S03.3: Social and cognitive implications for learning in the complex gargle calls of chickadees

Gargles are aggressive and sexual calls of many tits, studied mainly in the Black-capped Chickadee Parus (Poecile) atricapillus. Calls are about 0.5 sec long and composed of up to 13 different notes. A bird uses several gargles, shared by local birds. Birds 10 km away share some gargles, those 130 km away share none. Birds hand-reared in groups develop abnormal, shared gargles. This fact plus dialects means that gargling is mostly learned. Gargles are probably learned when a young bird disperses to a different dialect area and interacts in gargling duels with older, resident birds. Copy errors reject rote serial learning because the middle (core string) is best preserved rather than the start of the call. Simple ‘genetic’ types of errors (transposition, repetition, deletion, insertion, and substitution of single notes) almost never occur, but gargles are often truncated at beginning or end (or both). Finally, the same prefixed or suffixed notes can be added to different types of gargles. A principal aim of this contribution to the symposium is to show the relevance of field studies to social and cognitive issues in avian vocal learning. The aim is pursued through the example of ‘gargles,’ which are complex vocalisations used in close aggressive and sexual encounters by certain species of tits (Paridae). For brevity the verb ‘to gargle’ will mean ‘to utter gargle vocalisations,’ and ‘gargler’ will mean ‘the individual that utters the gargle vocalisations.’ I attempt to integrate all the germane information about gargles found in the literature, with emphasis on quantitative data derived from the award-winning senior thesis of Cortland K. Griswold (Hailman & Griswold 1996). The three major parts of this paper are: some necessary background about gargles and the birds that use them, social aspects of gargle learning, and cognitive implications of quantitative analyses.

The elements of useful background are provided in this major section. First presented is a brief history of the discoveries of gargles in the species known to use them. Then the complicated acoustic characteristics of gargles are explained. The section concludes with a discussion of the poorly understood communicative functions of gargles.

Gargles are one of three major classes of vocalisations used by tits of the Poecile group, treated by most past authors as a subgenus of Parus but recently elevated to generic status by the check-list committee of the American Ornithologists' Union (1997). The Poecile tits are all those North American species known as chickadees plus certain Eurasian species such as the Willow Tit P. montanus, Marsh Tit P. palustris, and Siberian Tit P. cinctus. All Poecile species whose vocal repertoire has been studied are known to possess gargle systems, which are similar among species, and some other kinds of tits appear to have combinatorial vocal systems that may be the general equivalents of gargles (Hailman 1989).

Poecile tits have three major kinds of vocalisations: gargles, highly combinatorial 'chick-a-dee' calls, and a species-specific vocalisation that is often termed the 'song' of the species (Hailman & Ficken 1996). Much controversy surrounds the interpretation of these 'song' vocalisations, and it has been argued that the three major vocalisations taken together function like true song of typical oscines (Hailman 1989). Gargles have been studied extensively only in the Black-capped Chickadee Parus (Poecile) atricapillus, whose 'song' is the 'fee-bee' vocalisation.

Gargles were described spectrographically beginning about 1970, but no homologies were recognised among species so when the calls were named, they were given different names in different species. Some early verbal descriptions of the Black-capped Chickadee's vocalisations probably refer to gargles (e.g. Odum 1942), but the first published sound spectrograms of a gargle appear to be those of Dixon and Stefanski (1970), who termed it a ‘fighting call,’ and Dixon et al. (1970), who termed it a ‘supplanting call.’ S. T. Smith (1972) showed sonograms of a variety of named calls of the Carolina Chickadee P. carolinensis, her ‘T-slink,’ ‘click-rasp’ and similar onomatopoetically named vocalisations all appearing to be gargles. Ludescher (1973) and later Romanowski (1978) showed figures from Willow Tit vocalisations of what may now be interpreted as gargles. The early spectrographic evidence for the Mountain Chickadee P. gambeli is less certain (Dixon et al. 1970, Dixon 1972), but this species certainly does have gargles (J.P.H. unpublished tapes, McCallum personal communication). Spectrograms of Marsh Tit vocalisations clearly include gargles (Ludescher 1973, Latimer 1977, Romanowski 1978). Similarly, sonograms of the Siberian Tit include calls now known to be gargles (Haftorn 1973), as probably do those from the Boreal Chickadee P. hudsonicus (McLaren 1976).

It remained for Ficken et al. (1978) to provide the name ‘gargle’ and to describe preliminarily the nature of these combinatorial calls as part of their pioneering repertoire study of the Black-capped Chickadee. That paper and Millicent Ficken's subsequent seminal papers on gargles, cited below, eventually allowed retroactive identification of gargle systems in the species mentioned above. In addition, at least one further Poecile species was shown to have a gargle system: the Mexican Chickadee P. sclateri (Ficken 1990). In all Poecile species gargles appear to be generally similar (Hailman & Ficken 1996), but as mentioned previously only the Black-capped Chickadee has been studied in detail. Nevertheless, enough is known about the Mexican Chickadee's gargles (Ficken 1990) to be fairly certain that they differ in details from those of the Black-capped Chickadee (see analysis in Hailman & Griswold 1996). Therefore, the similar gargle systems of different species are not necessarily identical. All statements about gargles hereafter refer specifically to the Black-capped Chickadee unless otherwise indicated.

