lable at ScienceDirect

Animal Behaviour 88 (2014) 125e135

Contents lists avai

Animal Behaviour

journal homepage: www.elsevier .com/locate/anbehav

Female zebra finches learn to prefer more than one song and frommore than one tutor

Marie-Jeanne Holveck a,b,*, Katharina Riebel a

aBehavioural Biology Group, Institute of Biology (IBL), Leiden University, The NetherlandsbBiodiversity Research Centre, Earth and Life Institute, Université Catholique de Louvain (UCL), Louvain-la-Neuve, Belgium

a r t i c l e i n f o

Article history:Received 23 June 2013Initial acceptance 25 July 2013Final acceptance 29 October 2013Published onlineMS. number: 13-00528R

Keywords:birdsongcondition dependencydevelopment of mating preferencesfemale song learninglearned biasesmultiple tutorsphenotypic plasticityzebra finch

* Correspondence: M.-J. Holveck, Université Catholand Life Institute (SST/ELI/ELIB), Biodiversity ResearLetter Box L7.07.04, Croix du sud 4-5, 1348 Louvain-la

E-mail address: marie-jeanne.holveck@uclouvain.b

0003-3472/$38.00 � 2013 The Association for the Stuhttp://dx.doi.org/10.1016/j.anbehav.2013.11.023

Birdsong is a culturally transmitted mating signal: young birds learn specific variants of species-specificsong(s) from conspecific models. Female song preferences are also learned early in life, but despite thepotential functional implications of such learned mating preferences, we still have a poor understandingof when and from whom females learn. This also holds true for one of the foremost models of vocallearning, the zebra finch, Taeniopygia guttata. Both male and female zebra finches memorize their tutor’ssong motif: as adults, males will sing and females prefer their tutor’s song. We here tested whetherjuvenile females would also learn to prefer the songs of several individuals, and whether the timing andpropensity of song preference learning were condition dependent. Young females raised and cross-fostered in experimentally manipulated brood sizes were exposed to several model songs: first theirfoster father’s song until nutritional independence (days 0e35) and then as subadults to playbacks oftwo different tutor songs (days 35e65). As adults, females preferred all three model songs over unfa-miliar songs. There were no interaction effects between females’ early rearing conditions (brood size)and preference strength for the different tutor songs. An additional live-tutored group had equally strongpreferences for the foster father’s song (only heard before day 35) as the tape-tutored females. Thecombined results demonstrate that subadult females memorize several song types during different timesof development and as adults prefer these songs over unfamiliar songs. These findings imply thatmultiple song learning needs to be taken into account for avian mate choice studies even in species thatlack song type repertoires but show individual differences in song.� 2013 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.

Learned mating preferences are increasingly recognized as anevolutionary force but to estimate their qualitative and quantitativeimpact on population dynamics, the strength and direction oflearned versus experience-dependent preferences need to be bet-ter documented (e.g. Verzijden et al., 2012) in order to improvemodelling approaches (Lachlan & Nowicki, 2012). In songbirds(oscines), (learned) song is an important mating signal (Catchpole& Slater, 2008) but while the process of (male) song learning isrelatively well studied, our understanding of female song produc-tion and preference learning lags behind in all aspects (Kroodsma,Vielliard, & Stiles, 1996; Riebel, 2003b; Riebel, Hall, & Langmore,2005) although the ‘when’ and ‘from whom’ females learn theirpreferences have turned out to be crucial parameters in theoreticalmodels for the evolution of learned birdsong (Lachlan & Nowicki,2012; Ritchie, Kirby, & Hawkey, 2008). Zebra finches, Taeniopygiaguttata, are one of the few species in which song acquisition in

ique de Louvain (UCL), Earthch Centre, Carnoy Building,-Neuve, Belgium.e (M.-J. Holveck).

dy of Animal Behaviour. Published

females has seen some systematic study (for reviews see Riebel,2003a, 2009). Complex courtship song in zebra finches is pro-duced only by adult males: each male sings an individuallydistinctive song motif that closely resembles an adult song model(the ‘tutor song’). As shown by sequential exposure to differenttutors, the peak of the sensitive phase for (male) song productionlearning is between 35 and 65 days posthatching (Slater, Eales, &Clayton, 1988). Although females do not sing, several experi-mental studies have shown that females learn about songs as ju-veniles too: after maturation, which can be as early as 100 days(Zann, 1996), adult females have been shown to prefer the song ofthe adult male with which they had been housed between 35 and65 days posthatching over the song of an unfamiliar male, showingcultural transmission of song preferences is possible along nonge-netic lines (Clayton, 1988, 1990; Miller, 1979; Riebel, Smallegange,Terpstra, & Bolhuis, 2002). Song preferences in female zebrafinches have been validated as predictors of live male preferencesand mating decisions (for a review see Riebel, 2009). Cross-fostering between the subspecies T. g. guttata and T. g. castanostishas shown that females not only learn preferences for specificsongs but also generalize some of the subspecies’ structural

by Elsevier Ltd. All rights reserved.

M.-J. Holveck, K. Riebel / Animal Behaviour 88 (2014) 125e135126

features (Clayton, 1990); however, how generalization processes oflearned preferences affect variation at the within-population levelis poorly understood (Riebel & Smallegange, 2003). Early exposureto song playback alone (‘tape tutoring’) is sufficient to induce astable (repeatable) preference in females for that particular songover unfamiliar songs (Riebel, 2000) which provides us with amethod with excellent stimulus control to expand on these earlierstudies. Although experimental studies have provided good evi-dence that adult song preferences develop at least partly bylearning from models (Riebel, 2003a), the details regarding whenand from whom female zebra finches learn are still poorly under-stood. For one, there is no clear delineation of the duration of thesensitive phase as there has not been the same experimental workin females as in males (Riebel, 2003a). For males, memorization ofsong can begin as early as 25 days posthatching but these early songmemories translate into adult song (which fully crystallizes atmaturation at around 100 days posthatching, Zann,1996) only if themale is experimentally deprived of further song input during theremainder (and peak) of the sensitive phase (35e65 days post-hatching; Böhner, 1990; Jones, ten Cate, & Slater, 1996). Outsidesuch experimental situations, males of this colonial species are,however, continuously exposed to song (Zann, 1996) and willpreferentially learn those songs heard during the sensorimotorlearning phase (35e65 days) as evidenced by the young male’ssong resemblance to these later rather than the earlier songs(Slater, Richards, & Mann, 1991). Females can also start memorizingsong as early as day 25 as evidenced by increased calling of nes-tlings towards familiar over unfamiliar songs (Clayton, 1988), butwhether these memories persist into adulthood if females aresubsequently exposed to other songs has not been tested. More-over, experiments on juvenile song preference learning haveexposed females to only one type of song during the (putative)sensitive phase from 35 to 65 days posthatching. However, in morenatural settings young females are likely to hear more than onetype of song during this phase. Up until 35 days posthatching birdsare cared for by both parents and will have predominantly heardtheir father’s song, but when zebra finches become nutritionallyindependent at around 35 days posthatching they will startflocking (Zann,1996). From this stage onwards they are exposed notonly to the songs of their father but also to more than one singingmale throughout the sensitive period for song learning.

Our first aim in this study was therefore to test the hypothesisthat females could simultaneously learn more than one tutor songduring the putative peak of the sensitive phase postindependence(i.e. 35e65 days posthatching, see Riebel, 2003a) if exposed to re-cordings of two different songs throughout this phase. If both songswere memorized, as adults females should show preferences forboth tutor songs over unfamiliar songs. Our second aimwas to testwhether females would also learn several songs sequentially, morespecifically whether females would also prefer the song they hadonly heard before independence (i.e. until 35 days posthatching)even if thereafter they heard only other songs. Again we predictedthat if all sequentially heard songs were memorized, as adults, fe-males should show preferences for all tutor songs over unfamiliarsongs. Third, because effects of rearing condition on cognitivefunctions within mate choice contexts are increasingly hypothe-sized but rarely tested empirically (Boogert, Fawcett, & Lefebvre,2011; Buchanan, Grindstaff, & Pravosudov, 2013; Riebel, 2011;Ritchie et al., 2008) we tested whether female condition wouldaffect the timing and/or quantity of learned songs. If adversedevelopmental conditions affect memorization then, as adults, fe-males from low-quality rearing environments should show weaker(or even no) preferences for the tutor songs over unfamiliar songsthan females from high-quality rearing environments. To this end,all tests were run with females from an experimental brood size

manipulation that in past studies and in our birds had inducedsignificant differences between birds from small and large broodsizes in morphology, physiology, adult song and social-learningpreferences (Holveck & Riebel, 2010; Riebel, Naguib, & Gil, 2009;Riebel, Spierings, Holveck, & Verhulst, 2012; Verhulst, Holveck, &Riebel, 2006). All females first heard the song of their foster fa-ther while being reared in experimental small (two to three chicks)or large (five to six chicks) brood sizes. Upon reaching nutritionalindependence at day 35 posthatching, females from differentrearing nests were experimentally matched to form pairs consist-ing of one female from a large and one from a small brood. Fromthen onwards each pair of tape-tutored females was exposed toplaybacks of two different songs for the remainder of the sensitivephase for song learning. This allowed us later on to compare indi-vidual pairs of females of different rearing backgrounds for thequantity and strength of song preference learning by testing howmany of the early tutor song types they would prefer (and howstrongly) over unfamiliar songs in adulthood. No previous studyhas investigated whether early live tutoring (until nutritional in-dependence at day 35 days posthatching) preceding the tapetutoring might bias song memorization of females towards the livetutor of the early juvenile phase. The timing and model selection inmale song learning can be affected by when and how malesencounter a live tutor (Böhner, 1990; Jones et al., 1996) and thismight also hold for female preference learning. We hypothesizedthat the absence of a live tutor postindependence could potentiallybias preferences towards the preindependence tutor in tape-tutored females more so than in live-tutored females. Therefore,when females were tested for preferences for the early live tutor(i.e. their foster father) an additional group of females from thesame brood size manipulation but raised in live-tutored, mixed-sexgroups (Holveck & Riebel, 2010) was also tested for preferences forthe songs they had heard before day 35 posthatching.

