Single Axon Analysis of Pulvinocortical

Connections to Several Visual Areas

in the Macaque†

KATHLEEN S. ROCKLAND,1* JON ANDRESEN,1 ROBERT J. COWIE,2,3

AND DAVID LEE ROBINSON2

1Department of Neurology; Division of Behavioral Neurology and Cognitive Neuroscience,University of Iowa, Iowa City, Iowa 52242–1053

2Laboratory of Sensorimotor Research, Section on Visual Behavior, National Eye Institute,Bethesda, Maryland 20892

3Department of Anatomy, College of Medicine, Howard University, Washington, D.C. 20059

ABSTRACTThe pulvinar nucleus is a major source of input to visual cortical areas, but many

important facts are still unknown concerning the organization of pulvinocortical (PC)connections and their possible interactions with other connectional systems. In order toaddress some of these questions, we labeled PC connections by extracellular injections ofbiotinylated dextran amine into the lateral pulvinar of two monkeys, and analyzed 25individual axons in several extrastriate areas by serial section reconstruction. This approachyielded four results: (1) in all extrastriate areas examined (V2, V3, V4, and middle temporalarea [MT]/V5), PC axons consistently have 2–6 multiple, spatially distributed arbors; (2) ineach area, there is a small number of larger caliber axons, possibly originating from asubpopulation of calbindin-positive giant projection neurons in the pulvinar; (3) as previouslyreported by others, most terminations in extrastriate areas are concentrated in layer 3, butthey can occur in other layers (layers 4,5,6, and, occasionally, layer 1) as collaterals of a singleaxon; in addition, (4) the size of individual arbors and of the terminal field as a whole varieswith cortical area. In areas V2 and V3, there is typically a single principal arbor (0.25–0.50mm in diameter) and several smaller arbors. In area V4, the principal arbor is larger (2.0- to2.5-mm-wide), but in area MT/V5, the arbors tend to be smaller (0.15 mm in diameter). Sizedifferences might result from specializations of the target areas, or may be more related to theparticular injection site and how this projects to individual cortical areas.

Feedforward cortical axons, except in area V2, have multiple arbors, but these do notshow any obvious size progression. Thus, in areas V2, V3, and especially V4, PC fields arelarger than those of cortical axons, but in MT/V5 they are smaller. Terminal specializations ofPC connections tend to be larger than those of corticocortical, but the projection foci are lessdense. Further work is necessary to determine the differential interactions within andbetween systems, and how these might result in the complex patterns of suppression andenhancement, postulated as gating mechanisms in cortical attentional effects, or in differentstates of arousal. J. Comp. Neurol. 406:221–250, 1999. Published 1999 Wiley-Liss, Inc.‡

Indexing terms: arbor size; attention; cortical layers; cortical uniformity; lateral pulvinar

†This report continues to use the classical nomenclature of Olszewski(1952) to designate the subdivisions of the pulvinar nucleus. This nomencla-ture is based on architectonic analysis and is not congruent with eitherfunctional (Bender, 1981; Ungerleider et al., 1983, 1984) or neurochemicalborders (Gutierrez et al., 1995; Gutierrez and Cusick, 1997; Stepniewskiand Kaas, 1997). Our results, however, do not directly address the issue ofpulvinar subdivisions, and we consider that the older nomenclatureprovides a convenient standard to which the several more current map-pings can be referred.

Grant sponsor: NIMH; Grant number: MH53598.

*Correspondence to: Kathleen S. Rockland, Ph.D., Department ofNeurology, University of Iowa, 200 Hawkins Drive, Iowa City IA 52242–1053. E-mail: rockland@blue.weeg.uiowa.edu

Received 13 March 1998; Revised 28 October 1998; Accepted 3 November1998

THE JOURNAL OF COMPARATIVE NEUROLOGY 406:221–250 (1999)

PUBLISHED 1999 WILEY-LISS INC. ‡This article is a USgovernment work and, as such, is in the public domain in theUnited States of America.

The visually dominated subdivisions of the pulvinar (thelateral and inferior pulvinar, PL and PI, respectively) arepart of a complex connectional network, which includesthe superior colliculus and several visual cortical areas(Benevento and Standage, 1983; Kaas and Huerta, 1988;Mesulam, 1990; Chalupa, 1991; Colby, 1991; Robinson andCowie, 1997). Cells in the retinotopically organized regionsof PL and PI1 are visually responsive, and have relativelysimilar properties, namely, small receptive fields and shortresponse latencies (reviewed in Robinson and Cowie, 1997).The functional importance of the visual pulvinar, however,has remained elusive. Lesions of the visual pulvinar haveproduced behavioral deficits that are surprisingly mildand hard to characterize. Early studies suggested that thenucleus was involved in the visual control of eye move-ments or the perception of size constancy (Ungerleider etal., 1977). More recent work has proposed that the pulvi-nar is implicated in mechanisms of visual salience orattention (Crick, 1984; Petersen et al., 1987; Robinson andPetersen, 1992). This hypothesis is based in part onpharmacological manipulations such as the injection ofdrugs related to gamma-aminobutyric acid (GABA) intothe pulvinar. These result in selective enhancement ofresponses to visual stimuli that are of behavioral signifi-cance to the organism, or in suppression of those that areirrelevant (reviews in Colby, 1991; Robinson and Petersen,1992; Desimone and Duncan, 1995; Robinson and Cowie,1997).

In considering the functional role of the pulvinar com-plex, the extensive interconnectivity between pulvinar andcortex is likely to be important (Guillery, 1995). Theseconnections have been the subject of a number of investiga-tions, which address questions of general topography andlaminar organization. Tracer experiments have demon-strated that pulvinocortical connections are dense andwidespread to both striate and extrastriate areas, but withsome variability (Benevento and Rezak, 1975, 1976; Ogrenand Hendrickson, 1976, 1977, 1979; Trojanowski andJacobson, 1976, 1977; Wong-Riley, 1977; Lund et al., 1981;Ungerleider et al., 1983, 1984; Cusick et al., 1993; Levitt etal; 1995; Gutierrez and Cusick, 1997). There are variationsin layers of termination. Connections to area V1 are knownto terminate preferentially in layers 1 and 2 (Beneventoand Rezak, 1975; Ogren and Hendrickson, 1977; Wong-Riley, 1977; Curcio and Harting, 1978; Rezak and Ben-evento, 1979). In extrastriate areas, autoradiography stud-ies describe some involvement of layer 1, but the bulk ofterminations are reported to be in layers 3 and 5. Varia-tions are also reported in density, both in pulvinocorticaland corticopulvinar connections. Retrograde tracer experi-ments suggest that pulvinocortical connections to V1 areless dense than those to area V2 (Kennedy and Bullier,1985), and that the connections from area V1 to thepulvinar are less dense than those from area V2 (Levitt etal., 1995). These findings raise the interesting possibilitythat pulvinocortical connections might have other area-specific features, not identifiable by the older anterogradetechniques, such as arbor size and the overall configura-tion of individual axons.

Functionally significant structural features have beendemonstrated in other connectional systems, such as thegeniculocortical (Blasdel and Lund, 1983; Freund et al.,1989). Similarly, the cortical connections from V1 to the

middle temporal area (MT or V5) exhibit structural special-izations such as large caliber axons, large terminations,and bistratified terminations which have not so far beenfound in other corticocortical systems (Rockland and Virga,1990; Rockland, 1992, 1995). With these examples in mind,the present study applies single axon analysis to investi-gate the detailed configuration and possible variability ofconnections from one subdivision of the pulvinar, PL, toseveral cortical visual areas (V1, V2, V3, V4, and V5/MT).An additional goal was to provide a basis for more quanti-tative comparisons between pulvinocortical and corticocor-tical connections, as one approach to understanding howthese systems might interact. Some of these results havebeen presented previously in abstract form (Andresen etal., 1997).

MATERIALS AND METHODS

For the sake of more accurate localization within thepulvinar, injections were preceded by physiological map-ping. This was carried out by using techniques as previ-ously published (Petersen et al., 1985), in accordance withNIH guidelines for animal research and under protocolsapproved by the NIH Institutional Animal Care and UseCommittee. Briefly, monkeys were first trained to fixate alight target in return for a liquid reward. They were thenprepared for sterile surgery (tranquilized with ketaminehydrochloride, 11 mg/kg, intubated, and maintained underhalothane general anesthesia), during which a craniotomywas performed over the general location of the pulvinar(Anterior 1 3.5, Lat. 10.0 mm). A scleral eye coil wasimplanted, and stainless steel recording chamber andheadholder affixed with dental acrylic. One to two weekspostsurgery, the monkey resumed testing sessions in con-junction with physiological recording. Tungsten microelec-trodes were lowered through a stainless steel guidetubewithin a plastic grid which allowed electrode placementsin stereotaxically coordinated 1.0-mm increments (Crist etal., 1988). Pulvinar locations were mapped first in relationto the lateral geniculate nucleus, and then by recordingthe progression of visual field responses. In one animal(B42), a presurgical magnetic resonance imaging (MRI)was obtained in order to confirm the placement of therecording chamber and grid, and to facilitate physiologicallocalizations.

