Brain, Behavior and EvolutionEditor-in-Chief: R. Glenn Northcutt. La Jolla. Calif.

ReprintPublisher: S.Karger AG, BaselPrinted in Switzerland

Daniel Margoliash"Eric S. Fortune'Mitchell L. SutterhAlbert C. Yu'B. DavidWren-Hardin''A . Dave '

' Department of Organismal Biology andAnatomy, The University of Chicago,Chicago, Ill.;

b Center for Neuroscience,University of Califomia, Davis, Calif.;

" Committee on Neurobiology, TheUniversity of Chicago, Chicago, Ill., USA

Brain Behav Evol 1994:44:247264

Distributed Representation in theSong System of Oscines:Evolutionary lmplications andFunctional Consequences

Key WordsHierarchicalParallelDistributedRedundantSensorimotor integrationForebrain evolutionBird song system

AbstractThis paper reviews the organizational principles and implications that haveemerged from the analysis of HVc, a forebrain nucleus that is a major site of sen-sory, motor, and sensorimotor integration in the song control system of oscinepasserine birds (songbirds). Anatomical, physiological, and behavioral data sup-port the conclusion that HVc exists within a hierarchically organized systemwith parallel pathways that converge onto HVc. The organization of HVc is dis-tributed and redundant, and its outputs exhibit broad divergence. A similar pat-tern of connectivity exists for neostriatum adjacent to HVc. This and other datasupport the hypothesis that the song system arose from an elaboration or dupli-cation of pathways generally present in all birds. Spontaneous and auditoryresponse activity is strongly correlated throughout HVc, with auditory responsesexhibiting strong temporal modulation in a synchronized fashion throughout thenucleus. This suggests that the auditory representation of song is encoded in thesynchronized temporal pattems of activation, and that the predominant selectiv-ity for the individual's own song that is observed for HVc neurons results frominteractions of auditory input with central pattem generators for song. Most, orall HVc neurons are recruited during singing. The auditory response and motorrecruitment propefiies of individual HVc neurons have no simple relationship,and the spontaneous activity in HVc may build up in the seconds preceding asong. To the extent HVc participates in perceptual phenomena associated withsong, production and perception are not tightly linked in adults but may belinked by shared developmental processes during periods of sensorimotor leam-itrg.

Dr. Daniel MargoliashDepartment of Organismal Biology md AnatomyThe University of Chicago1025 E. 57rh St.Chicago, IL 60627 (USA)

O 1994 S. KugerAG, Basel

Abbreviations

A ArchistriatumAm Nucleus ambiguusAS 'Autogenous'song

cDLP Nucleus dorsolateralis posterior thalami pars caudalisCPGs Central pattern generatorsDIP NucleusdorsointermediusposteriorthalamiDLM Nucleus dorsolateralis anterior thalami, pars medialisDLP Nucleus dorsolateralis posterior thalamiDM Dorsomedial subdivision of nucleus intercollicularisDMP Nucleus dorsomedial is posterior thalamiHV Hyperstriatum ventraleIMAN Lateral magnocellular nucleus of the anterior neostriatummMAN Medial magnocellular nucleus of the anterior neostriatumN NeostriatumNC Neostriatum caudaleNCM Caudomedial neostriatumNd Dorsal caudal neostriatumNI Neostriatum intermediumNIDL NeostriatumintermediumparsdorsolateralisNIf Nucleus interfacialisNIVL NeostriatumintermediumparsventrolateralisnXIIts Nucleusnervihypoglossi,tracheosyringealcomponentOM Tractus occipitomesencephalicusOv Nucleus ovoidalisRA Nucleus robustus archistriatalisRAm Nucleus retroambigualisSMS Syllable of maximum synchronizationTCS TemporalcombinationsensitiveTMS Time of maximal synchronizationUva Nucleus uvaeformtsWTS White+hroated sparrowsX Area X

lntroduetion

The richness of behavioral phenomena associated withsong leaming, production, and perception in oscine passer-ine birds [see Kroodsma and Miller, 1982] is supported bya richness of neurobiological complexity of the song con-trol system. Although the analysis of song system circuitryat the cellular level is incomplete, for some nuclei there isnow sufficient information to besin to relate neuronal struc-

1 TheHVc is anucleus originally identified as ventral caudal hyper-striatum [Nottebohm et al.,1976] but is now thought to be in the dor-sal caudal neostriatum [Nottebohm, 1987]. The suggested revision'High (or Higher) Vocal Center' (HVC) retains the acronym for con-sistency with the substantial quantity ofprevious literature but is inad-equate because it has funcitonal implications, and because it does notdescribe the position, shape, or size of the nucleus or its cellular con-stituents. We propose adopting the acronym as the proper name. Thisis not inconsistent with old terminology (e.g. field L) nor with somenew termhology (e.9. area X).

ture with neuronal function. In particular, the nucleus HVclis a major site of integration in the song system where sen-sorimotor transformations related to syllable and songstructure probably occur. This paper will review recent datademonstrating a distributed, time-domain representation forsong in HVc, including anatomical observations, physio-logical recordings in anesthetized preparations, and physio-logical recordings and lesion studies in singing birds. Theanalysis of HVc's anatomy and physiology has yieldedinsights into the organization of episodic behaviors, audi-tory temporal pattem processing, sensorimotor integration,and the relationship between production and perception ofsong. Additionally, we reinterpret the forebrain circuitry inand around HVc from an evolutionary perspective, con-cluding that the caudal forebrain components of the songsystem are an elaboration of a system which was probablypresent in ancestral forms.

HVc within the Context of the Song Control SystemThe motor system for song production comprises dis-

crete forebrain, midbrain, and brainstem nuclei that sub-serve a limited set of discrete episodic behaviors (songs andcalls) (fig. 1). The HVc is a major site of integration withinthe song system and participates in the two major songsystem forebrain pathways. One forebrain pathway in-cludes nucleus interface (NIf), HVc, and the robust nucleusof the archistriatum (RA) in the caudal telencephalon andthe caudal subdivision of the posterior thalamic nucleus(cDLP) in the thalamus - previously called nucleus uvae-formis (Uva); see below and Wild tl993al - and ultimatelyprojects to the brainstem hypoglossal nucleus, which inner-vates the syrinx, the vocal organ in birds. Lesion andchronic recording experiments demonstrate that these nu-clei are involved in moment-to-moment control of songproduction [Nottebohm et al., 1976, 1982; McCasland,1987; Williams and Vicario, 19931. A second, more rostralpathway involves projections between HVc, area X (X), thelateral magnocellular nucleus of the anterior neostriatum(MAN), and RA in the telencephalon and the medial divi-sion of the dorsolateral thalamic nucleus (DLM) in the tha-lamus. Lesion studies demonstrate X. IMAN. and DLM arenot necessary for song maintenance in adult birds, but theyimplicate these nuclei in juvenile song development [Bott-jer et al., 1984; Scharff and Nottebohm, 1991; Sohrabji eta1., 19901. Lesions of the rostral pathway may also result insong degradation in open-ended leamers, such as canaries(Serinus canarius), during seasonal periods of adult songdevelopment [Nottebohm et al., 1990], or in other speciesduring experimentally induced periods of adult song plas-ticity [Monison and Nottebohm, 1993].

Margoliash/Fotune/Sutter/YuAVren-Hardin/Dave

Organization of Song System:

Functions and Origins

248

RAm Am nXIIts

Telencephalon

Dorsal thalanursMidbrainBrainstem

IIII

I

Fig. 1. Schematic of the song system.Arows indicate the direction of major pro-jections. Subdivisions of nuclei (esp. Ov,Field L, and RA) are not represented. Theprojection of RA onto DLM lwild, 1993b] isdifiicult to confirm with retrograde tracers.Note HVc is the major site of convergence,and RA is the major output structure [see alsowi ld .1994t .

The unidirectional, and mostly unilateral connectivity ofsong system nuclei supports a hierarchical organizationscheme, within which HVc has a special position. HVc isthe major site of convergence in the telencephalon, and itgives rise to two efferent routes to RA, the major telence-phalic output structure (fig 1). One efferent pathway is thedirect projection of HVc caudally and ventrally onto RA

[Nottebohm et a1.,I976]; the second pathway is the indirectprojection to RA via the rostral pathway, HVc-+areaX+DLM -+ IMAN-+RA [Nottebohm et al., 191 6, 1982;Okuhata and Saito, 1987; Bottjer et al., 19891. The indirectpathway has also been called 'recursive'

lWilliams, 1989;Nottebohm, 19911, however multiple but separate routesfrom the same source (HVc) to the same target (RA) doesnot constitute a feedback system. More recently Wild

U993bl presented preliminary evidence that RA may pro-ject to DLM, which would provide an anatomical substratefor possible feedback pathways within the forebrain, if notthe highly specific form of feedback implied by recursion.The reinterpretation of nucleus Uva as homologous to thecaudal division of DLP in non-oscines highlights a sourceof multimodal input to the song system, potentially includ-ing proprioceptive feedback fOkuhata and Nottebohm,19921. Multimodal activity in cDLP, which projects toHVc, may also help to explain visual flash evoked re-sponses recorded in HVc [Bischof and Engelage, 1985].

Physiological and behavioral evidence also suggest thatHVc acts as a major integrative site for sensory, motor, andsensorimotor processing in the song system. The auditoryresponse properties of HVc develop during the sensorimo-tor phase of song leaming fVolman, 1993), and it has beenhypothesized that HVc neurons are actively involvedduring development in shaping the motor program for songvia auditory feedback fMargoliash, l99ll.In adults, audi-tory neurons in HVc are selective for the individual bird'sown song, exhibiting stronger responses to playback of

own song than to conspeci{ic songs [see Margoliash, 1987].Such response selectivity implies hierarchical organizationand a high degree ofconvergence, organizational principlesgenerally observed in auditory systems [e.g. Konishi, l99l;

Carr,19921. Furthermore, HVc is a nexus for auditory inputto the song system: lesions of HVc abolish auditory songselectivity otherwise seen in both of HVc's efferent targets,area X and RA fDoupe and Konishi, l99l:' Vicario andYohay, 19931. The concept of a hierarchical organization in

the song system is also supported by motor properties ofHVc. In zebra finches (Taeniopygia guttata), HVc stimula-

tion during singing causes a re-setting of the pattem of songsyllables, whereas stimulation in RA affects only the acous-tic structure of the syllable being produced [Vu et al.,19931. On average, there is a sequential recruitment ofnuclei at the onset of singing, with earliest recruitment of

the NIf neurons, followed in order by HVc, RA, and thetracheosyringeal subdivision of the hypoglossal nucleus(nXIIts) lMcCasland, 1987]. This temporal order followsthe caudal pathway's connectivity.