METHODS

Subjects and Housing Conditions

The subjects were a subsample from a large brood size manipu-lation experiment conducted in two breeding rounds in 2 differentyears (2004 and 2005) in the zebra finch breeding colony at LeidenUniversity (Holveck, de Castro, Lachlan, ten Cate, & Riebel, 2008;Holveck & Riebel, 2010). To break correlations between parentalquality and brood size and possible genetic links between fathers’songs and females’ preferences, chicks (N ¼ 113) of first-timebreeders were assigned to nests of unrelated foster parents at3� 2 days posthatching (mean� 1 SD) to form either small (two tothree chicks per nest) or large broods (five to six chicks) assigningsiblings to different foster nests in different brood size categories(Holveck et al., 2008). Chicks were housed with their foster parentsin standard laboratory cages (80� 40 cm and 40 cm high) on a13.5:10.5 h light:dark schedule (lights on 0700e2030 hours CentralEuropean Time) at 20e22 �C and 35e50% humidity. Cages had solidsidewalls and were stacked three cages high in three rows along thelength of the roomandwith other birds in earshot and in sight at 2 macross the aisle. Throughout, birds had ad libitum access to a com-mercial tropical seed mixture (Tijssen, Hazerswoude, Holland),drinking water and cuttlebone. This was supplemented thriceweekly with egg food (Witte Molen, B.V., Meeuwen, Holland), twicewith millet branches and once with germinated seeds. After thestudy, we returned the experimental birds to the colony.

All procedures followed to the ASAB/ABS Guidelines for the Useof Animals in Research and Dutch laws and were approved byLeiden University’s Animal Experimentation Committee (Dier-experimentencommissie Universiteit Leiden, permit DEC 04090).

M.-J. Holveck, K. Riebel / Animal Behaviour 88 (2014) 125e135 127

When reaching nutritional independence at 34 � 3 days post-hatching, the juvenile birds were removed from their foster nestgroups: 20 females were assigned to 10 ‘tape-tutoring’ groups (eachconsisting of two females, one from a large and one from a smallbrood), and 26 females and 26 males (the birds from Holveck et al.,2008; Holveck & Riebel, 2010) were assigned to mixed-sex ‘live-tutoring’ groups (each consisting of four experimental birds, onefrom each sex*brood size category combination; housed with anadult tutor and his mate in the same group cage within the colonyroom). After the tutoring phase (34e70 days posthatching; seeTable 1), tape- and live-tutored females were treated alike: thetutee groups were split and all experimental females were nowhoused inmixed-treatment, mixed-tutoring, same-sex groups (fourto five birds) in laboratory cages (at least 80 � 40 cm and 40 cmhigh) stacked along the walls in a large colony room with otherbirds in sight and in earshot until song preference tests began inadulthood (see below and Table 1). The tape-tutored femalesreceived four different preference tests. The live-tutored femalesonly participated in test 4 (note that they had also received threepreference tests prior to the fourth and final test as part of adifferent study reported in Holveck & Riebel, 2010; see Table 1 foroverview).

The brood size manipulation had affected birds’ structural size(Holveck et al., 2008; Holveck & Riebel, 2010; Holveck, Geberzahn,& Riebel, 2011) and this was also true for the subsample in thisstudy: at 180 days posthatching females from large broods hadshorter tarsi than those from small broods (large broods: mean � 1SD ¼ 15.2 � 0.5 mm; small broods: 15.6 � 0.4; c2

1 ¼ 8:1, P ¼ 0.004;N ¼ 22 and 19 successfully tested females, respectively, see below;linear mixed models with year, hatching nest, foster brood andtutor group ID as random factors) but did not differ in absolutemass (large broods: 17.5 � 2.3 g; small broods: 17.0 � 2.0 g;c21 ¼ 0:1, P ¼ 0.8) or size-corrected mass (calculated as standard

residuals of the linear regression of mass on tarsus size; c21 ¼ 1:2,

P ¼ 0.3). Morphological traits (mass, size, size-corrected mass) didnot differ between the subsamples assigned to the differenttutoring procedures (live versus tape) or for the manipulated broodsize*tutoring procedure interaction (all 0:01 � c2

1 � 2:3,0.1 � P � 1.0).

Tape-tutoring Procedure

The 20 females were grouped into 10 pairs: eight pairs in whichone female originated from a large and one from a small brood andanother two pairs in which both females originated from large

Table 1Timeline of experimental procedures

Experimental phase Age (dph) (�1 SD)

Nestling phase 3 (�2)Brood size manipulation 3 (�2)e34 (�3)Song tutoring Tape Live

34e70 (�4) 34e70 (�3)

Song preference tests Tapez LivexTest 1 207�39 164�15Test 2 230�39 182�14Test 3 241�38 197�18Test 4 254�36 211�19

Live: live-tutored birds; Tape: tape-tutored birds; dph: days posthatching; UFam: unfami* These were 53 males and 54 females and six unsexed chicks that had died before sey There were 54 birds raised in small and 59 in large broods; sex ratio at this stage wz Only 17 of the 20 females were successfully tested (see text for details).x Only 24 of the 26 females were still alive at this stage.

** The first three tests of the live-tutored females took place within a separate study testibrood sizes (for details see Holveck & Riebel, 2010).

broods (because there were overall more females from largebroods). Within a pair, females never originated from the samenatal or foster nest. Each pair was housed in a cage (70 � 30 cm and45 cm high) placed in a separate sound-attenuated chamber(100 � 200 cm and 220 cm high) and tape-tutored with a differentset of two different songs simultaneously (for four examples seeFig. A1 in the Appendix): one originating from a male from thezebra finch colony (but never from a foster father) at Leiden Uni-versity (from now on referred to as ‘natal’ colony) and one from thecolony at Utrecht University (‘non-natal’ colony). Songs from twodifferent colonies were chosen in order to detect potential un-learned biases for own colony song (Riebel, 2003b) by comparingpreference strengths for both tutor songs in each female. Choosingthe two tutoring songs from two colonies also ensured that the twotutor songs were as different as possible to avoid generalizationduring learning and testing (songs within one colony are moresimilar to each other than between two colonies, van Heijningen,2012). During tutoring, playbacks of the two songs werecontrolled and data logged by two custom-built minicomputerswith an Oki MSM6388 (Tokyo, Japan) sound chip (Houx & ten Cate,1999), each connected to one of two identical loudspeakers (Blau-punkt CB 4500, Hildesheim, Germany) placed behind the wiremesh at each side, outside the cage. The two tutor songs were eachplayed back randomly at a mean rate (�1 SD) of 5.1 � 0.1 songs/hresulting in a total of 10.3 � 0.3 songs/h for both songs (N ¼ 10 � 2different tutor songs) between 0815 and 2115 hours (lights on0800e2130 hours) starting 1 h after rehousing (Riebel, 2000). Thissong rate is within the range of diurnal singing activity of males inour colony (Jesse & Riebel, 2012), has been shown to induce stableand repeatable song preferences in tape-tutored females (Riebel,2000) and is above the minimal song rate needed for male songlearning (Tchernichovski, Lints, Mitra, & Nottebohm, 1999).Throughout the whole tutoring period, songs were switched be-tween the two loudspeakers every night. Amplitude was set to70 dB peak amplitude at 30 cm from the loudspeakers (re 20 mPa;CEL-231 sound level meter, fast response F and low range A LOsettings, Lucas CEL Instruments Ltd, Hitchin, Herts, U.K.). Tutoringstopped at 70 � 4 days posthatching (N ¼ 20), the age at which thelive-tutored females (70 � 3 days posthatching, N ¼ 26) were alsoremoved from their tutors and returned to the colony (see Table 1).