Mapping sessions were carried out in two female rhesusmonkeys, 9.4 kg (S43) and 9.5 kg (B42), with the goal ofidentifying the location and extent of the visual fieldrepresentation in the region of the lateral and inferiorpulvinar. Once the target was physiologically identified, itwas reapproached with the needle tip of a modified Hamil-ton syringe (fitted with a recording wire), which had beenfilled with 16% biotinylated dextran amine (BDA; in0.01 M phosphate-buffered saline, pH 7.0). The needle tipwas lowered to the predetermined depth within the pulvi-nar, visual responses again confirmed, and 0.6–1.0 µl oftracer was pressure ejected over 5–10 minutes. Twenty-one (S43) or 23 (B42) days postinjection, the animals weretranquilized with ketamine, deeply anesthetized with alethal dose of Nembutal (75 mg/kg), and perfused transcar-dially with 4% paraformaldehyde. Brains were blocked inthe coronal plane, removed, and processed histologicallyfor BDA as described previously (Rockland, 1996).

222 K.S. ROCKLAND ET AL.

Fig. 1. Outlines of coronal sections, at four anteroposterior (AP)levels, depicting the location and extent of injection sites in case S43(A) and B42 (B). Shading depicts needle track and ‘‘halo’’ zonesurrounding the smaller dense region of actual uptake. The sections inB42 have been reversed to facilitate comparison with those from S43.The short dashed line in each section represents an approximation ofthe horizontal meridian as recorded immediately prior to the injectionof tracer. The plus and minus symbols indicate the respective locations

of upper and lower visual field representations. The numbers belowthe section outlines denote individual tissue sections, and largernumbers are progressively anterior. BrSC, brachium of the superiorcolliculus; LG, lateral geniculate; MG, medial geniculate; Pdm, dorso-medial pulvinar; PI, inferior pulvinar; PL, lateral pulvinar; PM,medial pulvinar; PT, pretectum; R, reticular nucleus of the thalamus;SC, superior colliculus.

AXON ANALYSIS OF PULVINOCORTICAL CONNECTIONS 223

Fig. 2. A, B: Outlines of coronal sections from cases S43 (A) andB42 (B) to depict the location of pulvinocortical foci (asterisks;snowflake symbol denotes lower density). In both cases, larger sectionnumbers are progressively more anterior. Short lines indicate theborder between V1 and V2. C: The projection focus in the IOS of S43was charted through 100 serial sections and represented as a flattenedview (dashed lines signify a gap of about 0.5 mm of tissue lost in

cutting). The projections form 2–3 spatially discrete clusters, whichmerge into stripes with an anteroposterior orientation along thelateral bank. (Clusters are shown to scale as ovals within the stripes.)CF, calcarine fissure; IOS, inferior occipital sulcus; IPS, intraparietalsulcus; LS, lunate sulcus; STS, superior temporal sulcus. Scale bar 51 mm in C.

224 K.S. ROCKLAND ET AL.

The location of intrapulvinar injection sites was verifiedby reference to published maps (Bender, 1981; Petersen etal., 1985); and electrode and injection tracks were ana-

lyzed histologically and compared to records of pre-injection recording sites. The locations of pulvino-cortical foci and individual arbors were established withreference to published maps of extrastriate visual areas(Zeki, 1978; Van Essen, 1985; Felleman and Van Essen,1991; Hof and Morrison, 1995; Felleman et al., 1997).Of these, area V2 could be confidently identified, in theupper bank of the inferior occipital sulcus (IOS) and theposterior bank of the lunate sulcus (LS), by its relationshipto the border with V1, and the area V4 and MT/V5 complexcould also be identified in relation to sulcal patterns.(Possible subdivisions or satellite areas, such as superiortemporal areas located medially [MST] and at the fundus[FST], would not be identifiable.) Area identities were harderto assign for small foci anterior to V2 in the IOS, LS, andannectent gyrus, especially as there is still disagreement as toboth the nomenclature and interpretation of the topographicmaps (Krubitzer and Kaas, 1993; Felleman et al., 1997).

Fig. 3. Photomicrographs of dense biotinylated dextran amine(BDA)-labeled projections. A and (higher magnification) B: Pulvinocor-tical projection in the inferior occipital sulcus (IOS), primarily layer 3(section S-261). Hollow arrows indicate two clusters of moderate andhigher density. Inset: Example of occasional darker and larger

terminations. C and (higher magnification) D: For comparison, cortico-cortical projections from area V1 to V2 in the lunate sulcus. Hollowarrows in C indicate separate clusters within the focus. L., layer. Scalebars 5 100 µm in A, C; 20 µm in B, D, and inset.

Fig. 4. Dot cluster diagrams of number of boutons per counting box(63 3 63 3 50µm). A: Dense field of pulvinocortical terminations inlayer 3 (from S-288, lunate sulcus [LS]: 157 boutons; cf. Fig. 3B).B: Dense field of corticocortical terminations in layer 4 (case I11 -section 167; 236 boutons in area V2 of the LS; see Fig. 3D). C: Sparsefield of pulvinocortical terminations from a single arbor, in layer 3, inthe LS (S-283: 30 boutons; Fig. 6).

AXON ANALYSIS OF PULVINOCORTICAL CONNECTIONS 225

Analysis

Tissue was first analyzed by scanning at 1003 and 2003to identify the location and extent of cortical projectionfoci. During this process, individual arbors were notedfor further serial section analysis. Axon reconstruc-tions were achieved by a camera lucida microscope at-tachment. Parallel sets of drawings were made at1003 and 4003, and ambiguous or intricate points werefurther verified at 1,0003 magnification. Corticallayers were usually identifiable by the presence of largepyramids in layers 3 and 5. These were visible from somemeasure of background staining, and occasional sec-tions were counterstained with thionin as further confirma-tion.

In B42, and to a lesser extent in S43, sporadic retrogradefilling of corticopulvinar neurons occurred, mainly in lay-ers 5 or 6 of the lunate sulcus. Labeling in all cases wasconfined to the cell body. Because of this and because of thesmall number of filled neurons, we think it unlikelythat there was any collateral filling. The possibility thatour results were contaminated by collateral filling isfurther unlikely because most of our reconstructions werecarried extensively into the white matter toward thepulvinar.

For two axons, three-dimensional reconstructions ofcamera lucida drawn axons were carried out, as follows: aphotographic negative of the master camera lucida draw-ing was scanned via Imapro QCS 3200 flatbed scanner(Imapro Corp., Ottawa, Ontario), using an Adobe Photo-shop plug-in module (Adobe Systems Inc., Mountain View,CA) running on a Macintosh 7100/80AV computer (AppleComputer, Inc., Cupertino, CA). Two-dimensional regionsof interest (ROIs) of the axons were created using VTracesoftware (Image Analysis Facility, University of Iowa)running on Silicon Graphics computer workstations (Sili-con Graphics Inc., Mountain View, CA). Points along theROIs were marked to indicate known depths along theoriginal reconstruction space. Three-dimensional ROIswere created by interpolating the changes in known depthalong the original ROI, and converted to DXF (Autodesk,Inc., San Rafael, CA) format files using ROIto3D software(Image Analysis Facility). Three-dimensional images andmovies were created by extruding surfaces along thethree-dimensional axon paths and then rendering theresults using Alias/Wavefront Studio software (Alias/Wavefront, Toronto, Ontario).

Fig. 5. Photomicrographs of pulvinocortical terminations in layer 1of area V1. Arrows indicate corresponding features in A and (highermagnification) B. Arrowheads in B and C point to terminal specializa-tions. (B is from section S-186; C from S-184). Scale bars 5 100 µm inA; 50 µm in B, C.

Fig. 6. Camera lucida reconstruction through 40 coronal sections(2.0 mm AP) of axon LS-S2, labeled by biotinylated dextran amine(BDA) anterogradely transported from lateral pulvinar (PL). As shownin the low magnification view (inset), the axon has a principal arbor inthe middle layers (asterisk), two secondary arbors (hollow arrows) inthe lower bank of the lunate sulcus (LS), and a collateral in layer 1that partially converges with the principal arbor. One branch (shortdouble lines at 264) is incomplete. Numbers in this and subsequentfigures refer to individual tissue sections (smaller are more posterior).Dashed lines indicate the white matter (WM) border at two AP levels(sections 283 and 292); thicker lines indicate part of the LS at twolevels (sections 294 and 304). The proximal end of the axon, continuingto the injection site, is shown by dashes and an arrow. (The sameconventions are used for all the following axon reconstructions.) Atleft, higher magnification of the individual arbors to show boutondistribution and fine detail. Asterisk, hollow arrows, and large andsmall arrowheads mark corresponding arbors at low and high magnifi-cation. L, layer; PIA, pial surface.