Multiple Pathways through HVc:Gonvergence and Divergence

We were interested in characterizing the anatomical or-ganization of auditory input to HVc because it is an essen-

tial component for an understanding of the physiologicalproperties of HVc. Neither field L, which was the presump-

tive source of auditory input to HVc, nor HVc itself, had

been defined in cytoarchitectonic terms. Technically, suchdefinitions are needed to determine the extent of tracerinjections and the resulting pattems of labelling. Also, inthe absence of such cytoarchitectonic definitions numerousauthors have adopted a variety of definitions for field L andHVc, which has led to some confusion conceming bound-

/ t trespfatory laryngea-l synnx

4 , 4 0

aries, nomenclature, structure, and afferents of these nuclei.We attempted to resolve these issues in the zebra finch

[Fortune and Margoli ash, 1992, 1 994, unpubl. observ.].Field L is the primary thalamorecipient telencephalic

auditory strucfure in birds, receiving input from ovoidalisand surrounding areas [Karten, 1968; Bonke et al.,1979a;Kelley and Nottebohm,19791. Brauth et a1.,1987; Hdusler,1988; Bell et al., 1989; Wild et al., 1990; Durand et al.,1992; Wild et al., 19931. In oscines, we showed that field Lis a complex of 4 or 5 subdivisions - Ll, L2a,L2b, L3 andL, with the distinction between LZb and L uncertain [For-tune and Margoliash, 19921. These subdivisions wraparound and insert into each other in a complex, three-dimensional organization. They can be differentiated by themorphology of Nissl stained cells and arrangement of cellclusters. Four types of Golgi-stained neurons occur in fieldL in the zebra finch and correspond to those described forthe European starling (Sturnus vulgaris) [Saini and Leppel-sack, 19811. Types 1 and 2 neurons, which are the largestneurons in field L, are the major projection neurons.

We also showed that male HVc has 3 cytoarchitectonicregions, a ventral caudomedial region with densely packedsmall and medium-sized cells, a dorsolateral region withlarge oriented oblong or ovoid cells and cell clusters ar-ranged into rows, and the commonly recognized centralregion with large, darkly staining cells and cell clusters

[Fortune and Margoliash, 1994, unpubl. observ.]. Theoriented cell region had not been previously described. All3 regions of HVc in males project to area X and to RA,although the ratio of X-projecting to RA-projecting cellsmay vary across regions. Interestingly, all major cytoarchi-tectonic features of HVc's subdivisions may result from theaction of hormones early in life: a female treated with hor-mones early in life (generously donated by M. Konishi) hadall 3 subdivisions, whereas HVc in untreated females can-not be subdivided and is distinct from all the male subdivi-sions.

Originally, it was thought that field L projected to the'shelf' ventral to HVc and that an area rostral and lateral tofield L projected directly into HVc [Kelley and Nottebohm,l9l9l. That report, however, was based on an incompletecytoarchitectonic definition of field L - including L2abfiprobably excluding Ll and L3. Subsequently, the projec-tion of NIf onto HVc was discovered [Nottebohm et al.,19821. Since NIf is apposed to Ll and L2a, Nottebohm etal. [1982] postulated that the direct projections into HVcfrom rostrolateral field L described by Kelley and Notte-bohm [1979] were the projections of NIf and not of field L.Thus, a direct connection between the auditory telencepha-lon and the song system had not been established.

Afferents of HVcOur recent experiments using fluorescent and biotinyl-

ated dextrans injected into HVc have demonstrated a directprojection of field L into HVc, as well as other connections

[Fortune and Margoliash, 1994]. All projections were con-firmed with anterograde injections, although in some casesfibers-of-passage prevented final conflrmation. Since struc-tures ventral to HVc also receive inputs from field L (see

below), it was necessar:y to make some small tracer injec-tions that were confined to HVc. Injection sites that ap-peared to be restricted to HVc labeled axons within RA;injections that appeared to encroach upon the neostriatumventral to HVc labeled axons around as well as within RA.This finding will be retumed to later. We used the distribu-tion of axons around RA as a second criteria for determin-ing the extent of an injection site for our retrograde experi-ments.

Field L.Injections of biotinylated dextrans restricted toHVc labeled cells in subdivisions L1 and L3. Those cellsthat filled well enough to be classified were either type 1 ortype 2 as defined in Golgi material [Fortune and Margoli-ash, 19921. There was no apparent topography of labeledcells within Ll or L3. Labeled cells were distributedthroughout L1 and L3, but more cells were labeled in Llthan in L3 in all cases. Injections of biotinylated dextransrestricted to HVc retrogradely labeled small numbers ofcells in the fleld L complex (< 100), whereas injections offluorescent dextrans restricted to HVc labeled somewhatmore (> 100). Since field L is tonotopically organized (see

below), the sparse labeling throughout Ll and L3 subdivi-sions implies that there is broad convergence of frequencyinformation onto HVc. We found several types of Golgi-impregnated axons within HVc that have broad arboiza-tions, but none that have small, dense spatially-restrictedarborizations as would be expected in a topographicallyorganized projection. This suggests that inputs to HVc pro-ject broadly throughout HVc.

Nucleus interfacialis. The NIf, which is apposed IoLZaand L1 near the lateral edge of the field L complex, is acytoarchitectonically distinct nucleus that projects to HVc

[Nottebohm et al., 1982]. Restricted dextran injections ofHVc densely labeled cells throughout NIf [Fortune andMargoliash, l994l.The projection of NIf onto HVc also didnot appear to be topographic. Small injections into HVclabeled fewer cells in NIf than did larger injections, but inall cases labeled cells were seen throughout NIf with a scat-tered distribution.

Most of the labeled cells in NIf had the morphologicalcharacteristics of type 5 cells as seen in Golgi material

[Fortune and Margoliash,1992), which have dendrites re-

Margoliash/Fortune/S utter/YuAVren-Hardin/Dave

Organization of Song SystemFunctions and Orisins

250

stricted to NIf, but HVc injections also labeled cells alongthe rostral border of NIf that had extensive dendritic arbor-izations extending into adjacent Ll. Such NIf cells had fusi-form somata, not oblong as is typical for type 5 neurons.These cells may represent a separate morphological type ofNIf neuron that receives auditory input from Ll and pro-jects to HVc. Thus, Ll may modulate HVc activity indi-rectly through NIf.

Neostriatum Ventral to HVc. Caudal to field L and theectostriatum there is a large, poorly characterized regionknown as the caudal neostriatum (NC). Various nameshave been applied to potentially overlapping parts of thisregion, including Nd (dorsal neostriatum) and NCM (cau-domedial neostriatum). This area includes neostriatum ven-tral to HVc, and also includes the 'shelf', which we con-sider separately (see below). Ventrally located HVc neu-rons commonly have dendrites that extend outside of HVc,in some cases for hundreds of microns. Thus, axons synaps-ing near to but not within HVc may be potential sources ofinput to HVc. The neostriatum caudal to the field L com-plex and ventral to HVc is auditory [Scheich et al., 1979;Katz and Gumey, 1981; Miiller and Leppelsack, 1985;Stripling et al., 7994; Margoliash, unpubl. observ.l, andmay form yet another pathway for auditory input into HVc.

Injections restricted to neostriatum ventral to HVc aredifficult to interpret, because with the standard dorsal ap-proach the pipette passes through HVc, and because theinjection will intemrpt fibers projecting tolfrom HVc. In ourstudies, we compared the pattem of labeling of injectionsrestricted to HVc with the pattem of labeling of injectionsthat encroached ventrally beyond HVc. We confirmed eachputative projection with an injection of complementary(anterograde or retrograde) tracer into the putative projectiontarget. The data support the conclusion that Ll and L3 of thefield L complex project onto neostriatum ventral to HVc[Fortune and Margoliash, 1994]. This confirms Kelley andNottebohm's |l979l result, although labeling is not re-stricted to the shelf (see below). Neostriatum ventral toHVc also receives inputs from the ventral hyperstriatum(HV), which is reciprocally connected with field L. Field Linput to HV is primarily from Ll and L3 [Fortune and Mar-goliash, 19941, and HV projections terminate primarily with-inL2a [Bonke et al., 7979a; Wild et al., 19931. The ventralhyperstriatum is auditory [Heil and Scheich, 1991; Miillerand Leppelsack, 1985; Margoliash, 19861, thus neostriatumventral to HVc may receive auditory input from field Ldirectly and indirectly from HV. There are several parallelpathways, therefore, whereby field L may influence HVc.

Additionally, tracer injections centered in caudal HVretrogradely labeled cells in a shelf region that surrounds

Uva [Fortune and Margoliash, 1994]. If Uva is homologousto caudal DLP of non-oscine species (see below), then thearea surounding Uva that projects to HV may represent asepa.rate subdivision of DLP. We did not determine thepositions and possible overlap of the regions of HV thatreceive from the area suffounding Uva, that are reciprocallyconnected with field L, and that project immediately ventralto HVc. All 3, however, are in caudal HV.

The Shelf. Some controversy surrounds the deflnition ofthe structure immediately ventral to HVc, the 'shelf'. Theshelf was originally described as an a.rea medial and poste-rior to ventral HVc, based on pattems of anterograde label-ing in autoradiographs after injection of tritiated aminoacids into field L of canaries [Kelley and Nottebohm,19791. The shelf as described by cytoarchitecture, however,is distinct along the lateral and anterior borders of ventralHVc but fades out medially and posteriorly [Fortune andMargoliash, 1994, unpubl. observ.l. That is, the pattem oflabeling described by Kelley and Nottebohm [1979] is notco-extensive with the fiber-rich area just ventral to HVc. Itis not uncommon in the literature to find the cytoarchitec-tonic and hodological definitions of the shelf used inter-changeably.

There is also some question as to whether the shelf pro-jects to HVc. In our restricted dextran injections of HVc,we did not label any neurons in the shelf, though we rou-tinely labeled more distant cells in field L, NIf, Uva and themedial magnocellular nucleus of the anterior neostriatum(mMAN). The severe flbers-of-passage problem aroundHVc and apposition of the shelf to HVc make it particularlydifficult to assess this putative projection with anterogradelabeling from extracellular injections restricted to the shelf.Unfortunately, the intracellular data are also unclear. Invivo intracellular labeling showed that the axons of shelfneurons ramify within HVc [Katz and Gumey, 1981], butsubsequent experiments using living brain slices, whichpermit direct visual guidance of the electrode, showed thataxonal arborizations of shelf neurons avoid HVc lL. Katz.reported in Margoliash, 19871.

Intrinsic Connections and OutputsAny injection site that impinged on any of the 3 regions

of HVc labeled cells in all 3 regions of HVc. We probablylabeled cells via their intrinsic projections because they areby far a larger component of axons within HVc than areHVc's extrinsic projections. In each case there were differ-ences in the distribution of labeled cells in HVc, and therewas no consistent topographical pattem across cases. In allcases, axons were also extensively labeled throughout HVc[see also Katz and Gumey, 1981]. This web of connections

251

could support considerable communication between allparts of HVc.

