HBW 6 - Foreword on avian bioacoustics by Luis Baptista and Don Kroodsma

View foreword in PDF format: 

Birds are delightfully noisy animals, and we humans have appreciated their sounds in so many ways. Poets have immortalized the songs of Old World birds like nightingales and skylarks, for example, just as songs of thrushes and wrens have enchanted New World naturalists. This aesthetic beauty of bird song was championed by Charles Hartshorne (1973), who compared the musical abilities of birds throughout the world. This music in bird song (Elliott 1999) is heard and celebrated by many of us, as  it was by Mozart himself (West & King 1990). We humans also incorporate bird song into our own music, the relationship between music and bird song being a special interest of the late Luis Baptista (Dreifus 2000). Songs are appreciated in other ways, too, such as by those who love to watch and identify birds; birders value bird sounds not only for their outright aesthetic appeal but also as an aid in identifying different species. Rare is the person who could not appreciate the beauty of a nightingale's songs, who could not marvel at the vocal vigor with which birds greet the dawn, or who could not be enchanted by some of the most exquisite sounds of our natural world.

  Given all this appeal, then, it is no wonder that inquisitive humans began asking questions about these sounds. Why does the nightingale sing in the first place, and why are the songs so beautiful? Who is listening, and for what purpose? Does he seek to impress other males, or perhaps females, or perhaps himself? As we hear him deliver song after song, we ask why successive songs are so different, and how many different songs he can sing. And just how does he make those sounds? What specialized voice box or neural or hormonal machinery might he have that enables him to produce such exquisite songs? Over the lifetime of this individual, how did he come to sing this particular way? How much of his song and singing style is dictated by genes he has inherited, and how much must be learned from his elders, as a cultural tradition? Over the lifetime of this particular species, how did this behavior evolve? How did the ancestral nightingale sing, and how or why has the behavior of this lineage changed? We notice simpler vocalizations from this nightingale, too, as he interacts at close range with other nightingales, and we repeat all of our questions as we ponder how he uses his entire vocabulary. Indeed, questions abound about every possible aspect of what we hear.

  Such questions are the province of avian bioacoustics, the study of how birds use sound to communicate with each other. No question is too small, no question too big. Some questions are easily answered, such as how many different songs a nightingale sings (up to 200 for the Common Nightingale Luscinia megarhynchos--see Todt & Hultsch 1996). Answers to other questions, such as for whom the male sings, are more elusive. Given increasing interest in avian bioacoustics, the primary literature grows exponentially, as do attempts to review, synthesize, and integrate the facts and ideas with broader issues of how all animals communicate. Our growing knowledge is revealed in volumes appearing during the 1960's (Thorpe 1961; Armstrong 1963; Greenewalt 1968; Hinde 1969), 1970's (Thielcke 1970b; Hartshorne 1973; Jellis 1977; Sebeok 1977; Smith 1977), 1980's (Kroodsma & Miller 1982a, 1982b), and especially the 1990's (McGregor 1992; Catchpole & Slater 1995; Hauser 1996; Kroodsma & Miller 1996; Bradbury & Vehrencamp 1998; Hopp et al. 1998; Hauser & Konishi 1999; Pepperberg 1999).

  In this foreword, we2 offer our personal perspective on the field of avian bioacoustics, which has been the focus of our passion for a combined 70 or so years. As a tribute to the late Luis Baptista, this personal perspective is supplemented by questions submitted by a number of colleagues who were asked “To what question in avian bioacoustics would you (perhaps most) like an answer?” The questions are placed appropriately in the margins of the text, and are intended to reveal the diverse research interests of avian bioacousticians and some of the great unknown questions that drive them. In the following text, we begin with a brief discussion of the technological advances that have helped enable the rapid growth in the field. Then, in turn, we discuss how birds produce and perceive their sounds, how an individual develops its vocalizations, what the functions of those vocalizations are, and a discussion of the evolution of sound. We conclude with a look forward, discussing conservation, the increasing importance of sound archives, and the limitless possibilities for future study.

Bioacousticians and their machines

Although all modern scientists use an increasing array of gear, the study of bird sounds did not await a technological revolution. In 1773, for example, Barrington (1773: 249) presented to the Royal Society his experiments and observations on “a subject that hath never before been scientifically treated”. Barrington had reared young birds of many species in his home, and ramifications of his discoveries continue to be the focus of current studies. Barrington described, for example, how young birds must practise their songs, much like “the imperfect endeavour in a child to babble” (p. 250). After 10 to 11 months, the song becomes “fixed, and is scarcely ever altered” (p. 251). “Notes in birds are no more innate, than language is in man” (p. 252), as Barrington showed that young songbirds could routinely learn songs of other species. How young birds developed their songs was studied by others, too, such as Scott (1901) in the New World. Other “research” had been done over the past several centuries, too, as bird-fanciers, with nothing but their unaided ears and aesthetic tastes, conducted a marvellous experiment in artificial selection, showing that certain qualities of canary (Common Canary Serinus canaria) songs can breed true in different lineages (Güttinger 1985; Mundinger 1995).

  In the field, too, much was learned with patience, hard work, and a good ear. Saunders (1929) developed a system of note-taking that allowed him to compare singing behaviors of a great diversity of species. The fine details in songs of Song Sparrows (Melospiza melodia) were appreciated by Margaret Morse Nice (1943). She learned that males in her Ohio (USA) population had repertoires of up to 10 or so different songs, and she studied how the males used those songs in different contexts. She was fascinated by how young birds acquired their songs, clearly learning some of the details from certain adults in the population. Another early study was that by Craig (1943), who studied the singing behavior of Eastern Wood Pewees (Contopus virens), describing important differences between dawn and daytime singing, a topic that persists as crucial for understanding the function of song (see below). Even today, much can still be learned by keen observers who simply watch and listen to birds carefully; no machines are needed for so many kinds of studies, such as determining the contexts in which birds use certain vocalizations, or discovering examples of interspecific vocal learning in nature, or mapping geographic song variants that are distinctive to our ears.

  Although these studies achieved considerable understanding of birds and their sounds, further advances would require the ability to dissect and analyze the sounds. Perhaps the most important breakthrough for the field of bioacoustics was the invention of the magnetic tape recorder by German scientists during World War II (Baptista & Gaunt 1994). The tape recorder, along with improved microphones and parabolic reflectors (Kellogg 1962), enabled sounds to be “captured” and stored for later analysis (see also Wickstrom 1982). Current advice on both recording equipment and techniques can be found at the web site of the Library of Natural Sounds at Cornell University (birds.cornell.edu/LNS/).

  Especially crucial was the invention of the sound spectrograph machine. This device, developed by The Bell Laboratories prior to World War II, converts recorded sounds to pictures of frequency versus time, enabling scientists to measure the details of songs and produce objective, rather than anecdotal, descriptions of bird repertoires (e.g. Borror & Reese 1953; Collias 1991). Complementing sonagrams are waveforms, which graph amplitude versus time, and amplitude spectra, which graph amplitude versus frequency. Although these visual representations of sounds look like “hen scratchings” to the uninitiated, a little familiarity with how to read them reveals how useful they can be (see Fig. 2).

  Improvements in these basic instruments have helped revolutionize the study of bird (and other animal) sounds. The original sound spectrograph took what seemed an eternity (actually only several minutes) to produce a sonagram of a few seconds' duration, but the continuous spectrum analyzer could capture sounds in real time (Hopkins et al. 1974), so that exciting songsters with large and varied song repertoires could be studied. We learned of Northern Mockingbirds (Mimus polyglottos), for example, that males have a hundred or so different songs, and that repertoire size may influence reproductive success (Howard 1974). It was downright exhilarating to graph hours of a male Brown Thrasher's (Toxostoma rufum) effort and discover that he could sing well over a thousand, and perhaps 2000 different songs (Kroodsma & Parker 1977); with the advent of the continuous spectrum analyzer, some of us felt like kids in a candy store, eager to deploy this new machine to help understand what these vocal virtuosos were doing.

  More recently, advances in computer technology have provided special programs dedicated to sound analyses, e.g. SIGNAL from Engineering Design (Beeman 1996) and CANARY from the Cornell Laboratory of Ornithology (Charif et al. 1995), and a variety of other excellent programs. Digital technology (e.g. see Beeman 1998; Clements 1998) now enables animal sounds to be acquired, analyzed, and even synthesized with an ease unimagined only a few years ago, though a few pitfalls do await those who fail to understand some of the basics of how this technology works (e.g. see Stoddard 1998 for the importance of understanding how to use filters for animal sounds). Computers now also allow us to interact with birds in a more realistic fashion (Dabelsteen 1992), as we can now respond to a bird with any sound that had been stored in the computer's memory.

How birds produce and perceive sounds

Sounds that birds use in communication can be produced either by “instruments”, such as special feather structures, or by the vocal organ, the syrinx. The signal must then be transmitted through a medium (usually air) to the recipient, which must then extract relevant information in the signal from ambient noise. How the nervous system controls both production and perception of the signal, and how the environment influences signal transmission are topics in which exciting advances are being made.

1) Producing instrumental sounds

Specialized feather structures that produce sounds have evolved in many unrelated lines of birds (for review, see Prum 1998). Perhaps the most well-known instrumentalist is the Common Snipe (Gallinago gallinago; Fig. 3). Its outer tail feathers are stiffer because they are held together by more barbule hooklets than the other rectrices. During courtship the male flies high into the sky and then dives earthward with tail spread wide. The tail produces a pulsed humming sound, rendered as “wu-wu-wu-wu. . . ”, each “wu” corresponding to one wing beat as air is forced past these specialized feathers. This is a “long delicate eerie sound. It begins softly and increases in loudness and frequency as a dive progresses, reaching a crescendo just before a dive ends” (Miller 1996, p. 243). Proof that the humming is produced by the outermost rectrices comes from the experiments of Reddig (1978). He placed these specialized rectrices in a wind-tunnel and showed with spectrograms that sounds emanating from the tunnel were similar spectrally to those recorded during courtship in the wild.

  Another instrumentalist is the male Broad-tailed Hummingbird (Selasphorus platycercus), which has narrow outermost primaries that emit shrill whistles as the male performs territorial flight displays. This flight display can be silenced if the notch in the spread narrow primaries is blocked with glue. Birds then become less aggressive and lose their territories to other males. Removing the glue with acetone restores the wing whistles and with these the performer's territorial rights (Miller & Inouye 1983).

  Evidence from another species, the Flappet Lark (Mirafra cinnamomea), reveals that some instrumental sounds are learned from adults. By clapping their wings during aerial displays, males produce trains of snapping noises. Neighbors have more similar flapping rhythms than do non-neighbors, thus revealing that these local rhythms are apparently learned by young birds that enter a given breeding neighborhood (Payne 1978a), much like young songbirds of some species learn the local song dialect in their breeding neighborhood (see below).

  At least one species produces its instrumental sounds with a “manufactured tool”.  During courtship, the male Palm Cockatoo (Probosciger aterrimus) breaks a branch from a tree and shapes it into a “drum stick”. As part of his courtship, he then holds the stick with his foot and drums on a hollow log with a preferred resonance (Wood 1984).

  A large group of master instrumentalists, of course, is the woodpeckers. They drum on selected trees that seem to maximize resonance and sound transmission (Eberhardt 1997), with different species in a community typically distinguishable by unique rhythms. In North America, for example, the drumming sounds of the closely related and phenotypically similar Downy and Hairy Woodpeckers (Picoides pubescens and P. villosus) are distinctive, as are those of the Nuttall's and Ladder-backed Woodpeckers (P. nuttalli and P. scalaris--Stark et al. 1998). Further study of woodpecker drumming may reveal community-wide patterns of divergence or convergence. Exactly how carefully woodpeckers choose their drumming substrates is open to study, as suggested by the observation that a male Williamson's Sapsucker (Sphyrapicus thyroides) produced sounds of seven different pitches and tonal qualities by drumming on seven different sites situated on two different snags (Baptista & Keister 2000). It even seemed that he was matching his neighbor's drumming with a drum of the same pitch at a selected site, much like neighboring territorial songbirds use identical songs in their repertoires when they interact (i.e. matched countersinging--see Beecher et al. 2000 for review).