226 K.S. ROCKLAND ET AL.

Figure 6

AXON ANALYSIS OF PULVINOCORTICAL CONNECTIONS 227

Fig. 7. A: Camera lucida reconstruction through 40 coronal sections (2.0 mm AP) of axonLS-S1. The axon has two larger arbors (1 and 2), and a collateral in layer 1 (arrowhead in B).Two branches (?) are incomplete. Dashed line and arrow denote proximal portion, traveling

back to the injection site. B: Higher magnification of the individual arbors. Numbered arrows(in this and subsequent figures) point to corresponding features at lower and highermagnifications. L., layer; WM, white matter; PIA, pial surface.

Fig. 8. Camera lucida reconstruction through 28 coronal sections (1.4 mm AP) of axonLS-B2. A: Higher magnification of terminal portions to show bouton distribution and finerdetails. B: The axon has a principal arbor in the middle layers (asterisk), two secondary arbors

(3 and 5), and three collaterals in layer 1 (1, 2, and 4). Dashed line and arrow denote proximalportion, traveling back to the injection site. L., layer; LS, lunate sulcus; WM, white matter.

Fig. 9. Photomicrographs of pulvinocortical arbors. A and (highermagnification) B: Portion of axon LS-B2 in layer 5 of the lunate sulcus(section 267). C and (higher magnification) D: Portion of arbor in layer

3 and upper layer 4 of the annectent gyrus (case S, section 162).Arrows indicate corresponding features in A and B, and C and D. L.,layer; WM, white matter. Scale bars 5 100 µm in A, C; 30 µm in B,D.

Fig. 10. A, B: Sequential sections (238, 239; axon LS-B1) of an arbor in layer 3 and upper layer 4 of thelunate sulcus. Arrows point to matching segments in the two sections. Arrowhead in B indicates regionshown at higher magnification in C. Scale bars 5 50 µm in A; 30µm in in B; 20µm in C.

230 K.S. ROCKLAND ET AL.

Fig. 11. Camera lucida reconstruction through 27 sections (1.35 mmAP) of axon A-S1, primarily in the annectent gyrus. The axon has a diffuseprincipal arbor (asterisk), possibly with two subcomponents; one of these is

in register with an arbor in layer 5 (hollow arrow). There are two spatiallyseparate secondary arbors (arrowheads) in the depth of the lunate sulcus(LS). Dashed line (at 216) and arrow denote proximal portion of the axon.

AXON ANALYSIS OF PULVINOCORTICAL CONNECTIONS 231

Fig. 12. A: Camera lucida reconstruction through 73 sections (3.65 mm AP) of axonIOS-S1. The axon has a principal arbor (asterisk) in the depth of the inferior occipital sulcus(IOS), in layer 3, and three spatially separate secondary arbors (hollow arrows). B: Higher

magnification of individual arbors to show bouton distribution and finer details. L., layer;WM, white matter; PIA, pial surface. Dashed line and arrow denote proximal portion of theaxon.

RESULTS

Injections sites

As stated in footnote 1, the organization of the pulvinarcomplex is still under investigation, and several ways ofdrawing the subdivision boundaries have been proposedon the basis of visuotopic maps (Bender, 1981) or neuro-chemical distribution and cortical connectivity (Unger-leider et al., 1984; Gutierrez et al., 1995; Gutierrez andCusick, 1997; Stepniewska and Kaas, 1997). According tothe classical nomenclature (Olszewski, 1952) followed inthis report, the injection sites in the two monkeys wereboth primarily at the lateral margin of ventral PL, withslight involvement of the lateral edge of PI in monkey S43(Fig. 1). This corresponds to PIL and/or PIL-S of Gutierrezand Cusick, 1997 (also: Gutierrez et al., 1995), or PL ofStepniewska and Kaas (1997).

In monkey S43, the effective injection site (representingthe merging together of two injections) encompassed avolume of tissue about 2.5 mm in the anteroposterior (AP)plane, 2.0 mm dorsoventral (DV), and 0.5–2.0 mm medio-lateral (ML). The injection in monkey B42 was sited moreanterior, and was smaller (about 2.0 mm AP 3 2.0 mmDV 3 0.5–1.0 mm ML). According to previously publishedmaps and the coordinates mapped during physiologicalrecording sessions, the injections are near the representa-tion of the horizontal meridian. The injection in S43 lieswithin the upper visual quadrant; that in B42 may involvethe lower visual field (Bender, 1981; Ungerleider et al.,1983; Peterson et al., 1985; Gutierrez and Cusick., 1997).

Projection foci

The location of cortical projection foci was overall similarin both cases, and involved several extrastriate areas(Fig. 2). Projections were denser in case S43, probablybecause of the larger size and more medial spread of theinjection, and the following description is based mainly onthis case.

The largest and densest projection was to area V2 in theupper bank of the IOS and the posterior bank of the LS(Fig. 2). Projections to the IOS extended over 5.0 mm APand 2.0 mm DV, along the lateral wall of the sulcus. Atsome levels, the projection formed two or three spatiallydistinct clusters, each about 0.2–0.3 mm in diameter; atother levels, these seemed to merge into one larger patch,1.0–1.2 mm across (Figs. 2C and 3). The band of denseterminations varied in width from 0.35 mm (correspondingto the deep part of layer 3, the upper part of layer 5, andthe intervening layer 4), to 0.20 mm (corresponding to thedeep part of layer 3 and the upper part of layer 4).

Projections to area V2 in the LS were dense, althoughthey extended over a slightly smaller territory than thosein the IOS (about 4.0 mm AP and 1.3–2.3 mm DV, along theposterior wall of the sulcus). As in the IOS, these formedeither separate clusters or a single larger patch.

Secondary foci occurred in both the IOS and LS, spa-tially separate from the main projection field in area V2,and anterior to it. These locations probably correspond toseparate areas, variously termed V3v and V3d, VP and V3,or VP and DM, respectively (Gattass et al., 1981; Krubitzerand Kaas, 1993; Felleman et al., 1997). Dorsally, projec-tions were found both in the annectent gyrus and thedepth of the LS, and these were confluent at more posteriorlevels. These secondary foci are smaller, and more consis-

tently form two or three spatially separate clusters (Figs. 2and 3). These are about 0.2–0.3 mm in width, and extendover about 1.0 mm AP.

There were sparse projections in area V4, dorsally adjoin-ing and within the anterior tip of the IOS, as well asscattered projections within the superior temporal sulcus(STS). Terminations in area V1 occurred over a zonemeasuring about 6.5 mm AP 3 3.0 mm DV, on the dorsolat-eral surface, at a site probably corresponding in part to therepresentation of the horizontal meridian.

Projections in case B42 also occurred in two foci in theLS, probably corresponding to areas V2 and V3 (also calledDM; see discussions in Krubitzer and Kaas, 1993; Fellemanet al., 1997); in the annectent gyrus; in the vicinity of theIOS (but shifted more laterally, to the part of area V2 thatlies between the IOS and the LS); and in the STS (Fig. 2).Projection foci were smaller and sparser than in case S43,perhaps because of differences in the size and exact locationof the two injection sites. This may account as well for theapparent lack of projections to area V1 in case B42.

Terminal density

In order to evaluate the density of the projection foci (Fig.4), numbers of boutons were counted (3 1,000 magnifica-tion) within a standardized 63 3 63 3 50 µm box (‘‘dotcluster routine,’’ Eutecticsy version 5.18). Twenty-five adja-cent fields were counted in case S43 from the primary fociin the IOS and LS. Two sections from each area were chosenthrough the densest portion of the projection focus (asqualitatively determined). Counts were started at themidpoint of the focus (within layer 3), and the counting boxwas moved sequentially and laterally within layer 3 towardthe low-density fringe. In the IOS, the bouton numberranged from 15 to 135. Four fields had # 63 boutons(average 41 6 20 boutons per box), and were considered tobe low density. The average for the remaining higherdensity fields was 105 6 19 boutons per box. In the LS, therange was 34–157 boutons. Six fields had # 73 boutons(average 56 615 boutons per box), and were considered tobe low density. The average for the remaining higherdensity fields was 114 6 20 boutons per box.