Gargles are remarkably short (about 0.5 sec) calls composed of many extremely brief notes. A ‘note’ is a continuous trace in a sound spectrogram (or simultaneous traces at different acoustic frequencies). The mean number of notes per call was between 5.7 and 7.3 at five sites in southeastern Wisconsin (Ficken & Weise 1984), with a similar finding in south-central Wisconsin (Hailman & Griswold 1996). The range of composition was found variously as 2 to 9 notes per call (Ficken et al. 1978), 2 to 13 (Ficken & Weise 1984), 1 to 11 (Ficken & Popp 1992), 1 to 9 in a spectrographed sample and 1 to 13 in a total sample (Hailman & Griswold 1996). To the human listener gargles sound like one or two slurred notes, hence the verbal names such as T-slink given by S. T. Smith (1972) for gargles of the Carolina Chickadee. The experiments of Dooling (e.g. 1982) on various species suggest that the superior temporal resolution of birds allows chickadees to hear the separate notes of gargles that our auditory systems cannot resolve. Indeed, if the birds could not perceive the separate notes, it would seem impossible for them to learn the gargles of a local dialect (discussed below).

The ways in which notes compose unit gargle calls show some trends of constraint. Gargles are ‘typically composed of an array of different notes’ (Ficken & Popp 1992: 683), where ‘typically’ can be specified as 81% of 84 gargle types sampled having all different note types (Hailman & Griswold 1996). The rare repetition (Ficken & Popp 1992) is in fact somewhat more frequent than ‘rare’ might suggest: 14.3% of 84 gargle types showed reduplication of at least one note (Hailman & Griswold 1996). Nevertheless, these reduplications occurred only in prefixed notes, a concept explained later in this contribution. The number of different note types used in a local population has been variously determined as 14 (Ficken et al. 1978), 16, 17 and 23 at three different sites in southeastern Wisconsin (Ficken & Weise 1984), and 37 in a population in south-central Wisconsin (Hailman & Griswold 1996). The component notes show a ‘trend for descending pitch within the call’ (Ficken & Popp 1995: 683), evaluated quantitatively to mean that 88.1% of 84 gargle types showed a strictly descending acoustic frequency of successive notes (Hailman & Griswold 1996). Finally, Ficken & Popp (1995: 683) stated that ‘trills, if they occur in the call, [are] towards the end,’ and all nine gargle types having trills were found to have them as the terminal note (Hailman & Griswold 1996).

Every unique sequence of notes composing a gargle is a ‘gargle type,’ and an individual chickadee possesses a repertoire of gargle types much like many songbirds possess repertoires of song types. The maximum repertoire size of gargle types reported by Ficken et al. (1987) was 29; Hailman and Griswold (1996) found a bird with 24 gargle types. These large repertoire figures are misleading, however, because not all gargles are equally distinct. Indeed, the major discovery that reorients much interpretative thinking about gargles is that they can be sorted objectively into groups of similar structure (Hailman & Griswold 1996). Nevertheless, individual chickadees still have repertoires of gargle groups, the maximal size found being seven (Hailman & Griswold 1996).

The contextual uses of and the information encoded in gargles remain to be clarified fully. An aggressive nature of these calls was obvious from their earliest spectrographic characterization as reflected in the approbations ‘fighting call’ (Dixon & Stefanski 1970) and ‘supplanting call’ (Dixon et al. 1970). Ficken et al. (1978) avoided functional names for vocalisations and presumably coined the term ‘gargle’ because of an imagined resemblance to the sound of a person gargling. That paper seems to contain the first explicit statements that gargles are given in both agonistic and sexual contexts, and that the gargling bird is always close to a conspecific that is apparently the intended receiver of the vocal signal.

Both sexes gargle but males clearly gargle more than do females. Ficken et al. (1987) found that 79.0% of 62 individual chickadees recorded gargling were males, and this figure replicated remarkably closely in our population where 78.3% of 23 sexed garglers were known males (Hailman & Griswold 1996). The evident sex bias fits the interpretation that gargles are used in both aggressive and sexual contexts because male chickadees are generally more aggressive than females, and because the gargling in precopulatory sequences seems to be done mostly if not exclusively by the male (JPH personal observations).

The first direct data on the aggressive use of gargles were reported by Ficken et al. (1987), and the important subsequent research has come from experimental work in the laboratory of Myron Baker. Ficken et al. (1987) found that in 98.9% of 846 cases of agonistic encounters at a feeder, the bird that gargled was the winner. Furthermore, within a dominance hierarchy, high ranking birds gave more gargles than did those of low rank.