Injections of anterograde tracers into HVc demonstratedthat the outputs of HVc exhibit massive divergence. Axonsof HVc neurons ramify extensively throughout their targetsites, either X or RA [Fortune and Margoliash, 1994; Bott-jer et al., 19891. Single axons from HVc extend over largeparts of areaX, giving rise to sparse terminations through-out. Single HVc axons appear to ramify extensively oncethey reach RA, although the extent of the arbors is difflcultto estimate because even small extracellular injections la-beled numerous axons. The projections of HVc onto area Xand RA did not appear to be topographic [Fortune and Mar-goliash, I994l.In contrast, there is some evidence for a top-ographic projection of RA onto the hypoglossal nucleus

[Vicario, 1991], and the hypoglossal nucleus has at leastsome degree of myotopic organization lVicario and Notte-bohm, 19881. This suggests that single HVc axons gener-ally contribute to activation of all syringeal muscles. Thesedata are consistent with a hierarchical scheme of organiza-tion of the motor system for song.

SummaryIn summary, we find a massive convergence of inputs

onto HVc, with multiple parallel pathways including thefield L complex, NIi cDLP, neostriatum ventral to HVc,and intrinsic HVc. Each of these projections may besources of auditory input to HVc, and none of these inputsappear to be topographically organized. Additionally, thefield L projection involves at least 2 subdivisions and 2 celltypes from each subdivision, and there is evidence for 2 celltypes in NIf that project onto HVc. These data supporlseparate lines of inquiry, one regarding the functional im-plications of the convergence/divergence of informationflow through HVc, a second regarding the evolutionary ori-gins of this pattem of connections. We will address the sec-ond question first.

Interspecific Gomparisons:Evolution of the Song System

Our recent anatomical studies give some insight into theevolution of the song system. There is compelling evidencethat vocal leaming has evolved independently at least threetimes, in the oscines, psittacines, and in some of the hum-mingbirds [Nottebohm, 1972]. This implies the song sys-tem arose as an independent adaptation of oscine passerinebirds. All oscines apparently have a song system [e.g.DeVoogd et a1., 19931, including area X, HVc, IMAN,

mMAN, NIf, and RA in the forebrain. (The distinctionbetween the medial and lateral aspects of MAN may not bepresent in all songbirds.) This set of cytoarchitectonicallydistinct forebrain nuclei has been recognized only in theoscines, and similar nuclei are not readily visible in subos-cine brains fNottebohm, 1980; Kroodsma and Konishi,19911. The relationship of song system nuclei to forebrainstructures in non-oscine birds is not well understood.

Parrots and allies, the other extensively-studied exampleof vocal leaming [see also Baptista and Schuchmann,19901, also have cytoarchitectonically distinct forebrainnuclei related to vocal production. Originally, it was con-cluded that the similarity in the connections of analogousnuclei labeled 'HVc' and 'RA'

[Paton et al., 1981] implieda similar evolutionary origin for the forebrain structuresinvolved in vocal control in psittacines and oscines. Thisview has been recently challenged by an analysis of thebudgerigar ( M e lop sittacus undulatus,) vocal control system,which showed that many details of budgerigar forebrainconnections were different than those in oscines [Striedter,1994]. For example, 'HVc' and 'RA' of the budgerigar donot receive direct or indirect input from field L [Hall et al.,1993; Striedter, 1994f. These significant differences be-tween the vocal control system in songbirds and the bud-gerigar has motivated a change in the nomenclature of thebudgerigar forebrain nuclei away from the implied homol-ogy fStriedter, 1994]. Some elements of the budgerigartelencephalic system involved in vocal control may be in-dependent adaptations of non-homologous pathways, whileothers may be independent adaptations of homologouspathways also used for vocal control by oscines.

In our own studies, we noted that injections of HVc thatincluded neostriatum ventral to HVc anterogradely labeledan area surrounding RA [Fortune and Margoliash, 1994].Injections of X (possibly also involving tissue surroundingX) retrogradely labeled cells in neostriatum ventral to HVcas well as in HVc proper. Thus, the outputs of one or moreareas immediately ventral to HVc are parallel to the outputsof HVc itself. Moreover, we established that field L pro-jects both to HVc and areas ventral to HVc. The similarityof connections of HVc and neostriatum adjacent to HVcraises the possibility that HVc arose as an elaboration orduplication of an ancestral structure in the caudal neostria-tum that is common to birds.

ForebrainThe song system follows the general pattem of reptile-

bird forebrain organization lUlinski and Margoliash, 1990]and the general architecture for the flow of sensory infor-mation in the forebrain of birds. Figure 2 shows this pattem

Margoliash/Fortune/Sutter[uAVren-Hardin/Dave

Organization of Song System:

Functions and Origins252

I H V I

{IlI

l-ruua -l

I I

Fig' 2. Comparisonoftheconnectivi tyofHVc(a),NCadjacenttoHVc(b),andNd/NIVL(c)innon-songbirdspe-cies. a Afferents and efI'erents of HVc as demonstrated in the zebra finch. b Afferents and efferents of NC adjacent toHVc [see Fortune and Margoliash, 1994]. Parentheses indicate a region adiacent to a structure. The dotted line indicatesthat the (Uva) recipient area of HV may not be equivalent to the area of HV that projects to (HVc). c Connectivity ofNd/NIVL as demonstrated in non-songbird species (see text). The terms NIVL and NIDL are used exclusively in bud-gengars.

fbr the auditory system by summarizing our recent resultsand those from previous studies of the ascending auditorypathways of non-songbirds. This general architecturebe_qins with a projection of a dorsal thalamic sensory nu-cleus onto a structure between the intemediate and caudalneostriatum. This connection in the auditory system is theprojection of nucleus ovoidalis (Ov) and associated areasonto the field L complex [Karten, 1968; Bonke er al.,1919a; Brauth et al., 1987; Durand et al., 19921, and in rhevisual system it is the projection of rotundus onto the ecto-striatum [Karten and Hodos, 1970]. The somatosensorypathway is not as well described, though the emerging pic-ture is that DLP, a multisensory nucleus fKorzeniewskaand Giinttrktin, 19901, projects onto a discrete areabetween the field L complex and the ectostriatum fGamlinand Cohen, 19861.

The second projection in this general pathways is fromthe neostriatal structure onto the caudal neostriatum. Boththe field L complex and the ectostriatal complex are knownto project to the dorsal areas of the caudal or lateral neo-striatum [Bonke et al.,l9]9a; Shimizu and Kar-ten, l99lal.The projections of the neostriatal somatosensory area are

not known. There is an overall topography, such that theefferents of the field L complex are medial to the efferentsof the ectostriatum [Ulinski, 1982]. Topography withineach of these projections has not been addressed, except theprojection of field L onto HVc in zebra finches is non-ropo-graphic [Fortune and Margoliash, 1994].

The last projection in this general pathway is from thecaudal neostriatum onto the archistriatum. In pigeons, thevisual and auditory areas in the caudal neostriatum projectonto the archistriatum [Shimizu and Karlen, l99la, Wild etal., 19931.In the zebra finch, we have shown that the cau-dal neostriatum projects onto the archistriatum [Forluneand Margol iash. 19941.

The HVc, which resides in the caudal neostriatum, is sit-uated in the middle of this topographical series of projec-tions, apparently at the junction of all 3 sensory pathways.The visual projections from the ectostriatum are lateral toHVc, and auditory projections from the field L complex aremedial and ventral to HVc. Since we expect the somatosen-sory projections to terminate between the visual and audi-tory projections, HVc may be near these inputs as well. Inour laboratory, Mr. Peter Bell originally suggested that

253

nucleus Uva in zebra finches had similar position and effer-ents as did DLP (which is multisensory but primarily soma-tosensory) in other species.

Uva and DLPWe find that the morphology of cells in Uva, and the

relationship of Uva to other dorsal thalamic nuclei, are sim-ilar in many respects to nucleus DLP in pigeons (Columbalivia) lsee also Wild, l993al.In pigeons, DLP projects ontothe intermediate neostriatum [Kitt and Brauth, 1982; Gam-lin and Cohen, 19861. A reconstruction of flgure 8A-D ofGamlin and Cohen [1986], which shows the projection areaof DLP on the neosftiatum, is similar in shape to NIf shownin figure 12 of Fortune and Margoliash [1992]. The areawhere DLP fibers terminate is also cytoarchitectonicallydistinct. The photomicrograph of the DLP projection areain pigeons (fig.9,A. of Gamlin and Cohen t19861) is similarto the appearance of NIf shown in figure 4E of Fortune andMargoliash ll992l.In addition, Gamlin and Cohen t19861stated that injections of retrograde tracers into HV andmore dorsal structures labeled a few cells in DLP. They didnot describe the distribution of these cells, but the result isnot dissimilar to that obtained from injections into HV ofthe zebra finch, where cells were labeled along the dorsal,caudal, and rostral edges ofUva.

Korzeniewska and Giintiirktin [1990] showed that DLPin pigeons contains neurons that respond to somatosensory,visual, and auditory stimuli. If we assume that Uva ishomologous to DLP, then this can help to explain auditoryactivity in Uva lOkuhata and Nottebohm, 1992]. The Uvacould be a source of auditory input to HVc as well as thepathway for the visual responses in HVc described byBischof and Engelage [1985].

We have been unable to determine the location of DLPin zebra finches. The position of DLP as shown in theStokes et aI.ll914l canary atlas is largely inconsistent withthat shown in the pigeon atlas fKarten and Hodos, 1967l.Inthe canary atlas DLP is shown along the dorsal surface ofthe occipitomesencephalic tract (OM). In pigeons, DLPdoes not contact OM at all. Stokes et al. ll974l recognizedthis uncertainty - in their plate 2l DLP is shown in twolocations, medial to OM and the dorsal intermediate poste-rior thalamic nucleus (DIP), and lateral to the dorsal medialposterior thalamic nucleus (DMP). In the pigeon atlas, DLPis lateral and caudal to DIP. In L3.00 of the pigeon atlas,there is a dark band which appears to be marked as DLP inthe accompanying line drawing. This dark band is similarin appearance and position to Uva seen in zebra finches.Lastly, Wild [1993a] showed that the afferents of DLP andUva are identical in several songbird and non-songbird spe-

cies. Wild U993al concluded that Uva is homologous tothe caudal subdivision of DLP (cDLP).

A Hypothesis for the Evolution of the Song SystemFigure 2 shows a comparison of the pathways described

in this repofi and those seen in previous reports, includingthose of Bonke et al. ll919al, Kelley and Nottebohm

19791, Brauth et al. [1987], and Brauth and McHale

t19881. The primary differences between the patterns seenin the zebra flnch and those in other species are the projec-tions of mMAN onto HVc, HVc onto area X, and field Lonto a 'cup' sunounding RA. In other respects, however,the projections are similar. These similarities include thepathways from DLP through the neostriatum (N) to caudalneostriatum (NC), from Ov to NC through both the field Lcomplex and HV, and the projections of dorsal NC and thefield L complex onto the archistriatum (A). These pathwaysare parallel to those of the song system (fig.z).