Projection foci consist of many convergent arbors. By wayof comparison, four equivalent fields (63 3 63 3 50 µm)were sampled through single pulvinocortical arbors in theIOS and LS. For these, the range was 26–30 boutons(average: 29 6 4 boutons per box). As a further comparison,eleven fields were counted through a dense focus in layer 4of corticocortical projections from area V1 to V2 (in twoother animals, both with BDA injections). The range in oneanimal was 168–225 boutons (average: 199 6 15 boutons),and in the other was 172–236 (average: 205 6 19).

Laminar termination

The laminar pattern of termination was similar in areasV2, V4, and the secondary foci in the IOS and annectentregion.. At the densest portions of a focus, terminationsextended rather uniformly through deep layer 3, all of layer4, and superficial layer 5 (Fig. 3). In our material, there wasno indication of separate bands in layers 3 and 5. As a rule,however, terminations were densest in the adjoining partsof layers 3 and 4, with scattered terminations occurring in

AXON ANALYSIS OF PULVINOCORTICAL CONNECTIONS 233

layer 5 and, at even lower density, in layer 6. The actualband of terminations in layers 3 and 4 was frequentlyunderlain by a field of preterminal axons in layers 5 and 6.Some terminations occurred sporadically in layer 1, particu-larly of area V2, but these were not dense and did not forma distinct band. In area MT/V5, terminations again concen-trated in layer 3, but there was more involvement of layer6. One axon (STS-B1; see below) had terminations only inlayer 6.

Terminal specializations appeared generally the sameacross the several extrastriate foci. Terminations wereboth beaded and stalked, with a preponderance of beadedprofiles. A few axons were larger in caliber and had slightlylarger terminal specializations (Fig. 3). These measuredbetween 1 and 2µm in diameter, and occurred intermingledwithin the general field of terminations, especially in the IOSand LS. Overall, pulvinocortical terminations were largerthan corticocortical (Fig. 3), except in MT/V5. In the graymatter, most axons were about 1.0 µm in width.

In area V1, as reported previously (Benevento andRezak, 1975; Wong-Riley, 1977; Curcio and Harting, 1978;Rezak and Benevento, 1979), terminations were concen-trated in layer 1, where they did form a distinct band.Typically, axon segments coursed parallel to the pia, and in

their terminal portions, were studded with both beadedand stalked terminations (Fig. 5). In individual sections,terminal segments measuring 0.35–0.50 mm in lengthwere common. These appeared to be portions of evenlonger axon segments, with a DV orientation (parallel tothe coronal plane of sectioning). Occasional clusters wereseen in layers 2 and 3, but no terminations were detectablein any of the other layers.

Axon reconstructions

Our sample consists of 25 relatively complete axons thathave been analyzed in extensive reconstructions (continu-ing at least 1.0 mm into the white matter, back toward theinjection site) and 12 partial axons, for which analysis waslimited to single arbors or was not continued into the whitematter. Because of the extensive reconstructions and theconsistency of the sample, we are fairly confident thatthese axons originate specifically from neurons within theinjection site. Some possibility remains, however, thataxons from more medial or ventral parts of the pulvinarmay have been labeled by passing through the injectionsite.

Axon reconstructions of other connectional systems havefrequently succeeded in demonstrating standard and recog-

Fig. 13. A: Camera lucida reconstruction through 73 coronalsections (3.65 mm AP) of axon IOS-S3. The axon has a large, diffuseprincipal arbor (asterisk), possibly with subcomponents (numberedarrows 4–6), in the depth of the inferior occipital sulcus (IOS), and

three spatially separate secondary arbors (numbered arrows 1–3).B: Higher magnification of individual arbors to show bouton distribu-tion and finer details (see also Fig. 25). L., layer; WM, white matter; PIA,pial surface. Dashed line and arrow denote proximal portion of the axon.

234 K.S. ROCKLAND ET AL.

nizable characteristics (geniculocortical: Blasdel and Lund,1983, Freund et al., 1989; corticopulvinar: Rockland, 1996).Some characteristic features could be identified for pulvino-

cortical axons, although these are relatively subtle. Theseaxons tend: (1) to terminate in several different layers, (2)to have several spatially dispersed arbors, and (3) to have

Figure 13 (Continued)

AXON ANALYSIS OF PULVINOCORTICAL CONNECTIONS 235

Fig. 14. A: Camera lucida reconstruction through 50 coronalsections (2.5 mm AP) of axon IOS-S2. The axon has a principal(asterisk and hollow arrow) and secondary arbors (numbered arrows4, 5) in the middle layers, and several collaterals in layer 1 (numbered

arrows 1–3). B: Higher magnification of individual arbors to showbouton distribution and finer details. L., layer; WM, white matter;PIA, pial surface. Solid arrow (367) denotes proximal portion of theaxon.

236 K.S. ROCKLAND ET AL.

terminal specializations that are rather regularly arrayed(Figs. 15, 18). For most of these axons, there is both a‘‘principal’’ arbor, distinguished by its larger size andgreater number of boutons, and several smaller arbors.The principal arbor is overall diffuse or ‘‘weedy,’’ and canalso be viewed as a composite of a small, dense core andseveral sparser outliers. Arbors of axons terminating inthe STS are more similar to each other, and the pattern ofprincipal and secondary arbors may not hold in this field.

Seven axons were analyzed from the LS (4 from caseS43, 3 from case B42). These are from the lower bank, or atthe junction of the lower bank with the depth of the sulcus,and are from area V2 or possibly V3, respectively. Detaileddescriptions are given for four of these axons. (See alsolegends for Figs. 6–8).

Axon LS-S2 (Fig. 6) has a single principal arbor, which isabout 1.2 mm DV 3 1.0 mm AP. This can be viewed asconsisting of a central arbor, concentrated in layer 3 andmeasuring about 0.25 mm in diameter, with several associ-ated terminal ‘‘spokes.’’ These rise toward layer 1 and arespaced about 0.50 mm apart. There are two smaller,secondary arbors, lying primarily in layer 5. These areslightly anterior to the principal arbor, by 0.5–1.0 mm. Thetwo smaller arbors are spaced about 1.5 mm apart, and aredisplaced about 2.0 mm medially from the principal arbor.Finally, there is a small, straight arbor in layer 1. This isabout 0.5 mm long and converges back toward the princi-pal arbor by a collateral branch off one of the secondaryarbors.

Axon LS-S1 (Fig. 7) similarly has three or four spatiallyseparate arbors, with components mainly in layers 3, 4, or5. Some terminations impinge on layer 6, and there is

again a straight segment (1.0 mm long) in layer 1. AxonLS-B2 (Figs. 8 and 9) has a single principal arbor (0.25 mmAP 3 0.65 mm ML) with components in layers 3–5, andthree spatially separate smaller arbors distributed overlayers 3, 4, or 5. There are two straight arbors in layer 1,each about 1.0-mm-long, and about 0.5 mm apart AP.

Axon LS-B1 has one diffuse principal arbor, with compo-nents in layers 3 and 5 (Fig. 10 illustrates the histology ina single section; the camera lucida drawing is not illus-trated). This can be viewed as a single structure, about 1.0mm in diameter, or as two small clusters (around sections254 and 257), associated with two wider spokes (extendingto sections 251 and 277). A smaller arbor (0.25 mm indiameter) is spaced about 1.0 mm posterior and 1.5 mmventral in the sulcus, and there is a second small arborspaced another 2.5 mm distant. No terminations occurredin layer 1.

Axon LS-S4 (not illustrated) had a spherical arbor (0.5mm in diameter) in layers 3, 4, and 5, and a separate,elongated arbor about 0.6 mm long, located 2.0 mmanterior, mainly in layer 3. Another axon (LS-B3; notillustrated) had two arbors in layer 3, spatially separate byabout 0.5 mm. One of these continued into layer 1. Thisaxon was apparently simpler than the other five, but mayhave been incomplete.

One axon (A-S1) was analyzed from the annectent gyrus.This axon (Fig. 11) has the characteristic pattern ofextensive branching and multiple, spatially separate ar-bors. The principal arbor is loosely spread over an areaabout 0.60 3 0.75 mm. (Alternately, this can be seen as tworelatively small clusters, each about 0.25 mm in diameter,in layers 3 and 5, with multiple associated sprigs.) Second-

Fig. 15. Photomicrographs of pulvinocortical terminations fromaxon IOS-S2. A and (progressively higher magnifications from arrows)C, D: Terminations in layer 1 (section 371). Arrowhead in A indicates a

segment of the main trunk. B: Portion of arbor descending from layer 1into layers 3 and 4 (section 364). L., layer. Scale bars 5 50 µm in A– C;20 µm in D.

AXON ANALYSIS OF PULVINOCORTICAL CONNECTIONS 237

ary arbors, in layers 3 and 5, occur in the depth of the LS,about 0.70 mm distant (probably in V3 or DM of Krubitzerand Kaas, 1993). There are no terminations in layer 1.