Stimulus Songs

Spontaneous undirected songs of 40 males (aged at least 100days posthatching) were recorded under standardized conditions

Details and sample sizes

Cross-fostering (N¼113 chicks)*Rearing in foster nestsy

Tape LivePlaybacks of 2 songs(N¼20 F in 10 FF groups)

1 male tutorþfemale(N¼26 F in 13 FFMM groups)

Tapez LivexTutor 1�UFam 1 LFam�SFam**Tutor 2�UFam 2 LUFam�SUFam**Tutor 1�Tutor 2 LUFam�SUFam**Foster father�Fam Foster father�UFam

liar; Fam: familiar; F: female; M: male; L: large foster broods; S: small foster broods.xing.as 1:1.

ng condition dependency of female preference for songs of males from small or large

M.-J. Holveck, K. Riebel / Animal Behaviour 88 (2014) 125e135128

at Leiden and Utrecht Universities. In Leiden, all foster fathers(N ¼ 25) and an additional five males were first placed singly in acage (70 � 30 cm and 45 cm high) in a sound attenuation chamberand then recorded from a distance of 75 cm (see Holveck et al.,2008) with a Sennheiser MKH40 microphone onto a PC’s harddisk (CDX-01 soundcard; Digital Audio Labs), using the ISHMAELsoftware (v. 1.0.2, freely available at http://www.pmel.noaa.gov/vents/acoustics/whales/ishmael/) with the following settings:automatic energy detection 2000e10 000 Hz, detection threshold1, detection limits 0.2e100 s, buffer 3 s. In Utrecht, all recordedmales (N ¼ 10, originally purchased from a commercial breeder)were kept and recorded under similar conditions as in Leiden, i.e.while housed individually in sound-proof chambers (115 � 115 cmand 205 cm high), but with Sennheiser MKH 50P48 directionalmicrophones with the recorder of Sound Analysis Pro (v. 1.056,http://ofer.sci.ccny.cuny.edu) and under a 16:8 h light:darkschedule (this was 2.5 h more daylight than for the Leiden birds).

From these recordings, we edited 25 foster father stimuli, 10natal colony (Leiden) and 10 non-natal (Utrecht) tape-tutor songs.Five of the natal tutor songs originated from the pool of recordingsof the foster fathers, but for tutoring we always chose songs frommales that were unfamiliar and unrelated to the tape-tutored fe-males. For each stimulus, we chose one four-motif song per male(natural range is one to eight, Zann, 1993), because songs of thisduration have been validated as a reward in the operant set-up andto predict preferences for livemales (Holveck & Riebel, 2007; Riebelet al., 2002). We digitally deleted introductory elements that werenot part of the core motif (Holveck et al., 2008), high-pass filteredthe songs at 500 Hz (smoothing ¼ 100 Hz) to remove low-frequency background noise and root mean square-equalized am-plitudes (peak digitally scaled to 1) with the PRAAT software (4.2.07for Windows, http://www.praat.org). Within a pair of stimuli, tape-tutor songs were matched for duration as much as possible (meanmotif duration per song ¼ 0.76 � 0.20 s, N ¼ 20 tape-tutor songs;mean difference between average motif lengths of two songs in thesame stimulus set ¼ 0.08 � 0.07 s, N ¼ 20 sets).

Song Preference Tests

Females were tested in four operant two-choice preference testswith (mean � 1 SD) 9 � 3 days of rest in their home cage betweentests (for details see below and Table 1). Testing started at 164 � 15days posthatching for live-tutored females (Holveck & Riebel, 2010)and 207� 39 days posthatching for tape-tutored females (Table 1).The tape-tutored females were significantly older (Table 2), but ageat testing did not differ between foster brood size categories(Table 2).

Preference Tests for Tape-tutor Songs (Tests 1e3)

In each of the first two tests of the tape-tutored females, femalescould choose between one of their tutor songs and one unfamiliarsong from the same colony (¼a tutor song from another tutorgroup). The design was fully balanced with respect to song originand test order: half of the females received songs from their natal

Table 2Comparison of tape- and live-tutored females’ age at the start of preference testing

Model terms Estimate�1SE LRT c21 P

Tutoring 0.148�0.051 6.8 0.009Brood size 0.001�0.027 0.002 0.96Tutoring*Brood size 0.060�0.051 1.3 0.26

Rejected terms are in italics, significant one in bold. LRT: log-likelihood ratio test.Model ¼ Age at testing startw Tutoring*Brood size. GLMM with Poisson distribu-tion and random factors year, hatching nest, foster brood and tutor group ID.

colony (their natal tutor and one natal unfamiliar song) in their firsttest and then non-natal colony songs in their second test; the otherhalf of the females were tested in reverse order. Because no pre-vious study had tutored females with more than one song simul-taneously (Riebel, 2003a, 2009), we could not formulate a prioriexpectations with respect to whether females would later in lifeprefer only one or both of the songs. Therefore, in the third test, allfemales were given the choice between their two tutor songs. Thiswas to test whether simultaneous tutoring with two songs resultedin two equal preferences or whether females had a favourite songamong the two tutor songs and whether this was specific withrespect to the origin of the song or test order in adulthood.

Females from the same tape-tutored group remained pairedwithin the design throughout testing. Each pair of females receiveda unique set and order of stimuli for the first three tests (sameprocedure as for the live-tutored females, Holveck & Riebel, 2010).Moreover, each tape-tutor song was also used once as unfamiliarsong with another pair of females. Hence, each song was a tape-tutor song for one tape-tutored group and an unfamiliar song foranother tape-tutored group thereby controlling in the design forpotential intrinsic differences in song attractiveness.

Foster Father versus Unfamiliar Song (Test 4)

The fourth test investigated whether females had also acquiredpreferences for their foster father’s song they had heard only before35 days posthatching and before the actual tutoring phase from 35to 65 days posthatching. In this test, in addition to the tape-tutoredfemales a control group of live-tutored females was also tested.These live-tutored females had also been in three different operantsong preference tests before test 4 (but this was during a differentstudy, see Holveck & Riebel, 2010, and timeline in Table 1). In thisfourth preference test, each female could choose between the songof her foster father and an unfamiliar song (which was always thesong of the foster father of the other female in her tutor group). Thisresulted in a fully balanced paired design for the stimulus songassignment with 21 different stimulus sets all used once (with theexception of one stimulus set that was used for two tutor groupswith the same foster fathers).

Operant Set-up for Preference Testing

Song preferences were tested in an operant set-up described andvalidated previously (Holveck & Riebel, 2007, 2010; Riebel, 2000;Riebel et al., 2002). Briefly, females were moved individually intoan experimental cage placed in a sound attenuation chamber wherethey remained until the end of each test. Cages were furbished withdispensers providing ad libitum water and food and five perches. Ifperched at the far end of the two outmost perches, females couldpeck two small red buttons in the back panel of the cage. For eachtest, each of the two test stimulus songs in a set was assigned to oneof the buttons, and with each peck to either button, females couldtrigger one playback of the song assigned to that side from a loud-speaker placed behind a central opening of the back wall. Becauselearning in some females was rapid the training and the first testwere combined and training startedwith the stimulus songs for test1 (rather than separate training songs). When moved into theexperimental cage for the first time, several females discovered bythemselves (N ¼ 7) that pecking the red keys yielded a song reward,which meant that their testing could start the day after (see below).The 13 females that had not started to peck the keys on the secondday received daily training sessions (see Holveck & Riebel, 2007) ofabout 20 min from then onwards. During training, care was taken touse both songs equally often as the reward. The training resulted inanother four females learning the operant task within the first

100

50

0 50Song preference strength

for familiar tape tutor in test 1

Son

g p

refe

ren

ce s

tren

gth

for

fam

ilia

r ta

pe

tuto

r in

tes

t 2

100

Large brood: Natal tutor 1st test songLarge brood: Non-natal tutor 1st

Small brood: Non-natal tutor 1stSmall brood: Natal tutor 1st

Figure 1. Tape-tutored females’ preference strength (% choices) for the familiar tutorsong against an unfamiliar song from the same colony (Leiden ¼ natal colony orUtrecht ¼ non-natal colony) in test 1 versus test 2 per brood size category. Eachsymbol is an individual female value and shows the origin of the males whose songswere first heard in test 1. The dashed lines show random preference.

M.-J. Holveck, K. Riebel / Animal Behaviour 88 (2014) 125e135 129

round. In this first training and testing round, females stayed onaverage (�1 SD) 6.5 � 2.1 days (range 3e11, N ¼ 20) in the operantset-up. Then training was suspended for 7.9� 6.7 days for thenonlearners (N ¼ 9) for a rest in their home cage (with three or fourother females) in the colony room. Training was then resumed andtwo females learned the operant task in the second training session,one in the third and three in the fourth one (5.6 � 1.9 days in op-erant set-up, range 3e10, N ¼ 9). These additional training phaseswere always alternated with rest periods (9.2 � 7.2 days). After this,females had either learned the operant task (N ¼ 17) or wereomitted from further testing (N ¼ 3). No further training wasrequired after the first preference test, as once females have learnedthe operant task they can be retested even after considerable timespans (e.g. 6 months later, Riebel, 2000). For these subsequent tests(tests 2e4) females spent on average 3.2� 0.5 days (range 3e5days, N ¼ 17 tape-tutored females in tests 2e4 and an additional 24live-tutored females in test 4) in the operant set-up per test. Theadditional live-tutored females in test 4 did not require training asthey had learned the operant task during their earlier preferencetests preceding this study (Holveck & Riebel, 2010).