Baker et al. (1991) paired males not previously known to one another and found that the birds established relative dominance through fighting, with the larger bird winning. Beginning early in the interactions, the dominant bird began accompanying aggressive acts with gargling and the subordinate's gargling was suppressed. Experiments then tested the relative effectiveness of optical and acoustic stimuli in forcing a bird away from its partner's cage and in deterring it from approaching a food source. Visual stimuli had little effect on a dominant's spatial movements, but playback of the subordinate's gargles (both with and without the subordinate being present in the adjacent cage) deterred the dominant from food. By contrast, sight of the dominant (with or without playback of his gargles) caused the subordinate to move away from the adjacent cage whereas playbacks alone had no effect. Curiously, sight of the dominant coupled with playbacks of his gargles was the only treatment effective in deterring subordinates from food. A simple, straightforward interpretation of these interesting results seems elusive.

In another study Baker et al. (1996) used video and well as audio playback and eliminated entirely live birds as stimuli. In this study the combination of optical and acoustic stimuli was the most effective threat. Perhaps more importantly, when the audio playback had three kinds of gargles, the threat was absolutely more effective than when it had just one kind of gargle although the result was not statistically significant.

Very little indeed is understood about the sexual use of gargles. Ficken et al. (1985) proposed that different kinds of gargles were used in aggressive and precopulatory contexts, but their evidence was little more than suggestive (see Hailman & Griswold 1996). My laboratory tried to evaluate this hypothesis by comparing gargles given within pairs in the spring, including a few sequences believed to be precopulatory, with gargles recorded by Griswold in flocks the previous fall and hence presumably agonistic in nature. All of the gargles given in these ‘sexual’ contexts could be assigned to gargle groups recorded in the ‘agonistic’ contexts, so we failed to find anything unique about sexual gargles (JPH unpublished data). More detailed studies, however, are needed for a conclusive resolution of the ‘sexual gargle’ hypothesis.

Other proposals by Ficken and Weise (1984) concerning special note sequences having special meanings were rejected by the analyses of Hailman and Griswold (1996), leaving as the last issue the possible function of suites of gargles being used for local identification. Ficken and Weise (1984) found some differences in gargling between chickadee populations as close as 5.7 km, thus qualifying gargles as having microgeographic variation in the sense of Mundinger (1982). Ficken and Weise (1984) used the ‘coincidence index’ of Dice (1945) to compare similarities in gargling among areas. This index has mathematical and logical problems (Hailman & Griswold 1996), but without access to their original data, no better comparison can be made. Ficken and Weise compared note types and gargle types, but as mentioned previously (and developed in the last major section below), gargle types are misleading because that concept treats variants within one gargle group as qualitatively different kinds of gargles. Therefore, the only valid common currency between the studies of Ficken and Weise (1984) in southeastern Wisconsin and Hailman and Griswold (1996) in south-central Wisconsin is the note type.

Integrating the data from the two studies just cited gives a preliminary picture of geographic variation of gargles. Coincidence indices for note types among feeder locations within a field station in southeastern Wisconsin were 0.97-1. The south-central Wisconsin study population comprised three areas along a primarily east-west arc with the maximum distance between extremes being slightly less than 2 km. The centers of adjacent areas 1 and 2 are about 1 km apart, likewise for adjacent areas 2 and 3, with about 1.5 km separating the centers of non-adjacent areas 1 and 3. The coincidence indices for note types in the three comparisons were between 0.83 and 0.85. Back in southeastern Wisconsin indices between field station feeders and a site 5.7 km distant were 0.67-0.75. Between the feeders and a site 9.8 km away, the indices dropped to 0.63-0.67. Finally, between the sites in southeastern and south-central Wisconsin, the index is zero. It thus appears that the number of note types shared between chickadee populations is a monotone function of their distance apart, with complete uniqueness being reached somewhere between roughly 10 and 100 km of separation.

Probably more relevant communicatively than geographic variation in note types is the difference in gargle groups. Information is available only for south-central Wisconsin, but it is very clear (Hailman & Griswold 1996). Five of nine gargle groups occurred in all three areas, two occurred only in adjacent areas 1 and 2, and two others occurred only in adjacent areas 2 and 3. (None occurred only in non-adjacent areas 1 and 3, and no gargle group was unique to an area.) These results mean that locations only about 1 km apart have a different suite of gargle groups, with about 70% of groups in common and about 30% different. We have no idea whether chickadees use suites of gargle groups to identify conspecifics as locals or ‘foreigners,’ but the requisite information is available to them.

Ever since Pepperberg (1985) proposed a theory of social models in avian vocal learning, it has become increasingly obvious that social interactions are important in learning song (e.g. Petrinovich 1988). Furthermore, compelling evidence exists that social interactions can be of critical importance in learning contact calls as well (e.g. Farabaugh et al. 1994). This major section reviews a chain of evidence implicating autumnal dyadic interactions as the behavioural context within which gargles are learned. First, it is established that major features of gargles are learned. Next, it is shown that young birds cannot possibly learn gargles from their parents before dispersing. Last, the hypothesis is presented as to when, where and how young chickadees learn their gargles.

