.1006/anbe.1999.1364, available online at http://www.idealibrary.com on ARTICLES Acoustics, context and function of vibrational signalling in a lycaenid butterfly–ant mutualism MARK A. TRAVASSOS & NAOMI E. PIERCE Museum of Comparative Zoology, Harvard University (Received 8 April 1999; initial acceptance 29 July 1999; final acceptance 17 November 1999; MS. number: A8282) Juveniles of the Australian common imperial blue butterfly, Jalmenus evagoras, produce substrate-borne vibrational signals in the form of two kinds of pupal calls and three larval calls. Pupae stridulate in the presence of conspecific larvae, when attended by an ant guard, and as a reaction against perturbation. Using pupal pairs in which one member was experimentally muted, pupal calls were shown to be important in ant attraction and the maintenance of an ant guard. A pupa may use calls to regulate levels of its attendant ants and to signal its potential value in these mutualistic interactions. Therefore substrate-borne vibrations play a significant role in the communication between J. evagoras and its attendant ants and pupal calls appear to be more than just signals acting as a predator deterrent. Similarly, caterpillars make more sound when attended by Iridomyrmex anceps, suggesting that larval calls may be important in mediating ant symbioses. One larval call has the same mean dominant frequency, pulse rate, bandwidth and pulse length as the primary signal of a pupa, suggesting a similarity in function. 2000 The Association for the Study of Animal Behaviour From the foot-drumming of a banner-tailed kangaroo rat, calls is ant attraction. DeVries concluded: ‘under selection Dipodomys spectabilis, in the presence of a snake (Randall for symbiotic associations, the calls of one insect species & Matocq 1997) to the coordinated group chorusing of have evolved to attract other, distantly related insect nymphal treehoppers (Cocroft 1996), vibrational signal- species’ (DeVries 1990, page 1106). ling is a widespread form of communication, functioning Both larvae and pupae of some species of Lycaenidae primarily in defence, mate attraction and displays of can produce sound (Dodd 1916; Downey 1966; DeVries aggression. Instances of such communication between 1990). Larvae produce vibrations that are primarily sub- unrelated species, however, are relatively rare. Neverthe- strate borne, although they may have a slight airborne less, recent work indicates that vibrational communi- component (DeVries 1991a; M. Travassos, personal cation may play a vital role in butterfly–ant mutualisms, observation), whereas pupae produce signals with both wherein caterpillars and pupae use an intricate combi- vibrational and airborne components (Hoegh-Guldberg nation of chemical, behavioural and secretory signals to 1972; Downey & Allyn 1978). In pupae, a file of teeth on maintain a retinue of ants that protect them from pred- the anterior side of the sixth abdominal segment grates ators and parasites. DeVries (1990) showed that cater- against an opposing plate on the posterior side of the fifth pillars that interact with ants are capable of producing abdominal segment (Downey 1966). Such a plate may be vibratory ‘songs’. These larvae are found exclusively in made up of either tubercles, reticulations, or ridges. The the butterfly families Lycaenidae and Riodinidae. In com- mechanism of sound production in lycaenid larvae, how- paring ant attendance levels between control larvae of the ever, has proved more elusive. Hill (1993) proposed a riodinid Thisbe irenea and larvae that had been experi- possible stridulatory organ similar to that found in pupae: mentally muted, calling T. irenea caterpillars were tended a file of teeth and an opposing plate. Caterpillars in the by more ants, indicating that one function of riodinid family Riodinidae, the sister taxon to the Lycaenidae (Kristensen 1976; D. Campbell, A. Brower, N. Pierce, Correspondence: N. E. Pierce, Museum of Comparative Zoology, unpublished data), signal by beating vibratory papillae Harvard University, 26 Oxford Street, Cambridge, MA 02138, against epicranial granulations; lycaenids lack such a U.S.A. structure (e.g. Cottrell 1984). 