2) Producing vocal sounds

The vocal sounds of birds originate in the syrinx, which consists of paired vocal organs near the junction of the two primary bronchi and the trachea (Fig. 4; for recent reviews, see Gaunt & Nowicki 1998; Suthers 1999a). The syringes of all birds have at least one pair of rather membranous walls that can be distended into the airstream in the bronchi (i.e. internal tympaniform membranes), and others have denser but still flexible pads that can also extend into the airstream (e.g. internal and external labia). These membranes and pads are tightened or relaxed by external muscles (e.g. the sternotrachealis) that can change the length of the trachea, thereby controlling how much these membranes or pads vibrate in the airstream. Finer control is accomplished by the more complex syringes, as in songbirds, which have additional muscles (internal musculature) that are believed to control the membranes and tissues of the syrinx directly. Although all birds except New World vultures have syringes, the diversity in syringeal structure among avian orders is considerable (King 1989), and syringeal complexity is not obviously correlated with the complexity of sounds produced (Baptista & Trail 1992).

  One of the most fascinating consequences of the paired syringeal morphology is that birds can produce their own “internal duet”. The first convincing demonstration of this two-voice theory was by Greenewalt (1968), who based his conclusions on careful study of spectrograms and waveforms of the sounds that the birds produced. Conclusive proof of this two-voice ability in birds awaited the development of micro-techniques that allowed measuring respiratory air flow and pressure and the simultaneous activity of syringeal and respiratory muscles (see Suthers 1999a, 1999b).

  What has been learned from these micro-techniques is enough to make a grown avian bioacoustician giddy, because the two syringes interact in different ways that allow each species to achieve its own type of vocal virtuosity. In some species, such as the Grey Catbird (Dumetella carolinensis) and Brown Thrasher, the left and right syrinx make roughly equal overall contributions to the hundreds (Kroodsma et al. 1997) or thousands (Boughey & Thompson 1981) of songs that these birds have mastered (Fig. 5; e.g. Suthers et al. 1996). In other species, such as the domesticated Waterslager Canary (Serinus canaria), the left syrinx is responsible for about 90% of the song (Suthers & Goller 1998), thus confirming what had been deduced two decades before when the controlling nerve to the left syrinx had been severed (Nottebohm & Nottebohm 1976). Predominant use of only one syrinx may restrict the Waterslager Canary to the characteristic low-frequency notes in its song, but other species achieve much broader frequency ranges by using both syringes. This point is illustrated well by the songs of the Northern Cardinal (Cardinalis cardinalis), which have pure whistled notes sweeping over a broad range of frequencies, from roughly 1 to 7 kHz (Lemon & Chatfield 1971); in each of these sweeping whistles, the two syringes are used sequentially, the left syrinx producing sound below about 3.5 to 4.0 kHz and the right syrinx producing sounds above that range (Suthers & Goller 1998). Because of this exquisite coordination of the two syringes, the whistle sweeps seamlessly throughout its broad frequency range, with no trace of the separate origins of the upper and lower frequency components. Perhaps most impressive is the song of the Brown-headed Cowbird (Molothrus ater), with its abrupt changes in frequency over 1 to 13 kHz (King et al. 1980); the frequency contrasts in successive notes are achieved by rapidly alternating use of the left and right syrinx, with the final high whistle note produced by the right syrinx (Allan & Suthers 1994).

  Exactly how sounds are produced, however, remains something of a mystery. It was thought, for example, that the sounds in the syrinx were produced by vibration of the membranous walls that extend into the airflow (e.g. Miskimen 1951; Greenewalt 1968). That hypothesis was challenged only recently, when it was shown that destruction of these membranes has little effect on the quality of song produced (Goller & Larsen 1997); instead, it appears that the medial and lateral labia within each syrinx are extended into the airflow, and these labia vibrate as air is forcefully exhaled through the slit between the two labia. The sounds are modulated by the vocal tract (Nowicki 1987), too, and the trachea in some birds is elongated, perhaps by sexual selection in an attempt to exaggerate the size of the vocalizing bird (Fitch 1999). Beak motion also modulates the sounds (Westneat et al. 1993), and bill morphology may determine the range of sounds that birds are able to produce, as suggested for Darwin's finches (J. Podos 2000); thus, the food that is eaten may dictate the morphology of the beak, which in turn may restrict the range of possible vocalizations, so that morphology and behavior are inextricably interrelated in sexual selection. Given all that is involved in producing the intricate details of each vocalization (e.g. Lavenex 1999), it is not surprising that Lambrechts & Dhondt (1988) and Lambrechts (1996) proposed that birds physically “tire” from producing one song type repeatedly, and that neuromuscular fatigue has been one reason for the evolution of larger song repertoires (see below). However these vocalizations are produced, the resulting sounds are the culmination of an extraordinary coordination of syringeal and respiratory muscles (see Vicario 1993) and the vocal tract, a feat that becomes all the more impressive as more is learned about the complexity of bird sounds and how masterfully and precisely they are produced and reproduced. It is no wonder that Donald Borror, a pioneer in North American avian bioacoustics, referred to some of the more remarkable songsters as “vocal gymnasts” (Borror & Reese 1956).

3) Signal transmission

Those who listen carefully to birds realize that birds sound different in different environments. In open country, for example, when we are close to a singing bird, songs are often sharp and crisp. In the forest, however, sounds are typically softer, lacking the crisp quality of birds in more open environments. Although the types of sounds that birds make in open and closed habitats may differ somewhat (see below), one of the main reasons we hear them as different is because of the “echoes” (i.e. reverberation) that are caused by dense vegetation. The echo results from sound bouncing off the vegetation, so that the sound reaching our ears directly from the bird's bill is slurred by the sound reaching our ears just a split second later, as it bounces off vegetation. This “blurring” of sound is readily seen in sonagrams, where the onset of a sound is typically sharp but the offset is not, with the notes on the sonagram trailing off in a smear.

  Reverberation has implications for how we try to capture bird song on our tape recordings, too. The crisp qualities of a song are best preserved by placing the microphone as close as possible to a singing bird; the greater the distance from which a singing bird is recorded, the more useful is a parabolic microphone, because this highly directional microphone limits the recording of sounds (i.e. echoes) that arrive at the parabola from directions other than the singing bird. Sometimes recordings are so good that they sound unnaturally shrill, but the recordings sound “unnatural” to us only because our ears are highly directional and we are therefore accustomed to hearing sound after it has been degraded by the environment.

  The medium between the signaler and receiver is air, of course, and as the signaler and receiver birds become more distant from each other, their sounds are altered by the environment in predictable ways (for an excellent review, see Bradbury & Vehrencamp 1998). Sound is not instantaneous, for example, but travels at 344 m/sec, so that birds at greater distances will experience delays. Sound intensity diminishes with distance from the source, too, this “spreading loss” roughly equal to the square of the distance from the source. Higher frequencies are attenuated more over distance than are lower frequencies. Reverberation varies by habitat, such that rapid amplitude modulations would be obscured in forests (because the echo of one pulse off the trees obscures the next pulse from the bird) but not in open country. Temperature differences between air at and above the ground (e.g. in the canopy) can produce either sound channels, in which sound is trapped and propagated longer distances, or sound shadows, areas in which sound is greatly attenuated. The signaling animal must also contend with background noise, such as from wind, rain, or other animals. Indeed, to be effective, long-distance communication must contend with many environmental factors.

  These and other features of sound propagation have implications for the vocalizing bird, which (normally) tries to communicate as effectively and efficiently as possible. Considerable progress is being made in understanding the effects of the environment on signal transmission and the structure of signals themselves. Pioneering these efforts was Morton (1975), who surveyed Neotropical birds and discovered that open-country birds were more likely to use rapid frequency modulations than were forest birds; this result was largely confirmed for North American habitats by Wiley and colleagues (Wiley & Richards 1978, 1982; Wiley 1991). Even within species, songs vary among populations, and it is possible that songs are locally adapted to transmit best through particular habitats (Gish & Morton 1981); the songs that are best adapted to a local habitat might be culturally transmitted by young birds, which would preferentially learn songs the fine details of which were best preserved between the signaler and the receiver (Hansen 1979; Morton et al. 1986). The amount of reverberation or degradation in a signal can also be used as an indication of the distance of the vocalizing bird (Richards 1981; Naguib et al. 2000), the implications of which remain controversial (Morton et al. 1998). Our understanding of the relationship between bird song and habitat structure has been aided by studies of a number of species, such as the Rufous-collared Sparrow (Zonotrichia capensis--e.g. Nottebohm 1969; Handford & Lougheed 1991; Lougheed & Handford 1992), the Great Tit (Parus major--Hunter & Krebs 1979), the White-throated Sparrow (Zonotrichia albicollis--Wasserman 1979; Waas 1988), and paruline warblers (Lemon et al. 1981). We also know that birds vocalize less under suboptimal conditions (e.g. during rain or wind), or avoid masking each other by alternating songs with singers of other species (Ficken et al. 1974). All of these evolutionary adjustments in song quality or singing behavior are made in an attempt to maximize the ability of birds to transmit relevant signals in a noisy world (Klump 1996).

4) Signal perception

The complex sound wave that arrives at a bird's ears is converted to nerve impulses by the peripheral auditory system (review in Saunders & Henry 1989). Sound is first concentrated by the external ear (auditory meatus), which is covered by feathers. Sound waves vibrate the tympanic membrane, which transmits these vibrations to the inner ear by a single bone, the columella. Exactly how the fluid-filled cochlea of the inner ear converts these complex vibrations to meaningful nerve signals is unknown, but nerves eventually project the extracted signals to the auditory region in the forebrain, called Field L, which controls how birds hear sounds. How this auditory region interacts with the forebrain's motor regions (i.e. regions that control how birds produce sounds) is an exciting area of inquiry (see next section).

  The hearing ability of most birds falls within the range of human hearing (Dooling 1982; Fay 1988). The range of maximum sensitivity for songbirds is roughly 1 to 5 kHz, for example, and non-songbirds, which tend to be larger, hear better at slightly lower frequencies. For some species, hearing acuity may be sharpest in the narrow range where each species produces its sounds (Okanoya & Dooling 1988). Most birds, it seems, do not hear sounds of either very low or very high frequencies outside the range of human hearing.

  Despite of having small heads with short distances between the ears, birds are able to determine direction and distance rather well. Rufous-sided Towhees (Pipilo erythrophthalmus), for example, determined distance to within 7% accuracy and azimuth within 5 degrees (Nelson & Stoddard 1998). Distance clues are provided in a sound by the amount of reverberation (see above), but birds also use the amplitude of the sound to estimate distance (Phillmore et al. 1998; Nelson 2000).

  Some species clearly have special hearing abilities. The Rock Pigeon (Columba livia) can hear in the 1 to 10 Hz range (Kreithen & Quine 1979), a range that we humans call “infrasound” because it is well below our range of hearing. Oilbirds (Steatornis caripensis) and Cave Swiftlets (Collocalia linchi) probe the dark with brief (1 to 20 msec), audible clicks of 2 to 15 kHz, using the echoes to navigate (Medway & Pye 1977; Konishi & Knudsen 1979). Common Barn-owls (Tyto alba) are renowned for their ability to locate prey in complete darkness (Payne 1971); these special auditory skills are achieved, in part, by an asymmetrical arrangement of the two ears, with the left facing downward and the right facing upward, enabling the owl to pinpoint sound sources in both the horizontal and the vertical plane (Knudson 1981). Thirty to 40 years ago it was thought that birds had special abilities to hear rapidly occurring sounds (i.e. an excellent “temporal resolving power”), perhaps up to 10 times better than we could hear (e.g. see Schwartzkopff 1973). When birds were tested in the laboratory, however, the results were surprising: birds appeared to be no better than us humans (Dooling 1982). New testing procedures now favor the birds again, it seems. Unless they can hear far better than we can, it is certainly a mystery how young birds are able to copy the precise details of songs, whose details which we cannot begin to appreciate with the unaided ear. Clearly, much remains to be learned about how well birds hear and the role of hearing in their lives.

5) Neural control of bird song

Given the impressive abilities of songbirds to learn and produce large repertoires of songs and the apparent importance of those abilities in sexual selection (see below), we should expect complex neural control of these feats, too. During the last 30 years, neurobiologists probing the brains of songbirds have revealed discrete neural pathways for producing and hearing songs (Fig. 6; sample reviews in Konishi 1989; Nottebohm 1989, 1999; Brenowitz & Kroodsma 1996; Brenowitz et al. 1997). What has been revealed in this area of research is exciting and is progressing rapidly (Marler & Doupe 2000).