At least three kinds of evidence make it virtually certain that major features of gargles, particularly the types and sequences of notes, are learned. First, chickadees hand reared in groups developed gargle vocalisations whose abnormal notes had no resemblance to gargle notes used in the population from which the hand-reared birds had come (Ficken et al. 1985). Furthermore, the birds within each of the two rearing groups gave nearly identical gargles whereas those gargles were completely different between the groups.

A second kind of evidence is the very existence of the microgeographic variation in gargles (described above) coupled with information about dispersal distances and directions. Dispersal data from Weise and Meyer (1979), from the same field station where Ficken did her gargle studies, show random directions of movement. The median distance of movement from the natal territory to the place where the young birds spend the first winter or breed in their first spring of life exceeds a kilometer. These data are markedly biased toward short distances, however, because the natal sites are in the field station where observers are many and chickadees are baited into the area at feeder-traps. Weise and Meyer (1979) did find one disperser 11.2 km away from the natal territory. Robbins et al. (1986) reported average dispersals of females at 2.3 km and males at 1.3 km. In our study area in south-central Wisconsin, we banded nestlings for a couple of years and had only one disperse within the study tract. The nesting cavity where it was hatched was nearly 2 km from where it was found in winter, the farthest distance a bird hatched on the study tract could disperse and still be within the tract. Remembering that birds may share only about 70% of gargle groups with chickadees in an adjacent area 1 km away, it is clear that dispersal distances usually exceed distances over which local dialects change noticeably. Therefore, if gargle characteristics were wholly ‘encoded genetically,’ local dialects would soon be obliterated by birds dispersing into a local area from all directions. Put differently, local distinction of genetically determined characters cannot be maintained in the face of massive immigration and emigration.

A third, related but not so compelling, piece of evidence comes from sampling of gargles in one locality over a period of years (Ficken & Popp 1995). Although it would be desirable to reanalyse the original data with regard to gargle groups (sensu Hailman & Griswold 1996), the comparison of note types and other characteristics suffices to make a point. Gargles were recorded in the period 1970 to 1981 and then again in 1988 to 1989. The researchers found that ‘no common note types were added in the 19-year period and none disappeared’ (Ficken & Popp 1995: 687). The emphasis here must be on ‘common,’ as it is the disappearance of rare old note types and the appearance of rare new ones that shows drift in the local dialect. Quantitative changes in the frequencies with which note types were used also occurred, as did complicated changes in gargles as grouped in a subjective manner. Although dialect drift in genetically determined vocalisations could theoretically occur through mutation, the existence of such drift is more parsimoniously taken as evidence for vocal learning that suffers copy errors.

Several lines of evidence reject the hypothesis that young chickadees learn their gargles from their parents. First, young chickadees are not known ever to gargle while still associated with their parents. Clemmons and Howitz (1990) followed vocal development from hatching to dispersal (about 32 to 36 days posthatch). They describe nothing in the calls of nestlings and fledglings that is a precursor of gargling, and when birds suddenly began giving crystallised adult-like vocalisations just before dispersing, the vocalisations included chick-a-dee calls, fee-bee songs, and subsong, but no gargling.

The fact that young chickadees do not utter gargles before dispersing does not of course mean that no latent learning took place. They could hear and learn the parents' gargles but reproduce them vocally only later. This possibility of latent learning seems remote because parents almost never gargle around the nest. Clemmons (1995) provided quantitative data on vocalisations given by parents within and outside of the nesting cavity where the young birds could hear them. Seven kinds were fairly common (squawk, twitter, faint fee-bee, A-complex, broken dee, composite, and chick-a-dee) and two were very rare (variable see and gargle). There is no equivalent study of parental vocalisations during the fledgling period, but I have never heard adults with dependent young give gargles. They probably do so rarely of course (as when encountering a neighbouring pair) but not reliably enough to promote latent learning in their fledglings.

Finally, the same argument used against ‘genetic coding’ of gargles may be applied to the hypothesis of learning gargles from the parents. When a chickadee leaves its natal territory (forever), it disperses in a random direction for a distance that probably usually exceeds 2 km, bringing it into a locale where it will probably spend the rest of its life. This new locale is sufficiently far from the natal territory that local gargle groups are probably at least 30% different from those of the parents. Therefore, if young chickadees learned gargles from their parents, they would be bringing into given dialect areas all sorts of new gargle groups from every compass direction. Nothing like this can be detected in early fall (Griswold unpublished data), and such carriers of parental gargles would soon swamp out local dialects, which also does not happen.

The foregoing facts secure the conclusions that the gargle note types and their sequences are learned and cannot be learned from the parents, which are usually the only adult chickadees they encounter prior to dispersing. Therefore, it must be that gargles are learned after dispersal, and the timing of onset of gargling supports this hypothesis. As noted above, young birds do not gargle prior to dispersing. Unbanded birds newly arrived at my feeder-trap in the late summer and fall do gargle; these birds are confirmed as young of the year when trapped and aged by morphological characters. The hand-reared chickadees reported by Ficken et al. (1985) did not begin giving their (abnormal) gargles until 56 to 59 d of age, well after the usual dispersal age of 32 to 36 d. These birds could have been unusually slow to begin gargling, especially considering they were in a captive environment in which there were no adult chickadees giving gargles. Nevertheless, the result is consistent with field data showing that chickadees begin gargling only after they have dispersed.