The actual preference test started on the morning after the dayfemales were observed to have learned the association betweenkey pecking and song reward. For this we monitored the data logsof the custom-built minicomputer that was controlling the songplayback as well as keeping a time-stamped data log of all peckingactivity. Learning was assessed by the following criteria: the datalogs showed a transition from not pecking at all to occasionalpecking and then to a steep increase in pecking activity at bothkeys. Throughout training and testing, assignment of songs to thetwo buttons was switched during the lights-off period. After thefirst day on which females showed consistent and regular peckingactivity, the actual test was scheduled to start the next morning.Each preference test then lasted for 2 days during which femaleshad unlimited access to song. Assignment of stimulus songs wasfully balanced with regard to presentation on the left and rightsides, and stimulus presentation was side-reversed on the secondtest day. For each preference test, we assessed two response vari-ables: (1) the total number of keypecks over the 2 test days and (2)song preference strength, defined as the proportion of keypecks forthe tutor song versus an unfamiliar song (tests 1 and 2), or non-natal versus natal colony tutor (test 3) or the foster father’s songversus an unfamiliar song (test 4).

Three tape-tutored females (two from large broods in amatchedpair and one from a small brood) did not learn to perform theoperant task. This left song preference data for a total of 17 tape-tutored females for the analyses: seven pairs with one femalefrom a small and a large brood eachwith identical tape tutoring andthree tape-tutored females that were not matched for brood sizemanipulation (one pair of large-brood females with identical tapetutoring and one unmatched female from a large brood). All sta-tistical analyses of the song preference datawere run oncewith andonce without the three tape-tutored females unmatched for broodsize manipulation. As the results did not change qualitatively whenthe three unmatched females were excluded, we only report theanalyses of the complete data set.

Statistical Analyses

Within each test, the two response variables (song preferencestrength and keypecks) showed no pronounced correlations (tape-tutored females: �0.22 < Spearman rS < 0.23, N ¼ 17, all P > 0.3;live-tutored females: rS ¼ 0.15, N ¼ 24, P ¼ 0.5), and were thus bothkept for analyses. For each test, we tested the departure of songpreference strength (following arcsine transformation to ensurestabilization of the variance) from chance level (50%) with one-

sample Student t or Wilcoxon signed-ranks tests. We analyseddata with two-tailed (a ¼ 0.05) generalized linear mixed models(GLMM) with Poisson and binomial distribution for the number ofkeypecks and preference strength, respectively, in R v.3.0.1 underthe lme4 package (R Development Core Team, 2013), which allowedus to integrate the pecking activity into the preference analyses withthe ‘cbind’ function (number of pecks for one song, number of pecksfor the other song).We added an observation-level random effect tocapture the overdispersion when needed (e.g. Elston, Moss,Boulinier, Arrowsmith, & Lambin, 2001). We always started withthe full model and used a stepwise backward selection procedure onfixed factors until we reached the minimal adequate model andalways kept the model with the best fit based on the log-likelihoodratio test. We report the estimate (�1 SE) of each fixed factortogether with the chi-square statistics of the comparison betweenthe models with and without the tested factor.

In analyses of repeated measures within longitudinal data (i.e.the tape-tutored females’ preferences measured in successivetests), we fitted, as fixed factors, all two-way interactions betweentest number (1e4), the females’ foster brood size (small versuslarge), and the origin of the male stimulus pair (Leiden ‘natal’versus Utrecht ‘non-natal’) or the origin of the males whose songswere first heard in test 1 (to check for potential priming effects). Inaddition, female identity was included as a random factor nested inyear, hatching nest, foster brood and tutor group ID, and studied ininteraction with test number; a significant interaction betweenfemale ID and test number (both random factors) would indicatesome between-individual variation in the response to test number.In analyses of a single test (i.e. test 3 or 4), the random factors wereyear, hatching nest, foster brood and tutor group ID only; weincluded brood size in interaction with the origin of the males intest 1 to analyse preference strength in test 3; we added females’tutoring procedure (live versus tape) in interaction with brood sizeand age at testing as fixed factors in analyses including both tape-and live-tutored females (test 4).

M.-J. Holveck, K. Riebel / Animal Behaviour 88 (2014) 125e135130

RESULTS

Preference for Tape-tutor Songs

The first two song preference tests (tests 1 and 2) both testedadult females’ preferences for a familiar tutor song versus an un-familiar song from the same colony. In both tests, females showedan overall preference for familiar tutor songs over unfamiliar songs(one-sample Student t test: test 1: mean preference strength � 1SD ¼ 61.4 � 12.4%, t16 ¼ 3.7, P ¼ 0.002; test 2: 67.3 � 12.7%,t16 ¼ 5.3, P < 0.001; Figs 1, 2a). Both songs seemed equally welllearned by all females, as there was no difference in preferencestrength for familiar tutor songs between tests 1 and 2 (no effects oftest number, males’ origin, females’ brood sizes or any of the two-way interactions between these three factors; see preference intests 1 and 2 in Table 3).

In test 3, females were given a choice between the two familiartape-tutor songs (always one from their natal colony in Leiden andone from the non-natal colony in Utrecht). Although females hadpreferred both tutor songs over unfamiliar songs equally strongly intests 1 and 2 (see above), in test 3 when choosing between their

100Tape-tutoredLive-tutored

Large broodSmall brood

(a)

(b)

50

Son

g p

refe

ren

ce s

tren

gth

Tota

l n

um

ber

of k

eyp

ecks

0

1100

550

0

T1Tape

tutor1

T2Tape

tutor2

T3Non-natal

tutor

T4Fosterfather

T4Fosterfather

T1Tape

tutor1

T2Tape

tutor2

T3Non-natal

tutor

T4Fosterfather

T4Fosterfather

Figure 2. Females’ (a) song preference strength (mean % � 1 SD) and (b) key-peckingactivity (mean � 1 SD) for the two tape tutors (T1 and T2), for non-natal colony Utrechtversus natal colony Leiden tutors (T3) and for their foster father (T4) in relation tofemale tutoring procedure (tape- or live-tutored) and/or foster brood size (large orsmall). The dashed line indicates random preference. T1 ¼ first preference test for tutorover unfamiliar song (both from the natal colony in Leiden or both from non-natalcolony in Utrecht), T2 ¼ second preference test for tutor over unfamiliar song,T3 ¼ third preference test for Utrecht tutor song versus Leiden tutor song, T4 ¼ fourthpreference test for foster father over unfamiliar song.

natal and non-natal tutor songs, they now showed an overallpreference for the non-natal over natal songs (one-sample Studentt test: mean preference strength� 1 SD for Utrecht versus Leidentutor songs ¼ 64.4 � 20.8%, t16 ¼ 2.8, P ¼ 0.01; Fig. 2a). This pref-erence for the non-natal songs was stronger in females from largebroods (large: 70.7 � 15.7%, N ¼ 10; small: 55.4 � 24.9%, N ¼ 7; seepreference in test 3 in Table 3), but was not affected by the origin ofthe males whose songs were first heard in test 1 or by the inter-action between females’ brood size and males’ origin in test 1 (seepreference in test 3 in Table 3).

Preference for Foster Father Songs

In the fourth test we tested whether females also had a pref-erence for (and thus had also retained the memory of) their fosterfather’s song they had heard only up until 35 days. Females fromboth live- and tape-tutoring procedures showed an overall prefer-ence (mean � 1 SD ¼ 59.0 � 18.5% and 59.1 �14.3%, N ¼ 24 and 17,respectively; Fig. 2a) for the familiar foster father songs over un-familiar songs (always the song of another female’s foster father)which was significantly different from chance (59.0 � 16.7% of allkeypecks; Wilcoxon signed-ranks test: V ¼ 38, N ¼ 40, P < 0.001).Neither preference strength nor activity level depended on tutoringprocedure, brood size or their interaction (Table 4). However, olderfemales had a stronger preference for their foster father’s song(Table 4) and pecked less than younger females (Table 4, Fig. 3).