Nine axons were reconstructed from the IOS, all fromcase S43. Only one of these axons had terminations inlayer 1, but otherwise this group is similar to the axonsdescribed above from the LS; that is, they exhibit a highdegree of branching and form a principal arbor and severalspatially separate smaller arbors, each with terminationsin several layers. Three representative axons are illus-

trated from this group (Figs. 12–14), and briefly describedbelow.

Axon IOS-S1 (Fig. 12) has a single principal arbor, witha dense core measuring about 0.25 mm in diameter (or 0.75mm across, if the sparser outliers are included). This isconcentrated in layer 3. There are three smaller secondaryarbors, separated by intervals of 2.0–2.5 mm. These are allin the deeper layers, 5 or 6.

Axon IOS-S3 (Figs. 13 and 25) is similar, but overalllarger: six arbors are distributed over a volume of 27 mm3.

Fig. 16. A: Camera lucida reconstruction through 88 coronalsections (4.4 mm AP) of axon V4-S1. This axon has a single largeprincipal arbor (asterisk) and a second, smaller arbor lying ventraland posterior in the prelunate gyrus. One branch (short double lines at

408) may be incomplete. Dashed line and arrow (at 392) denote theproximal portion of the axon. B: The principal arbor is redrawn athigher magnification to show bouton distribution and finer details. L.,layer; WM, white matter; PIA, pial surface.

238 K.S. ROCKLAND ET AL.

Figure 16 (Continued)

AXON ANALYSIS OF PULVINOCORTICAL CONNECTIONS 239

Of these, two are relatively large (cluster 2: about 1.0 mmML in layers 3, 4, and 5; and cluster 3: about 1.0 mm).Clusters 4 and 5 are close together, and appear to form asingle, larger principal arbor, 1.5–2.0 mm in diameter.There are two small outlier clusters, with terminations inlayers 3, 4, and 5 (cluster 6) or layers 5 and 6 (cluster 1).

Axon IOS-S2 (Figs. 14 and 15) has a diffuse appearance.The principal arbor extends over 0.5 mm AP 3 1.0 mm ML,with components in layers 3, 4, and 5. There is a smallerarbor (0.45 3 0.25 mm) in layer 3, about 1.5 mm anterior tothe principal arbor, and there are five separate collateralsin layer 1, each between 0.25- and 0.50-mm-long.

Two axons were analyzed from area V4. One of these(Fig. 16) is relatively simple, with two distinct, relativelylarge arbors. The single principal arbor is 2.5 mm AP 3 2.0mm DV. The secondary arbor, located 1.5 mm posterior and1.0 mm ventral to the principal arbor, is about 1.0 mm indiameter. For both of these arbors, most of the termina-tions are in layer 3, although there are sparse collateralsin all layers.

Axon V4–2 had an even larger field, encompassing over5.0 mm AP and 6.0 mm ML (Figs. 17 and 18). This field wasmade up of multiple foci, with some suggestion of a 1.0-mmperiodicity in the AP plane. Of these foci, two arbors wereon the lateral surface of the prelunate gyrus. These wereboth about 0.6-mm-wide, spaced 1.0 mm apart center-to-center. One of these, located ventral and anterior, wasidentifiable as a principal arbor by its larger number ofboutons. Two other foci were located in the anteriormostportion of the IOS, and there was another small, isolatedarbor in the IOS. These arbors had terminations mainly inlayer 3, and secondarily in layer 5, but a few terminationsinvaded other layers.

Six axons were analyzed from the STS (5 from case B42and one from case S43). (It is important to note that thezone involved by our injection sites is only one source ofpulvinocortical projections to MT/V5, which also originatefrom more medial subdivisions of the pulvinar; Unger-leider et al., 1984; Cusick et al., 1993.) Two of these(STS-S1 and STS-B4, Figs. 19–22) are very similar. Theyhave multiple arbors, but these are distinctly small (, 0.20mm wide) and mainly in layer 3. Both axons are relativelyflat in the AP plane. Axon STS-S1, but not STS-B4, has twosmall collaterals in layers 5 and 6. A third axon (STS-B1)also has small arbors, but these are confined to layer 6 (notillustrated). For this group, the size of the individualarbors and of the terminal field as a whole is conspicuouslysmall.

Two other axons (one of which is illustrated in Fig. 23)had two separate arbors, both mainly in layer 3, but thesewere slightly larger than for the axons described above.The terminal fields were relatively flat (about 0.7 mm AP).Another axon of this group (STS-B5, Fig. 24) had twoclusters, widely separated by 4.0 mm DV and 1.0 mm AP.These may be targeting area MT/V5 and one of its satelliteareas, such as FST or MST. The dense center of each arboris about 0.2–0.3 mm in diameter, mainly in layer 3, andthere are smaller collaterals in layers 5 and 6.

Number of boutons

The total number of boutons was counted for 20 pulvino-cortical axons (see Table 1). The two large axons terminat-ing in V4 had the highest number of boutons, and most ofthe axons terminating in the STS had smaller numbers. Asample of arbor volumes was also calculated. For this

purpose, arbors were considered as ellipsoids, and thevolume calculated according to the standard formula(4/3)p abc, where a, b, and c are 1⁄2 lengths of the principalaxes x, y, and z.

Three-dimensional shape

Pulvinocortical axons have a complicated three-dimen-sional (3-D) structure. Multiple arbors of a single axon aredistributed in space and are rarely in the same plane.Moreover, these axons frequently take complicated trajec-tories, with distant branch points converging to the samelocus (Figs. 6, 13, 24). Trajectories can be contorted (Fig. 23),or meandering, even through gray matter. In an attempt todepict spatial relationships in a 3-D environment, a Quick-Time movie of axon IOS-S3 (Fig. 13) has been prepared (ashort version can be viewed at www.medicine.uiowa.edu/RocklandLab/). On the printed page, some 3-D sense canbe conveyed by multiple views of an axon rotated throughseveral angles. Selected views are presented for axonIOS-S3 (Fig. 25). This axon is reproduced together withSTS-B1, in order to depict the dramatic difference in scalebetween their two terminal fields.

DISCUSSION

Using the technique of single axon reconstruction, wereport four observations about the morphological organiza-tion of pulvinocortical connections; namely, pulvinocorticalaxons in extrastriate areas have multiple, spatially sepa-rate arbors; the arbors of a single axon are distributed todifferent layers; individual arbors and terminal fields as awhole vary in size in different visual areas; and someaxons, with a distinctly larger caliber, may constitute aseparate subpopulation. These findings and their broaderimplications are discussed below.

Multiplicity of arbors

Pulvinocortical axons in extrastriate areas have mul-tiple, spatially separate arbors. This is consistent withprevious results from experiments using double retro-grade tracers. In these experiments, cortical injections oftwo tracers spaced up to 1.5 mm apart produced a largenumber (about 20%) of double-labeled thalamocorticalneurons (Kennedy and Bullier, 1985; Steriade et al., 1997),as would be expected if single axons arborized over largedistances. This feature is concordant with a general ten-dency for cortical inputs to have multiple arbors, andsuggests relevant comparisons with other thalamocorticaland corticocortical connections.

Terminal fields have been demonstrated in the macaque,at the single axon level, for thalamocortical connections toprimary visual (Blasdel and Lund, 1983; Freund et al.,1989), auditory (Hashikawa et at., 1995), and somatosen-sory cortices (Garraghty and Sur, 1990; Shinoda et al.,1993; see review in Steriade et al., 1997). In layer 4,parvicellular and slowly adapting axons (in visual andsomatosensory areas, respectively) are reported to termi-nate in single arbors, whereas magnocellular and rapidlyadapting axons typically have multiple arbors. A recentstudy in the rat of thalamocortical connections from theventrolateral and ventral posterolateral nuclei describesaxons with multiple arbors that diverge over a widecortical territory (Aumann et al., 1998).

Another comparison is with corticocortical connectionsto extrastriate areas. These, as shown by morphologicalstudies, exhibit a range in the number and size of their

240 K.S. ROCKLAND ET AL.

Fig. 17. Camera lucida reconstruction through 100 coronal sec-tions (5.0 mm AP) of axon V4-S2. There is a large principal arbor(asterisk) and a smaller, secondary arbor, both on the prelunate gyrus.The smaller arbor is dorsal and drawn in red, as is the pial surface(PIA) for the most closely corresponding level. Ten other arbors can be

distinguished (hollow arrows). Three of these (drawn in black, red, andblue at the right) appear to overlap, but are actually offset in the APaxis, as can also be inferred from the section outlines (Fig. 2). WM,white matter. Dashed line and solid arrow (377) denote proximalportion of the axon.