After leaving the natal territory, a young chickadee seems to do one of three things (S. M. Smith 1991). Some birds may wander alone for weeks while others may band together to form temporary flocks of young birds. These groupings seem to be commonest in young produced early in the nesting season (Odum 1941). Typically, though, young chickadees join (or at least try to join) gatherings of older birds that are themselves in the process of forming flocks or are already in formed flocks. It is at this time, in late summer and early fall, that gargling is especially prevalent (Griswold unpublished data) so likely to be involved with flock formation. A second, lesser peak in late fall is more likely to reflect interactions between formed flocks. It seems an inescapable conclusion from the timing involved that young chickadees learn their gargles during this period of flock formation. It is possible that they simply remember the gargles they hear given by others, but I think it more likely that they learn and perfect their gargles during direct interactions with older chickadees.

Close encounters with older birds in early fall might be interpreted as purely agonistic interactions like counter singing between males of other kinds of oscines, but the situation may not be that simple. Two things are imperfectly understood about chickadee behaviour. One is the nature of a chickadee flock, which some authors consider to be stable associations of about a half-dozen birds. I agree with Susan Smith (1991) that flocks of this kind do exist and that this is not the whole story. She states (p. 171) that ‘I contend that Black-capped Chickadee flocks, at least over most of the species' geographic range, are made up of mated pairs.’ We have evidence from the analysis of feeder visits that pairs normally do stay together in the same flock, and weaker evidence suggesting that when a bird leaves one flock and joins another, its mate does, too. The other contentious subject of chickadee behaviour is when (and how) pair formation occurs. Unlike migratory passerines where males arrive in spring and set up territories for attracting potential mates, the permanent-resident chickadees are already paired when spring arrives and the pair carves out its breeding territory in part of the general home range of its winter flock. Again, my experience coincides with Smith's (1991:91) that both widowed adults and young birds ‘most often form pairs in the fall as part of winter flock formation.’ In sum, we are faced with the possibility that pair formation and acceptance in a flock are closely related and simultaneous phenomena for a young chickadee. Therefore, all this early autumnal sorting out of social relationships probably involves both agonistic and sexual close encounters. Obviously, much scope exists for further research.

No widely agreed-upon definition of animal cognition exists, much less an operational definition. A general underlying notion is that whenever one is forced by data to postulate an intervening variable because a stimulus-response model is inadequate, cognitive processing may be at work (e.g. Vauclair 1996). Cognitive intervening variables may be called mental representations, and the challenge in animal studies is to show their existence and determine their characteristics.

The obvious stimulus-response model for learning an ordered string such as the notes composing a gargle vocalisation is serial learning by rote. We human beings use this form extensively in memorising ordered words that we wish to mimic exactly, as in a favourite poem, a soliloquy from Shakespeare, or a moving speech such as Lincoln's Gettysburg address. In general, the reliability of reproduction of anything learned by serial rote degenerates as a function of the distance from the beginning. Early parts are reproduced faithfully whereas copy errors increasingly intrude as one moves through the memorised string. That this error distribution expected under the rote serial learning model does not apply to gargles forces us to begin search for kinds cognitive processing that might underlie gargle learning. First, the operations behind finding gargle groups are explained, and then types of copy errors that rule out serial rote learning are discussed.

The gargle group has been referred to earlier in this contribution as the real ‘type’ of gargle so it is time to relate how these groups were discovered and how they are identified objectively. Intuitively, what we have called gargle groups (Hailman & Griswold 1996) should probably be called gargle types, but that designation was preempted by Ficken and coworkers in the many papers cited previously. By their operational definition every unique sequence of note types composing a unit gargle is a separate gargle type. What Griswold discovered while laboriously determining note-type sequences of more than 3000 taped gargles was that all gargle types were separate, but some were more separate than others. He sorted gargle types into preliminary groups based on a subjective assessments of similarities in their component strings of note types. Subsequently, we developed computer programs (especially StringFinder©1996 Jack P. Hailman) to parse gargle types into gargle groups objectively. The remainder of this section is all derived from Hailman and Griswold (1996) without further attribution.

Gargle groups were determined in the following way. First, (1) all note types were compiled by eye from printed sonograms and each type given a two-digit number such as 04 or 29. From the spectrographed sample of 3146 gargles, 37 note types were thus identified. (2) Every gargle was next represented by its note-type sequence, this labeling identifying 84 unique sequences (gargle types sensu Ficken). (3) The list of these gargle types was then searched to determine the longest string that any two types had in common. For example, this longest sequence (termed the ‘defining string’) was the eight-note sequence 24-24-08-09-11-04-05-17 shared by two gargle types. (One of those types was that exact string and the other had an additional note #24 at the beginning.) Finally, (4) the computer then found every gargle type that had a string of at least three notes in common with the defining string. In this case, three gargle types were found: one missing the first note #24, one missing both of the #24 notes, and one severely truncated to the sequence 11-04-05-17. The search then returned to step (3) to find the next longest string that was not a subset of the first defining string, and repeated step (4) to find all gargle types that shared a string of at least three note types with the new defining string. Recycling through steps (3) and (4) continued until no further defining strings existed.