Tape-tutored Females’ Preference and Activity Levels

When combining the three tests involving a familiar versus anunfamiliar song (i.e. tests 1, 2 and 4) in a single analysis, we foundthat each female’s preference strength for the familiar songs did notchange over the course of the three tests (see preference in tests 1, 2and 4 in Table 3) and, overall, females’ preference strength did notdepend on the test number (see preference in tests 1, 2 and 4 inTable 3) or on males’ origin (grand mean preference strength � 1SD for natal colony Leiden songs ¼ 62.5 � 11.1, non-natal colonyUtrecht ¼ 62.6 � 10.6, both N ¼ 17; Table 3), suggesting femaleshad learned equally well from their three song models (their fosterfather and two tape tutors from either origin; Fig. 2a). Females’preference strength for the familiar songs in tests 1, 2 and 4 wasalso independent of females’ brood sizes (grand mean preferencestrength � 1 SD for females from small broods ¼ 63.2 � 8.5, largebroods ¼ 62.1 � 9.4, N ¼ 7 and 10, respectively; see preference intests 1, 2 and 4 in Table 3) and of any of the two-way interactionsbetween test number, females’ brood size and males’ origin(Table 3).

Over all four tests, preference strength (see preference in tests1e4 in Table 3) did not vary with the origin of the males whosesongs were first heard in test 1, the test number and their inter-action, showing an absence of a priming effect on preferencethroughout testing. There was also no effect of females’ brood sizethroughout testing or of the interactions between brood size andorigin of songs heard in the first test or between brood size and testnumber (see preference in tests 1e4 in Table 3). Pecking activitywas lower in females from large broods than in females from smallones (grand mean � 1 SD large broods ¼ 290.6 � 219.8, smallbroods ¼ 444.4 � 135.4, N ¼ 10 and 7, respectively; Table 5, Fig. 2b)and the total number of keypecks decreased throughout testing(Table 5, Fig. 2b). This effect was due to the lower pecking activity intest 4 than in the three preceding tests (mean � 1 SD for test 1:350.2 � 263.6; test 2: 435.4 � 341.7; test 3: 452.4 � 409.2; test 4:177.7 � 211.8; all N ¼ 17; Fig. 2b; see post hoc tests in AppendixTable A1). Note that variation in activity level was controlled for inall preference analyses (with ‘cbind’, see above).

Table 3GLMM analyses of tape-tutored females’ song preference strength in relation to test number, foster brood size, stimulus song origin and origin of first-heard songs (¼stimulusorigin T1)

Model Factor Model terms Estimate�1SE LRT c21 P % Variance

explained

Preference in T1 and T2wTest no.*Brood sizeþTest no.*Stimulusorigin in T1 and T2þBrood size*Stimulus origin in T1 and T2*

Fixed Test no. 0.276�0.181 2.2 0.14Stimulus origin �0.243�0.172 1.9 0.17Brood size 0.058�0.215 0.1 0.79Test no.*Brood size �0.475�0.321 2.1 0.15Test no.*Stimuli origin �0.329�0.426 0.6 0.44Brood size*Stimuli origin 0.020�0.326 0.004 0.95

Random Test no.jFemale ID 0.5 0.47 15.4

Preference in T3wBrood size*Stimulus origin in T1y

Fixed Brood size �0.773�0.343 4.1 0.04Stimulus origin in T1 0.856�0.463 3.0 0.09Brood size*Stimulus origin in T1 0.437�0.671 0.4 0.53

Preference in T1, T2 and T4wTest no.*Brood sizeþTest no.*Stimulusorigin in T1, T2 and T4þBrood size*Stimuli origin in T1, T2 and T4*

Fixed Test no. �0.061�0.061 1.0 0.32Stimuli origin �0.104�0.173 0.4 0.55Test no.*Stimulus origin 0.398�0.281 1.8 0.18Brood size 0.035�0.205 0.03 0.87Test no.*Brood size �0.079�0.116 0.5 0.50Brood size*Stimuli origin �0.113�0.330 0.1 0.73

Random Test no.jFemale ID 0.2 0.62 1.7

Preference in T1, T2, T3 andT4wTest no.*Brood sizeþTest no.*Stimulus origin in T1þBrood size*Stimulus origin in T1*

Fixed Stimulus origin in T1 0.318�0.224 1.9 0.17Brood size �0.186�0.219 0.7 0.40Test no. �0.041�0.069 0.3 0.56Test no.*Brood size �0.164�0.137 1.4 0.24Test no.*Stimulus origin in T1 0.118�0.137 0.7 0.39Brood size*Stimulus origin in T1 �0.015�0.443 0.001 0.97

Random Test no.jFemale ID 0.2 0.67 1.4

Rejected terms are in italics, significant one in bold. Test no.: test number; T1 ¼ first preference test for tutor over unfamiliar song (both from Leiden or both from Utrecht);T2 ¼ second preference test for tutor over unfamiliar song; T3 ¼ third preference test for Utrecht tutor song versus Leiden tutor song; T4 ¼ fourth preference test for fosterfather over unfamiliar song; LRT: log-likelihood ratio test.

* GLMM with binomial distribution and random factors female ID nested in year, hatching nest, foster brood, tutor group ID and interaction between female ID and testnumber.

y GLMM with binomial distribution and random factors year, hatching nest, foster brood and tutor group ID.

10

ber

Tutoring procedureLiveTape

M.-J. Holveck, K. Riebel / Animal Behaviour 88 (2014) 125e135 131

DISCUSSION

In this experiment we set out to test whether young zebra finchfemales exposed to more than one song and from more than onetutor would prefer only one or several of these songs in adulthoodandwhether the ‘when’ and ‘fromwhom’ they learnedwas affectedby juvenile condition. The results show that female zebra finchescan learn to prefer several songs when young and that they canlearn from a succession of tutors as well as from two tutors heardsimultaneously. More specifically, as adults, females preferred (1)the song of their foster father they had heard only until 35 days ofage and (2) the two tutor songs they had heard from loudspeakers

Table 4Females’ preference and activity level in test 4 in relation to females’ tutoring pro-cedure (live or tape), foster brood size and age at testing (in days)

Responsevariable

Model terms Estimate�1SE LRT c21 P

Preferencestrength

Tutoring procedure �0.420�0.298 1.9 0.16Age at testing 0.009�0.004 4.3 0.04Brood size �0.223�0.204 1.2 0.28Tutoring procedure*Brood size

0.345�0.409 0.7 0.41

Total numberof keypecks

Age at testing �0.014�0.004 11.4 <0.001Tutoring procedure �0.448�0.299 2.1 0.15Brood size 0.330�0.265 1.5 0.22Tutoring procedure*Brood size

0.064�0.474 0.02 0.89

Rejected terms are in italics, significant ones in bold. T4 ¼ fourth preference test forfoster father over unfamiliar song; LRT: log-likelihood ratio test. Model ¼ Y inT4w Tutoring procedure*Brood size þ Age at testing. GLMM with binomial distri-bution for preference and Poisson distribution for activity level, and random factorsyear, hatching nest, foster brood, and tutor group ID.

from 35 to 65 days of age but not thereafter. Despite the brood sizemanipulation affecting phenotypic quality, behaviour and physi-ology (Holveck et al., 2008; Holveck & Riebel, 2010; Holveck et al.,2011; Riebel et al., 2012; Verhulst et al., 2006), we found an effect of

5

0150 250

Age at testing (days posthatching)

Log-

tran

sfor

med

tot

al n

um

of k

eyp

ecks

in

tes

t 4

350

Figure 3. Tape- and live-tutored females’ log-transformed total number of keypecks intest 4 as a function of their age at testing. Each symbol is an individual female value.Linear regression line (plain) is shown.

Table 5Tape-tutored females’ total number of keypecks per test in all preference tests (tests1, 2, 3, 4) in relation to test number, foster brood size and origin of first-heard songs(¼stimulus origin T1)

Factors Model terms Estimate�1SE LRT c21 P % Variance

explained

Fixed Test no. �0.248�0.085 7.9 0.005Brood size 0.572�0.249 3.8 0.05Stimuli origin in T1 0.048�0.276 0.0 0.99Test no.*Stimulusorigin in T1

�0.104�0.172 0.4 0.55

Brood size*Stimulusorigin in T1

�0.213�0.497 0.2 0.70

Test no.*Brood size �0.073�0.172 0.2 0.67Random Test no.jFemale ID 0.0 0.99 0.0

Rejected terms are in italics, significant ones in bold. Test no.: test number; T1 ¼ firstpreference test for tutor over unfamiliar song (both from Leiden or both fromUtrecht); T2 ¼ second preference test for tutor over unfamiliar song; T3 ¼ thirdpreference test for Utrecht tutor song versus Leiden tutor song; T4 ¼ fourth pref-erence test for foster father over unfamiliar song; LRT: log-likelihood ratio test.Model ¼ Keypecks in T1, T2, T3 and T4w Test number*Brood size þ Test num-ber*Stimuli origin in T1 þ Brood size*Stimulus origin in T1. GLMM with Poissondistribution and random factors female ID nested in year, hatching nest, fosterbrood, tutor group ID and interaction between female ID and test number.

