of reinforcement or stimulus sequences (e.g., Leak and Gibbon, 1995; Meck and Church, 1982, 1984; Pang and McAuley, this volume; Rousseau and Rousseau, 1996). It is unclear, however, whether there are specialized systems for the perception and production of serial-ordered behavior that rely on the same interval timing mechanisms that are engaged by the types of stimuli that are typically presented in laboratory studies of timing and time perception. One such system might be the vocal learning circuit of songbirds, which is highly structured and specific.

To date, the study of interval timing and birdsong acquisition has remained largely detached (but see Hills, this volume; Weisman et al., 1999). Laboratory research on interval timing has traditionally used rats, pigeons, and humans. In contrast, the songbird may be an ideal organism for studying certain aspects of interval timing because the neural pathways that are involved in song learning and production are well described from both a molecular and electrophysiological perspective. Although this description is by no means complete, the interaction between song learning and production coupled with the neurobiological findings makes the songbird amenable to bio-behavioral study. Consequently, this chapter will begin with a brief introduction to the temporal patterns of behavior occurring in birdsong and a hypothesis of a connection to interval timing. The latter part of this chapter will explore the neural systems underlying birdsong in the context of what is known about the neurophysiology of interval timing in mammals. This is followed by the proposal of an alternative solution for song learning that relies on the same interval timing mechanisms studied in other species.


Songbirds comprise nearly half of the known living avian species and thus make up the largest suborder of birds. Ethologists divide songbird vocalizations into songs and calls. In general, songs are longer and more syntactically complex sounds relative to calls, although this is not an absolute distinction. Songs are organized by a temporal hierarchical structure. The simplest individual sounds that a bird can make are termed notes. A bird can string together several notes without a silent interval, in which case it is called a syllable. These syllables are further arranged into sequences to form song types or motifs. The durations of song types can also vary among species from the 10-sec utterances of the male winter wren to the 1-sec hiccup of the winter Henslow. Characteristically, an individual songbird will produce several song types, called its repertoire. However, the repertoire number can vary greatly from species to species.

Typically, a songbird will produce bouts of singing by arranging discrete song types into a sequence. The song types are separated by variable intervals of time that may be as short as 5 sec during an intense bout of singing. The bouts themselves also vary among species. A bout will typically contain many vocalizations whereby a bird will systematically cycle through its song types (e.g., A, A, A, B, B, B, B, C, C, C, ... ). The number of songs in a bird's repertoire, and how they are presented, is thought to be an important functional aspect of communication (Hartshorne, 1956; Krebs, 1977; Kroodsma, 1980). Accordingly, birdsong is a complex motor sequence that brings to light a dilemma that commonly arises in the study of serial-ordered behavior (Lashley, 1951). How do songbirds learn and express motor programs that are marked by rigid ordinality and variable tempo while maintaining a consistent hierarchical structure?

16.2.1 The Development of Avian Vocal Behavior

Up to this point, all studies conducted on the oscine suborder of songbird species have revealed that song is learned (Ball and Hulse, 1998; Kroodsma and Baylis, 1982). Although the specifics in the development of birdsong can vary among species, some useful generalizations have emerged. Vocal learning in the songbird can be divided into two stages, a sensory phase and a sensorimotor integration phase (Konishi and Nottebohm, 1969; Marler, 1970).

In the sensory phase, a young songbird is exposed to conspecific utterances from either its father or other males that are within hearing distance. The juvenile songbird is believed to form a sensory template that represents the memorized vocalizations from the tutor so that it may be used to guide song development. Before the sensorimotor integration phase begins, the bird goes through a period called subsong. Subsong is characterized by amorphous and unstructured vocalizations akin to a soft babbling and is considered nonimitated vocal motor practice. This leads to the sensorimotor integration phase, where subsong progresses to plastic song and many of the elements that are common to the memorized template begin to appear; however, the final organization is incomplete (Marler and Peters, 1982). Recent findings suggest that juvenile zebra finches that are not exposed to song will begin vocalizing by instinctively producing a string of back-to-back syllable prototypes. The acquisition of an auditory representation permits different syllables to emerge from these prototypes (Tchernichovski et al., 2001). Lastly, the final form of the birdsong develops, which represents the adult crystallized song type.