  Several studies have now revealed that the number of different songs a bird sings is positively correlated with “brain space”, or the volume of certain clusters of neurons (i.e. song nuclei) in the forebrain. This relationship was first discovered in the Common Canary, in which the repertoire of a male's song syllables is correlated with the volume of two forebrain nuclei, HVC and RA (Nottebohm et al. 1981). Similar relationships have now been found in other species, such as the Marsh Wren (Cistothorus palustris--Canady et al. 1984), several Neotropical duetting wrens (Brenowitz & Arnold 1986), and, indeed, in a broad comparative analysis of 41 songbird species in 9 different families (DeVoogd et al. 1993). Needing more brain tissue with a larger repertoire must be costly, but this increased investment in song-control centers is apparently needed so that a bird can learn and manage its larger song repertoire. Because the volume of the song nuclei in Marsh Wrens does not increase as more songs are learned, it seems that the size of the song nucleus with which a young bird is endowed dictates how many songs he will be able to learn (Brenowitz et al. 1995). Exactly how or where songs are stored in the brain is unknown, but the volume of song-control nuclei is apparently influenced by both the number and the size of the neurons in these song-control centers, and the density of dendritic spines does increase in certain nuclei as more songs are learned (Airey et al. 2000).

  Another intriguing feature of these forebrain song nuclei is their remarkable seasonal plasticity (Nottebohm 1981), as several of them are up to 70% larger in breeding birds than in non-breeding birds (review in Brenowitz & Kroodsma 1996). One exciting aspect of this plasticity is that adult songbirds generate new neurons that migrate to and are incorporated into these song-control centers (e.g. Álvarez-Buylla et al. 1990). But why these brain features change with the seasons remains a mystery. Initially it was proposed that the seasonal fluctuations were related to the annual learning of new songs, such as by the Common Canary, but seasonal plasticity also occurs in species that do not learn to sing new songs each year (Brenowitz et al. 1991).

  So much more remains to be learned about how the brain controls song perception and production in birds (e.g. see Brenowitz & Kroodsma 1996). Just where and how, for example, are the sensory models of songs stored in the brain? What role do all of the song-control centers play during development (Bottjer & Arnold 1997), as the young male first memorizes his songs and then, even while sleeping, practises them (Dave & Margoliash 2000)? The discrete vocal-control nuclei are remarkably similar but are believed to have evolved independently in the brains of songbirds, parrots and hummingbirds, suggesting that the brain is constrained in how it controls these learning tasks (see Jarvis & Mello 2000; Jarvis et al. 2000); what other homologies or analogies might be found as vocal learning is discovered in other groups, too? How are the neural and endocrine systems interrelated in their control of song (Ball & Hulse 1998; Ball 1999)? What is the neural basis for how non-singing females learn to recognize different songs, and for singing females what factors regulate the development of this neural control system (Nottebohm & Arnold 1976; Arnold et al. 1996)? As species other than the convenient laboratory canary and Zebra Finch (Taeniopygia guttata) are studied, what might we learn about how neural and endocrine systems are adapted to different life histories?

How sounds develop in individuals

From the embryo to an accomplished vocalist (or instrumentalist), a rather remarkable transformation occurs. The sounds used by individuals change over time, but, at all stages of life, the appropriate sounds are available to help manage a social environment, and largely for selfish gain. From the best mimics to the most selective of learners (to those which do not learn at all), each species seems to have its own genetically-endowed blueprint as to how to proceed through this developmental process (e.g. see Mundinger 1999). Understanding how birds acquire their sounds and the ability to use them has received much attention, as we try to understand the role of nature and nurture, the ecological or social circumstances that might favor different kinds of vocal development, and so many other features of how birds acquire the sounds they use to communicate. In this section, we review several of these areas.

1) Nature versus nurture

As a young bird matures, it becomes competent in using the vocal repertoire of adults. Exactly how that youngster becomes competent has been a focus of considerable debate (e.g. Johnston 1988), but the broad issues are rather simple, with developmental styles perhaps best viewed along a continuum. At one end of the continuum are sounds that develop normally in an individual regardless of what it experiences as it matures. At the other end of the continuum are sounds that the young bird must imitate from adults, and, without proper “tutoring” by adults, the sounds that the maturing bird produces are highly abnormal.

  Among birds, one can readily find examples at both ends of the continuum and at all points in between (e.g. see Slater 1989). An example of songs that are largely genetically inherited is that of a New World suboscine flycatcher, the Eastern Phoebe (Sayornis phoebe--Kroodsma 1985; Kroodsma & Konishi 1991). Each male of this species has two different songs, the fizz-bew and the fitz-bew, and the young birds, regardless of their auditory experience, develop seemingly normal songs. Songs develop normally even if young birds are deafened so that they cannot hear themselves practise. (Depriving a young bird of this “auditory feedback” during development is the crucial test to determine the extent to which songs are encoded directly in the genes, or, put another way, the extent to which songs must be learned (see following paragraphs).

  At the other end of this developmental spectrum are numerous examples of how songbirds must learn the details of their songs from accomplished adults. One of the earliest demonstrations of this phenomenon was by Barrington (1773) over two centuries ago (see above). It is William Thorpe (e.g. 1958, 1961), however, who is considered the pioneer of the scientific study of song-learning in birds. Thorpe demonstrated that young Chaffinches (Fringilla coelebs) would imitate songs from loudspeakers, that young birds isolated from adult songs would develop aberrant songs, and that the learning was concentrated in the first year of life. Other biologists soon followed, with Thorpe's student Peter Marler moving to the New World and beginning his classic studies of song-learning among White-crowned Sparrows (Zonotrichia leucophrys--Marler & Tamura 1962, 1964; Marler 1970b). And students of Marler, in turn, would continue the quest for understanding how young songbirds learn their songs (e.g. Konishi 1963, 1965; Nottebohm 1968, 1970). Elsewhere, in Germany Immelmann (1969) began his studies of the Zebra Finch, and Thielcke (1965) his studies of creepers (Certhia spp.). Laboratories everywhere were soon exploring these remarkable song-learning abilities of songbirds (e.g. see Kroodsma & Baylis 1982).

  What was learned from these early studies suggested a two-step model for song-learning (reviewed in Nottebohm 1999). First the young bird memorizes the required sound and stores it in the brain. Some time later, from days to months depending on the species, the maturing bird begins to practise that sound, learning to match its voice with the stored template in the brain. Initial attempts are unrecognizable warblings (i.e. subsong), but gradually the practice improves (plastic song), until the song produced is a good match of the memory in the brain. Once the songs have reached their stable adult form, they often remain essentially unchanged throughout the life of the individual (e.g. Ewert & Kroodsma 1994). This stability seems to be maintained in some species even without auditory feedback (i.e. when the singer is deafened--Konishi 1965), but in other species the song begins to deteriorate immediately after loss of hearing (Bengalese Finch  Lonchura striata--Okanoya & Yamaguchi 1997). Thus, the extent to which adult birds are constantly “relearning” their songs as they sing may differ among species.

  The characteristics of songs that need to be learned from adults vary among species. At one extreme are species like the White-crowned Sparrow and Anna's Hummingbird (Calypte anna), which must learn frequency, syllable structure, and rhythm of the song (Marler 1970b; Baptista & Petrinovich 1984; Baptista & Schuchmann 1990). In other species, certain characteristics such as the rhythm of the song's components might develop independently of experience (e.g. European Greenfinch Carduelis chloris--Güttinger 1979). In European creepers, the song's notes seem to develop normally without experience, because they are derived from the begging calls of the juvenile; to sing the local dialect, however, juveniles must learn the permutations and combinations of the different “inherited” call notes from adults in the population (Thielcke 1970a). The features of the song that must be learned from other adults are minimized for species in which individuals improvise, or make up, their song repertoires. Young Sedge Wrens (Cistothorus platensis) and Grey Catbirds can acquire hundreds of different songs in this fashion (Kroodsma, Houlihan et al. 1997, Kroodsma, Liu et al. 1999); because no fine details of songs are imitated from other individuals, the song repertoires of different individuals are highly distinctive, yet fall within the inherited guidelines dictated by an internal song-generator. In the two-step model of song-learning, these improvisers perhaps do not need the first step, that of memorizing model songs; maybe they still require auditory feedback during song practice, as they learn to match their song output to this internal pattern-generator.

2) Ecology of vocal learning in songbirds

Questions about “how much is learned” are quickly followed by a host of other questions, such as “Where, from whom, and when are songs learned?” Consider a juvenile male White-crowned Sparrow in the coastal chaparral of California, for example. His father has a song typical of the local dialect area, which consists of several hundred males all singing the same song pattern (Marler & Tamura 1962; Baptista 1975). As the young male hears his father's song, the species-typical introductory whistle seems to be a cue to learn what follows (Soha & Marler 2000); he memorizes the details of that song, to an extent that he could use that memory later to develop his own song. When this young male disperses from his natal territory to his breeding territory, however, he might encounter a sharp boundary between two different dialect areas. If he chooses to stay and breed within his own dialect area, he can simply perfect the song in his memory (perhaps refining the song to match the details sung by immediate neighbors--Trainer 1983), but if he crosses the boundary to the new dialect area, he must learn a new song in order to fit in there. The implications of his choice are important, because the relationship between song dialects and genetic population structure is at stake, and whether or not song-learning and song dialects tend to isolate neighboring populations and hasten speciation. How to gather good data on this topic and how to interpret them has been hotly debated (see Baker & Cunningham 1985 and accompanying commentaries).

  The answers to these questions are far from complete and certainly differ among species, but some answers are in hand. We have learned, for example, that answers to these kinds of questions must be pursued in nature, because how young birds behave in the highly artificial laboratory environments can differ substantially from how they behave under natural conditions (e.g. see Beecher 1996; Baptista 1999; Liu & Kroodsma 1999; Liu 2001). When young marked birds have been followed in nature, they do cross dialect boundaries and settle in neighborhoods with songs that differ from those of their father, whether the species be Bewick's Wrens (Thryomanes bewickii--Kroodsma 1974) in Oregon, USA, or Saddlebacks (Philesturnus carunculatus--Jenkins 1978) in New Zealand. Young White-crowned Sparrows disperse across dialect boundaries, too (Baker & Mewaldt 1978; Petrinovich et al. 1981; DeWolfe et al. 1989; Bell et al. 1998). After moving to a new song neighborhood, new song memories are acquired, and songs of the new location are practised and eventually produced (O'Loghlen & Rothstein 1995), although traces of the father's songs are sometimes still detectable (Kroodsma 1974; Bell et al. 1998).

  For some species, song-learning “rules” can be identified. Field studies of the sedentary Song Sparrow in Seattle, Washington, have revealed that a young male during his hatching summer copies entire songs (up to 10) from individual males in a small neighborhood where he will eventually establish his territory; a young male preferentially learns songs that are shared among his neighbors, too, as if to maximize the probability that he will share songs with those neighbors that survive to his first (and subsequent) breeding season (see Beecher 1996; Beecher, Stoddard et al. 1996, Beecher, Campbell et al. 2000; Nordby et al. 1999). Essentially these same rules are followed by the migratory Chipping Sparrow (Spizella passerina) in western Massachusetts (Fig. 7). This sparrow also learns his single song at his breeding location, either during his hatching summer (like the Song Sparrow) or after migration the following spring (Liu 2001). Migratory young male Indigo Buntings (Passerina cyanea) also learn songs of adult neighbors after their first migration (Payne 1996).

  Although learning songs at the breeding location seems to be a general pattern (see also Payne & Payne 1997), there are exceptions. One is found among the Galapagos finches, in which sons sing primarily the songs of their fathers (Millington & Price 1985; Gibbs 1990; Grant & Grant 1996). Another possible example is the Zebra Finch (Slater & Mann 1990; Zann 1997). Why species differ in their song-learning strategies remains largely a mystery, in part because the functions of songs and “song-matching” (an exchange during which males respond to each other with their identical, learned songs) are so poorly understood (see below).

  Other aspects of song development must have an ecological basis, too, such as the extent to which males imitate or improvise their songs. For this question, New World wrens in the genus Cistothorus provide a glimpse of an answer. North American Marsh Wrens imitate precise details of their large song repertoires and tend to be site-faithful, with males forming stable singing communities in which their imitated songs can be hurled incisively at each other during matched-countersinging duels (review in Kroodsma & Verner 1997). In contrast, North American Sedge Wrens improvise their large repertoires, and populations are not site-faithful, such that a singing male wren cannot expect to interact with the same other wrens for more than a month or so during his life (Kroodsma, Liu et al. 1999). When populations of this Sedge Wren are sedentary, however, as in the Neotropics, more song-learning occurs, local dialects exist, and males countersing with matching song types, suggesting that in these Cistothorus wrens site-fidelity promotes song imitation and lack of site-fidelity promotes song improvisation (Kroodsma, Sánchez et al. 1999).