The song system itself is an autapomorphy of the oscinepasserines [Kroodsma and Konishi, 19911, but the path-ways that are parallel to the song system may be generallypresent in other species of birds (fig.2).These data suggestthe hypothesis that the song system of oscine passerinesevolved from pathways that are plesiomorphic - specifl-cally, that HVc and RA evolved as an elaboration or dupli-cation of pathways that are common to birds [see also Bre-nowitz, I99Ial. For example, during evolution HVc mayhave arisen from a subdivision of caudal neostriatum thatbecame cytoarchitectonically distinct with increased func-tional specialization, presumably associated with auditory-feedback mediated song learning.

The caudal neostriatum may generally exhibit integra-tive mechanisms involved in perceptual phenomena (for arelated discussion, see Shimizu and Karten [1991b]). Thisis consistent with the general flow of sensory information inavian forebrains, in which NC receives input from primarysensory areas and in turn projects to archistriatum. Recentexperiments support this hypothesis, implicating areas ofNC other than HVc in conspecific song recognition in os-cine birds [Mello etal.,1992].In pigeons, the output of Nd,the part of NC that receives from L1 and L3, projects ontoareas of the archistriatum that give rise to bilateral projec-tions to areas surrounding thalamic and midbrain auditorystructures [Wild et al., 19931. This suggests the hypothesisthat the ancestral caudal neostriatal structure that gave riseto HVc originally had an auditory function, and that part ofthe specialization involved acquiring projections to archi-striatum involved in motor control.

Margoliash/Fortune/Sutter/YuAVren-Hardin/Dave

Organization of Song System:Functions and Origins

254

(a)

-1.0 "0.s 0.0 0.5 1.0Gonspecific song selectivity

Fig. 3. Response selectivity of 329 zebrafinch HVc single unitsrecorded in 24 urethane-anesthetized birds. a One or more conspecificsongs and the bird's own song (autogenous song - AS) were pre-sented, resulting in 378 pair-wise comparisons. Response is defined asaverage spikes/s over an entire song minus spontaneous rate. Conspe-cific selectivity is the response to conspecific song/response to AS.For example, selectivity index = 1.0 implies salne response to conspe-

(b)

n -rF- -0.5 0.0 0.5 1 . 0

Reverse song selectivity

cific song and AS. Note that the average response to conspecific songsis close to the spontaneous rate (0.0); many conspecific songs elicitoverall inhibition. b The response to forward and reversed AS wascompared for 99 single units. Response is defined as in a. In almost allcases, forward song elicited a stronger response. For both a and b,extreme values are collapsed into the outermost bins of each graph.Songs were typically presented for I 0 or 20 repetitions.

20

oc 5 0o.9(E t^cL -v

tro(, Booo'=

i e o'6o.ct 10z

oo.9,$CLtroO { noo

(ECL

itz

Selectivity for Autogenous Song:Distributed Representation within A Hierarchy

The unidirectional flow of information between most orall of the forebrain constituents of the song system, theposition of HVc within those pathways, and the conver-gence and divergence of information flow through HVcsuggests that HVc is relatively high within a sensory hierar-chy and that information within HVc may be organized in adistributed fashion. The functional realization ofthese ideasis explored in the next three sections. First, auditory re-sponses in HVc are compared with those in field L. In thesecond and third sections, auditory and motor data are pre-sented that provide evidence for a distributed representationfor song, and suggest that the distributed representationencodes song in the time domain.

Auditory Responses in HVc and in Field LIn several species the projection ofovoidalis onto field L

has been shown to be topographic [Bonke et al., \979a;Bigalke-Kunz et al., 1987; Brauth et al., 1987; Bell et al.,19891. A tonotopic organization exists for fleld L, varying

both with subdivision and species fBonke et al., I9j9b;Heil and Scheich, 1985, I99l; Mtiller and Leppelsack,1985; Miiller and Scheich, 1985; Rijbsamen and Dcjrr-scheidt, 1986; Lim, 19931. In general, L2a neurons havehigh spontaneous rates and robust responses to tone bursts,whereas neurons in Ll and L3 have lower spontaneousrates and, in some cases, weak responses to tone bursts [e.g.Bonke et al., I9l9b; Scheich et al., 19791. Neurons of Lland L3 also tend to exhibit greater selectivity for species-specific vocalizations [Leppelsack and Vogt, 1976; Scheichet al., 1919; Leppelsack, 1983; Mori and Striedter, 19921,which may be based in part on broad tuning curves, re-quirements for several spectral peaks, or temporal require-ments. Thus, the field L input to HVc is likely ro exhibitsome specialization for complex, species-specific sounds,although the physiology of the neurons that project to HVchas not been explicitly described.

In contrast, there is no evidence of tonotopic organiza-tion in HVc fMargoliash, 1987]. The HVc neurons tend torespond poorly to tone bursts (see below) but are selectivefor the individual bird's 'autogenous' song (AS) [Margo-liash, 1986; Margoliash and Konishi, 19851, a property that

255

results from tuning to parameters of AS [Margoliash, 1983,1986; Margoliash and Fortune, l992l.In a recent analysis,329 single units in the HVc of 24 wethane-anesthetizedzebra finches were tested with AS, conspecific songs, toneand noise bursts, and other stimuli [D. Margoliash and E.S.Fortune, unpubl. observ.l. (The combination sensitive prop-erties of a subset of these cells has been previously re-ported, where details of the experimental procedures aredescribed; see Margoliash and Fortune tl992l.) The re-sponses to conspecific songs were almost always weakerthan the responses to AS (fig.3a). The responses to tonebursts were also generally quite weak. For only 13 of 69neurons was the peak response (30 ms bin) to any toneburst stronger than the peak response to AS. (Since tonebursts are much shorter duration than songs, peak rates area more meaningful comparison than average rates.) If any-thing, the actual distribution is even more skewed, sincetone burst tuning is not rewarding in HVc and was notalways attempted for weakly responding cells. Selectivityfor AS may obtain generally in the HVc of oscine birds -

similar results have been obtained in multiunit recordingsin white-crowned sparrow s (Zonotric hia leuc op hrys) [Mar-goliash and Konishi, 1985; Margoliash, 19861 and in pre-liminary single unit data in white-throated sparrows (Zono-trichia albicollis) [Margoliash, unpubl. observ.], and multi-unit data in canaries and mockingbirds (Mimus polyglottus)

lMcCasland and Konishi, 1981].The selectivity for AS apparently first emerges at the level

of HVc - neurons with comparable propenies are not recog-nized in field L [Margoliash, 1986; Lim, 1993], nor in pre-liminary data from Uva [Okuhata and Nottebohm, 1992] orneostriatum ventral to HVc [D. Margoliash, unpubl. observ.].The auditory responses of NIf may exhibit some degree ofsong selectivity [M.L. Sutter, A. Dave, and D. Margoliash,unpubl. observ.l. The broad distribution of field L neuronsthat project to HVc make this conclusion particularly difflcultto establish for field L. In perhaps the most direct comparisonto date, Margoliash t19861 examined the multiunit propertiesof both HVc and field L in 3 adult white-crowned sparrows.The HVc recordings exhibited highly statistically significantstronger responses for AS compared to conspecific songs,reversed AS, frequency-shifted AS, and two other parametricmodifications of AS. Two of the birds had been reared in thelaboratory; HVc recordings in those birds exhibited strongerresponses to AS than to the tutor songs. In contrast, the fieldL recordings in the same 3 birds with the same stimulus rep-erloires exhibited responses to AS that were weaker or com-parable to the responses to other stimuli, with the exceptionthat a weak preference for forward AS compared to reversedAS was observed.

Temporal Order Sensitivity of HVc Auditory Neurons.Stimulus selectivity can result from spectral specificity,temporal specificity, or both. Because frequency is mappedin the auditory periphery, spectral flltering has been empha-sized as a mechanism for achieving selectivity. Recentcomputational work, however, has demonstrated that anauditory neuron's time-domain response property - forexample, the phasic and tonic components of response -

can more strongly affect the neuron's response to complexstimuli (in a least-squared-error sense) than static parame-ters of tone burst responses such as the frequency tuningcurve [Bankes and Margoliash, 1993]. In HVc, there is evi-dence that analysis in the temporal domain is crucial. Forexample, HVc neurons systematically prefer forward AScompared to reversed AS, two stimuli with different time-varying features that share identical overall spectra (fig.3b).The song selective properties of HVc units also suggest theimportance of time-domain processing. Most single neu-rons in the HVc of zebra finches were highly selective forAS. When the response rate (spikes/s) of the population ofHVc single units was corrected for spontaneous activity,the average response to conspecific songs was near zero,and many songs elicited overall inhibition (fig.3a). If HVccontributes to song recognition [e.g. Margoliash, 1986;Brenowitz, 1991b1, these data suggest that the temporalpattem of response to conspecific songs, rather than theabsolute strength of response, is likely to be the most salientcue for recognition.

Studies of HVc temporal processing have focused onsensitivity to combinations of sounds. Temporal combina-tion sensitive (TCS) neurons have been described in white-crowned sparrows [Margoliash, 1983] and zebra finches

[Margoliash and Fortune, 1992;Lewicki and Doupe, 19931.In general, non-linear summation of excitation to temporalsequences of notes [Suga et a1., 1978; Margoliash, 1983;McKenna et al., 1989; Olsen, 19941 or to spectral compo-nents of a note [Fuzessery and Feng, 1982; Langner et al.,1981; Sutter and Schreiner, 1997; Margoliash and Fortune,19921 is a widely recognized phenomenon in the auditorysystem used to generate higher-order response properties.In most cases, combination sensitivity is closely associatedwith processing for species-specific vocalizations.

In white-crowned sparrows, all TCS neurons respondedto a sequence of two syllables but not to either syllablewhen presented in isolation [Margoliash, 1983]. The TCSneurons were highly selective, responding most strongly toAS, more weakly to other songs from the same dialect, andtypically not at all to songs from other dialects or to simplearlificial stimuli. Thus these neurons were termed 'song-

specific'. The response selectivity of song-specific neurons

Margoliash/Fortune/Sutter/YuAVren-Hardin/Dave

Organization of Song System:Functions and Origins

256

was based, in parl, on specificity to parameters of AS. Thiswas best demonstrated in cases where it was possible tosynthesize complex artificial sounds that mimicked the nat-ural signal yet permitted parametric modification. In thesecases, changing the frequency of either the leading or fol-iowing element in the sequence reduced the responsestrength. The TCS neurons were quite insensitive to theduration of the first element of the sequence or to the dura-tion of the interval between the elements. In general, theseneurons had long integration times (up to 1 s), longer thanfield L neurons lSchafer et al., 1992].

Recent single unit studies of the zebra finch have con-firmed and extended these results. In the zebra finch, approx-irnately I57o (481329) of the cells were strictly temporal-combination sensitive (TCS), responding with excitation to asequence of sounds but not to the individual sounds pre-sented in isolation [Margoliash and Fortune, 1992]. In allcases, TCS cells responded much more strongly to AS thanto any conspecific song tested. It should be noted that thelong periods of time needed to unambiguously establish theTCS property suggests that l5o/o is probably a significantunderestimate of the size of the population.