The sorting process just described classified more than three quarters of all gargle types into only nine gargle groups with a single ambiguity: one gargle type objectively qualified for two different gargle groups having similar structure. Gargle groups thus defined were labelled with capital letters. The unclassified gargle types were then sorted according to any two-note sequence they shared with the defining strings of the gargle groups, and this process assigned eight types uniquely to one or another of the groups; two more gargle types qualified for two different gargle groups. In the end, nine gargle types remained, mostly apparent fragments, and these were simply termed ‘ungrouped’ gargle types. The nine gargle groups into which most of more than 3000 gargles could be placed objectively were considered the true kinds of gargles. The various gargle types within a group were interpreted as variants of one basic kind of gargle.

The question arises as to whether gargle groups are collections of variants on some Aristotelian idea or model. To attempt answering this question two indices of frequency of use were employed to find the commonest gargle types, and an explicit procedure identified two types from group A, four from group B, and one each from groups C, D, and G. (Groups F, H and I are tiny, so only group E did not contain a particularly frequent gargle type by the criteria used.) The two most frequently used gargle types in group A were 24-08-09-11-04-05-17 and that string without the initial note #24. These are both close to the defining string, which as mentioned above began with two notes #24. Similarly, the four frequently used gargle types from group B were the defining string itself and three types close to it, and the single most frequent gargle type of groups C, D, and G were the defining strings or one note different from them. Although no gargle type from group E met criteria as particularly frequent, the most frequently used type within the group was exactly the defining string.

One can conclude from the foregoing analysis that the defining string or something quite close to it is the archetypical gargle of each group. Therefore, other gargle types in the group may be interpreted as imprecise copies of the archetype. They ways in which these variants fail to match the ideal are the copy errors, and analyses of the nature of these errors tell us something about the way in which gargles are mentally stored or accessed.

The first thing discovered after sorting gargles into basic groups was the existence of prefixed and suffixed notes (Hailman & Griswold 1996). Certain note types tended to occur at the beginning of gargles (prefixes), and less commonly these same or other note types tended to occur at the end of gargles (suffixes). What characterises most of these note types is their similar occurrences in different gargle groups. For example, note type #24 is the first or second note of one or more gargle types in groups A, B, C, and H. Similarly, note type #06 is the last note of one or more gargle types in groups C, D, F, and H. Sometimes these prefixed or suffixed notes were part of the defining string of a group and sometimes they were not. Sometimes note types that appeared as prefixes or suffixes were also integral components of the middle parts of gargles and sometimes they were not. No operational way of distinguishing prefixes and suffixes from gargles sensu stricto has been offered so the reality of the concept remains uncertain. Nevertheless, as reported near the outset of this contribution, when rare repetition of note types occurs in gargles, these are always in notes identified subjectively as prefixes.

The simplest copy errors that might arise in a string of notes composing a gargle are genetic-like errors, of which five may be specified. First, a note can be repeated, which as mentioned does not occur except rarely in prefixes. Other types of simple errors include switching the order of two adjacent notes (transposition), dropping a note from the string (deletion), adding a new note in the middle of a string (insertion), and replacing one note with another of a different type (substitution). The search for these five types of simple copy errors failed almost completely (Hailman & Griswold 1996).

What is the fundamental nature of variation within gargle groups? The straightforward answer is truncation. Defining strings, which as shown above are the archetypes of gargle groups or close to those models, are most faithfully reproduced in the middle. For example, the defining string of group E is 14-15-16-17-18-12, of which the terminal #12 may be a suffix. Some gargle types in this group include 14-15-16-17, 15-16-17, and 15-16-17-18-12. Disregarding the complexities of poorly delimited prefixes and suffixes, most gargle groups include gargle types that truncate the defining string at its start, its end, or both. Group A has examples of truncations only at the start and group G only at the end, both groups being small. Even tiny group F containing only four gargle types shows terminal truncation of the defining string.

The fact that the most faithful replication of a gargle is the middle of its defining string implies that gargles are not acquired by rote serial learning. Copy errors, including truncation, should become commoner as one moves through a string learned serially by rote, and this phenomenon simply does not occur in gargles. At least one similar example has been reported in song. Adret (1993) tutored male Zebra Finches Taeniopygia guttata with a taped song of eight notes (this species sings just one song type), creating something like the natural situation in which males learn their father's song by an imprinting-like process. The tutored birds learned the middle of the song (notes c-f, lettered successively) better than either end, and some further truncated the core string of c-d-e-f to c-d-e, d-e-f, or e-f. Adret (p. 156) noted that ‘this result appears to go against the evidence that birds identifying conspecific song attend primarily to the initial elements.’