0003–3472/00/070013+14 $35.00/0 13 2000 The Association for the Study of Animal Behaviour 14 ANIMAL BEHAVIOUR, 60, 1 At least half of all lycaenids interact with ants, varying abdominal segment, larvae have a pair of eversible ten- from facultative interactions where juveniles are found tacle organs (TOs), believed to release a volatile chemical with or without ants, and often with many species, that alerts attendant ants if a larva is alarmed or the DNO to obligate ones where juveniles cannot live without is depleted (Henning 1983; Fiedler & Maschwitz 1988b). ants and usually associate with only one or two closely Late-instar larvae and pupae of J. evagoras stridulate related species (Pierce 1987; Fiedler 1991). Not only are when disturbed (Pierce et al. 1987). Like other lycaenids myrmecophilous lycaenids protected against ants them- (Downey 1966), pupae have a file-and-plate stridulatory selves, which might otherwise be threatening predators organ between the fifth and sixth abdominal segment. (Malicky 1970), but it has been shown experimentally The pupal plate extends the length of the intersegmental that attendant ants also protect juveniles from predators region and has a series of ridges against which the teeth and parasites (e.g. Pierce et al. 1987; Fiedler & Maschwitz grate. DeVries (1991a, page 17) found that pupae produce 1988a; DeVries 1991c; Cushman et al. 1994; Wagner 7.5 ‘metallic click-like pulses’ per second, each with a 1994). Dodd (1916) and DeVries (1990) suggested that frequency of 2300 Hz, and that larvae of J. evagoras lycaenid caterpillar calls, which are essentially substrate- produce a drumming call resembling a ‘khen-khen-khen- borne vibrations, may be important in mediating ant khen’ at a rate of 7 calls/s and a mean frequency of interactions, because ants, although nearly deaf to air- 1700 Hz. Our study aimed to investigate the parameters borne sounds, are sensitive to vibration (Fielde & Parker of these calls in greater detail and to explore their func- 1904; Ho¨lldobler & Wilson 1990). Lycaenid larvae are tional significance. We focus on the vibrational compo- able to produce several different vibrational signals. nents of these sounds, measured with an accelerometer. DeVries (1991a) noted that some lycaenid sounds have Throughout, except where noted otherwise, we use the two components: a low background sound accompanied terms ‘sounds’ and ‘calls’ for simplicity in describing by a louder pulsing. Leptotes cassius, for example, has ‘a these substrate-borne vibrations. Nothing is known of the ticking background and an irregular, galloping series of sensitivity of lycaenid larvae and pupae to airborne versus trills’ (DeVries 1991a, page 17). No experimental work has substrate-borne signals; ants are nearly deaf to airborne examined the function of lycaenid larval calls. sound, but sensitive to vibrations (Fielde & Parker 1904; Research on the function of pupal calls has been incon- Ho¨lldobler & Wilson 1990). clusive. Functional explanations for lycaenid pupal sig- nals have speculated about their role in defence (Hinton GENERAL METHODS 1948; Downey & Allyn 1978), conspecific attraction in the formation of aggregations (Prell 1913), and The study was conducted in the Museum of Comparative myrmecophily (Downey 1966; Elfferich 1988; Brakefield Zoology Laboratories from June 1996 to February 1997. et al. 1992). The acoustic characteristics of such signals We collected J. evagoras eggs from field sites in Ebor, are better known. Lycaenid pupae produce three distinct New South Wales (3024S, 15219E), Mount Nebo, signals, each distinguishable by amplitude level (Downey Queensland (2724S, 15247E), and Canberra, Australian & Allyn 1978). The loudest pulse, the primary signal, is Capital Territory (3521S, 14856E), Australia. Queen- also the longest. Secondary signals are briefer and are right colonies of Iridomyrmex anceps maintained in the often found in pulse trains. Tertiary signals have only laboratory were collected from Mount Nebo, Canberra, been recorded in large pupae and consist of irregular click and Griffith, Australia (2733S, 1533E). Ant colonies trains not much louder than background noise. Although were fed an artificial diet (Bhatkar & Whitcomb 1970) and Downey (1966) and Hoegh-Guldberg (1972) reported a chopped crickets daily. Larvae of J. evagoras were reared correlation between primary signal production and on Acacia melanoxylon and A. irrorata raised from seed movement between the fifth and sixth abdominal seg- purchased from the Queensland Forestry Department. ments, no observations of abdominal movement have All calls were recorded in an experimental arena been made regarding the production of secondary and measuring 13266 cm and 132 cm high with two verti- tertiary signals. cal surfaces covered in black construction paper to reduce We analysed the acoustic properties, context and func- the effects of external stimuli such as sunlight. To allow tion of sound production in juveniles of the Australian for manipulation of the set-up, the other two vertical lycaenid Jalmenus evagoras. Larval and pupal J. evagoras surfaces remained uncovered. The room was maintained associate with ants of several species in the genus Irid- at a constant temperature of 22C and had fluorescent omyrmex. In return for producing nutritious