3) And so much more

Many other questions, too, have been the focus of studies of vocal development, especially in laboratory settings. One such question has focused on the timing of vocal learning, and several studies revealed that young birds concentrate their song-learning into a “sensitive period” early in life. A young Chaffinch does not learn new songs after the early spring of its second year (Thorpe 1958), for example, and a young White-crowned Sparrow learns most readily during its first 50 days of life (Marler 1970b). The Marsh Wren's peak of learning sensitivity occurs from roughly day 20 to day 60 (Kroodsma 1978a). The convenient Zebra Finch has been the focus of a flurry of studies, too (Immelmann 1969; Slater et al. 1988), showing that song-learning is concentrated into the first 2 to 3 months of life. In most species, it is as if young birds are “eager” to begin memorizing their songs (even though changes may be made later), as if organizing the neurons and acquiring early memories were crucial for later success. The timing of this early song memorization is relatively impervious to external factors, such as photoperiod, as in White-crowned Sparrows (Whaling et al. 1998). In the appropriate social circumstances, however, learning by these species can occur beyond these “most sensitive” periods (Pepperberg 1985).

  In other species, new songs are routinely learned throughout life. Male Common Canaries continue to add new components to their song each year (Nottebohm et al. 1986), as do Northern Mockingbirds (Derrickson 1987) and the ubiquitous Common Starling (Sturnus vulgaris--Eens et al. 1992; Eens 1997; Hausberger 1997). Many other species probably learn throughout life, too, but collecting information on this phenomenon in nature is difficult, because a marked bird is not easy to follow and record throughout its life.

  Exactly when each species learns its songs is best considered as falling on a continuum that begins early in life and ends at death. Roughly where each species falls on this continuum can be determined in the laboratory, but one should keep in mind several caveats. A species may occupy a broad stretch of the continuum, for example, and not a single point, because strategies of individuals even within a population may differ; a nestling male hatching early in a breeding season might complete its learning during its hatching summer, whereas a male raised late in a breeding season and hearing no songs during his hatching summer will be more likely to defer learning until the following spring (Kroodsma & Pickert 1980). Where a species falls on this continuum can also be influenced by (sometimes unknown) factors in our highly artificial laboratory conditions, even by the color of the leg-bands on the particular Zebra Finches used in an experiment (Pearson et al. 1999) or by a variety of other, sometimes subtle conditions (Slater et al. 1988).

  One highly artificial condition in the laboratory, of course, is the social environment in which the young bird is learning its song (Pepperberg 1985). A loudspeaker is a highly unnatural tutor, and several studies have shown that young birds are more responsive to live tutors than to loudspeakers (Baptista & Petrinovich 1984; Kroodsma & Pickert 1984). For some paruline warblers, which use different songs in different contexts (e.g. Staicer 1996a), strategies for developing the two songs differ (Lemon et al. 1994); in Chestnut-sided Warblers (Dendroica pensylvanica), males apparently need a social tutor more for songs used in intense male-male interactions than they do for songs used in solo singing to attract females (Byers & Kroodsma 1992). The overall literature seems to confirm this enhanced effect of social tutors (Baptista & Gaunt 1997b; but see Nelson 1997; Payne & Payne 1997).

  Another fascinating feature of song development is its selectivity, or lack of it. Renowned mimics occur throughout the world (Baylis 1982): the Neotropical Lawrence's Thrush (Turdus lawrencii--Hardy & Parker 1997), the Nearctic Northern Mockingbird (Borror 1964), the Australian lyrebirds (and other species--Chisholm 1946; Robinson 1974, 1975), the Common Starling (e.g. West et al. 1983; West & King 1990), and African species, too (Harcus 1977a). Mimicry by the Marsh Warbler (Acrocephalus palustris) is especially fascinating, because the birds mimic songs from both their European breeding sites and their African overwintering sites (Dowsett-Lemaire 1979). Some interspecific vocal learning is far more focused, with close relatives or potential competitors learning each others' songs, a number of examples occurring throughout Europe (Helb et al. 1985) and elsewhere (e.g. the Great Plains of North America--Baker & Boylan 1999). When these species learn each others' songs, they often defend territories against each other and sometimes attract females of the opposite species, too; how these cases of interspecific song-learning will play out over evolutionary time must await future studies.

  In contrast, young males of other species select only a narrow range of species-typical songs (e.g. the White-crowned Sparrow--Marler 1970b; Nelson & Marler 1993). This natural preference for learning conspecific song (Konishi 1985) can sometimes be overridden, however; when social tutors are provided, young male White-crowned Sparrows then can learn Song Sparrow songs (Baptista & Petrinovich 1986; see also Pepperberg 1997). White-crowned Sparrows can also be tricked to learn songs of other species if the alien song is preceded by the species-typical White-crown whistle, which apparently serves as a cue for song-learning (Soha & Marler 2000).

  A special case of interspecific song-learning occurs among the viduine finches (Vidua spp.) of Africa (Nicolai 1964; Payne 1973; Payne et al. 1998). These finches are brood parasites, and each viduine depends on a specific host waxbill species (Estrildidae) to rear its young. Male viduines learn the songs of their waxbill hosts and use those songs during courtship. In areas where several sibling species of viduines occur sympatrically, the mimicked song of the host is an important cue for females to identify males of their own species.

  Another intriguing feature of vocal learning is the “overproduction” that occurs as young birds practise their songs. During the first step of vocal learning, when birds memorize songs, more songs are committed to memory than will eventually be used. These extra songs are practised by the young bird, but he eventually discards them as he hones the final song(s) he chooses to sing (Marler & Peters 1982; Marler 1997). How the young male chooses which songs to discard and which to keep is probably influenced by what other males, especially adults, are singing adjacent to his breeding territory (Nelson & Marler 1994; Beecher et al. 1997; Nelson 1999).

  Other fascinating examples of vocal learning occur, too. Among cardueline finches, for example, the male and female of a pair may learn each other's call notes, such that they are distinctive and easily identified in a flock (Mundinger 1970). A tanager, the Thick-billed Euphonia (Euphonia laniirostris), mimics the calls of other passerines, and apparently uses those calls to rally those birds to help mob predators (Morton 1976; Remsen 1976). In the Brown-headed Cowbird, the female seems to play some role in directing what songs the male will learn to sing (West & King 1988).

  The above overviews, unfortunately, gloss over all of the astonishing details of how young males of particular species learn their songs, so we end this section by highlighting some of those details in relation to how Common Nightingales learn their large repertoires (Fig. 8; review in Todt & Hultsch 1996). Beginning with the PhD thesis of Henrike Hultsch (1980), Hultsch and Todt have revealed how a male learns the precise details of up to 200 different songs from singing males around him. In the laboratory, it can be shown that learning begins at about day 15, with a peak of learning sensitivity during the first three months, during which time the youngster needs to hear songs from a live social tutor (even a human being with a speaker around the neck will suffice, if that person fed the babies since about 6 days of age). Interestingly, the social factor appears less important later in life, when the birds can still learn from loudspeakers alone and do so. And these birds are rapid learners: 10 different song types can be learned in only 10-20 exposures, and 60 different types if heard only once a day for 20 days. Relatively short sequences of several songs are learned intact from different tutors, and the birds themselves will segment longer tutor sequences into “packages” of 3 to 5 song types that almost always occur together during a singing performance. The remarkable abilities of these skilled songsters and the means by which these songs are learned have been honed over evolutionary time, but for what purpose? Issues of function we turn to in the next section.

The function of sounds

Birds use different vocal or instrumental sounds in their repertoire to influence each other, i.e. to communicate with each other. The nature of this influence has been the focus of much discussion, such as whether the communication is honest or dishonest, whether animals manage or manipulate each other, and whether the communication is “true” or whether the recipient was unintended (recent review in Bradbury & Vehrencamp 1998). So much information is broadcast that entire networks of listeners can take advantage of each vocalization. These considerations must permeate any discussion of function, and they will recur (though not be stressed) throughout the thoughts presented in this section.

  Consider first our use of the words “song” and “call”, words that so far in this foreword have been used rather casually, as if they are two discrete categories (Spector 1992). What has happened over evolutionary time is that many species have one particular sound that has come to be used in a similar context across taxa. These sounds are typically loud, complex, prolonged, and/or delivered from high perches, almost always by the male (temperate regions) but sometimes by the female, too (especially in tropical regions), as if the sounds were being used to impress others or to advertise one's presence. This evolutionary trend is epitomized by the “songbirds”, the highly successful (almost half of all bird species!) suborder of the order Passeriformes in which vocal learning has enabled extreme elaboration of these “songs”. Almost by default, the other (non-song) vocalizations of birds become “calls”.

  This quasi-functional distinction between songs and calls is useful, although messy. Some songbirds, for example, seem to use no sound that we would identify as a song (e.g. the Cedar Waxwing Bombycilla cedrorum--Witmer et al. 1997), and the gargle call of the Black-capped Chickadee (Parus atricapillus) is far more complex than is its simple whistled song, the fee-bee-ee (Ficken 1981). Why a cross-species classification of songs and calls cannot be rigid is perhaps best illustrated by crows (e.g. American Crow Corvus brachyrhynchos--Brown & Farabaugh 1997). When in close contact with other crows within a group, both males and females “sing” a diverse array of clicks, rattles, caws and coos; for long-distance defense of territory, however, the crows use their loud caws, which we think of as “calls”. Crows thus seem to have reversed what we think of as the usual function of songs and calls. Other taxa, such as the 1000 or so suboscines, remind us that songs need not be especially elaborate or complex, either, as most of these species appear not to learn their vocalizations; they get along just fine with a diverse repertoire of simpler vocalizations, with the distinction between “calls” and “songs” unclear in many species. These few examples illustrate that we should use the terms “song” and “call” cautiously, and beware how our initial labelling might influence how we study the functions of various sounds that birds use.

1) The function(s) of calls

Using calls to influence others begins early in life. About 48 hours prior to hatching, embryos of Japanese Quail (Coturnix japonica) and Northern Bobwhite (Colinus virginianus) produce clicking sounds. Embryos that can hear each other hatch more synchronously than do embryos separated two days before the expected hatching time (Vince 1969). Embryonic chicks communicate with parents, too. Budgerigars (Melopsittacus undulatus) begin to vocalize about two days before hatching, and playback of these recorded chick sounds into a nestbox stimulates intense incubating behavior by the parents (Berlin & Clark 1998). Young grebes in the egg call as they become cool, undoubtedly stimulating parents to incubate (Brua et al. 1996).

  Once hatched, nestlings of all species call to communicate their relative hunger to adults; parents monitor these calls and increase feeding rates accordingly (as demonstrated by experimental playbacks with Red-winged Blackbirds  Agelaius phoeniceus and Yellow-headed Blackbirds Xanthocephalus xanthocephalus--Burger & Trout 1979; Price 1998). Parents are so attuned to serving their young that unsuspecting parents can be exploited by the exaggerated vocal displays of the nestlings of brood parasites, such as the Common Cuckoo (Cuculus canorus, Kilner et al. 1999). The begging calls of the cuckoo are not learned, but simply exaggerated to exploit the predispositions of their foster parents. In contrast, a young cockatoo raised in the nest of another cockatoo species exploits its foster parents by learning the begging calls of its foster siblings, though the young cockatoo's alarm calls seem more resistant to learning (Rowley & Chapman 1986). In most species, the differences between signals for food and for alarm seem to be known instinctively by young birds (e.g. Buitron & Nuechterlein 1993).

  These calls can be the social glue on which avian societies are based. Through these calls, parents and their offspring of certain species can identify each other, as revealed, for example, in a nice series of experiments on swallows (Hirundinidae--Stoddard & Beecher 1983; Beecher 1990). Within flocks, Black-capped Chickadees reveal flock membership by converging on the same “chick-a-dee” calls (Nowicki 1989; Hughes et al. 1998), the four notes of which are combined in various ways to produce an enormous number of different call-types with great potential for conveying information (Hailman et al. 1985, 1987). Patrilines are identifiable in Stripe-backed Wrens (Campylorhynchus nuchalis), because young males learn their call repertoires from older male relatives; the birds can thus easily recognize others within their own family unit, but they can also recognize calls of relatives which have dispersed to other groups (Price 1999). Dialects occur in calls, too, such as in the “gargle” of chickadees (Miyasato & Baker 1999; Baker et al. 2000), the “rain call” of Chaffinches (Baptista 1990), and the flight whistle of the Brown-headed Cowbird (Dufty & Hanson 1999). For the cowbird, a male's flight whistle might be an honest indicator of a male's age and therefore a good cue for a fertile female (Rothstein & Fleischer 1987), but the dialectal influence of most calls on bird societies is largely unknown.