M.-J. Holveck, K. Riebel / Animal Behaviour 88 (2014) 125e135132

the early rearing condition only on key-pecking activity (femalesfrom small nests pecked more frequently for the song reward), butnot on the direction or preference strength for the tutor songs (butfor effects on preference for unfamiliar songs in females from thesame cohort see Holveck & Riebel, 2010). This is in line with earlierbrood size manipulation experiments that also only found effectson preference strength in tests involving unfamiliar songs, but noeffects of the treatment on direction and strength of preference forearly learned songs (Holveck & Riebel, 2010; Riebel et al., 2009).This suggests that the song acquisition process in young females isextremely robust whereas adult selectivity for novel songs seemsmore vulnerable to developmental conditions (see also Schmidt,McCallum, MacDougall-Shackleton, & MacDougall-Shackleton,2013). The observation that females from small broods weremore active and pecked significantly more than females from largebroods is suggestive of an effect of condition on mate-samplingstrategies (Cotton, Small, & Pomiankowski, 2006; Riebel, Holveck,Verhulst, & Fawcett, 2010; Woodgate, Bennett, Leitner, Catchpole,& Buchanan, 2010). However, this effect was less pronounced intwo larger data sets involving (younger) females frommanipulatedbrood sizes (Holveck & Riebel, 2010; Riebel et al., 2009; see Ap-pendix Table A2) suggesting that this aspect needs further study.

Our study not only tested whether the songs heard from 35 to65 days posthatching (postindependence) would be preferred butalso whether females would prefer songs they had heard onlyduring the fledgling stage preceding the actual tutoring phase (i.e.before independence). Both tape- and live-tutored femalespreferred the songs of their foster father compared to unfamiliarsongs although they had heard their foster father’s song only until35 days posthatching and although both groups had had contin-uous exposure to other songs from day 35 onwards. This suggeststhat preindependence learning must be relatively robust to sub-sequent conditions that differed substantially between the live-and tape-tutored groups in our experiment. During the (putative)sensitive phase for song learning, the tape-tutored females werehoused in pairs in sound attenuation chambers in isolation fromother zebra finches and only heard the same two songs fromloudspeakers. The live-tutored groups were housed in mixed-sexpeer groups consisting of two young males and two young femaletutees, an adult male tutor and his female and were housed in opencages in the colony room. These substantial differences in theavailability of social models at this later stage seem not to have

affected the memorization and recall of the songs heard only pre-independence as evidenced by the overall preference for thesesongs over unfamiliar songs. This seems in contrast with male songproduction learning where the availability of social tutoring duringthe sensitive phase seems to overwrite early acquired model songs(Böhner, 1983; Slater et al., 1991; see below).

Our results confirm earlier work that showed that young zebrafinches could remember songs before 35 days posthatching asshown by calling assays with subadult birds at that age (Clayton,1988). Our results show that these early memories carry over intoadulthood even after a succession of tutoring events throughoutthe sensitive phase (in both the live- and tape-tutored group).While our laboratory study does not allow us to conclude at thisstage from whom birds would learn in a more complex socialsetting in the wild it shows that they are able to learn from socialcompanions (future work will have to identify whether these arecaregivers, other adults in the colony or peers) at this age and afterindependence continue to add additional song memories. Irre-spective of who the tutors might be in the wild, this laboratoryexperiment demonstrates that both the early juvenile and latersubadult phases contribute to the formation of adult song prefer-ences. The neuromodulatory effects of juvenile song preference areprobably lifelong (Hauber, Woolley, Cassey, & Theunissen, 2013;Hernandez, Phillmore, & MacDougall-Shackleton, 2008; Terpstra,Bolhuis, Riebel, van der Burg, & den Boer-Visser, 2006). Females alsolearn songs of male companions and mates when adult and preferthese over unfamiliar songs (Clayton, 1988) and this also affectsneuronal activation (Woolley & Doupe, 2008). In contrast, shorttape-tutoring episodes at this age failed to induce preferences forthe familiar over unfamiliar songs (Riebel, 2000). This suggests thatin females, as in males, reward schemes, social interactions andquantity of song interact in complex manners, an issue not eventouched on in the relatively young research into female songpreference development (Riebel, 2003b; Riebel et al., 2009). Futureworkwill have to address howmany songs and fromwhom femaleslearn if raised in complex social groups (Freed-Brown & White,2009) and whether early song preference learning is qualitativelydifferent from adult preference learning, because the mechanismsunderlying model choice and ensuing generalization processes(Clayton, 1988, 1990; Miller, 1979; Riebel, 2003b) will affect signaland preference evolution (Verzijden et al., 2012). More study intothe underlying learning mechanisms might also address possiblecauses of the unexpected difference in preference strength be-tween the (natal) Leiden and (non-natal) Utrecht songs when thesetwo were presented in the same preference test. Visual inspectionof the spectrograms showed substantial variation both within andbetween colonies (see Fig. A1 in the Appendix for examples).However, this need not exclude the possibility that subtler differ-ences such as the relative frequency of elements or theirsequencing, which differ between colonies (see van Heijningen,2012), might have affected female preferences. However, giventhe relatively small number of songs and the lack of experimentalcontrol of the variation in song structure, we can only speculatewhether the preference for the (non-natal) Utrecht songs in thethird test arose because structural differences between the twocolonies made the songs from the Utrecht (non-natal) colony moreattractive per se (because of systematic structural differences orbecause the sample accidentally contained more attractive songstimuli or because of the difference in photoperiod (see Methods)between the two colonies which can affect song parameters; e.g.Bentley, Spar, MacDougall-Shackleton, Hahn, & Ball, 2000; Dereg-naucourt, Saar, & Gahr, 2012). Alternatively, there might have beena preference for the tutor songs fromUtrecht not because theyweremore attractive in phonology but because they differed more fromthe early heard natal (foster father) tutor song than the natal Leiden

M.-J. Holveck, K. Riebel / Animal Behaviour 88 (2014) 125e135 133

test and tutor songs. In this case the preference would indeed haveresulted from the learning process rather than song structure perse. Teasing apart these different hypotheses would require addi-tional two-phase tutoring experiments using songs of differentorigin (or systematic experimental manipulations of song struc-ture). An experience-dependent preference for non-natal song infemales would be the opposite, though, from how pre-independence familiarity affects male song learning, where youngmales prefer to learn tutor songs more similar to the songs hearduntil day 35 (Clayton, 1987).

Further comparison of our result with the existing literature onsong acquisition in male zebra finches suggests that sequentialexposure to several tutors during the song acquisition phase hasdifferent consequences for female song preference learning and formale song production learning: experimental studies exposingbirds to a succession of tutors show that males normally sing a songheard between 35 and 65 days posthatching but not before, unlessthey are deprived of adult tutor song from 35 days posthatchingonwards (Böhner, 1983; Jones et al., 1996). Males that experienceseveral songs (i.e. one song up until 35 days posthatching andanother from 35 to 65 days posthatching) normally learn to sing thesong heard after day 35 (Slater et al., 1991). This has led to thesuggestion that, at least for song production learning, the experi-ences later in the sensitive phase overwrite the early song mem-ories (Slater et al., 1991). Although this is suggestive of afundamental difference in the song acquisition process betweenmales and females, one should keep in mind that males of severalspecies have been shown to memorize more songs than they sing(Geberzahn, Hultsch, & Todt, 2002;McGregor & Avery,1986; Nelson& Marler, 1994). Moreover, male zebra finches also show a prefer-ence for the song they have heard when young (Riebel et al., 2002)independently of how well they learned to sing it (Terpstra, denBoer-Visser, & Bolhuis, 2004) and as young birds learn to discrim-inate several songs (Braaten, Petzoldt, & Cybenko, 2007; Clayton,1988). No study has tested whether male zebra finches tutoredwith several songs when young would show preferences for thesesongs as adults, something that could be relatively easily done asadult males are also willing towork for song exposure (Adret, 1993;Riebel et al., 2002). Given that zebra finches have been instru-mental in the study of song development and sexual brain differ-entiation, comparative studies of the male and female songacquisition process can be highly informative with respect to howsong production might interfere with its perception (Riebel et al.,2002; Woolley, 2012).

Our study adds to the growing experimental evidence thatshows that adult females’ song preferences are greatly influencedby learning (for a review see Riebel, 2003b). Across species, thestriking variety in the timing and quantity of male song learning iswell documented, but even for males the ecological correlates ofthis variety are still poorly understood (Beecher & Brenowitz,2005). We have barely begun to map out the extent and mannerof song learning in females, which are the intended receivers ofthese signals (Riebel et al., 2005). The much rarer studies of theontogeny of female song preferences in repertoire species suggestthis process to be as diverse in its timing and codependence onsocial factors as male song acquisition (Bertin, Hausberger, Henry, &Richard-Yris, 2009; Freed-Brown & White, 2009; Nagle & Kreutzer,1997; Riebel & Slater, 1998; Searcy, Nowicki, Hughes, & Peters,2002; Yamaguchi, 1999). The striking variety in song productionversus preference learning in females offers opportunities forcomparative research (Riebel, 2003b). An improved understandingof female song learning should help to disentangle the mechanismof production versus perception learning (Riebel et al., 2002;Woolley, 2012), to develop hypotheses regarding the evolutionaryprocesses that have led to the origin and maintenance of vocal

learning in birds and other vocal learners (Beecher & Brenowitz,2005; Janik & Slater, 1997) and to model the consequences oflearned preferences on sexual selection processes (Verzijden et al.,2012).