Different classes of zebra finch TCS neurons requireddifferent temporal sequences of song elements. Some neu-rons required a sequence of 2 parts (notes) of a single sylla-ble. Other neurons required a sequence of 2 discrete syl-lables of song. Most remarkably, other neurons required asequence of 3,4, or in some cases even 5 discrete syllablesof AS in the proper order. Additionally, in some cases theTCS response at a given syllable of song required priorexcitation 3 or more syllables preceding the TCS excita-tion, but the prior excitation itself did not require a tempo-ral sequence of sound elements (i.e., was not TCS). In allcases tested, the interval between syllables could be ex-tended by hundreds of milliseconds without compromisingthe TCS response. Thus, there was a broad range of types ofTCS neurons in the zebra finch HVc, with individual neu-rons exhibiting integration times of up to I s and up to 5syllables of AS.

Moclels of Temporal Order Sensiriviry. Specific circuitmodels have been proposed to explain temporal order sen-sitivity in HVc. In no small way, these are of interestbecause they can potentially describe a neural circuitdirectly involved in song learning. For ease of exposition,rve define cell C1 which projects to cell C2; C2 exhibitstemporal order sensitivity to the stimulus sequence Sl fol-lowed by 52. Margoliash [1983] proposed an 'inhibitory-

rebound' two-cell model for TCS neurons in white-crowned sparrow HVc based on extracellular data. In thatmodel, C1 inhibited C2 such that rebound excitation of C2

from release of C1 inhibition at the offset of 51 combinedwith excitation of C2 by 52 to produce suprathreshold exci-tation of C2. Selectivity for autogenous song was proposedto result from unspecified mechanisms that tuned Cl selec-tively for Sl and tuned C2 selectively for 52. This modelwas one of potentially many that could account for all theknown dynamical properties of TCS neurons in white-crowned sparrows, especially very broad tuning to theduration of 51 or the interval between 51 and 52, phasicresponse near the onset of 52, maintenance of the TCSproperty even in cases when 51 and 52 are identical, andlack of response if 51 and 52 were presented simultane-ously [Margoliash, 1983]. It is notewofihy that the longintegration times in HVc precludes the use of delay lines, ashas been proposed for TCS neurons in bats [Kuwabara andSuga, 19931.

Recent intracellular recordings in zebra finch haveextended the analysis of temporal order sensitivity in HVc.Lewicki and Doupe ll993l presented preliminary evidencefor the 'inhibitory-rebound' model of Margoliash [1983] aswell as for the two other mechanisms. In the 'long-short'

model, long-lasting depolarization produced by 51 com-bined with a shofter-duration depolarization produced by52 to achieve suprathreshold excitation. In the 'burst-firing'

model, depolarization by 52 produced a burst if precededby a hyperpolarizing potential produced by 51; 52 aloneproduced few spikes. Interestingly, the burst-firing modelproduces a stronger non-linear response than the two othermodels. Below we note that the bursting activity of HVcneurons is probably synchronized throughout HVc.

The models of Margoliash [1983] and Lewicki andDoupe ll993l propose possible mechanisms for the final(or non-linear) stage of various circuits that exhibit tempo-ral order sensitivity, but they do not explain how specificityfor syllables of autogenous song is achieved. From the per-spective of circuit analysis of song leaming, this limitationimplies lack of insight into the neuronal elements modifiedduring song leaming. Each bird sings a different song; HVcneurons are tuned to the parameters of the bird's own song.Which elements of the circuit reflect the idiosyncratic prop-erlies of HVc auditory neurons? From a modeling perspec*tive, this limitation is particularly significant in the zebrafinch, where some HVc TCS neurons integrate a sequenceof up to 5 complex syllables before exhibiting excitation[Margoliash and Fortune, 1992].

The logical extension of the Margoliash [1983] two-cellmodel that accommodates a requirement for multiple syl-lables to elicit the TCS response is to add more cells inseries, with the last cell exhibiting the TCS response. Thisis an unattractive solution because: (1) it is not parsimoni-

25'�7

ooh\qo.Yd z v(no(E

rnto 1 0o(!

oo.o

g,

woL'*,

,uo(ELa

o5

8 1 0o

oO.a

Fig.4. Relationships between spontaneous and driven (re-

sponse) properties of zebra finch HVc single units. The spontaneousrate is calculated over 30 s of spontaneous activity (whenever thosedata were collected) or over 1 s preceding each repetition of AS.a Spontaneous data for 267 tnits. The index of bursting measuresspike 'clumping' in time: 0 implies equal inter-spike intervals for allspikes; 1 implies all spikes occurred simultaneously (a limit condi-tion). High spontaneous rate neurons tend to be bursty whereas a clear

ous; (2) it predicts the numbers of TCS neurons should godown with sequence length (more 2-syllable than 3-syllableTCS neurons, etc.), but the data indicate there are roughlyequal numbers of TCS neurons in all different classes, and(3) such an extension still does not predict many other sa-lient features of HVc activity. For example, in HVc highspontaneous rate neurons (>2.5 spikes/s) exhibit a burstingpattem of spontaneous activity (fig.4a). When stimulatedby AS, these neurons respond tonically throughout song(fig.ab). In contrast, virtually all TCS neurons are lowspontaneous rate neurons (< 2.5 spikes/s); Margoliash andFortune ll992l). The TCS neurons may or may not burstspontaneously (fig.4a), but they almost always have phasicresponses to song (fig.ab). Presumably, phasic TCS neu-rons derive their response properties by convergence fromthe more numerous tonic/bursting neurons.

The spontaneous bursting of individual HVc neuronsreflects greater complexity (more time constants) than thehighly regular spontaneous activity commonly seen in RA

[Mooney, 1992]. One possible interpretation of this com-plexity is in terms of central pattem generators (CPGs) -

HVc neurons participate in circuits that encode 'higher-

0.1 0.2 0.3 0.4 0.5

Normalized vector strength

pattem does not emerge for low spontaneous rate neurons. b Stimulusis AS, presented to 293 units. The vector strength (phasicness) ofspontaneous-corrected response was calculated over duration of eachsyllable. Normalized vector strength is the average of the vectorstrength ofresponses to all syllables weighted by the number ofspikeseach syllable elicited. Most temporal combination sensitive neurons(see text) have low spontaneous rates and respond in a phasic mannerto song.

level' temporal features of song. This is consistent with thehierarchical organization of the motor system as judged byelectrical stimulation of HVc and RA during singing [Vu etal., 79931. The intrinsic properties of these sensorimotorHVc circuits are exhibited, in part, during spontaneousbursting. From this perspective, the analysis of TCS andother song-selective properties in HVc relates to how audi-tory input modulates CPGs.

Global Synchrony of Spontaneous and Auditory Activity.The properties of HVc neurons are influenced by correlatedactivity observed throughout HVc. This can be visualizedmost easily in multiple electrode experiments. Simultane-ous recordings from two electrodes demosntrate that spon-taneous bursting in HVc is remarkably similar throughoutthe nucleus. For example, figure 5 shows 30 s of spontane-ous activity in two sites in HVc, separated by 850 pm thatwere recorded in a urethane-anesthetized zebra finch. Thepattem of multiunit bursting is highly conelated, althoughnot all bursts are well correlated (compare traces at14-16 s). This degree of correlation occurs regularlythroughout HVc, although we have yet to analyze it syste-matically or at the single unit level.

Margoliash/Fortune/Sutter/Yunvren-Hardin/Dave

Organization of Song System:Functions and Origins

' o * B o ̂o ^oOo ",o"

o - ;o3oi l

(b)

oo

o

o

258

Fig. 5. Correlated spontaneous activityin HVc. Spontaneous activity from tworecording sites in HVc separated by approxi-mately 750 pm. The first 15 s of data areshown in the top panel, the following 15 s ofdata are shown in the bottom panel. Note thatmany bursts of activity occur in synchrony.

I

1 5

The global temporal correlation of spontaneous HVcactivity suggests that stimulus-driven activity might alsoexhibit temporal correlation. This suggestion was con-firmed in an investigation of the spatial distribution ofextracellular responses to AS in the HVc of zebra fincheslsutter and Margoliash, 1994]. Within each of 7 urethane-anesthetized birds, multiunit and single unit responses toAS were similar across the entire spatial extent of HVc (upto 1.3 mm). This effect was not apparent in an 8th controlbird with recordings just outside of HVc. For each bird, anarrow range of times elicited the strongest peak responses(calculated with 25 ms bins), called the time of maximumsynchronization (TMS). Within a bird, 34-75Vo of record-ing sites exhibited the same TMS. The position of theTMS was generally unchanged as the window of integra-tion was varied from 10-150 ms, demonstrating the TMSeffect was robust. Each TMS was associated with a syllableof maximum synchronization (SMS). The positions of theSMS varied considerably across birds. In 4 birds, the SMSwas in the first motif (a motif is a temporal sequence of syl-lables that is repeated one or more times to form a song); in2 birds, the SMS was the introductory note of song, and inI bird the SMS was the second syllable of the last (third)motif. There were no apparent acoustical features of theSMS or the preceding syllable (morphological type, dura-rion, amplitude) that could account for the global synchro-

nous response to song. Typically, syllables of the same typeas the SMS occuning in other motifs elicited much weakerresponses, and some syllables of different type than theSMS elicited strong responses but only in non-SMS motifs.Thus, there was strong and consistent temporal modulationof the auditory response throughout HVc that was not pre-dicted by any obvious local acoustical features.

The intra-individual similarity of response propertiesthroughout HVc provides a logic for examining the aggre-gate activity of HVc [Sutter and Margoliash, 1994]. Ameasure of the aggregate (population) response of thenucleus was derived by summing multiunit data across allrecording sites of each bird ('group PSHT' - post-stimulustime histogram). The aggregate response exhibited phasicexcitation at most syllable onsets and offsets. Additionally,a recovery process after the end of response to AS wasobserved with a long inhibitory component. The duration ofthe recovery varied more than an order of magnitude acrossindividuals (300 ms- 6 s). Features of the temporal mod-ulation of the response were related to the duration and rel-ative strength of the recovery. This implies a long-timecourse process across the nucleus helps to shape theresponse to AS, and that the duration and magnitude of theprocess varies from bird to bird. Thus is consistent with theview that oscillators or other pattem-generating circuits thatare broadly distributed help shape HVc auditory responses.

259

We conclude the auditory representation of song in thezebra finch HVc is distributed and redundant [Sutter andMargoliash, 19941. The pattem of excitation and inhibitionin response to AS is strongly temporally modulated, and isglobally similar because of strong correlation of activityacross HVc. A temporally-encoded representation may pro-vide a useful reference framework for difference detectionin relationship to conspecific songs or for error detection inrelationship to auditory feedback.