  As adults, birds tend to use their repertoires of relatively simple calls in specific contexts. A Jungle Fowl (Gallus gallus), for example, has about nine different sounds in its repertoire (Collias 1987), and uses a long drawn-out “baaaaawk” when a hawk flies overhead, but a “Baak-buk-buk-buk, etc.” when it sights a ground predator, such as a raccoon. Other chickens respond appropriately, looking around and skyward after hearing the “aerial-predator alarm” and around but not up for the “ground-predator alarm”, thus illustrating the functions of these particular calls (Evans & Marler 1991; Marler & Evans 1996).

  Aerial and ground predators are identified with different calls by songbirds, too (Marler 1955). Alarms for ground predators are pulsed and easily locatable (“chinks”), but aerial alarm calls tend to be whistles of high frequency, delivered on a sustained pitch that fades in and fades out and is therefore ventriloquial in quality (“seeets”). The difference in locatability of these two sounds was confirmed by testing two captive aerial predators, a Red-tailed Hawk (Buteo jamaicensis) and a Great Horned Owl (Bubo virginianus), for the ability to locate these two calls of the American Robin (Turdus migratorius). Response to playback was recorded by video cameras, revealing that mean orientation error was greater to chinks than to seeets (51·5 degrees versus 124·5 degrees--Brown 1982).

  Use of these predator “alarms” might not always be honest, however. Among Neotropical birds, as in Peru, foraging flocks often consist of five to ten species in the same home range throughout the year. In one flock system, the White-winged Shrike-tanager (Lanio versicolor) is a “sentinel”, giving alarm calls when danger threatens, and other species rely on the tanager's vigilance. On some occasions, however, the tanager appears to give false alarms: when the tanager and another individual converge on the same insect, an alarm call by the tanager causes the competitor to hesitate long enough for the tanager to rush in and capture the insect (Munn 1986). Whether or how often individuals might deceive each other for selfish gain is a fascinating area of study.

  Another specialized call used by a number of songbirds is the “nest-departure call”, which is given by a female as she leaves the nest. Sample species include the Red-winged Blackbird (Yasukawa 1989; Yasukawa & Searcy 1995) and White-crowned Sparrow (Hill & Lein 1985). An explanation of the exact function of this call remains elusive, but one hypothesis is that it signals the male to be especially alert for predators as his mate is foraging away from the nest.

  In general, what we know about the functions of calls in most species is based largely on educated guesswork, although flexible playbacks that enable the investigator to interact with birds are likely to enable great strides in the future (e.g. see Nielsen & Vehrencamp 1995; Dabelsteen & McGregor 1996; Smith & Smith 1996). Years of watching Winter Wrens (Troglodytes troglodytes), for example, led Armstrong (1955) to conclude that they use about 14 different calls. Most are given by both sexes, in predictable situations, but Armstrong admitted that the repertoire of different calls is probably more extensive, and that the many calls he classified undoubtedly intergraded with each other and even with song. As with most species, no sonagraphic analyses are available for the call repertoire of this wren, in part because the function of the more complex songs has been so much more attractive to study (see next section).

2) The function(s) of song

So much information is contained in a song. Information about the species is the most obvious (Bremond 1976; Becker 1982), as all species have songs recognizably different from those of other species (Nelson 1989). One reason species sing differently, of course, is that we base our classifications, at least in part, on song, so that groups with different songs are classified as different species. Beyond this circular aspect of species differences, however, we must beware of a hidden assumption in our thinking: the fact that we humans use song to identify different species does not necessarily mean that songs are different so that individuals of a species can recognize each other and prevent hybridization. It is entirely possible that species are distinct only because of intraspecific forces of sexual selection, and that song features of a species exist irrespective of the song features of other species. This possibility seems rather plausible, given the absence of any good examples of character displacement at species boundaries.

  Beyond identifying the species, song also routinely identifies the population from which the bird came (DeWolfe & Baptista 1995), as well as the sex (Farabaugh 1982) and individual (Falls 1982; Stoddard 1996). The singer's motivation, too, can be conveyed by patterns of delivery (Kramer et al. 1985; Highsmith 1989; Smith 1991), and perhaps age (Nottebohm & Nottebohm 1978), status or overall health (Nowicki, Peters & Podos 1998) by the repertoire size. Selection pressures on song must clash, because, for example, being distinctive as an individual conflicts with adhering to a stereotyped signal for identifying the species (e.g. Nelson 1989).

  Presumably all of this information is used to accomplish what has been widely assumed to be the two main functions of songs, “to repel rival males from their territory and to attract and stimulate females to breed with the male” (Catchpole 1989, p. 1046). We consider evidence for each in turn (see also Kroodsma & Byers 1991), then conclude this section with a brief discussion of female song.

Repelling rivals, i.e. Territoriality. Indirect evidence that song is used to interact with and repel rival males is abundant. Males often countersing with each other, and a taped song played to a territory-owner will elicit approach, aggressive displays (e.g. wing-quivering, aggressive trilling) and song, revealing that the owner treats the recorded song as if it were from an intruding competitor. Response to this form of playback is greatest at the center of the territory and least at the edge (Ickes & Ficken 1970; Melemis & Falls 1982).

  Experimental evidence of a territorial function for male song has been provided by muting the territory-owner. Early experiments provided suggestive evidence for Red-winged Blackbirds (Smith 1979) and Brown-headed Cowbirds (Dufty 1986), but these muted males could neither sing nor call, so the inter-male function of all vocalizations, not just song, was addressed in those experiments. Better evidence for a territorial function of song alone was provided by McDonald (1989), whose muted Seaside Sparrows (Ammodramus maritima) could call but not sing. The songless sparrows were delayed in acquiring territories and hampered in maintaining them.

  Also providing evidence for a territorial function of song are “speaker-replacement experiments”, in which males are removed from their territory and replaced with loudspeakers broadcasting their songs. Beginning with work on Thrush Nightingales (Luscinia luscinia) by Goransson et al. (1974), a series of studies has demonstrated that song alone can initially deter males from settling on a vacant territory (Krebs 1977; Falls 1988; Searcy 1988; Nowicki, Searcy & Hughes 1998). Deterrence is short-lived, however, as potential settlers eventually encroach on and claim the vacant territory.

  Nevertheless, there must be more to the function of song than mere defense of territory. Perhaps the most obvious reason is that, during the non-breeding season, both males and females of some species can use simple call notes to defend territories (Holmes et al. 1989), thus showing that song is not the only sound that can be used for territorial defense. This simple observation suggests that song is a special vocalization used by males (and females) to assess one another during crucial mating choices (see next section).

Song and female choice. Much of the recent interest in bird song has been related to sexual selection, i.e. to how song might be used by females to assess and choose sexual partners (see early review in Marler 1960; more recent review in Searcy & Yasukawa 1996). Although songs can be used in quiet “negotiations”, we more often think of sexual selection and males displaying, perhaps at full aerobic capacity, in an attempt to influence a mating partner. In this section, we review evidence that females are attracted to or stimulated by these loud male songs, beginning with indirect evidence and ending with more direct evidence.

  Indirect evidence that song is used to attract females to a territory is also abundant. Bachelor males typically sing much more than do mated males, for example, and naturally or experimentally widowed males sing more than do paired individuals (e.g. Wasserman 1977). In some species, such as the European Sedge Warbler (Acrocephalus schoenobaenus) and California Towhee (Pipilo crissalis), males cease singing altogether once paired (Marshall 1964; Catchpole 1973), demonstrating that song probably serves to attract females.

  Many studies have also found correlations between male singing behaviors and reproductive success (summary in Searcy & Yasukawa 1996). A male's repertoire size is correlated with his harem size in the Red-winged Blackbird (Yasukawa et al. 1980), for example, as it is in the Great Reed Warbler (Acrocephalus arundinaceus--Catchpole 1986). Among certain species of paruline warblers, males which successfully attracted a female used a greater diversity of songs than did unpaired males (Spector 1992). An exciting recent correlation was found by Hasselquist et al. (1996), who showed that song repertoire size in the Great Reed Warbler was correlated with the male's suitability as an extra-pair partner and with the survival of his offspring (Fig. 9).

  There is danger in accepting correlations as showing cause and effect, however, as demonstrated by a closer look at the above four examples. For the first two studies, male repertoire size was found to be highly correlated with age or territory quality, such that the statistical effects of repertoire size on harem size were no longer significant. For the warblers, at any given time the vocal behavior of a male is correlated with his mating status, but pairing causes a larger repertoire to be used, not vice versa. Only in the study by Hasselquist et al. (1996) are there no known confounding explanations.

  More direct evidence of the effects of song on females can be obtained from testing female responses to male song (Searcy 1992). One widely used technique is to play different song stimuli to females and measure the number or intensity of the copulation-solicitation displays that the vocalization elicits (King & West 1977). This technique has been used frequently to address two particular questions: whether a female responds differently to songs of her own and those of a foreign dialect, and whether the female responds more to a larger song repertoire (see additional discussion below). Although the quality of these playback experiments continues to suffer from lack of adequate replication of the playback stimuli (McGregor et al. 1992; Kroodsma et al. 2001), most studies reveal that females are attentive to the details of song stimuli: they respond more to songs of their own dialects and to “more variable” singing (though not necessarily larger song repertoires, because most experimental designs cannot distinguish between different levels of variability in the songs--Kroodsma 1990). Playbacks can also be used in the field; Logan (1983) observed, for example, that male Northern Mockingbirds sing intensely as females begin a new nesting cycle, and females renested earlier if they heard playback of male song (Logan et al. 1990).

  Several studies suggest that male quality, or the quality of his territory, might be honestly encoded in how he sings. In Willow Warblers (Phylloscopus trochilus), the quality of a territory seems to determine the rate at which a male is able to sing, and males which sing faster attract females earlier (Radesater et al. 1987; Radesater & Jakobsson 1989). Female Barn Swallows (Hirundo rustica) also seem to prefer males which sing at faster rates (Moller et al. 1998), and faster rates may directly reflect male health (Saino et al. 1997). In the Neotropics, the rate of singing also seems important for a lekking Ochre-bellied Flycatcher (Mionectes oleagineus), as his song rate is correlated with how many females (and other males) visit him (Westcott 1992). In another Neotropical lekking species, the Long-tailed Manakin (Chiroxiphia linearis), how well two cooperating males match the frequency of each other's song predicts how many females will visit them (Trainer & McDonald 1995). Male health might also be encoded in the duration of songs during the dawn chorus, because song duration can predict the source of extra-pair offspring in Blue Tits (Parus caeruleus--Kempenaers et al. 1997). The act of singing itself might not be energetically expensive (see Horn et al. 1995; Eberhardt 1996; Gaunt et al. 1996), but honesty may well be a feature of many aspects of male song (and other vocalizations), thus enabling females readily to compare males (Gibson & Bradbury 1985; Rothstein & Fleischer 1987; Staicer 1996b).

  The most convincing evidence of an influence of male song on females comes from experiments in which male song is manipulated and mating success is changed accordingly. One approach is to increase the amount of singing by free-living males by providing extra food; such an experiment with Pied Flycatchers (Ficedula hypoleuca) showed that food-supplemented males sang at higher rates and attracted females earlier than did control males (Alatalo et al. 1990). Similar approaches, in which only features of song (and not other features, such as the quality of the territory in the Pied Flycatcher example) are manipulated, are needed to test effects of song on females. One possibility is that different vocal behaviors could be “assigned” at random to different males in a laboratory setting. Repertoire size can be manipulated in Marsh Wrens from eastern North America, for example, with males learning as few as 5 or as many as 45 different songs (Brenowitz et al. 1995); if a sufficiently naturalistic laboratory setting could be established, the differential attractiveness of these males to females would be the most conclusive kind of test for an effect of repertoire size on mating preferences (Kroodsma & Byers 1991). This type of manipulative test, in which behaviors are assigned at random to different males, would provide the strongest test for how a particular male vocal behavior influences female choice.

  Exactly how male song can influence females is revealed by a number of experiments. Song can stimulate hormonal activity, for example (Brockway 1965; Tchernichovski et al. 1998). In African Collared-doves (Streptopelia roseogrisea) male songs stimulate females to vocalize and the female's own vocalizations stimulate oogenesis (Cheng 1992).  In White-crowned Sparrows, song acts synergistically with photoperiod in stimulating gonadal growth (Morton et al. 1985): females exposed to taped male song and long photoperiods (e.g. LD14:10) had faster rates of gonadal growth than did females exposed to long photoperiods but deprived of song. In canaries (Leboucher et al. 1998), song stimulates nest-building activities, and even naive, inexperienced yearling females that hear songs have shorter egg-laying latencies and lay larger clutches than do yearlings which do not hear song. Songs with particular phrases (“sexy” phrases) are especially likely to elicit copulation-solicitation displays from females. These experiments with canaries demonstrate not only the overall effects of song quality but also the effects of particular song components on stimulating various reproductive activities in female canaries (Kreutzer et al. 1992, 1996).