16.2.2 Behavioral Parallels between Interval Timing and Birdsong

The presiding thought many years ago was that the auditory template guiding singing was tantamount to a motor tape so that each vocal production of a song type was considered invariant (e.g., Konishi and Nottebohm, 1969; Marler, 1970). However, a song type may be better thought of as a theme that underscores a number of vocalizations, some of these being improvisations that are scarcely repeated (Podos et al., 1992). As a result, in song sparrows it has been said that the "motor representation of song appears to be probabilistic, with each song type stored as an abstract average that carries with it probabilities describing an allowable range of within-type variation" (Podos et al., 1992, p. 104). Another way of interpreting this definition is that a song type may be defined by the probability of there being systematic behavioral regularities as a function of time.

Defining birdsong in this manner parallels some of the properties of interval timing (see Church, 1984, this volume). Scalar timing theory accounts for the decision of when to begin responding during an interval with a ratio comparison (e.g., Church and Gibbon, 1982; Gibbon, 1977; Gibbon et al., 1984). The comparison is made between two values. The first value is the current estimate of elapsed time since signal onset. The second value is the representation of the expected time of signal termination, selected from a normal distribution of expected times in memory. Once the current estimate of elapsed time exceeds a fixed fraction of the expected time to signal termination, the animal begins to respond in that trial. With regard to the development of birdsong, interval timing could mediate the juvenile songbird's perception of the temporal relationship among the syllables within the song during the sensory phase. Once an auditory representation of the adult song is formed, it may serve to impose temporal structure onto a motor sequence. This modulation of the vocal motor sequence would typically occur during the sensorimotor phase of development.

How would the auditory representation be formed in the first place? Each syllable in the song is equivalent to a subdividing stimulus that provides information about the hierarchical structure of the song. Within this structure, song completion serves as a common reference point, as there is evidence that it could serve as salient feedback (e.g., Adret, 1993; Stevenson, 1967, 1969). This arrangement is amenable to a system that mediates behavior where multiple timing mechanisms would operate from syllable onsets until song completion. The ability to independently time each perceived syllable within the song in relation to song completion may facilitate the emergence of specific patterns of motor production, as observed in other examples of simultaneous temporal processing (e.g., Meck and Church, 1984).

Because a syllable's duration can vary, there is a potential problem in determining what point in time within the syllable is credited with stimulus onset. This problem is avoided when one considers that categorical perception underlies a songbird's discrimination between syllables (Marler, 1989). By imposing a boundary on the syllable's continuum, discriminations among syllables within a given category are made ineffectual, while discriminations among syllables that straddle the boundary are made with relative ease. Therefore, in order for the syllable to be a reliable time marker a salient feature must be categorically perceived at some consistent time point within the syllable. The perception of this time marker within the syllable is a function of its overall dynamic spectra. How the songbird selects the conspecific syllable from the hodgepodge of sounds that pervade its auditory space is not known. There is evidence, however, that conspecific and heterospecific syllables are not equipotential as song learning stimuli (Marler and Peters, 1988). Moreover, some species of songbird learn songs by perceiving salient, introductory whistles that appear to draw attention to the song that follows (Soha and Marler, 2000).

There are indications that the syllable can be used as a reliable time marker in order to predict song completion. Canaries can learn to use perceived syllables as time markers during song playback if the song's offset is consistently paired with a noxious stimulus, like in a shock avoidance paradigm (Jarvis et al., 1995). The canaries first initiated an avoidance response at the onset of the song. However, with continued training, they gradually transferred the avoidance response to the final syllable in the song, right before electric shock was given. It was as if once the specific time of the shock's occurrence in relation to song onset was determined, it was used to guide the canary's decision of when to produce a fear response (cf. Gibbon, 1977).

Is it possible that the songbird's neural system evolved in such a way that it was able to exploit an ability to attend to brief durations of numerous auditory stimuli that simultaneously elapse (e.g., Meck and Church, 1984; Rousseau and Rousseau, 1996)? Would this allow the songbird to process the various auditory signals in parallel for the guidance of vocal development? In order to substantiate this hypothesis, a logical place to begin would be to first describe the similarities between the mammalian and avian brain. This would allow us to compare the properties of the various neural circuits implicated in interval timing in mammals with the avian neural circuits that have been found to contribute to birdsong.

0 0

Post a comment