The Organization of HVc during SingingStudies of HVc as a motor structure provide further evi-

dence for distributed represenlations in HVc. In one study,we examined the effects of small lesions on the songs ofwhite-throated sparrows (WTS) [Hardin and Margoliash,1992, unplbl. observ.l. Songs in white throated sparrowscomprise simple descending or ascending whistles particu-larly amenable to detailed analysis, typically three longintroductory whistles ('notes') followed by a series of shortwhistles organized into triplets. In the wild, the absolutefrequency of notes and frequency ratio of successive notesis very stable across songs [Borror and Gunn, 1965; Lem-mon and Harris, I974;Hurly et al., 19911. We focused ouranalysis on the first three notes of song. Bird's songs wereextensively recorded prior to and 6-28 days after lesions,which were made by small injections of ibotenic acid tominimize damage to fibers-of-passage. The extent of dam-age was calculated by comparing HVc volumes of lesionedand control sides of the brain in transverse sections. Birdssubmitted to left HVc lesions (size: 2-100Va,n= 14), rightHVc lesions (size:29*79Va,n= 3; this species is left later-alized ll-emmon [1973]), and 3 controls spared HVc.Lesions from < 267o of HVc (13 birds total) produced per-manent effects including a change in the frequency of song;lesions of I07o or smaller (4 birds) produced only tempo-rary effects, or no effects. Postoperatively, the variability inthe duration of notes also increased, although the meandurations did not change.

With a single exception (a bird singing an abnormalsong), birds sang all notes of song after the lesions. In2 of13 birds with permanent frequency shifts, all notes of songswere equally effected, and in 11 of 13 birds there were sig-nificant changes in the frequency ratios of successive notes(p < 0.01). The magnitude of frequency shift for each of thefirst three notes of song was strongly (and signiflcantly)positively correlated with the size of the lesion. Differentparts of song were not affected differently by the locationof lesions within HVc, although a unique effect of medio-caudal lesions on the initial note of song cannot be ruledout. This suite of observations is consistent with a distrib-

uted representation of AS within the level of resolution ofthe lesions.

Most birds exhibited a downward frequency shift, but 3of 4 birds with the lowest frequency songs exhibited anupward shift in frequency. We hypothesize the populationactivity of HVc provides 'drive' to RA. Loss of driveresults in lower tension of syringeal muscles (and intemaltympaniform membranes), which produces downward fre-quency shifts when the song is above natural modes ofvibration of the syringeal intemal tympaniform membrane,but upward frequency shifts when song is below the naturalmodes of membrane vibration. This implies a remarkablysimple code - the output of HVc is distributed throughoutthe part of RA that projects to the hypoglossal nucleus, withthe total output of HVc being a meaningful parameter. Ofcourse, many codes that have time-varying characteristicscould exhibit such a global behavior.

Some songs of lesioned birds were distorted. The num-ber of distoted songs significantly increased with the sizeof the lesion (R'�= 0.666, p < 0.01). Distortions were char-acterized as a loss of fine control of the frequency of whis-tles - an increase in frequency variability within a whistle.Typically, for individual whistles the effect was all or none- either a whistle was distorted or it was not. Although dis-tortions could begin with any of the first 3 notes, with theexception of 1 bird, the notes following a distorted notewere very likely to also be distorted (p<0.01, binomialtest). The degree of distortion tended to increase in asequence of distorted notes. Songs were more likely to ter-minate with distorted notes. Thus, the temporal dynamicsof the population of HVc neurons is resistant to loss oflarge parts of the population, but once an abnormaldynamic state is achieved, the system has difficulty recov-ering. The abnormal state increases the probability of com-plete loss of coordination. Such behavior would be ex-pected from a system that achieves and maintains its popu-lation dynamics through a high degree of correlated activ-ity, although other explanations cannot be ruled out.

The Behavioral Significance of AuditoryResponses in HVc: Funetional and EvolutionaryLinks Between Production and Perception

In the previous section we described the organization ofHVc, concluding that there is a distributed representationfor song both for auditory and motor states, that this repre-sentation is redundaht (similar throughout), and that theprincipal code for song is in the temporal rather than thespatial organization of neuronal activation. In this final sec-

Margoliash/Fortune/S utter/YuAVren-Hardin/Dave

Organization of Song System:Functions and Orieins

260

tion 'uve examine some of the implications of auditory

responses in HVc.

Tlrc Process of Song Perception

The HVc auditory neurons exhibit exquisite selectivity

tbr the individual bird's own song - the one song the bird

i: most unlikely to hear a neighbor sing. Auditory activity

in HVc is inhibited during singing [McCasland and Koni-

shi. 19811, and in any case most adult birds do not rely on

auditory feedback to maintain song [Konishi, 1965; Notte-

bolrrn, 1980; Marler and Sherman,1982: Price, 1979; Nor-

deen and Nordeen, 19921.What then is the functional role

of highly selective auditory responses in the adult HVc?

Birds can recognize individuals on the basis of song alone

[Falls, 1982]. In almost every species tested to date, the

bird's own song has a special status as an external stimulus,releasing a level of territorial behavior that is intermediate

between the effects of established neighbors and unknown

conspeciflc strangers [e.g. Brooks and Falls, 1975; McAr-

thur. 1986; see Falls, 19821. In playback experiments in

species with reperloires of different song types, individuals

n.ratch song type most consistently when challenged with

one of their own songs, a behavior thought to increase

information exchange between singers [e.g., Falls et al.,

1982; Falls, 19851. These data support the hypothesis that

adult songbirds perceive conspecific songs by comparing

them to an internal 'reference' of their own songs or other

known songs [Morton, 1982]. Further evidence to suppol-t

this hypothesis exists in the form of playback experimentsrvith acoustically degraded songs. Birds distinguished bet-

ter between degraded and undegraded songs when they

were familiar with those songs than when the songs were

novel [McGregor et al., 1983; McGregor and Falls, 1984;

McGregor and Krebs, 19841.

With these considerations in mind, Margoliash [1986,19871 proposed that the representation of AS in HVc acts as

a reference against which conspecilic songs are compared.

Support for a role of HVc in song recognition has come

fiom a few laboratory studies that show adverse effects on

discrimination following HVc lesions [Nottebohm et al.,

1990; Brenowitz, I99lbl. In general, these effects are not

strong or the results are somewhat ambiguous; for example,

the lesions of female canary HVc that resulted in loss of

discrimination between conspecific and heterospecific

songs fBrenowitz, l99lb] may have also involved more

rnedial aspects of NC. Thus, the role for HVc in song per-

ception has not been well established.If the motor system for song production is also involved

in song perception then this demonstrates a linkage be-

tu'een production and perception but does not explain the

form of that linkage. In humans, a motor theory of speechperception generally assefis that the articulatory mecha-nisms recruited during speech production are also recruitedduring the perception of like speech, with the pattem ofrecruitement being similar if in a somewhat vaguelydefined fashion [Liberman et al., 796]; Liberman and Mat-tingly, 19851. A parallel motor theory of song perceptionhas been proposed based on the claim that playback of dif-ferent types of song components selectively activates poolsof motor neurons in different parls of the hypoglossalnucleus [Williams and Nottebohm, 1985]. That observa-tion, however, results from a very small data set for a map-ping study (14 recordings in 4 animals). It is difficult tounderstand how a topographic arrangement of syllablesbased on acoustics is consistent with the myotopic organ-izat\on of the hypoglossal nucleus fVicario and Nottebohm,19881, especially when there is evidence that at least somesyringeal muscles are involved in producing most syllablesof song [Vicario, 1990]. Also, to date there is no knownpathway that leads from the hypoglossal nucleus back intothe brain, as would be required for auditory-related activityin the hypoglossal nucleus to have perceptual significance.

An alternative view of the relationship between produc-tion and perception arises from considerations of songlearning. During development, HVc auditory responseproperties are shaped during motor leaming fVolman,19931; this result had been predicted and hypothesized tocontribute to the process of matching the memorized songmodel with auditory feedback during the trial-and-errorphase of song development [Margoliash, 1987]. How doesauditory feedback become decoupled from singing duringdevelopment? In a recent set of experiments [Yu and Mar-goliash, 1992,1993, and unpubl. obseru.l, we were able toachieve direct recording of multiple and single HVc units insinging zebra finches, which is technically challengingbecause of the small size of songbirds and because ofmovement artifacts during singing. In our small sample (11multiunit clusters and 13 single units extracted using spikewaveform classification techniques recorded from 8 sites in7 birds), the neuronal recruitment PSTHs when the birdsang its song were completely different from the auditoryresponse PSTHs when the bird heard its song. In manycases the change from auditory to motor properliesoccurred over a period of about 5 s prior to singing, duringwhich the 'spontaneous' rates of neurons increased by5-10x. This suggests that auditory and motor systems actover different spatiotemporal regimes of neuronal popula-tions - have different state propefiies. This is not consistentwith the suggestion that a linkage between production andperception of song results from closely related states [cf.

261

Nottebohm et al., 19901. Rather, production and perceptionmay share developmental processes and thus exhibit link-ages in the adult fMargoliash, 1987]. In this scheme suchlinkages result only from shared changes during periods ofsong leaming.

Acknowledgments

The work reviewed here was supported in part by grants from theNIH (NS25677), rhe NSF (IBN-930977l), and the Whitehall Founda-tion (M9l-03), M.L.S. was supported by NIH NRSA (l F32DC00075), and A.C.Y. was supported by NIH NRSA (1 F31MH1015r ) .

References

Bankes, S.C., and D. Margoliash (1993) Paramet-ric modeling of the temporal dynamics of neu-ronal responses using connectionist architec-tures. J. Neurophysiol., 69: 980-991.

Baptista, L.F., and K.-L. Schuchmann (1990) Songlearning in the Anna Hummingbird (Calypteanna ). Ethology, 84 : 15 -26.

Bell, P.M., D. Margoliash, and P.S. Ulinski (1989)Cytoarchitecture and differential projections ofnucleus ovoidalis in the zebra finch. Soc. Neu-rosci . Abstr . , 15: 617.

Bigalke-Kunz, B., R. Riinsamen, and G.J. Ddn-scheidt (1987) Tonotopic organization andfunctional characterization of the auditory tha-lamus in a songbird, the European starling.J. Comp. Physiol. A, 16I:255-265.

Bischof, H.J., and J. Engelage ( 1985) Flash evokedresponses in a song control nucleus ofthe zebrafinch (Taeniopygia guttata castanotis). BlainRes. , J26: 370-3 '74.

Bonke, D.A., D. Bonke, and H. Scheich (1979a)Connectivity ofthe auditory forebrain nuclei inthe guinea fowl (Numida meleagrls). Cell Tis-sueRes.,200: l0 l -121.

Bonke, D., H. Scheich, and G. Langner (1979b)Responsiveness of units in the auditory neo-striatum of the Guinea fowl (Numida mele-agris) to species-specific calls and syntheticstimuli. L Tonotopy and functional zones. J.Comp. Physiol., I 32 ; 243-255.

Boror, D.J., and W.W.H. Gunn (1965) Variationin white{hroated sparrow songs. Auk, 82;26-61.