  Perhaps one of the greatest unexplored phenomena that should help us understand how song is used to influence females (and other males) is the dawn chorus (Mace 1987; Staicer et al. 1996). During the 30-60 minutes just before sunrise, the intensity of singing, at least in temperate zones, is extraordinary. Males sing rapidly, frequently switching song types (indicating high motivation--see below), sometimes filling the void between songs with call notes (e.g. Spector 1991), as if the intensity of the behavior dictated success. Sometimes singing males of otherwise territorial species gather at dawn in lek-like arenas, as do Chipping Sparrows, with up to 4 birds from neighboring territories singing in intense face-offs on the ground within a few metres of each other (Fig. 10; Liu 2001). The singing interactions among males at dawn may be monitored by females, whose (extra-pair) mating decisions might be based on who wins or loses the dawn vocal duels (Otter, Chruszcz & Ratcliffe 1997, Otter, McGregor et al. 1999). If extra-pair mating opportunities are reduced, the dawn chorus might be reduced, too, as in synchronously nesting tropical species (Morton 1996a). There certainly seems to be no better time of day than dawn, after males have fasted all night, for females to demand a performance on which they will base their decisions. The dawn chorus undoubtedly has much to reveal about the functions of song.

Female song. The study of singing in females has been a long-neglected field (Ratcliffe & Otter 1996), as reflected by the focus of our review. This neglect is probably due largely to the geographic locations where most bioacousticians study birds, in the temperate zone. If we lived in the tropics, our emphases would be different, as that is where females of many species do sing, such as wrens (Levin 1996), waxbills (Estrildinae--Güttinger 1976), grassquits (Tiaris--Baptista 1978), and many others (e.g. Farabaugh 1982).

  Even in the temperate zone, however, females of some songbirds sing. The female Northern Cardinal routinely sings (Lemon & Chatfield 1971; Ritchison 1986; Yamaguchi 1998), with songs somewhat more variable in structure than the male's; she apparently can use song to communicate with her mate about parental duties at the nest (Halkin 1997). Females of a number of other species in the temperate zone sing less regularly, or at times other than the peak of the breeding season. Song Sparrow females sing in the fall, perhaps a manifestation of female/female competition (Arcese et al. 1988), and sedentary female White-crowned Sparrows sing in the fall and winter, perhaps to aid their mates in defending territories against floaters of both sexes (Baptista et al. 1993). Females might also use song to defend their own winter territories, as in Townsend's Solitaires (Myadestes townsendi--George 1987) and European Robins (Erithacus rubecula--Hoelzel 1986).

  Females of species other than songbirds sing, too. Among columbiforms, females of many species sing (Cheng 1992; Baptista & Gaunt 1997b), as also do some female hummingbirds, such as the Blue-throated Hummingbird (Lampornis clemenciae--Ficken et al. 2000) and the Anna's Hummingbird (Calypte anna--Schuchmann 1979). Among the many duetting species listed by Farabaugh (1982) are many other female singers, too.

  These examples of “female song” again raise the question of what is a “song”. This dilemma is illustrated well by females of the genus Thryothorus. Females of some tropical species, such as the Buff-breasted Wren (T. leucotis), sing highly coordinated duets with their mates, with songs of the female and male both learned and equally complex, and with essentially equal investment in forebrain tissue for learning and controlling those songs (Farabaugh 1982; Brenowitz et al. 1985). In contrast, the female of the North American Carolina Wren (T. ludovicianus) uses a simple chatter to duet with her mate (e.g. Morton & Shalter 1977); this chatter is probably not learned and probably requires little, if any, investment in forebrain tissue (Nealen & Perkel 2000), but this simple chatter probably functions in much the same way as do the more complex, learned “songs” of her tropical relatives (most likely the joint defense of a year-round territory and mate--see below). The female Buff-breasted Wren clearly “sings”, but we are inclined to label the simpler vocalization of the Carolina Wren “only a call”, though for no truly objective reason. There is no obvious solution to this classification problem, of course, but none needs exist either, as use of the categories “call” and “song” is merely for our convenience and does not reflect any universal classification that the birds themselves use (see Spector 1994).

  The functions of female song are poorly known (Langmore 1998). Because of the odd time of year at which some temperate-zone females sing, it was thought that they perhaps were singing only because of a hormone imbalance (Thorpe 1961). More recently, however, it has been realized that female song is under fine neuroendocrine control, as in the female European Robin, which has elevated testosterone levels as it defends a winter territory (Kriner & Schwabl 1991). Defense of territory or a mating situation may be one function of female song (e.g. Baptista et al. 1993). Female song in the impressive duetting species of the tropics probably also functions in joint defense of territory and in intrasexual defense of a mate (Hooker & Hooker 1969; Thorpe 1972; Harcus 1977b; for discussions of function, see also Todt et al. 1981; Farabaugh 1982). Especially fascinating is the song of the female Alpine Accentor (Prunella collaris), which is polygynandrous; females use their songs to attract males (Langmore et al. 1996), just as males of typically monogamous or polygynous species use their songs to attract females. With a growing interest in studying female song, not only in resident tropical species but also in migratory species outside the breeding season, we shall begin to understand better the role of female song.

  Our focus on song production by females largely ignores another important field, that of song perception by females (Ratcliffe & Otter 1996). In most species, the female does not sing, yet it is she who chooses a mate, based in large part, we believe, on what a male sings. Just what does the non-singing female know about song, how does she know it, and to what use does she put that knowledge? How does she listen to her suitors? These are certainly some of the most crucial questions for understanding the evolution of song, yet we have so much to learn about how the female perceives her world.

3) The function(s) of song variation

Two particular features of song variation have been the object of frequent study. One is the number of different songs that an individual sings, and the other is how songs vary over geographic space. These two phenomena are largely a consequence of vocal learning, in that learning has permitted the development of large song repertoires and learning has resulted in micro-geographic song variation, or “dialects”. Although both phenomena result from song-learning, it should be stressed that one cannot therefore conclude that natural selection favored song-learning in some avian groups so that repertoire sizes could increase and dialects could form (see The evolution of sounds, below).

Repertoire size. Species differences in song repertoire sizes are considerable (for an attempt to standardize cross-species descriptions of songs, see Thompson et al. 1994; reviewed in Catchpole & Slater 1995). One of the simplest songs is that of the male Chipping Sparrow; his song consists of a repeated series of a single, brief syllable, and songs appear to vary only in overall duration because of different numbers of these repeated syllables that comprise each song (Borror 1959). Each song of a Swamp Sparrow (Melospiza georgiana) male is similarly simple, but careful listening reveals that the male first sings a series of songs based on one particular syllable, then a series based on another syllable, and so on, eventually revealing a repertoire of several different songs (3 or 4 is most typical; e.g. see Clark et al. 1987). An increasingly complex song repertoire is that of the Song Sparrow; males still tend to repeat a string of one particular kind of song before switching to another, but there are now roughly 5 to 14 different “song types”, each of which consists of several different phrases that vary in seemingly minor ways from rendition to rendition (Mulligan 1966; Searcy & Nowicki 1999). Song repertoires can be much larger, even huge. Western Marsh Wrens sing over 100 song types (review in Kroodsma & Verner 1997), Sedge Wrens 300-400 songs (Kroodsma, Liu et al. 1999), and Brown Thrashers over 1000 (Boughey & Thompson 1981).

  Species also differ in how they present their song repertoires (e.g. Hartshorne 1956). Species with relatively few songs often sing with “eventual variety”, producing one song type several times before switching to another. Species with larger song repertoires often sing with “immediate variety”, such that successive songs are different from and more sharply contrasted with each other. Birds that sing with immediate variety often sing rapidly or continuously, with successive songs especially different from one another (Verner 1976; Kroodsma 1978b; Whitney 1981), as if eager to display their song repertoire. Understanding the relationship between the similarity of successive songs (variety, or versatility) and the rate of singing (or continuity) may help us understand the evolution of large song repertoires, as in the European thrushes (Ince & Slater 1985).

  One important question to ask about these observations is whether or not our repertoire estimates mean anything to the birds themselves. Perhaps the best data are from Song Sparrows, the responses of which clearly show that what we identify as different song types (see Podos et al. 1992) are also distinctive to the birds, more so than the minor variations in renditions of each song type (Stoddard et al. 1992; Searcy et al. 1995). Other repertoire estimates seem reasonable, too, given that most are based on how the birds themselves learn and present their songs. What is often so impressive is that a songbird with hundreds of different songs in its repertoire (e.g. a Sedge Wren) can skillfully remember and reproduce the fine details of each song, so skillfully that sonagrams of thousands of songs unambiguously fall into a few hundred discrete categories. Songbirds are truly extraordinary in their ability to acquire (often by imitating) and store the details of these songs and in their ability to control the syrinx and vocal tract so that songs within a large repertoire are discrete, repeatable entities. All this is not to say, of course, that variation within each of these entities, or song types, is non-functional; the importance of this kind of variation has just received far less attention (Searcy & Nowicki 1999).

  Even though we can classify sounds into discrete categories, we should not, of course, be fooled into thinking that the variation within these categories is meaningless. Although we may classify the simple fee-bee-ee of a Black-capped Chickadee as a single song type, for example, the male pitch shifts that one form over a considerable range of frequencies (Ratcliffe & Weisman 1985). In all species, different renditions of what we label the same song type or call may vary in subtle to not-so-subtle ways, and perhaps we should think of repertoires of variation embedded within each of the types that we have defined, such that we have hierarchies of repertoires.

  What “good”, then, are these repertoires of different songs? What purpose do they serve? In the following paragraphs, we consider a few possible answers.

  One answer is that different songs can be used for different purposes, a phenomenon studied most thoroughly in the New World paruline warblers. In these species, one song (or group of songs) seems to be used primarily by unpaired males and in intersexual contexts after pairing; the other song (or group of songs) is used more in male-male countersinging, often in territorial conflicts (e.g. Ficken & Ficken 1967; Morse 1967, 1970; Lein 1978; Lemon et al. 1985; review in Spector 1992). A system in which songs have different functions can potentially convey more information; selection can also act independently on the two different song groups, creating, for example, geographically highly stereotyped songs in one group (female-attraction songs) or variable songs in the other (male-male interaction--see Byers 1996b; see Fig. 12 in next section; Staicer 1996a, 1996b).

  For most species, however, it seems that song repertoires are not functionally structured. Of what use, then, is a large song repertoire for these species if all of the songs have the same meaning? What advantage does a male have in singing a repertoire? One possibility is that motivational information can be encoded in how the repertoire is presented. A male that sings several different songs in succession is typically more highly motivated than is a male that sings several renditions of the same song during such a time span, as has been demonstrated for a number of species (e.g. the Song Sparrow--Kramer & Lemon 1983; but for an apparent exception to this general rule in the Banded Wren Thryothorus pleurostictus, see Molles & Vehrencamp 1999). Under high motivation, then, there is an increased probability that the next song will be different, whether the male is interacting with another male (Kroodsma & Verner 1978) or with a female (Searcy & Yasukawa 1990). Indeed, evolution of large song repertoires in some species may have occurred under ecological or social circumstances in which competition for resources was continually high, as in dense populations, where competing males interact frequently and intensely (Kroodsma 1999).

  Additional information can be encoded during presentation of a song repertoire if males learn their songs from each other. With “shared” repertoires, males can select particular songs so that they “match” each other during a countersinging performance (e.g. Northern Cardinal--Lemon 1968; Yellowhammer Emberiza citrinella--Hansen 1981; Common Nightingale--Todt 1981; or Western Meadowlark Sturnella neglecta--Falls 1985). It is as if a male names his opponent in a song duel (Armstrong 1963), and this “matched countersinging” has now been described for many songbird species (e.g. Krebs et al. 1981; see Whitney 1990 for an apparent example of how birds might avoid matching; see Beecher et al. 2000 for a special example of “repertoire matching” instead of “song-type matching”). These behavioral exchanges reach a feverish pitch in a species like the Marsh Wren of western North America (Fig. 11; Verner 1976): each male has 100-200 songs in his repertoire, and successive songs are different, so that a male must decide every few seconds which particular song he will sing next. He can choose to match a song that one of his neighbors just sang, or he can advance to the song that he anticipates next from a neighbor, or he can choose other songs, as if ignoring his singing neighbor. These escalated styles of interaction have evolved in a few other species, too, such as the Common Nightingale (Hultsch & Todt 1986; Todt & Hultsch 1996). The potential for conveying information in these countersinging duels is truly immense. Such exchanges, for example, could provide information about male dominance hierarchies or age, which would be relevant not only for territorial defense but also for prospecting females which eavesdrop on how males interact (e.g. Kroodsma 1979). These kinds of exchanges alone could have favored the evolution of large song repertoires in some species.