Fig. 18. Photomicrographs of pulvinocortical terminations in areaV4. A and (higher magnification) B: Portion of axon V4-S2 in layer 5(from section 374; Fig. 17). Arrows indicate equivalent points in A andB. Arrowheads in B point to regions illustrated at higher magnification

in the two insets. C: Portion of axon V4-S2, in layer 3 of the inferioroccipital sulcus (s. 404; Fig. 17). Note the semiregular spacing ofterminal specializations. L., layer. Scale bars 5 100 µm in A; 50 µm inB, C; 10 µm in inset (for both).

AXON ANALYSIS OF PULVINOCORTICAL CONNECTIONS 241

terminal arbors. Many systems, V2 to V4 (Rockland,1992), V1 to MT/V5 (Rockland, 1995), and V2 to MT/V5(Rockland, 1995), consistently have two to four arbors, andin this respect are roughly similar to pulvinocorticalconnections. An exception is the axons projecting fromarea V1 to V2, at least 50% of which have single arbors(Rockland and Virga, 1990). The reason for the apparentdifference between pulvinocortical and corticocortical ter-minations in V2 is not clear. One possibility is that the twosystems are differentially related to the functional architec-ture. Bulk labeling of the pulvinocortical connections toV2, for example, indicates that these may align with thecytochrome-oxidase-rich (CO) stripes (Livingstone andHubel, 1982; Levitt et al., 1995). Divergent corticocorticalconnections from V1 to V2 have also been inferred to alignwith the stripelike pattern (Rockland and Virga, 1990), butthe actual relationship is still unknown.

Because pulvinocortical terminations are concentratedmore in layer 3 than layer 4, it may be that these map moreclosely onto the system of patchy intrinsic connections,

which are concentrated in the same layer. These latterhave been reported to map largely orthogonal to the COstripes in V2 (Rockland, 1985; Malach et al., 1994; Levitt elal., 1995), and an interesting possibility is that there maybe a 3-way mapping of pulvinocortical, corticocortical, andintrinsic connections in area V2 and other extrastriateareas.

Laminar termination pattern

The present study confirms previous work, using intrap-ulvinar injections of 3H-amino acids or wheat germagglutinin-horseradish peroxidase (WGA-HRP), that ter-minations to extrastriate cortex target mainly layers 3and 5 (Benevento and Rezak, 1976; Curcio and Harting,1978), or the adjoining portions of layers 3 and 4 (Lund etal., 1981; Levitt et al., 1995). In our material, the densestfoci in extrastriate areas formed a band including deeplayer 3, layer 4, and the upper part of layer 5. At lowerdensities, terminations were restricted mainly to deeplayer 3 and the upper part of layer 4. Terminations

Fig. 19. A: Camera lucida reconstruction through 43 coronalsections (2.15 mm AP, of which 1.85 mm are in gray matter) of axonSTS-S1, in the superior temporal sulcus (STS). The axon has threesmall, spatially separate arbors in the supragranular layers, and two

others in the deeper layers (hollow arrows). B: Higher magnification ofindividual arbors.L., layer; WM, white matter; PIA, pial surface. Solidarrow (419) denotes proximal portion of the axon.

242 K.S. ROCKLAND ET AL.

occurred in layer 5 and even in layer 6, but these weresparser and most of the infragranular labeling was foundto be from preterminal axon segments. Terminations alsooccurred in layer 1, as previously described (Beneventoand Rezak, 1976; Ogren and Hendrickson, 1977; Curcioand Harting, 1978; Lund et al., 1981; Levitt et al., 1995);but, except in area V1, these were generally of onlymoderate density.

A novel result of the present study is the observationthat a single axon can have terminal arbors in severaldifferent layers. Commonly these are layers 3, 4, and 5, butoccasionally a single axon can target layers 1 and 6 as well.The arbors in different layers are usually offset spatiallyrather than being in register.

This pattern is highly dissimilar from that of geniculocor-tical axons terminating in area V1. When these havecollaterals, the targets are layers 4 and 6, and the arborstend to be in register at the different depths (Blasdel andLund, 1983). Terminations in layer 1 seem to derive from aseparate subpopulation without collaterals to layer 4.However, terminations from the koniocellular or W-cellpopulation, which terminate in layer 3, do have collateralsto overlying portions of layer 1 (Lachica and Casagrande,1992).

Corticocortical terminations within early visual areasless frequently have collaterals in several layers (anexception is the bistratified termination from area V1 toMT/V5; Rockland, 1995), and these do not typically involve

Fig. 20. Camera lucida reconstruction through 31 coronal sections (1.55 mm AP, of which 0.50 mm arein gray matter) of axon STS-B4. The axon has three small arbors (hollow arrows), all in layer 3 of thesuperior temporal sulcus (STS). These are redrawn at higher magnification below. L., layer; WM, whitematter. Dashed line and solid arrow (393) denote proximal portion of the axon.

AXON ANALYSIS OF PULVINOCORTICAL CONNECTIONS 243

layer 1. The reason for this difference is not clear; but itsuggests that pulvinocortical connections potentially tar-get a broader range of postsynaptic populations. They may,for example, converge directly with feedback connectionsin layer 1, as well as with feedforward and/or intrinsicconnections in the middle layers.

For corticocortical connections, as for the pulvinocorti-cal, arbors in different layers tend not to be in register.Perhaps, in contrast with primary visual cortex, extrastri-ate areas have a greater tendency for cross-columnarinteractions.

Differences in arbor sizeand terminal field size

Our results indicate that there are differences in the sizeof individual pulvinocortical arbors and the size of the

terminal field as a whole, and that these may be area-specific. In particular, pulvinocortical arbors and fields, atleast from the lateral PL, are large in area V4, and small inarea MT/V5, as compared with those terminating in areasV2 or V3. This result was unexpected, as corticocorticalarbors have not exhibited any comparable progression(Rockland and Virga, 1990; Rockland, 1992, 1995), al-though single axon analysis suggests that the number ofarbors may be different across areas. As a consequence, inarea V4, individual arbors and the terminal fields ofpulvinocortical axons are much larger than are corticocor-tical arbors and fields; in V2 and possibly V3 they aresomewhat larger, and in MT/V5, they are smaller.

In areas V1 and V2, it is well-established that there isorganization at multiple spatial scales (Shipp and Zeki,1989; Krubitzer and Kaas, 1990; Malach et al., 1994;

Fig. 21. Photomicrographs of pulvinocortical terminations in thesuperior temporal sulcus. A and C are lower magnifications of B andD, respectively; arrows indicate corresponding features. Region atarrowhead in D is shown at higher magnification in Figure 22. A is

from axon STS-S1 section 394 (Fig. 19); C is from axon STS-B4, section361 (Fig. 20). Numbers at the right of A and C represent cortical layers.Scale bars 5 100 µm in A, C; 30 µm in B, D.

244 K.S. ROCKLAND ET AL.

Yoshioka et al., 1996). Thus, it could be that pulvinocorti-cal connections in layer 3 might be less fine-grained thancorticocortical connections in layer 4. In this case, theapparent difference in arbor size and terminal densityindicated by our results may reflect different packingdensities. Corticocortical connections may preferentiallytarget the small, closely packed pyramids in layer 4, andthe pulvinocortical may target the larger, less closelypacked pyramids in layer 3.

Another possibility is that pulvinocortical connectionsdo not have the same function in all extrastriate areas. Forexample, one might speculate that the large pulvinocorti-cal fields in area V4 are related to the large suppressivesurrounds of the receptive fields of V4 neurons. Much ofthis suppression has been attributed to callosal connec-tions, and is reduced in both strength and extent by sectionof the corpus callosum (Desimone et al., 1993). There is,however, residual suppression, which could be of pulvinarorigin. The small size of pulvinocortical fields in MT/V5may imply that these connections have more focal interac-tions with other inputs in this area. Some caution ininterpretation is necessary, however, as our injections inboth cases were restricted to a small subdivision of thepulvinar complex, and in particular they did not involvethe medial components of PI which also give rise toprojections to MT/V5 (Ungerleider et al., 1984; Cusick etal., 1993). An alternate interpretation, therefore, might bethat axons originating from different subdivisions of thepulvinar have different arbor size and terminal fieldgeometry in a given cortical area (i.e., if they have ‘‘sustain-ing’’ or ‘‘modulatory’’ influences).