Chickadee gargles are more complicated than Zebra Finch songs, suggesting that fairly sophisticated information processing during learning should be sought. Our analyses suggest that the ‘anatomy’ of a gargle has five parts: prefix, early part of the gargle sensu stricto, core string, late part of gargle, and suffix. These parts have specifiable characteristics, such as prefixes being used in different gargle groups, early parts often being truncated, and core strings being best preserved. The existence of parts suggests that chickadees might learn gargles in ‘chunks,’ somewhat as Nightingales Luscinia megarhynchos use chunks in song learning (Todt & Hultsch, this symposium). If so, one possibility is that they use a cognitive mechanism to process different gargle parts into 'chunks', based on relative variability.

Using the gargle vocalisation of chickadees as the example, I have attempted to show how field studies can make important implications about social and cognitive aspects of vocal learning. The task involved not merely recounting studies of gargles but also drawing together disparate facts about chickadee biology. In integrating these diverse studies by various investigators it has of course been necessary to make some assumptions, which I have tried to keep intuitively reasonable, minimal in number, and explicitly stated.

The evidence makes virtually certain the conclusion that the types of notes and their sequences in composing gargles are learned. Almost as certain is that a young bird learns these gargles only after dispersing to the area that will likely be its home for the rest of its life. It is less certain but still probable that the birds learn the gargles from adults in the commonly occurring gatherings of chickadees in early fall — gatherings that apparently result in pairings of at least some unpaired birds and the formation of fall-winter flocks. Finally, the evidence suggests that the young birds learn gargles by participating in gargling duels with older birds. If we should be so lucky as to find that all these things are true, the next step would be to determine how the vocal learning takes place in these dyadic interactions. Does the older bird gargle while the younger one tries to listen and store the string of notes for later production of its own gargles? Or does the younger bird reply to the gargle, trying to mimic it with the answer? Or does the social interaction promote the vocal learning in still some other manner?

The nature of presumed copy errors in gargles has only begun to yield to analysis so implications for cognitive aspects of vocal learning are preliminary. We have only recently come to understand that most gargle variation represents imperfect versions of just a few local kinds of gargles (gargle groups) — instead of the welter of scores of separate and equal gargle types as believed for so long. The unequivocal conclusions that can be drawn at this stage are few in number and mainly negative in nature. First, simple copy errors of repetition, transposition, deletion, insertion, and substitution of single notes are exceedingly rare. Second, gargle variation definitely does not fit a pattern of most faithful reproduction occurring at the beginning and grading to increasingly higher error rates as one moves through the notes of the call. This negative finding seems to rule out the straightforward possibility that gargles result from rote serial learning, and so suggests that some kind of cognitive processes are at work. Finally, the evident feature of miscopying a gargle lies in truncating the core string at the beginning, the end, or both. The surprising finding is thus that the middle of gargles are best preserved, a phenomenon that seems to have been known from birds only through one study of song learning in Zebra Finches. Many pertinent questions remain. What is the significance of note types that can be more or less freely appended to the beginning (or more rarely the end) of calls with little regard as to which gargle group the call belongs? And why do these prefixed notes commonly show repetition whereas notes in the calls sensu stricto never do? What kind of a mental representation is made of a gargle such that its most reliably reproduced part is the middle of the call?

Adret, P. 1993. Operant conditioning, song learning and imprinting to tapes song in the Zebra Finch. Anim. Behav. 46:149-159.

American Ornithologists’ Union. 1997. Forty-first supplement to the American Ornithologists’ Union Check-list of North American Birds. Auk 114:542-552.

Baker, M.C., Boylan, J. T. & Goulart, C.A. 1991. Effects of gargle vocalizations on behavior of Black-capped Chickadees. Condor 93:62-70.

Baker, M.C., Tracy, T.T. & Miyasato, L.E. 1996. Gargle vocalizations of Black-capped Chickadees: test of repertoire and video stimuli. Anim. Behav. 52:1171-1172.

Clemmons, J.R. 1995. Vocalizations and other stimuli that elicit gaping in nestling Black-capped Chickadees (Parus atricapillus). Auk 112:603-612.

Clemmons, J. & Howitz, J.L. 1990. Development of early vocalizations and the chick-a-dee call in the Black-capped Chickadee, Parus atricapillus. Ethology 86:203-223.

Dice, L.R. 1945. Measures of the amount of ecological association between species. Ecology 26:297-302.

Dixon, K.L. 1972. Attack calls and territorial behaviour of the Mountain Chickadee. Proc. XVth Int. Ornithol. Congr.; pp. 640-641.

Dixon, K.L. & Stefanski, R.A. 1970. An appraisal of the song of the Black-capped Chickadee. Wilson Bull. 82:53-62.

Dixon, K.L., Stefanski, R.A. & Folks, F.N. 1970. Acoustic signals in the mating of Mountain and Black-capped Chickadees. Auk 87:322-328.