  The size of a song repertoire might also be an honest indicator of the singer's overall health (e.g. Nowicki, Peters & Podos 1998). In Sedge Warblers, a male's parasite load is negatively related to the size of his song repertoire and to his provisioning rate at the nest, so a female could use repertoire size as a cue to choosing a healthy mating partner (Buchanan et al. 1999). The number of songs that a Marsh Wren can learn appears to be related to the size of his song nuclei, which may be an honest indicator of his health (Brenowitz et al. 1995). In Bewick's Wrens, song repertoire size may be correlated with hatching date, and a young male hatching early in a breeding season may be healthier and have many advantages over young males hatching later in the season (Kroodsma 1974). A similar correlation between repertoire size and hatching date is found in Brown-headed Cowbirds (S. I. Rothstein, unpublished data). Indeed, song repertoire size is strongly correlated with longevity as well as with annual and lifetime reproductive success in one population of Song Sparrows (Hiebert et al. 1989), and repertoire size in Great Tits is also highly correlated with fitness (McGregor et al. 1981; Lambrechts & Dhondt 1986). The mechanisms by which repertoire size might influence success are unknown, however. One possibility is that males use larger repertoires directly to impress females or to acquire resources (e.g. a territory) critical for reproductive success; another possibility, of course, is that repertoire size is simply correlated with other male traits that enable success.

  All of this fascination with large song repertoires must be tempered with a discussion of why so many species have such small song repertoires, or no repertoire at all, i.e. a single song type. Given our belief in the power of natural selection, these species cannot be dismissed as inferior and incapable of responding to what we often think of as ubiquitous selection pressures for large repertoires; rather, for some reasons, these species must encounter different forms of selection for vocal variability, with necessary information (e.g. for female choice) encoded in one or a few song types. Evidence for selection limiting repertoire size comes from a number of species in which males practise much larger song repertoires than they eventually produce, showing that the brain and motor apparatus are certainly capable of singing larger song repertoires (Marler & Peters 1982). Song repertoires can be behaviorally constrained in individuals and in populations, as illustrated by the Black-capped Chickadee. From British Columbia to Nova Scotia, across the North American continent, male chickadees sing a highly stereotyped fee-bee-ee, a single song form that varies in frequency (Horn et al. 1992). In the laboratory, however, birds from these same populations achieve song repertoires of up to 4 different types, and groups isolated from each other form different dialects, just as repertoires and dialects form in certain isolated populations of this species (review in Kroodsma, Byers et al. 1999). Why repertoires (and interpopulation variability) are so constrained in some chickadee populations is unknown, though social factors involving female choice of mates is a likely suspect (Otter et al. 1998).

Geographic variation. How songs vary over geographic space also varies considerably among species. Among species that do not learn their songs, such as New World flycatchers (see above), geographic variation is low (Lanyon 1978); songs are “hard-wired”, and populations of individuals with similar genotypes have similar songs. Intraspecific song variation is also low in shorebirds, which likely do not learn their songs (Miller 1996). Indeed, if populational differences in song become too great among these non-learners, the populations are often considered to be separate species (Isler et al. 1997).

  When species do learn their songs, micro-geographic song variation often occurs, in the form of song “dialects”. Such dialects occur in a host of song-learning species, such as the sunbirds of Africa (Grimes 1974; Payne 1978b), the Winter Wren (Catchpole & Rowell 1993), Redwings (Turdus iliacus--Bjerke & Bjerke 1981), Corn Buntings (Miliaria calandra--McGregor 1980; McGregor & Thompson 1988), Neotropical hummingbirds (Gaunt et al. 1994), House Finches in North America (Carpodacus mexicanus--Mundinger 1975), Saddlebacks in New Zealand (Jenkins 1978), and, indeed, in almost every songbird that has been studied carefully (Krebs & Kroodsma 1980).

  The pattern of micro-geographic variation differs among species, and even among closely related species and among populations of the same species. Zonotrichia sparrows illustrate these patterns well, with sharp dialect boundaries occurring in the White-crowned Sparrows in the coastal chaparral of California (Marler & Tamura 1962; Baptista 1975) but not in Alaska (DeWolfe et al. 1974), perhaps a consequence of the short season for vocal development in Alaska (Nelson 1999). The congeneric Rufous-collared Sparrow of Central and South America has sharp dialect boundaries (Nottebohm 1969), but the North American White-throated Sparrow does not (Lemon & Harris 1974). Other contrasting patterns occur among populations of Melospiza sparrows (Searcy & Nowicki 1999), Cistothorus wrens, and Pipilo towhees (review in Kroodsma 1999). The literature on how bird songs vary over geographic space is truly immense (see reviews in Krebs & Kroodsma 1980; Catchpole & Slater 1995), and documenting how these song dialects occur (Lemon 1975) has certainly been a passion among field biologists.

  Showing how songs change over geographic space is much easier than determining why they change in the way they do. Any proposed explanations should, of course, focus on advantages that individuals might have in acquiring songs like or unlike those of other individuals, as the extent of geographic variation in song is a consequence of evolutionary “choices” made at the level of the individual.

  One possible influence on song could be the habitat in which song is used (Hansen 1979). Habitats do influence which types of sound transmit best (Wiley 1991), and, as habitats vary geographically, so too might song vary geographically (Bowman 1979; Gish & Morton 1981). Supporting data for this hypothesis come primarily from the Rufous-collared Sparrow (e.g. Nottebohm 1969; Handford & Lougheed 1991; Lougheed & Handford 1992). A match between song and habitat almost certainly does not, however, explain why the fine structure of songs changes at abrupt dialect boundaries in continuous habitat of other species, as with the White-crowned Sparrow in California chaparral.

  For Zonotrichia sparrows, another hypothesis has been proposed, the genetic adaptation hypothesis (Nottebohm 1972). Perhaps local birds that breed together maintain co-adapted gene complexes that favor local adaptations, and learning a local song dialect might help to maintain an optimal level of such inbreeding (e.g. Bateson 1978). This hypothesis predicts that young males and females will be influenced by their father and learn the songs of their natal dialect, that they will preferentially settle in and breed in the natal-dialect area, that dialect boundaries will inhibit dispersal, and that birds of adjacent dialects will differ genetically. Evidence supports some or all of these predictions, depending on who evaluates the data, and this issue was hotly debated in the mid-1980's (Petrinovich et al. 1981; Baker & Cunningham 1985; see review in Catchpole & Slater 1995). Although increasing evidence for several species shows that young birds can and do disperse to localities where songs are different from their father's (e.g. Payne & Payne 1997), we really do not know for any species the extent to which early experience with the song of the father influences the destiny of his offspring (see Grant & Grant 1996 for a possible example of sons learning the father's song and females mating with males which have songs unlike that of their father). Obtaining answers to these important questions about dispersal and mating patterns in nature is simply very difficult.

  Another explanation for why songs vary over space has been attributed to “social adaptation”. This hypothesis proposes that a young bird recruited into a local breeding population will have some advantage if he learns the local songs. This hypothesis is essentially a truism, given that young birds typically do learn local songs where they breed, so the real quest is to determine what the advantage might be. One possibility is that birds that use local songs convey their membership in the community (Feekes 1977); because learning the songs takes time or effort, a male singing the local songs indicates honestly his tenure within the community, and his songs might thus be a cue to prospecting females (Rothstein & Fleischer 1987). Perhaps learning a song from a dominant neighbor enables a young male to obtain an adjacent territory, especially if the dominant male has opportunities for extra-pair fertilizations in the mate of the younger male (Greene et al. 2000). A contrasting (though perhaps less likely?) explanation is that learning the song of a dominant, older male might convey some benefits to a younger male if the younger male might then occasionally be mistaken for the older (Payne 1981); the benefits result, then, from deception, not honesty.

  These three hypotheses aside, we almost certainly can use existing patterns in geographic song variation to inform us as to function. Consider, for example, the functionally structured song repertoires of the Chestnut-sided Warbler (Fig. 12; Byers 1995, 1996a, 1996b). The songs used in male-male interaction are shared only by immediate neighbors; these songs must be learned at the site where a young male will breed, because the local neighborhood is the sphere of influence for these songs. In contrast, the limited repertoire of four different songs used to attract females is highly stereotyped, and each song occurs throughout the geographic range of the species; these songs are thus highly conservative, both temporally and geographically, as a male apparently strives to attract a female from any geographic origin. For these two song systems, the sphere of expected influence for a song matches the pattern of geographic variation.

  Another example is provided by the Cistothorus wrens. A male Sedge Wren in North America improvises a large repertoire of species-typical songs, and songs of neighboring males seem to be no more alike than songs of distant males; the apparently uniform distribution of song characteristics throughout the geographic range enables a male to communicate with any male or female that might be encountered in the semi-nomadic lifestyle of this species. Strikingly different is the pattern of geographic song variation among resident male Marsh Wrens of western North America. A young male imitates the songs of his adult neighbors precisely, and they engage in highly coordinated, matched-countersinging duels (Verner 1976); neighboring males know each other, as they reside next to each other over a lifetime, and the sphere of influence of these particular learned songs would be expected to be more local.

  A corollary of this approach is that some patterns of geographic variation will occur by chance, because at increasing geographic distances selection for sharing of vocal characters will diminish. Sharing of vocal characters tends to decline rapidly over distance in resident populations, and less rapidly in populations that are migratory or less site-faithful (e.g. Ewert & Kroodsma 1994). At some distance, then, beyond the expected sphere of influence of songs, i.e. in the absence of selection for vocal similarities, accumulations of copy errors or improvisations may lead to differences in songs (Searcy & Nowicki 1999). Some features of songs could remain constant over large geographic distances, of course, even in the absence of selection, such as when some aspect of song structure is genetically inherited and phylogenetically conserved (e.g. Martens 1996); similarity of song characters over larger stretches of geographic space might then reveal past spheres of influence, or the phylogenetic history of the larger group.

  A final topic that begs for discussion is that of cultural evolution (Lynch 1996; Payne 1996). Learned songs within populations are cultural traditions, and these learned components (memes) vary not only over space but also over time. Populations of song characters change over time because of relationships between song tutors and their pupils and because of how and from whom an individual male chooses his particular songs to learn. For Indigo Buntings, the cultural survival of a song learned by a young male is estimated to be about three times the survival rate of the male himself (half-life survival of song is 4·2 years, that of male 1·3 years; Payne 1996). A special case of cultural evolution was described by Trainer (1989): males within a colony of Yellow-rumped Caciques (Cacicus cela) all have the same songs (Feekes 1977), but songs change during the lifetime of individual singers, so that males must constantly listen to each other and adjust their songs accordingly (much as happens within entire populations of humpback whales Megaptera novaeangliae--Payne et al. 1983). Common Starlings, too, adjust to different social situations by modifying their songs, so that songs within a colony change over time (Adret-Hausberger 1986; Adret-Hausberger et al. 1990; Hausberger 1997). Another example of this kind of cultural evolution is found among the suboscine Three-wattled Bellbirds (Procnias tricarunculata) in Costa Rica, where males throughout each of two dialects modify their songs over time to match each other (Kroodsma et al. 2001). How these various learned song traditions might influence genetic population structure, i.e. the coevolution of cultural and genetic traits, and how they might affect speciation are especially fascinating areas of inquiry (Lachlan & Slater 1999; see above).

4) Vocal variation and intelligence

In all of these discussions of calls, songs, and how and why they vary, we often think of birds as robots, with males and females responding in some reflex-like fashion to sounds in their environment. This thinking on our part is reinforced by the seemingly “stupid” behavior of birds in some circumstances (such as repeated attacking of a self-image in a window), but a surge of interest in animal cognition (e.g. Balda et al. 1998) is helping us take another look at the mental abilities of birds. Some communicative studies of birds reveal capabilities that are truly extraordinary, suggesting that these birds use considerable mental capacities when interacting with other birds (e.g. Hausberger 1993; Kroodsma & Byers 1998; Smith 1998).