Area-specific dimensional differences have been fre-quently reported in the system of horizontal intrinsicconnections (Yoshioka et al., 1992; Amir et al., 1993; Lundet al., 1993; Fujita and Fujita, 1996). The diameter andspacing of terminal patches (but not necessarily of indi-vidual arbors) increase progressively across several of theearly visual areas. There are similar reports of a progres-sive increase in the diameter of the basal dendritic field ofsupragranular pyramidal neurons (Lund et at., 1993;Elston et al., 1996; Elston and Rosa, 1997). This trend hasbeen interpreted as reflecting more global computational

processes in the later stages of a visual hierarchy. This isconsistent with the size increase of pulvinocortical arborsand fields from V2 and V3 to V4, but not with the smallerfields in MT/V5.

Different types of pulvinocorticalconnections?

Pulvinocortical projection neurons are known to be ofseveral types. Earlier Golgi investigations noted a range inthe diameter of their dendritic fields (150–600µm) and cellbodies (15–40µm), and differences in the number of proxi-mal dendrites (5–10) and in the overall geometry of thedendritic trees (Ogren and Hendrickson, 1979; Jones,1985; Lysakowski et al., 1986; Steriade et al., 1997). Morerecently, immunohistochemical studies have described sub-populations of parvalbumin- or calbindin- positive neu-rons. In particular, much of the visual pulvinar, includingthe lateral edge of PL (the ‘‘shell’’ region of PIL as desig-nated by Gutierrez et al., 1995), which largely coincideswith our injection sites, contains a small number of giantpulvinocortical projection neurons (with a 40-µm cell bodyand 600-µm dendritic field) that are positive for calbindinand for SMI32 (Gutierrez et al., 1995; Stepniewska andKaas, 1997; Steriade et al., 1997). An interesting possibil-ity is that this subpopulation of giant projection neurons,possibly associated with faster conduction velocity, givesrise to the small number of larger caliber, darkly stainingpulvinocortical axons in our material.

Reciprocity of pulvinocorticaland corticocortical connections

Connections between the pulvinar and visual corticalareas are known to be reciprocal at the level of areas andsubdivisions (Benevento and Rezak, 1976; Ogren andHendrickson, 1976; Wong-Riley, 1977; Lund et al., 1981;Cusick et al., 1993; Adams et al., 1995; Levitt et al., 1995;and reviews in Jones, 1985; Kaas and Huerta, 1988;Chalupa, 1991; Guillery, 1995; Robinson and Cowie, 1997;Steriade et al., 1997). This result, however, is based onlarge injections, which are not optimal for probing fineorganization, and it is not known how the pulvinar and itscortical targets are interconnected at the level of func-tional units or neurons. Indeed, as noted by Crick andKoch (1998), it is not particularly clear what constitutes a‘‘functional unit’’ within the pulvinar. Given that pulvino-cortical axons have multiple, spatially separated arbors, anatural question is whether reciprocating corticopulvinarconnections originate from all these loci or from just one,perhaps from the locus of the principal arbor? This ques-tion could be addressed by a very small intrapulvinarinjection of retrograde tracer. In the former case, we wouldpredict multiple, spatially separate clusters of corticopulvi-nar neurons; in the latter, if the injection were sufficientlyminute, there is more likely to be only a single cluster ofcells per cortical area.

It is also unknown how strictly reciprocal these connec-tions might be at the neuron-to-neuron level. In principle,there could be a two-neuron, direct corticopulvinocorticalloop: pulvinocortical terminations in the deeper layersmight contact proximal portions of corticopulvinar projec-tion neurons, and those in the more superficial layersmight contact this same group at more distal dendritic

Fig. 22. Higher magnification of Figure 21D (from arrowhead) toshow detail of pulvinocortical terminations; montage in two focalplanes. Scale bar 5 20 µm.

AXON ANALYSIS OF PULVINOCORTICAL CONNECTIONS 245

locations. This issue requires experimental investigationby electron microscopic or other techniques. One ultrastruc-tural study of the synaptic connectivity between the prefrontalcortex and the mediodorsal nucleus of the rat reports thatsome thalamocortical terminations do end directly onto cortico-thalamic neurons, and thus constitute a monosynapticallyrecurrent loop. Some of the terminations to layer 3, however,contacted callosally projecting neurons (Kuroda et al., 1998).Areasonable assumption is that the corticothalamocortical loopmay well include a monosynaptic component, but probably

has additional components whereby the excitation from tha-lamic afferent fibers can influence other pathways.

Functional implications

The pulvinar is one of several structures implicated inattentional processes or salience. The evidence for thisincludes electrophysiological experiments in animals (Pe-tersen at al., 1987; Robinson and Petersen, 1992), andbehavioral (Rafal and Posner, 1987) and functional imag-

Fig. 23. A: Camera lucida reconstruction through 127 coronal sections (6.35 mm AP, of which 0.85 mmare in gray matter) of axon STS-B3. The axon has two delimited arbors in the middle layers of the superiortemporal sulcus. B: Higher magnification of individual arbors. L., layer; WM, white matter. Dashed linesand solid arrow (430) denote proximal portion of the axon.

246 K.S. ROCKLAND ET AL.

Fig. 24. Camera lucida reconstruction through 62 coronal sections (3.1 mm AP) of axon STS-B5. Theaxon has two terminal clusters at the depth of the superior temporal sulcus (STS), at its upper and lowerbends. (Two branches marked by ‘‘?’’ are incomplete.) Terminal portions are redrawn at the margins of thefigure. L., layer; WM, white matter. Dashed line and solid arrow (460) denote proximal portion of the axon.

ing studies (LaBerge and Buchsbaum, 1990; Morris et al.,1997) in humans.

In one model of visual attention and invariant patternrecognition, pulvinocortical neurons are assigned an impor-tant role as ‘‘control neurons’’ that dynamically modify synap-tic strengths along the pathways between V1 and inferotempo-ral cortex (Olshausen et al., 1993). The model proposes severalgating mechanisms whereby control neurons might decreaseor nullify the efficacy of corticocortical synapses, or alternately,under an excitatory gating scheme, might work in concertwith these to boost gain and raise firing thresholds.

Realistic models at this level will require considerablymore experimental data concerning synaptic interactionsof single neurons and populations. However, two of ourfindings may be significant in the context of the proposedmodel. First, at the populational level, the spatially diver-gent pulvinocortical arbors reported in this study couldmediate different combinations of inhibitory-excitatoryinteractions, that is, selective suppression or enhancementof neighboring groups of neurons. Second, at the level ofdendritic microcircuitry, the present study indicates thatpulvinocortical terminations are slightly larger than thecorticocortical (feedforward, feedback, and horizontal in-trinsic) in all areas except MT/V5. This may correlate withphysiologically more effective synapses (if there is a largercontact zone and/or greater number of vesicles), or possiblyfaster conduction velocity. More work is needed on thesynaptic organization of these connections, especially onthe involvement of inhibitory elements.

ACKNOWLEDGMENTS

We thank Mitchell Smith for his expert assistance insurgery, injections, and perfusion; Diane Topinka for herexcellent help with manuscript preparation; Michelle Mac-Donald for photographic assistance; and Steve Beck (Uni-versity of Iowa Image Analysis Facility) for consultationson 3-D representation and software. We thank Dr. DanPollen for his helpful comments on an earlier version of themanuscript.

LITERATURE CITED

Adams MM, Webster MJ, Gattass R, Hof PR, Ungerleider LG. 1995. Inputsto areas V1, V2, and MT from the macaque pulvinar and theirrelationships to chemoarchitectonic subdivisions. Soc Neurosci Abstr21:904.

Amir Y, Harel M, Malach R. 1993. Cortical hierarchy reflected in theorganization of intrinsic connections in macaque monkey visual cortex.J Comp Neurol 334:19–64.

Fig. 25. Computer rotations, to scale, of axons IOS-S3 (Fig. 13) andSTS-S1 (Fig. 19). A: ‘‘Home’’ view, corresponding to camera lucidamaps. B: Ninety degree rotation out of the page. C: Further rotation.Solid arrows point to corresponding branch point in axon IOS-S3;hollow arrows point to corresponding arbor in axon STS-S1. A,anterior, D, dorsal; L, lateral.

TABLE 1. Bouton Number and Arbor Volume

AxonNumberof arbors

Totalnumber ofboutons

Number ofboutons/arbor1

Arborvolume(µm3)1

LS-S2 7 614 201 3.7 3 107

LS-S1 4 453 130 1.1 3 107

LS-B2 6 966 410 8.5 3 107

LS-B1 3 760 246 2.9 3 107

LS-B3 P2 3 282LS-S3 P2 2 181LS-S4 P2 2 428SA-1 4 657 257IOS-S1 3 577 200 5.2 3 106

IOS-S3 8 1,355 384 2.2 3 108

IOS-S2 6 665 192 1.9 3 107

IOS-S4 P2 1 1,394V4-S1 2 2,110 806 2.0 3 108

V4-S2 14 5,213 1,011 4.4 3 108

STS-S1 4 222 93 1.2 3 106

STS-B4 4 232 96 1.9 3 106

STS-B1 2 130 98 4.8 3 106

STS-B3 2 1,106 324 1.1 3 107

STS-B5 6 735 427 6.2 3 107

STS-B2P 1 130

1Principal arbor (or, for middle temporal area [MT]/V5, representative arbor.2P, partially reconstructed axon.

