called notes. (This usage of “motif” for starling songs is at slight variance
with motifs as defined for zebra finch songs in Section 2.3.3.1.) Notes can
be broadly classified by the presence of continuous energy in their spectrographic representations, and although several notes may occur in a given
motif, their pattern is usually highly stereotyped between successive renditions of the same motif. One can thus consider starling song as a sequence
of motifs, where each motif is an acoustically complex event. The number
of unique motifs that a male starling can sing (i.e., his repertoire size) can
be quite large, and consequently different song bouts from the same male
are not necessarily composed of the same set of motifs. This broad acoustical variation in their song provides several potential cues that starlings
might use when learning to recognize the songs of an individual conspecific
and while maintaining that recognition over time. One straightforward
recognition mechanism is the association of specific motifs with specific
singers. Although some sharing of motifs does occur among captive males
(Hausberger and Cousillas 1995; Hausberger 1997), the motif repertoires
of different males living in the wild are generally unique (AdretHausberger and Jenkins 1988; Eens et al. 1989, 1991; Chaiken et al. 1993;
Gentner and Hulse 1998). Thus, learning which males sing which motifs can
provide a discriminative cue for song classification.
Data from operant studies in starlings support the idea that recognition
is based at the level of the motif. Starlings trained to recognize individual
conspecifics by one set of song bouts can readily generalize correct recognition to novel song bouts from the same singers (Gentner and Hulse 1998;
Gentner et al. 2000; Fig. 7.4A, B). However, when these novel song bouts
have no motifs in common with the training songs, and when song exposure outside of the operant apparatus is restricted, recognition falls to
chance (Gentner et al. 2000; Fig. 7.4C). Likewise, starlings trained to discriminate among pairs of motifs will reverse the discrimination when transferred to the same motif sung by the opposite individual and perform at
chance when transferred to novel motifs sung by the training singers
(Gentner 1999). This failure to generalize correct recognition to songs composed of novel motifs, or to single novel motifs, is inconsistent with the use
of individually invariant source and/or filter properties (voice characteristics) for vocal recognition.
The data suggest that starlings learn to recognize the songs of individual
conspecifics by attending to information contained at (or below) the level of
the motif and by then associating distinct sets of motifs with individual
singers. If this is true, then once recognition is learned, it should be possible
to control it systematically by varying the proportions of motifs in a given
bout that come from two “vocally familiar” males. That is, recognition behavior ought to track the statistical distribution of motifs from two vocally
familiar males rather than the presence or absence of single motifs from
either male. Recent data confirm this prediction (Gentner and Hulse 2000;
Fig. 7.4) and thereby suggest that when starlings are compelled to
7. Neuroethology of Vocal Communication
359
with motifs as defined for zebra finch songs in Section 2.3.3.1.) Notes can
be broadly classified by the presence of continuous energy in their spectrographic representations, and although several notes may occur in a given
motif, their pattern is usually highly stereotyped between successive renditions of the same motif. One can thus consider starling song as a sequence
of motifs, where each motif is an acoustically complex event. The number
of unique motifs that a male starling can sing (i.e., his repertoire size) can
be quite large, and consequently different song bouts from the same male
are not necessarily composed of the same set of motifs. This broad acoustical variation in their song provides several potential cues that starlings
might use when learning to recognize the songs of an individual conspecific
and while maintaining that recognition over time. One straightforward
recognition mechanism is the association of specific motifs with specific
singers. Although some sharing of motifs does occur among captive males
(Hausberger and Cousillas 1995; Hausberger 1997), the motif repertoires
of different males living in the wild are generally unique (AdretHausberger and Jenkins 1988; Eens et al. 1989, 1991; Chaiken et al. 1993;
Gentner and Hulse 1998). Thus, learning which males sing which motifs can
provide a discriminative cue for song classification.
Data from operant studies in starlings support the idea that recognition
is based at the level of the motif. Starlings trained to recognize individual
conspecifics by one set of song bouts can readily generalize correct recognition to novel song bouts from the same singers (Gentner and Hulse 1998;
Gentner et al. 2000; Fig. 7.4A, B). However, when these novel song bouts
have no motifs in common with the training songs, and when song exposure outside of the operant apparatus is restricted, recognition falls to
chance (Gentner et al. 2000; Fig. 7.4C). Likewise, starlings trained to discriminate among pairs of motifs will reverse the discrimination when transferred to the same motif sung by the opposite individual and perform at
chance when transferred to novel motifs sung by the training singers
(Gentner 1999). This failure to generalize correct recognition to songs composed of novel motifs, or to single novel motifs, is inconsistent with the use
of individually invariant source and/or filter properties (voice characteristics) for vocal recognition.
The data suggest that starlings learn to recognize the songs of individual
conspecifics by attending to information contained at (or below) the level of
the motif and by then associating distinct sets of motifs with individual
singers. If this is true, then once recognition is learned, it should be possible
to control it systematically by varying the proportions of motifs in a given
bout that come from two “vocally familiar” males. That is, recognition behavior ought to track the statistical distribution of motifs from two vocally
familiar males rather than the presence or absence of single motifs from
either male. Recent data confirm this prediction (Gentner and Hulse 2000;
Fig. 7.4) and thereby suggest that when starlings are compelled to
7. Neuroethology of Vocal Communication
359