  One suggestion of “intelligence” in songbirds derives from their similarity to us and our immodest opinions of ourselves. We learn to speak, and these songbirds learn to sing; we babble as we practise, as do the young birds, although for them the babbling is called “subsong”. Both humans and songbirds must be able to hear in order to learn, and in both groups the learning leads to vocal dialects. Brains of both are lateralized for sound learning and production, too. Some warblers seem to learn not only their songs but also the contexts in which to use them (Kroodsma 1988b), which is what we do with our words, too. These oft-cited parallels between human speech and bird song (Marler 1970a) need have no bearing on intelligence in the birds, of course, but the similarities do force us to think about what intelligence is, and how we differ, or how much we differ, especially given the seeming absence of similar vocal learning among chimpanzees and other primates (Marshall et al. 1999; Mitani et al. 1999).

  One particular group of birds, the corvids, is widely regarded as intelligent (e.g. Heinrich 1989, 1995), but we know so little of their vocal behavior (e.g. see Thompson 1982). Crows, ravens and jays do not sing in the usual sense, but they have complex vocal repertoires (Hardy 1967; Elowson & Hailman 1991). Ravens in particular are renowned for their cunning as well as their diverse repertoire of squawks, rattles, chortles, yells, and so much more. As we find the courage to tackle the complex vocal communication systems of these corvids, we shall surely learn more about the minds of these intelligent creatures.

  Pepperberg's (1999) recent review of her studies with Alex, the African Grey Parrot (Psittacus erithacus), certainly force us to realize that not all birds are the creatures of instinct that we once thought they were. As Alex competes for attention with two humans, he learns how to use his human language in appropriate contexts (a technique developed by Todt 1975). By speaking in a “tongue” that we humans can understand, Alex has revealed that he knows far more about his world than we would have given him credit for if he had vocalized only in parrotese. He can identify numerous items in his environment, count them, and can request or refuse them. He can identify the color, shape or make-up (“matter”) of an item, and can group items by these three categories. More impressively, when asked what a group of items has in common, he can tell whether it is the shape, the color, or the matter from which they are made. He knows the concept of “none”, too, because when asked what numerous dissimilar items have in common, he responds with the word “none”. These abilities are revealed in a highly artificial laboratory environment with human social companions, but each grey parrot in nature must also have these same abilities. Just how might those abilities be used in nature? What kind of information is really conveyed between parrots with what seems like a series of amorphous, meaningless squawks? We do not know, but Pepperberg's work with Alex has clearly encouraged us to rethink the minds of the birds that we work with.

The evolution of sounds

What we know about how and why avian sounds have changed over evolutionary time is inferred largely from our studies of current song function among a diversity of species. A danger lurks here, however, because one cannot assume that current functions or characteristics of a vocalization necessarily reveal the reasons for origin of that behavior (Gould & Lewontin 1979). For example, although song-learning enables large song repertoires and regional dialects, we cannot assume that selection for repertoires or dialects favored the origin of vocal learning.

  One big question to which we would like an answer is, of course, “How or why did song evolve?” Because of presumed current functions, we often assume that complex male song arose via sexual selection to impress females, and the literature is replete with studies and theory about sexual selection and how song is used to impress other individuals. This pathway certainly seems plausible; a recent study of Zebra Finches, for example, showed that females preferred even longer and more complex songs than males routinely deliver (Neubauer 1999). The best we can do on evolutionary questions, however, is to generate plausible explanations, but we should also remind ourselves that our “song category” is highly artificial (Spector 1994). In sexual selection theory, we think of functional definitions of song, but then we should think more broadly, in terms of what behaviors (or other features, such as feather structures) function to impress members of the opposite sex. We then realize that all species must have a “song”, because mate choice occurs in all species, but that the functional “song” of some species might not be vocal.

  Another big question is “Why did vocal learning evolve?” Learning involves risks, and many examples of interspecific song-learning “mistakes” have been documented (e.g. Baptista & Morton 1981). Few hints are offered by surveying the distribution of vocal learning among bird groups. Songbirds learn their songs, of course, but members of their sister suborder within the order Passeriformes had been thought not to (Kroodsma 1988a). New data now show, however, that bellbirds (Procnias spp.), close relatives of the flycatchers, do learn their songs, thus confirming Snow's earlier conclusions (Snow 1977; Kroodsma et al. 2001). Other vocal learners include the parrots (Farabaugh & Dooling 1996; Wright 1996), certain hummingbirds (Snow 1968; Wiley 1971; Baptista & Schuchmann 1990), and probably a duck (Musk Duck Biziura lobata--McCracken 1999), with hints of some subtle degree of vocal learning in a few other avian groups (Sparling 1979). Explanations as to why vocal learning evolved range from protecting the inner ear during loud vocalizing (Nottebohm 1991) to an arms race over “ranging” (i.e. using sounds to estimate distance of conspecifics--Morton 1996b), but no current explanations are satisfactory.

  As this question is pondered, it is important to realize that social learning can involve not only imitating the signal but also learning how to use that signal in appropriate social contexts (e.g. West et al. 1997; Janik & Slater 2000), and distinguishing among the different forms of learning will help us understand more clearly their respective origins. We should also remember the surprising conclusion of the model developed by Lachlan & Slater (1999), that once song-learning has evolved in a lineage, it might be maintained by an “evolutionary trap”, and reverting to non-learned signals might be impossible. Whatever the evolutionary patterns of vocal learning, our knowledge of the world is far from complete, and additional surveys of poorly known groups are needed, especially in the New and Old World tropics (Kroodsma, Vielliard & Stiles 1996).

  Inferences about the evolution of male song characteristics are sometimes made from correlations between current behavior and ecological context. This comparative approach suggests that large song repertoires develop when songbird males compete intensely for mates or territories, as in dense populations (review in Kroodsma 1999). In none of the studies suggesting this correlation (Catchpole 1980; Kroodsma 1983; Catchpole & McGregor 1985), however, is the phylogeny of the birds known, so that identifying independent evolutionary events is impossible in reconstructing evolutionary pathways. Ideally, one would want to know phylogenies and even to estimate how males or females would respond to ancestral vocalizations (Ryan & Rand 1995). More studies like Irwin's (1990) are needed, in which she mapped repertoire size onto a phylogeny of New World blackbirds and concluded that directional selection cannot explain the evolution of different repertoire sizes among these species.

  The evolution of singing behaviors can also be inferred from patterns of geographic variation. The Greenish Warbler (Phylloscopus trochiloides) has relatively simple songs in the Himalayas, but in two independent radiations extending to the north its song becomes more complex. Irwin (2000: 998) concludes that “parallel south-to-north ecological gradients have caused a greater intensity of sexual selection on song in northern populations and that the stochastic effects of sexual selection have led to divergence in song structure.”

  Because vocalizations have diverged in different avian lineages, sounds can be used as characters in systematics (Lanyon 1969; Payne 1986; Miller et al. 1988; Martens 1996; Miller 1996). Vocalizations that are not learned, such as most call notes or songs of suboscines, are more conservative characters and therefore best track phylogenies. The non-learned song differences between the Alder and Willow Flycatchers (Empidonax alnorum and E. traillii), two sibling species, were used to identify them as species (Stein 1958), just as current studies of non-learned suboscine or non-passerine songs in the Neotropics are revealing additional species (Fig. 13; Isler et al. 1997, 1998; Robbins & Stiles 1999). Non-learned calls also identify evolutionary units, as in Winter Wrens, in which eastern and western populations have markedly different call notes (Garrett & Dunn 1998).

  Learned vocalizations can also offer clues to phylogenetic histories, but interpreting the data requires more caution. In North America, for example, the learned songs of eastern and western populations of the Marsh Wren are unmistakably different, even though a male is capable of learning songs of either tradition (Kroodsma & Canady 1985); characteristics of the learned songs have been preserved over space and time, thus preserving information about phylogeny. Similarly, sympatric Red Crossbills (Loxia curvirostra) with different learned calls appear to mate assortatively, so that different breeding populations can be identified by their call notes alone (Groth 1993). Sometimes phylogenetic information among learned vocalizations can be found in traits that are the more genetically based (e.g. see Baptista 1996). The ability to learn song repertoires may differ genetically among populations, for example, such that the number of songs a male sings, but not the qualities of the songs themselves, can provide a useful systematic character (Kroodsma & Canady 1985); other traits, too, may be relatively conservative because their variability is limited by morphology (Podos 1996, 1997, 2000).

Looking forward

We hope that this foreword begins to reveal the extraordinary world of bird sounds and the passions of those who thrive on it. The thriving and passion occur at all levels, of course, from the poets who just want to listen and connect with and celebrate our Planet Earth, to the neurobiologists who search for how songs may be stored in the brain's neurons. In every scientist who studies bird song, however, there lies a poet, too, as all who are privileged to know this world of bird sound are deeply affected by it, well beyond whatever the biological significance of our discoveries might be. Epitomizing this inevitable multi-faceted effect that bird sounds have on us was Luis Baptista, the poet, philosopher, musician, biologist, and all-around enthusiast. It is his enthusiasm and outright joy of knowing the beauty of bird sounds, at all levels, that this foreword celebrates.

  And where do we go from here? In so many different directions! Some of our initial concerns must be about conservation of birds, to preserve avian biodiversity on this planet. Use of bird sounds is crucial here (Baptista & Gaunt 1997a), not only to help determine the units of biodiversity we wish to preserve (e.g. Isler et al. 1997), but to help survey critical habitats (Parker 1991) or to monitor the number of migrants passing overhead (Evans 1994). Sound playback can be used not only to locate and identify individuals or populations, but also to help re-establish breeding populations, a technique used successfully with colonial seabirds (Kress 1992). Sound can also be used as a tool to identify individuals, which is valuable if catching or handling members of an endangered population would be detrimental (Telford 1993).

  An increasingly important role in conservation (as well as in other studies of bird sound) is being played by the major archives of bird sounds (Kroodsma, Budney et al. 1996), as the sounds archived there can aid studies of all kinds, from conservation to systematics to general behavior. A sampling of those archives would include the Library of Natural Sounds at the Cornell Laboratory of Ornithology, Ithaca, New York; the Wildlife Section of the National Sound Archive at The British Library in London; the Borror Laboratory of Bioacoustics at Ohio State University; the Bioacoustics Archives and Library at the Florida Museum of Natural History in Gainesville, Florida; the Australian National Collection of bird sound recordings at the National Film and Sound Archive, Canberra; the Fitzpatrick Bird Communication Library at the Transvaal Museum, Pretoria, South Africa; the Veprintsev Phonotheka of Animal Voices in Puschino, Russia; the Phonotek of Animal Voices, Lomonosov State University, Moscow, Russia; the Tierstimmenarchiv at Humboldt University, Berlin; the Fonoteca, Museu de Zoologia in Barcelona, Spain; the Laboratorio de Sonidos Naturales, Museo Argentino de Ciencias Naturales “Bernardino Rivadavia”, Buenos Aires, Argentina; and the sound archive at the Universidade Estadual de Campinas, Brazil. As these archives grow, they become increasingly valuable in many ways, and amateur as well as professional biologists are encouraged to contribute much-needed recordings of bird sounds to these archives (Bradbury et al. 1999).

  Beyond conservation, the possibilities are limitless, as revealed by the diversity of questions that have been contributed to this foreword by leading bioacousticians throughout the world. So much remains to be learned, from how the syrinx works, to how songs are stored and controlled in the brain, to why birds acquire and use their sounds the way they do, to how and why sounds have changed over evolutionary time. We hope that the collective passions of those who study bird sounds, not the least of which were those of the late Luis Baptista, will serve as a welcoming invitation to others who would join us in the quest to understand what bird sound is all about. 

Acknowledgements: Thank you, to friends and colleagues of Luis Baptista who contributed your questions as a tribute to Luis; to Eliot Brenowitz, Bruce Byers, Toby Gaunt, Frank Gill, Mort and Phylis Isler, Wan-chun Liu, Irene Pepperberg and William Munoz, Kathleen Berge and Helen Horblit, and David Sibley for your help with the figures; to HBW editors Andy Elliott and Josep del Hoyo and your staff for a class act; to Greg Budney, Bruce Byers, Eben Goodale, Elijah Goodwin, Henrike Hultsche, Steve Johnson, Wan-chun Liu, Jeff Podos, Gary Stiles, Nick Thompson and, especially, David Spector for your help on the manuscript; to Sandy Gaunt for helping from A to Z; to the National Science Foundation for research support (IBN-9408520); and a special thanks to those around the world who have helped make the study of bird sounds a lifetime passion for so many of us. My apologies to those who would have liked to participate in this tribute but whom I overlooked or could not reach in time.