Acknowledgments

We thank two anonymous referees and N. Sharani for theircomments on the manuscript, Sharon Gobes for providing songrecording from Utrecht colony and Henny Koolmoes for assistingwith animal care. This work was supported by the Research Councilfor Earth and Life Sciences (ALW) with financial aid from theNetherlands Organization for Scientific Research (NWO).

References

Adret, P. (1993). Operant conditioning, song learning and imprinting to taped songin the zebra finch. Animal Behaviour, 46, 149e159.

Beecher, M. D., & Brenowitz, E. A. (2005). Functional aspects of song learning insongbirds. Trends in Ecology & Evolution, 20(3), 143e149. http://dx.doi.org/10.1016/j.tree.2005.01.004.

Bentley, G. E., Spar, B. D., MacDougall-Shackleton, S. A., Hahn, T. P., & Ball, G. F.(2000). Photoperiodic regulation of the reproductive axis in male zebra finches,Taeniopygia guttata. General and Comparative Endocrinology, 117(3), 449e455.

Bertin, A., Hausberger, M., Henry, L., & Richard-Yris, M. A. (2009). Adult:young ratioinfluences song acquisition in female European starlings (Sturnus vulgaris).Journal of Comparative Psychology, 123(2), 195e203. http://dx.doi.org/10.1037/a0014050.

Böhner, J. (1983). Song learning in the zebra finch (Taeniopygia guttata): selectivity inthe choice of a tutor and accuracy of song copies. Animal Behaviour, 31, 231e237.

Böhner, J. (1990). Early acquisition of song in the zebra finch, Taeniopygia guttata.Animal Behaviour, 39, 369e374.

Boogert, N. J., Fawcett, T. W., & Lefebvre, L. (2011). Mate choice for cognitive traits: areview of the evidence in nonhuman vertebrates. Behavioral Ecology, 22(3),447e459. http://dx.doi.org/10.1093/beheco/arq173.

Braaten, R. F., Petzoldt, M., & Cybenko, A. K. (2007). Recognition memory forconspecific and heterospecific song in juvenile zebra finches, Taeniopygia gut-tata. Animal Behaviour, 73, 403e413. http://dx.doi.org/10.1016/j.anbehav.2006.08.009.

Buchanan, K. L., Grindstaff, J. L., & Pravosudov, V. V. (2013). Condition dependence,developmental plasticity, and cognition: implications for ecology and evolution.Trends in Ecology & Evolution, 28(5), 290e296. http://dx.doi.org/10.1016/j.tree.2013.02.004.

Catchpole, C. K., & Slater, P. J. B. (2008). Bird song: Biological themes and variations(2nd ed.). Cambridge: Cambridge University Press.

Clayton,N. S. (1987). Song tutor choice in zebrafinches.Animal Behaviour, 35, 714e721.Clayton, N. S. (1988). Song discrimination learning in zebra finches. Animal

Behaviour, 36, 1016e1024.Clayton, N. S. (1990). Subspecies recognition and song learning in zebra finches.

Animal Behaviour, 40, 1009e1017.Cotton, S., Small, J., & Pomiankowski, A. (2006). Sexual selection and condition-

dependent mate preferences. Current Biology, 16(17), R755eR765.Deregnaucourt, S., Saar, S., & Gahr, M. (2012). Melatonin affects the temporal

pattern of vocal signatures in birds. Journal of Pineal Research, 53(3), 245e258.http://dx.doi.org/10.1111/j.1600-079X.2012.00993.x.

Elston, D. A., Moss, R., Boulinier, T., Arrowsmith, C., & Lambin, X. (2001). Analysis ofaggregation, a worked example: numbers of ticks on red grouse chicks. Para-sitology, 122, 563e569. http://dx.doi.org/10.1017/s0031182001007740.

Freed-Brown, G., & White, D. J. (2009). Acoustic mate copying: female cowbirdsattend to other females’ vocalizations to modify their song preferences. Pro-ceedings of the Royal Society B, 276(1671), 3319e3325. http://dx.doi.org/10.1098/rspb.2009.0580.

Geberzahn, N., Hultsch, H., & Todt, D. (2002). Latent song type memories areaccessible through auditory stimulation in a hand-reared songbird. AnimalBehaviour, 64, 783e790. http://dx.doi.org/10.1006/anbe.2002.3099.

Hauber, M. E., Woolley, S. M. N., Cassey, P., & Theunissen, F. E. (2013). Experiencedependence of neural responses to different classes of male songs in the pri-mary auditory forebrain of female songbirds. Behavioural Brain Research, 243,184e190. http://dx.doi.org/10.1016/j.bbr.2013.01.007.

van Heijningen, C. A. A. (2012). On the production and perception of syntacticalregularities in zebra finches: Experimenting with ABBA, ACDC and others. Un-published doctoral dissertation. Department of Behavioural Biology, Institute ofBiology (IBL), Faculty of Science, Leiden University.

Hernandez, A. M., Phillmore, L. S., & MacDougall-Shackleton, S. A. (2008). Effects oflearning on song preferences and Zenk expression in female songbirds.Behavioural Processes, 77(2), 278e284.

Holveck, M. J., de Castro, A. C. V., Lachlan, R. F., ten Cate, C., & Riebel, K. (2008).Accuracy of song syntax learning and singing consistency signal early conditionin zebra finches. Behavioral Ecology, 19(6), 1267e1281. http://dx.doi.org/10.1093/beheco/arn078.

M.-J. Holveck, K. Riebel / Animal Behaviour 88 (2014) 125e135134

Holveck, M. J., Geberzahn, N., & Riebel, K. (2011). An experimental test of condition-dependent male and female mate choice in zebra finches. PLoS One, 6(8),e23974. http://dx.doi.org/10.1371/journal.pone.0023974.

Holveck, M. J., & Riebel, K. (2007). Preferred songs predict preferred males: consis-tency and repeatability of zebra finch females across three test contexts. AnimalBehaviour, 74(2), 297e309. http://dx.doi.org/10.1016/j.anbehav.2006.08.016.

Holveck, M. J., & Riebel, K. (2010). Low-quality females prefer low-quality maleswhen choosing a mate. Proceedings of the Royal Society B, 277(1678), 153e160.http://dx.doi.org/10.1098/rspb.2009.1222.

Houx, B. B., & ten Cate, C. (1999). Song learning from playback in zebra finches: isthere an effect of operant contingency? Animal Behaviour, 57, 837e845.

Janik, V. M., & Slater, P. J. B. (1997). Vocal learning in mammals. Advances in the Studyof Behavior, 26, 59e99.

Jesse, F., & Riebel, K. (2012). Social facilitation of male song by male and femaleconspecifics in the zebra finch, Taeniopygia guttata. Behavioural Processes, 91(3),262e266. http://dx.doi.org/10.1016/j.beproc.2012.09.006.

Jones, A. E., ten Cate, C., & Slater, P. J. B. (1996). Early experience and plasticity ofsong in adult male zebra finches (Taeniopygia guttata). Journal of ComparativePsychology, 110(4), 354e369.

Kroodsma, D. E., Vielliard, J. M. E., & Stiles, F. G. (1996). Study of bird sounds in theneo-tropics: urgency and opportunity. In D. E. Kroodsma, & E. H. Miller (Eds.),Ecology and evolution of acoustic communication in birds (pp. 268e281). Ithaca,NY: Comstock Publishing Associates.

Lachlan, R. F., & Nowicki, S. (2012). How reliable is song learning accuracy as a signalof male early condition? American Naturalist, 180(6), 751e761. http://dx.doi.org/10.1086/668010.

McGregor, P. K., & Avery, M. I. (1986). The unsung songs of great tits (Parus major):learning neighbour’s songs for discrimination. Behavioral Ecology and Socio-biology, 18, 311e316.

Miller, D. B. (1979). Long-term recognition of father’s song by female zebra finches.Nature, 280, 389e391.

Nagle, L., & Kreutzer, M. L. (1997). Song tutoring influences female song preferencesin domesticated canaries. Behaviour, 134, 89e104.

Nelson, D. A., & Marler, P. (1994). Selection-based learning in bird song develop-ment. Proceedings of the National Academy of Sciences of the United States ofAmerica, 91, 10498e10501.

R Development Core Team. (2013). R: A language and environment for statisticalcomputing. Vienna, Austria: R Foundation for Statistical Computing.

Riebel, K. (2000). Early exposure leads to repeatable preferences for male song infemale zebra finches. Proceedings of the Royal Society B, 267(1461), 2553e2558.