Bottjer, S.W., K.A. Halsema, S.A. Brown, andE.A. Miesner (1989) Axonal connections of aforebrain nucleus involved with vocal learningin zebn finches. J. Comp. Neurol., 279.'3 t 2 -326 .

Bottjer, S.W., E.A. Miesner, and A.P. Arnold(1984) Forebrain lesions disrupt developmentbut not maintenance of song in passerine birds.Science,224: 901 -903.

Brauth, S.E., and C.M. McHale (1988) Auditorypathways in the budgerigar. II. Intratelence-phalic pathways. Brain Behav. Evol., J2:193-20'7.

Brauth, S.E., C.M. McHale, C.A. Brasher, ard R.J.Dooling (1987) Auditory pathways in the bud-gerigar I. Thalamo-telencephalic projections.Brain Behav. Evol.. -10: l'74-I99.

Brenowitz, E.A. (199la) Evolution of the vocalconffol system in the avian brain. Neurosci-ences,3.' 399-407.

Brenowitz, E.A. (1991b) Altered perception ofspecies-specific song by female birds afterlesions of a forebrain nucleus. Science, 251:303 - 305.

Brooks, R.J., and J.B. Falls (1975) Individualrecognition by song in white-throated spar-rows. L Discrimination of songs of neighborsand strangers. Can. J. Zool., 5 3 : 879-888.

Can, C.E. (1992) Evolution of the cantral auditorysystem in reptiles and birds. 1n The Evolution-ary Biology of Hearing (ed by D.B. Webster,R.R. Fay. and A.N. Popper). Springer Verlag.New York, pp. 511-544.

DeVoogd, T.J., J.R. Krebs, S.D. Heady, and A.Purvis (1993) Relations between song reper-toire size and the volume ofbrain nuclei relatedto song: comparative evolutionary analysesamong oscine birds. Proc. Royal Soc. Lond.,Series B: Biol. Sci., 2 54 :'7 5 -82.

Doupe, A.J., and M. Konishi (1991) Song-selectiveauditory circuits in the vocal control system ofthe zebra finch. Proc. Natl. Acad. Sci. USA,88.'11339-t t343.

Durand, S.E., J.M. Tepper, and M.-F. Cheng(1992) The shell region of the nucleus ovoid-alis: A subdivision of the avian auditory thala-mus. J. Comp. Neurol., -?2.3: 495-518.

Falls, J.B. (1982) Individual recognition by soundin birds. 1r Acoustic Communication in Birds,2. Song Leaming and Its Consequences {ed byD.E. Kroodsma, and E.H. Miller), AcademicPress, NY, pp. 237 -278.

Falls, J.B. (1985) Song matching in westem mea-dowlarks. Can. I. Zool., 6 3 : 252O - 2524.

Falls, J.B., J.R. Krebs, and P.K. McGregor (1982)Song matching in the great tit (Parus major):the effect of similarity and familiarity. Anim.Behav., -10: 997 *1009.

Fortune, E.S., and D. Margoliash (1992) Cytoar-chitectonic organization and morphology ofcells of the field L complex in male zebrafinches (Theniopygia guttata). J. Comp. Neu-ro1. ,325' .388 -404.

Fortune, E.S., and D. Margoliash (1994) Parallelpathways and convergence onto HVc and adja-cent neostriatum of adult male zebra finches(Taeniopygia guttata). J. Comp. Neurol., inpress.

Fuzessery, 2.M., and A.S. Feng (1982) Frequencyselectivity in the aunuran auditory midbrain:single unit responses to single and multipletone stimulation. J. Como. Phvsiol.. 146:411-484.

Gamlin, P.D.R., and D.H. Cohen (1986) A secondascending visual pathway from the optic tec-tum to the telencephalon in the pigeon (Co-lumba livia). J. Comp. Neurol.,250:296-310.

Hall, W.S., P.L. Cohen, and S.E. Brauth (1993)Auditory projections to the anterior telenceph-alon in the budgerigar (Nelopsittacus undula-lus). Brain Behav.Evol.,4l : 97 -116.

Hardin, B.D., and D. Margoliash (1992) Effects onsongs of lesions in HVc. Proc. Third Int. Cong.Neuroethol., 344.

HZiusler, U.H.L. (1988) Topography of the thala-motelencephalic projections in the auditorysystem of a songbird (Sturnus vulgaris). InAuditory Pathways, Structure, and Function(ed. by J. Syka, and R.B. Masterton), PlenumPress, New York,pp. 197 -202.

Heil, P., and H. Scheich (1985) Quantitative analy-sis and two-dimensional reconstruction of thetonotopic organization ofthe auditory field L inthe chick from 2-deoxyglucose data. Exp.Brain Res., 58: 532-543.

Heil, P., and H. Scheich (1991) Functional organ-ization of the avian auditory cortex analogue. I.Topographic representation of isointensitybandwidth. Brain Res.,539: 1 10- 120.

Hurly, T.A., R.G. Weisman, L. Ratcliffe, and I.S.Johnsrude (1991) Absolute and relative pitchproduction in the song of the white-throatedsparrow ( Zo no tr ic hia albicolli s ). Bioacoustics,3 : 8 1 - 9 1 .

Kaften, H.J. (1968) The ascending auditory path-way in the pigeon (Columba livia). II. Telence-phalic projections of the nucleus ovoidalis tha-lami. Brain Res.,I I : 134-I53.

Karten, H.J., and W. Hodos (1967) A StereotaxicAtlas of the Brain of the Pigeon (Columba/ivia). Johns Hopkins Press, Baltimore.

Karten, H.J., and W. Hodos (1970) Telencephalicprojections of the nucleus rotundus in thepigeon (Columba livia). J. Comp. Neurol.,140:35-52.

Katz, L.C., and M.E. Gumey (i981) Auditoryresponses in the zebra finch's motor system forsong. Brain Res.,21 I : 192-197.

Kelley, D.B., and F. Nottebohm (1979) Projectionsofatelencephalicauditorynucleus FieldL -

in the canary. J. Comp. Neurol.,183:455-470.Kitt, C.A., and S.E. Brauth (1982) A paleostriatal-

thalamic-telencephalic path in pigeons. Neuro-science, 7 : 21 3 | -2'7 35.

Margoliash/Fortune/Sutter/YuAVren-Hardin/Dave

Organization of Song System:Functions and Origins

262

Korzeniewska, E., and O. Giinttirkiin (1990) Sen-sory properties and afferents of the N. dorsola-teralis posterior thalami of the pigeon. J.Comp. Neurol., 292: 457 -4'79.

Konishi, M. (1965) The role of auditory feedbackin the conffol of vocalization in the white-crowned sparrow. Z. Tierpsychol., 22:'770-783.

Konishi, M. (1991) Deciphering the brain's codes.Neu ra lComp . ,3 :1 -18 .

Kroodsma, D.8., and M. Konishi (1991) A subos-cine bird (eastem phoebe, Sayornis phoebe)develops normal song without auditory-feed-back. Anim. Behav.,42 : 4'77 -481 .

Kroodsma, D.8., and E.H. Miiler (1982) AcousticCommunication in Birds. Academic Press,New York.

Kuwabara, N., and N. Suga (1993) Delay lines andamplitude selectivity are created in subtha-lamic auditory nuclei: the brachium ofthe infe-rior colliculus of the mustached bat. J. Neuro-physio l . , 69 : 17 13 -1724.

Langner, G., D. Bonke, and H. Scheich (1981)Neuronal discrimination of natural and syn-thetic vowels in field L of trained mynah birds.Exp. Brain Res., 43: ll-24.

Lemmon, R.E. (1973) Nervous conhol of thesyrinx in white-throated sparrows (Zo notrichiaalbicollis). J. Zool. (London),171 : L3l-140.

Lemmon, R.8., and M. Hanis (1974) The questionof dialects in the songs of white-throated spar-rows. Can. J. Zool., 52 : 83 -98.

Leppelsack, H.-J. (1983) Analysis of song in theauditory pathway ofsong birds. ln Advances inVeftebrate Neuroethology (ed. by J.P. Ewert,R.R. Capranica, and D.J. Ingle), Plenum, N.Y.,pp.783-799.

Leppelsack, H.-J., and M. Vogt (1976) Responsesto auditory neurons in the forebrain of a song-bird to stimulation with species-specificsounds. J. Comp. Physiol., 107: 263-27 4.

Lewicki, M.S., and A.J. Doupe (1993) Synapticactivity of neurons in zebra finch song nucleusHVc in response to auditory stimuli. Soc. Neu-rosci . Abstr . , .19: 1016.

Liberman, A.M., F.S. Cooper, D.P. Shankweiler,and M. Studdert-Kennedy (1967) Perceptionof the speech code. Psychol. Rev., 74:431 -46 t .

Liberman, A.M., and I.G. Mattingly (1985) Themotor theory of speech perception revised.Cognition, 21: I -36.

Lim, D. (1993) Auditory response properties offield L in the zebra finch. Soc. Neurosci. Abstr.,19 : 415 .

Margoliash, D. (1983) Acoustic parameters under-lying the responses of song-specific neurons inthe white-crowned sparrow. J. Neurosci., 3:1039- 1057.

Margoliash, D. (1986) Preference for autogenoussong by auditory neurons in a song systemnucleus ofthe white-crowned sparrow. J. Neu-rosci.,6; 1643-1661.

Margoliash, D. (1987) Neural plasticity in bird-song leaming. In Imprinting and cortical plas-ticity, (ed. by J.P. Rauschecker, and P. Marier),John Wiley and Sons, New York.

Margoliash, D., and E.S. Fortune (1992) Temporaland harmonic combination-sensitive neuronsin the zebra finch's HVc. J. Neurosci., 12:4309-4326.

Margoliash, D., and M. Konishi (1985) Auditoryrepresentation of autogenous song in the song-system of white-crowned sparrows. Proc. Natl.Acad. Sci. USA, 82: 5997-6000.

Marler, P., and V. Sherman (1982) Structure insparrow song without auditory feedback: Anemendation of the auditory template hypothe-s is. J . Neurosci . , 3 : 517 -531.

McArthur, P.D. (1986) Similarity of playbacksongs to self song as a determinant of responsestrength in song sparrows (Melospiza melo-dia). Anim.Behav.. 34: 199-207 .

McCasland, J.S. (1987) Neuronal control of birdsong production. J. Neurosci., 7 : 23-39.

McCasland, J.S., and M. Konishi (1981) Interac-tion between auditory and motor activities inan avian song control nucleus. Proc. Natl.Acad. Sci . USA, 78:7815-7819.

Mcgregor, P.K., and J.B. Falls (1984) The responseof westem meadowlarks (Sturnella neglecta)to the playback of undegraded and degradedsongs. Can. J. Zool., 62 : 2125 -2128.

McGregor, P.K., and J.R. Krebs ( 1984) Sound deg-radation as a distance cue in great tit song.Behav. Ecol. Sociobiol., I 6 : 49 - 56.