Dooling, R.J. 1982. Auditory perception in birds. In: Kroodsma, D. & Miller, E. (eds), Acoustic Communication in Birds, vol. 1; New York; Academic Press: pp. 95-130.

Farabaugh, S.M., Linzenbold, A. & Dooling, R.J. 1994. Vocal plasticity in Budgerigars (Melopsittacus undulatus): evidence for social factors in the learning of contact calls. J. Comp. Psychol. 108:81-92.

Ficken, M.S. 1990. Vocal repertoire of the Mexican Chickadee. II. Song and song-like vocalizations. J. Field Ornithol. 61:388-395.

Ficken, M.S., Ficken, R.W. & Apel, K.M. 1985. Dialects in a call associated with pair interactions in the Black-capped Chickadee. Auk 102:145-151.

Ficken, M.S., Ficken, R.W. & Witkin, S.R. 1978. Vocal repertoire of the Black-capped Chickadee. Auk 95:34-48.

Ficken, M.S. & Popp, J.W. 1992. Syntactical organization of the gargle vocalization of the Black-capped Chickadee, Parus atricapillus. Ethology 91:156-168.

Ficken, M.S. & Popp, J.W. 1995. Long-term persistence of a culturally transmitted vocalization of the Black-capped Chickadee. Anim. Behav. 50:683-693.

Ficken, M.S. & Weise, C.M. 1984. A complex call of the Black-capped Chickadee (Parus atricapillus): I. Microgeographic variation. Auk 101:349-360.

Ficken, M.S., Weise, C.M. & Reinartz, J.A. 1987. A complex vocalization of the Black-capped Chickadee. II. Repertoires, dominance and dialects. Condor 89:500-509.

Haftorn, S. 1973. Lappmeisa Parus cinctus in hekketiden: forplantning, stemmeregister og hamstring av næring. Sterna 12:91-155.

Hailman, J.P. 1989. The organization of major vocalizations in the Paridae. Wilson Bull. 101:305-343.

Hailman, J.P.& Ficken, M.S. 1996. Comparative analysis of vocal repertoires, with reference to chickadees. In: Kroodsma, D. E. & Miller, E. H. (eds), Ecology and Evolution of Acoustic Communication in Birds Ithaca, N. Y. and London: Cornell Univ. Press: pp. 136-159.

Hailman, J.P. & Griswold, C.K. 1996. Syntax of Black-capped Chickadee (Parus atricapillus) gargles sorts many types into few groups: implications for geographic variation, dialect drift, and vocal learning. Bird Behav. 11:39-57.

Latimer, W.A. 1977. A comparative study of the songs and alarm calls of some Parus species. Z. Tierpsychol.45:414-433.

Ludescher, F.B. 1973. Sumpfmeise (Parus p. palustris L.) und Weidenmeise (P. montanus salicarius Br.) als sympatrische Zwillingsarten. J. Ornithol. 114:3-56.

McLaren, M. 1976. Vocalizations of the Boreal Chickadee. Auk 93:451-463.

Mundinger, P.C. 1982. Microgeographic and macrogeographic variation in acquired vocalizations of birds. In: Kroodsma, D. E. & Miller, E. H. (eds); Acoustic Communication in Birds, vol. 2; New York: Academic Press: pp. 147-208.

Odum, E.P. 1941. Annual cycle of the Black-capped Chickade - 2. Auk 58:518-535.

Odum, E.P. 1942. Annual cycle of the Black-capped Chickadee - 3. Auk 59:499-531.

Pepperberg, I.M. 1985. Social modeling theory: a possible framework for understanding avian vocal learning. Auk 102:854-864.

Petrinovich, L. 1988. The role of social factors in White-crowned Sparrow song development. In: Zentall, T. R. & Galef, B. G., Jr. (eds), Social Learning: Psychological and Biological Perspectives; Hillsdale, N. J; Lawrence Erlbaum: pp. 255-278.

Robbins, M.B., Braun, M.J. & Tobey, E.A. 1986. Morphological and vocal variation across a contact zone between the chickadees Parus atricapillus and P. carolinensis. Auk 103:655-666.

Romanowski, E. 1978. Der Gesang von Sumpf- und Weidenmeise (Parus palustris und Parus montanus) - Variation und Funktion. Die Vogelwarte 29:235-253.

Smith, S.M. 1991. The Black-capped Chickadee: Behavioral Ecology and Natural History; Ithaca, N. Y.; Cornell Univ. Press: 362 pp.

Smith, S.T. 1972. Communication and other social behavior in Parus carolinensis. Publ. Nuttall Ornithol. Club 11:1-125.

Vauclair, J. 1996. Animal Cognition: An Introduction to Modern Comparative Psychology; Cambridge, Mass. and London; Harvard Univ. Press: 206 pp.

Weise, C.M. & Meyer, J.R. 1979. Juvenile dispersal and development of site-fidelity in the Black-capped Chickadee. Auk 96:40-55.