Riebel, K. (2003a). Developmental influences on auditory perception in femalezebra finches e is there a sensitive phase for song preference learning? AnimalBiology, 53(2), 73e87. http://dx.doi.org/10.1163/157075603769700304.

Riebel, K. (2003b). The ‘mute’ sex revisited: vocal production and perceptionlearning in female songbirds. Advances in the Study of Behavior, 33, 49e86.http://dx.doi.org/10.1016/s0065-3454(03)33002-5.

Riebel, K. (2009). Song and female mate choice in zebra finches: a review. Advancesin the Study of Behavior, 40, 197e238. http://dx.doi.org/10.1016/S0065-3454(09)40006-8.

Riebel, K. (2011). Comment on Boogert et al.: mate choice for cognitive traits orcognitive traits for mate choice? Behavioral Ecology, 460e461. http://dx.doi.org/10.1093/beheco/arr003.

Riebel, K., Hall, M. L., & Langmore, N. E. (2005). Female songbirds still struggling tobe heard. Trends in Ecology & Evolution, 20(8), 419e420.

Riebel, K., Holveck, M.-J., Verhulst, S., & Fawcett, T. W. (2010). Are high-qualitymates always attractive?: state-dependent mate preferences in birds andhumans. Communicative & Integrative Biology, 3(3), 271e273.

Riebel, K., Naguib, M., & Gil, D. (2009). Experimental manipulation of the rearingenvironment influences adult female zebra finch song preferences. AnimalBehaviour, 78(6), 1397e1404. http://dx.doi.org/10.1016/j.anbehav.2009.09.011.

Riebel, K., & Slater, P. J. B. (1998). Testing female chaffinch song preferences byoperant conditioning. Animal Behaviour, 56, 1443e1453.

Riebel, K., & Smallegange, I. (2003). Does zebra finch (Taeniopygia guttata) prefer-ence for the (familiar) father’s song generalize to the songs of unfamiliarbrothers? Journal of Comparative Psychology, 117(1), 61e66. http://dx.doi.org/10.1037/0735-7036.117.1.61.

Riebel, K., Smallegange, I. M., Terpstra, N. J., & Bolhuis, J. J. (2002). Sexualequality in zebra finch song preference: evidence for a dissociation betweensong recognition and production learning. Proceedings of the Royal Society B,269, 729e733.

Riebel, K., Spierings, M. J., Holveck, M. J., & Verhulst, S. (2012). Phenotypic plasticityof avian social-learning strategies. Animal Behaviour, 84(6), 1533e1539. http://dx.doi.org/10.1016/j.anbehav.2012.09.029.

Ritchie, G. R. S., Kirby, S., & Hawkey, D. J. C. (2008). Song learning as an indicatormechanism: modelling the developmental stress hypothesis. Journal of Theo-retical Biology, 251(4), 570e583. http://dx.doi.org/10.1016/j.jtbi.2007.12.013.

Schmidt, K. L., McCallum, E. S., MacDougall-Shackleton, E. A., & MacDougall-Shackleton, S. A. (2013). Early-life stress affects the behavioural and neuralresponse of female song sparrows to conspecific song. Animal Behaviour, 85(4),825e837. http://dx.doi.org/10.1016/j.anbehav.2013.01.029.

Searcy, W. A., Nowicki, S., Hughes, M., & Peters, S. (2002). Geographic songdiscrimination in relation to dispersal distances in song sparrows. AmericanNaturalist, 159(3), 221e230.

Slater, P. J. B., Eales, L. A., & Clayton, N. S. (1988). Song learning in zebra finches:progress and prospects. Advances in the Study of Behavior, 18, 1e34.

Slater, P. J. B., Richards, C., & Mann, N. I. (1991). Song learning in zebrafinches exposed to a series of tutors during the sensitive phase. Ethology,88, 163e171.

Tchernichovski, O., Lints, T., Mitra, P. P., & Nottebohm, F. (1999). Vocal imitationin zebra finches is inversely related to model abundance. Proceedings of theNational Academy of Sciences of the United States of America, 96(22), 12901e12904.

Terpstra, N. J., den Boer-Visser, A., & Bolhuis, J. J. (2004). An analysis of the neuralrepresentation of birdsong memory. Journal of Neuroscience, 24(2), 4971e4977.

Terpstra, N. J., Bolhuis, J. J., Riebel, K., van der Burg, J. M. M., & den Boer-Visser, A. M.(2006). Localized brain activation specific to auditory memory in a femalesongbird. Journal of Comparative Neurology, 494(5), 784e791.

Verhulst, S., Holveck, M.-J., & Riebel, K. (2006). Long-term effects of manipulatednatal brood size on metabolic rate in zebra finches. Biology Letters, 2(3), 478e480. http://dx.doi.org/10.1098/rsbl.2006.0496.

Verzijden, M. N., ten Cate, C., Servedio, M. R., Kozak, G. M., Boughman, J. W., &Svensson, E. I. (2012). The impact of learning on sexual selection and speciation.Trends in Ecology & Evolution, 27(9), 511e519. http://dx.doi.org/10.1016/j.tree.2012.05.007.

Woodgate, J. L., Bennett, A. T. D., Leitner, S., Catchpole, C. K., & Buchanan, K. L. (2010).Developmental stress and female mate choice behaviour in the zebra finch.Animal Behaviour, 79(6), 1381e1390. http://dx.doi.org/10.1016/j.anbehav.2010.03.018.

Woolley, S. C., & Doupe, A. J. (2008). Social contexteinduced song variation affectsfemale behavior and gene expression. PLoS Biology, 6(3), 525e537. http://dx.doi.org/10.1371/journal.pbio.0060062.

Woolley, S. M. N. (2012). Early experience shapes vocal neural coding andperception in songbirds. Developmental Psychobiology, 54(6), 612e631. http://dx.doi.org/10.1002/dev.21014.

Yamaguchi, A. (1999). Auditory experience does not shape sexual preferences forsongs in female northern cardinals. Behaviour, 136, 309e329.

Zann, R. (1993). Variation in song structure within and among populations ofAustralian zebra finches. Auk, 110, 716e726.

Zann, R. A. (1996). The zebra finch: A synthesis of field and laboratory studies. Oxford:Oxford University Press.

M.-J. Holveck, K. Riebel / Animal Behaviour 88 (2014) 125e135 135

APPENDIX

10 kHz

(a)

(b)

(c)

(d)

(

(

(

(

Figure A1. Spectrograms and amplitude waveforms of four of the 10 stimulus sets used forpraat.org) software v. 5.3.32 (fast Fourier transformations with 1000 time and 250 frequencthe non-natal colony (at Utrecht University) and (eeh) songs from the natal colony (at Leid

Table A1Comparison of tape-tutored females’ total number of keypecks across tests 1e4

Comparison Estimate�1SE

T1 versus T2 0.201�0.243T1 versus T3 0.116�0.243T1 versus T4 �0.801�0.244T2 versus T3 �0.085�0.243T2 versus T4 �1.002�0.243T3 versus T4 �0.917�0.243

The post hoc test tests the null hypothesis that testn � testn�i ¼ 0. Significant terms are inboth from Utrecht); T2 ¼ second preference test for tutor over unfamiliar song; T3 ¼ tpreference test for foster father over unfamiliar song. Model ¼ Keypecks in tests 1e4w Teyear, hatching nest, foster brood and tutor group ID. Test category effect: c2

3 ¼ 18, P < 0

Table A2Between-study comparison of key-pecking activity for females from different brood size

Studies using operant songpreference testing

Testing age (dph) Operant tests Brood

Riebel et al. (2009) 145�18 2�4-day tests LargeInterSmal

Holveck et al. (2010) 164�15 4�2-day tests LargeSmal

This study 207�39 4�2-day tests LargeSmal

Grand means � 1 SD of keypeck means per female throughout testing are shown. dph: dfive to six chicks. Intermediate broods ¼ four chicks.

1 s

e)

f)

g)

h)

tape tutoring (one set per row). Spectrograms were calculated with the Praat (www.y steps, 0.005 window size, dynamic range 65 db, Hanning window). (aed) Songs fromen University).

Tukey post hoc test P

0.8 0.840.5 0.96

�3.3 0.005�0.4 0.99�4.1 <0.001�3.8 <0.001

bold. T1 ¼ first preference test for tutor over unfamiliar song (both from Leiden orhird preference test for Utrecht tutor song versus Leiden tutor song; T4 ¼ fourthst category. GLMMwith Poisson distribution and random factors female ID nested in.001. Post hoc test under the multcomp package in R.

s

size Grand mean�1 SD N % Pecking difference betweensmall and large broods

536.3�326.4 14 16mediate 584.3�297.8 19l 637.0�392.1 20

461.4�271.2 12 16l 551.1�369.1 12

290.6�219.8 10 35l 444.4�135.4 7

ays posthatching. In all studies small broods ¼ two to three chicks and large broods