McGregor, P.K., J.R. Krebs, and L.M. Ratcliffe( I 983) The reaction of great tits (Parus major)to playback of degraded und undegradedsongs: the effect of familiarity with the stimu-lus song type. Auk,.100: 898-906.

McKenna, T., N.M. Weinberger, and D.M. Dia-mond (1 989) Responses of single auditory cor-tical neurons to tone sequences. Brain Res.,481: 142-153.

Mello, C.V., D.S. Vicario, and D.F. Clayton (1992)Song presentation induces gene expression inthe songbird forebrain. Proc. Natl. Acad. Sci.USA,89; 6818-6822.

Mooney, R. ( 1992) Synaptic basis for developmen-tal plasticity in a birdsong nucleus. J. Neu-rosci . ,12: 2464-2417.

Mori, K., and G. Striedter (1992) Neurons in field Lof budgerigars prefer species specific calls towhite noise. Soc. Neurosci. Abstr., I 8 : 223.

Morrison, R.G., and F. Nottebohm ( 1993) Role of atelencephalic nucleus in the delayed songleaming of socially isolated zebra finches. J.Neurobiol., 24: 1045- 1064.

Morton, E.S. (1982) Grading, discreteness, redun-dancy, and motivation-structural rules. 1nAcoustic Communication in Birds, Vol. 1, Pro-duction, Perception, and Design Features ofSounds (ed. by D.E. Kroodsma, and E.H.Miller), Academic Press, N.Y., pp. 183-213.

Miiller, C.M., and H.-J. Leppelsack (1985) Featureextraction and tonotopic organization in theavian auditory forebrain. Exp. Brain Res.,59:587-599.

Miiller, S.C., and H. Scheich (1985) Functionalorganization of the avian auditory field L. J.Comp. Physiol., 156: l-12.

Nordeen, K.W., and E.J. Nordeen (1992) Auditoryfeedback is necessary for the maintenance ofstereotyped song in adult zebra finches. Behav.Neural Biol., 57: 58-66.

Nottebohm, F. (1912) The origins of vocal leam-ing. Amer. Nat. , 106: 116- 140.

Nottebohm, F. (1980) Brain pathways for vocalleaming in birds: a review of the first 10 years.Prog. Psychobiol. Physiol. Psychol., 9:85-r24.

Nottebohm, F. (1987) Birdsong. In Encyclopediaof Neuroscience I (ed. by G. Adelman), Birk-hiiuser Boston, Inc., Boston, pp. 133-136.

Noftebohm, F. ( 1991) Reassessing the mechanismsand origins of vocal leaming in birds. TrendsNeurosci., 14 : 206-211.

Nottebohm, F., A. Alvarez-Buylla, J. Cynx, J.Kirn, C.-Y. Ling, M. Nottebohm, R. Suter, A.Tolles, and H. Williams (1990) Song leamingin birds: the relation between perception andproduction. Phil. Trans. R. Soc. Lond. B,329:115-t24.

Nottebohm, F., D.B. Kelley, and J.A. Paton (1982)Connections of vocal control nuclei in thecanary telencephalon. J. Comp. Neurol.,207:344-351.

Nottebohm, F., T.M. Stokes, and C.M. Leonard(1976) Central control of song in the canary,Serinus canarias. J. Comp. Neurol., 165:457 -486.

Okuhata, S., and F. Norrebohm (1992) Single unitsin nucleus Uva respond to sound and respira-tion and are part of a motor loop in song pro-duction. Proc. Third Int. Cong. Neuroethol.,60.

Okuhata, S., and N. Saito (1987) Synaptic connec-tions of thalamo-cerebral vocal nuclei of thecanary. Brain Res. Bull., 18: 35-44.

Olsen, J.F. ( 1994) Medial geniculate neurons in thesquinel monkey sensitive to intercomponentdelays that categorize species-typical calls.Assoc. Res. Otolaryngol. Abstr., l 7 : 21.

Paton, J.A., K.R. Manogue, and F. Nottebohm(1981) Bilateral organization ofthe vocal con-trol pathway in the budgerigar, Melopsittacusundulatus. J. Neurosci., l: 1279-1288.

Price, P.H. (1979) Developmental determinants ofstructure in zebra finch song. J. Comp. Physiol.Psychol., 9J: 268-277 .

Rtibsamen, R., and G.J. Dtjrrscheidt (1986) Tono-topic organization ofthe auditory forebrain in asongbird, the European starling. J. Comp. Phy-siol. A,158; 639-646.

Saini, K.D., and H.-J. Leppelsack (1981) Cell typesof the auditory caudomedial neostriatum of thestarling (Sturnus vulgaris). J. Comp. Neurol.,198:209-23O.

Schafer, M., R. Riibsamen, G. Dtinscheidt, and M.Knipschild (1992) Setting complex tasks tosingle units in the avian auditory forebrain. II.Do we really need natural stimuli to describeneuronal response characteristics? Hear. Res.,57: 231-244.

Scharff, C., and F. Nottebohm (1991) A compara-tive study of the behavioral deficits followinglesions of various parts of the zebra finch songsystem: implications for vocal leaming. J. Neu-rosci.. 11: 2896-2913.

263

Scheich, H., G. Langner, and D. Bonke (1979)Responsiveness of units in the auditory neo-striatum of the Guinea fowl (Numida melea-

.gris) to species-specific cal1s and synthetic sti-muli. II. Discrimination of iambus-like calls. J.Comp. Physiol., I 32 : 257 -27 6.

Shimizu, T., and H.J. Karten (1991a) Cenffal visualpathways in reptiles and birds: evolution of thevisual system. 1n Vision and Visual Dysfunc-tion, 2. (ed. by R. Gregory, and J.R. Cronly-Dillon). Macmillan Press. New York.

Shimizu, 2., andH.J. Karten (199lb) Computatio-nal significance of lamination of the telence-phalon. 1n Visual Structures and IntegratedFunctions (ed. by M.A. Arbib, and J.-P.Ewert), Springer Verlag, Berlin, pp. 325-337.

Sohrabji, R., E.J. Nordeen, and K.W. Nordeen(1990) Selective impairment of song learningfollowing lesions of a forebrain nucleus in thejuvenile zebra linch. Behav. Neural Biol.,5J:51-63.

Stokes, T.M., C.M. Leonard, and F. Nottebohm(1974) The telencephalon, diencephalon, andmesencephalon ofthe canary, Serinus canaria,in stereotaxic coordinates. J. Comp. Neurol.,156:337 -374.

Striedter, G.F. (1994) The vocal control pathwaysin budgerigars diff'er from those in songbirds.J. Comp. Neurol., 34.1; 35-56.

Stripling, R., S.F. Volman, and D.F. Clayton( 1994) Electrophysiological responses to songpresentation in caudiomedial neostriatum ofzebra finches: links to ZENK gene induction.Soc. Neurosci. Abstr..20. in press.

Suga, N., W.E. O'Neill, and T. Manabe (1978)Cortical neurons sensitive to combinations ofinformation-bearing elements of biosonar sig-nals in the mustached bat. Science, 200:7'78-78r .

Sutter, M.L., and D. Margoliash (1994) Globalsynchronous response to autogenous song inthe zebra finch HVc. J. Neurophysiol., in press.

Sutter, M.L., and C.E. Schreiner (1991) Physiol-ogy and topography of neurons with multi-peaked tuning curves in cat primary auditorycortex. J. Neurophysiol.,65(5): 120'l -1226.

Ulinski, P.S. (1982) Dorsal Ventricular Ridge. ATreatise on Forebrain Organization in Reptilesand Birds. John Wiley and Sons, New York.

Ulinski, P.S., and D. Margoliash ( 1990) Neurobiol-ogy of the reptile-bird transition. 1n CerebralCotex (ed. by E.G. Jones, and A. Peters), Ple-num, New York,pp.2l7 -265.

Vicario, D.S. (1990) Contributions of syringealmuscles to respiration and vocalization in thezebra finch. J. Neurobiol.,22 :63-13.

Vicario, D.S. (1991) Organization of the zebrafinch song control system: II. Functionalorganization of outputs from nucleus robustusarchistriatalis. J. Comp. Neurol., 309:486-494.

Vicario, D.S., and F. Nottebohm (1988) Organiza-tion of the zebra finch song control system: I.Representation of syringeal muscles in thehypoglossal nucleus. J. Comp. Neurol., 271:346-354.

Vicario, D.S., and K.H. Yohay (1993) Song-selec-tive auditory input to a forebrain vocal controlnucleus in the zebra finch. J. Neurobiol.,24;488-50s.

Volman, S. (1993) Development of neural selectiv-ity for birdsong during vocal leaming. J. Neu-rosci., I 3 : 4737 -47 47 .

Vu,8.T., M.E. Mazurek, and Y. Kuo (1993) Hier-archical organization of brain areas mediatingzebra finch learned vocalizations. Soc. Neu-rosci . Abstr . , 19: 1015.

Wi1d, J.M. (1993a) Does the pigeon have a nucleusuvaeformis (Uva), or the songbird a nucleusdorsolateralis posterior thalami, pars caudalis(cDLP)? Soc. Neurosci. Abstr., .19: 41 5.

Wi1d, J.M. (1993b) Descending projections of thesongbird nucleus robustus archistriatalis. J.Comp. Neurol., 3 3 8 : 225 -247.

Wild, J.M. (1993c) The avian nucleus retroambi-gualis: a nucleus for breathing, singing and cal-ling. Brain Res., 606:319-324.

Wild, J.M. (1994) The auditory-vocal-respiratoryaxis in birds. BrainBehav. Evol.,44:192-209.

Wild, J.M., B.J. Frost, and H.J. Karten (1990)Some aspects of the organization of the audi-tory forebrain and midbrain in the pigeon. Soc.Neurosci. Abstr., 16 :'7 15.

Wild, J.M., H.J. Karten, and B.J. Frost (1993) Con-nections of the auditory forebrain in the pigeon(Columba livia). J. Comp. Neurol., 337:32-62.

Williams, H. (1989) Multiple representations andauditory-motor interactions in the avian songsystem. Ann. N.Y. Acad. Sci.,56-?: 148-164.

Williams, H., and F. Nottebohm (1985) Auditoryresponses in avian vocal motor neurons: a

. motor theory for song perception in birds. Sci-ence , 229 : 2'7 9 -282 .

Williams, H., and D. Vicario (1933) Temporal pat-terning of song production: participation ofnucleus uvaeformis of the thalamus. J. Neuro-biol.,24 : 903-912.

Yu, C.-H., and D. Margoliash (1992) Recording ofHVc in singing zebra finches. Proc. Third Int.Cong. Neuroethol., 342.

Yu, A.C.-H., and D. Margoliash (1993) Differ-ences between motor recruitment and auditoryresponse properties in zebra finch HVc neu-rons. Soc. Neurosci. Abstr...19: 1018.

Margoliash/Fortune/Sutter/YuAVren-Hardin/Dave

Organization of Song System:

Functions and Origins

264