248 K.S. ROCKLAND ET AL.

Andresen J, Rockland KS, Robinson DL, Cowie RJ. 1997. Pulvinocorticalaxons terminating in extrastriate cortex. Neurosci Abstr 23:844.

Aumann TD, Ivanusic J, Horne MK. 1998. Arborization and termination ofsingle motor thalamocortical axons in the rat. J Comp Neurol 396:121–130.

Bender DB.1981. Retinotopic organization of macaque pulvinar. J Neuro-physiol 46:672–693.

Benevento LA, Rezak M. 1975. Extrageniculate projections to layers VI andI of striate cortex (area 17). in the rhesus monkey (Macaca mulatta).Brain Res 96:51–55.

Benevento LA, Rezak M. 1976. The cortical projections of the inferiorpulvinar and adjacent lateral pulvinar in the rhesus monkey (Macacamulatta). An autoradiographic study. Brain Res 108:1–24.

Benevento LA, Standage GP. 1983. The organization of projections of theretino-recipient and nonretino-recipient nuclei of the pretectal complexand layers of the superior colliculus to the lateral pulvinar and medialpulvinar in the macaque monkey. J Comp Neurol 218:307–336.

Blasdel GG, Lund JS. 1983. Termination of afferent axons in macaquestriate cortex. J Neurosci 3:1389–1413.

Chalupa LM. 1991. The visual function of the pulvinar. In: Leventhal AG,editor. The neural basis of visual function. London: Macmillan. p140–159.

Colby CL. 1991. The Neuroanatomy and neurophysiology of attention. JChild Neurol 6:S90–S118.

Crick F. 1984. The function of the thalamic reticular complex: the search-light hypothesis. Proc Natl Acad Sci USA 81:4586–4590.

Crick F, Koch C. 1998. Constraints on cortical and thalamic projections: theno-strong-loops hypothesis. Nature 391:245–250.

Crist CF, Yamasata DSG, Komatsu H, Wurtz RH. 1988. A grid system and amicrosyringe for single cell recording. J Neurosci Meth 26:117–122.

Curcio CA, Harting JK. 1978. Organization of pulvinar afferents to area 18in the squirrel monkey: Evidence for stripes. Brain Res 143:155–161.

Cusick CG, Scripter JL, Darensbourg JG, Weber JT. 1993. Chemoarchitec-tonic subdivisions of the visual pulvinar in monkeys and their connec-tional relations with the middle temporal and rostral dorsolateralvisual areas, MT and DLr. J Comp Neurol 336:1–30.

Desimone R, Duncan J. 1995. Neural mechanisms of selective visualattention. Annu Rev Neurosci 18:193–222.

Desimone R, Moran J, Schein SJ, Mishkin M. 1993. A role for the corpuscallosum in visual area V4 of the macaque. Vis Neurosci 10:159–171.

Elston GN, Rosa MGP. 1997. The occipitotemporal pathway of the macaquemonkey: comparison of pyramidal cell morphology in layer III offunctionally related cortical visual areas. Cereb Cortex 7:432–452.

Elston GN, Rosa MGP, Calford MB. 1996. Comparison of dendritic fields oflayer III pyramidal neurons in striate and extrastriate visual areas ofthe marmoset: a lucifer yellow intracellular injection study. CerebCortex 6:806–813.

Felleman DJ, Van Essen DC. 1991. Distributed hierarchical processing inthe primate cerebral cortex. Cereb Cortex 1:1–47.

Felleman DJ, Burkhalter A, Van Essen DC. 1997. Cortical connections ofareas V3 and VP of macaque monkey extrastriate visual cortex. J CompNeurol 379:21–47.

Freund TF, Martin KAC, Soltesz I, Somogyi P, Whitteridge D. 1989.Arborization pattern and postsynaptic targets of physiologically identi-fied thalamocortical afferents in striate cortex of the macaque monkey. JComp Neurol 289:315–336.

Fujita I, Fujita T. 1996. Intrinsic connections in the macaque inferiortemporal cortex. J Comp Neurol 368:467–486.

Garraghty PE, Sur M. 1990. Morphology of single intracellularly stainedaxons terminating in area 3b of macaque monkeys. J Comp Neurol294:583–593.

Gattass R, Gross CG, Sandell JH. 1981. Visual topography of V2 in themacaque. J Comp Neurol 201:519–539.

Guillery RW. 1995. Anatomical evidence concerning the role of the thala-mus in corticocortical communication: a brief review. J Anat 187:583–592.

Gutierrez C, Cusick CG. 1997. Area V1 in macaque monkeys projects tomultiple histochemically defined subdivisions of the inferior pulvinarcomplex. Brain Res 765:349–356.

Gutierrez C, Yaun A, Cusick CG. 1995. Neurochemical subdivisions of thepulvinar in macaque monkeys. J Comp Neurol 363:545–562.

Hashikawa T, Molinari M, Rausell E, Jones EG. 1995. Patchy and laminarterminations of medial geniculate axons in monkey auditory cortex. JComp Neurol 362:195–208.

Hof PR, Morrison JH. 1995. Neurofilament protein defines regional pat-terns of cortical organization in the macaque monkey visual system: aquantitative immunohistochemical analysis. J Comp Neurol 352:161–186.

Jones EG. 1985. The thalamus. New York: Plenum Press.Kaas JH, Huerta MF. 1988. Subcortical visual system of primates. In:

Steklis HP, editor. Comparative primate biology, vol. 4. New York: Liss.p 327–391.

Kennedy H, Bullier J. 1985. A double-labeling investigation of the afferentconnectivity to cortical areas V1 and V2 of the macaque monkey. JNeurosci 5:2815–2830.

Krubitzer LA, Kaas JH. 1990. Cortical connections of MT in four species ofprimates: Areal, modular, and retinotopic patterns. Vis Neurosci 5:165–204.

Krubitzer LA, Kaas JH. 1993. The dorsomedial visual area of owl monkeys.Connections, myeloarchitecture, and homologues in other primates. JComp Neurol 334:497–528.

Kuroda M, Yokofujita J, Murakami K. 1998. An ultrastructural study of theneural circuit between the prefrontal cortex and the mediodorsalnucleus of the thalamus. Prog Neurobiol 54:417–458.

LaBerge D, Buchsbaum MS. 1990. Positron emission tomographic measure-ments of pulvinar activity during an attention task. J Neurosci 10:613–619.

Lachica EA, Casagrande VA. 1992. Direct W-like geniculate projections tothe cytochrome oxidase (CO) blobs in primate visual cortex: Axonmorphology. J Comp Neurol 319:141–159.

Levitt JB, Yoshioka T, Lund JS. 1995. Connections between the pulvinarcomplex and cytochrome oxidase-defined compartments in visual areaV2 of macaque monkey. Exp Brain Res 104:419–430.

Livingstone MS, Hubel DH. 1982. Thalamic inputs to cytochrome oxidase-rich regions in monkey visual cortex. Proc Natl Acad Sci 79:6098–6101.

Lund JS, Hendrickson AE, Ogren MP, Tobin EA. 1981. Anatomical organiza-tion of primate visual cortical area V-II. J Comp Neurol 202:19–45.

Lund JS, Yoshioka T, Levitt JB. 1993. Comparison of intrinsic connectivityin different areas of macaque monkey cerebral cortex. Cereb Cortex3:148–162.

Lysakowski A, Standage GP, Benevento LA. 1986. Histochemical andarchitectonic differentiation of zones of pretectal and collicular inputsto the pulvinar and dorsal lateral geniculate nuclei in the Macaque. JComp Neurol 250:431–448.

Malach R, Tootell RBH, Malonek D. 1994. Relationship between orientationdomains, cytochrome oxidase stripes, and intrinsic horizontal connec-tions in squirrel monkey area V2. Cereb Cortex 4:151–165.

Mesulam M-M. 1990. Large-scale neurocognitive networks and distributedprocessing for attention, language, and memory. Ann Neurol 28:597–613.

Morris JS, Friston KJ, Dolan RJ. 1997. Neural responses to salient visualstimuli. Proc Roy Soc Lond B 264:769–775.

Ogren MP, Hendrickson A. 1976. Pathways between striate cortex andsubcortical regions in Macaca mulatta and Saimiri sciureus: Evidencefor a reciprocal pulvinar connection. Exp Neurol 53:780–800.

Ogren MP, Hendrickson AE. 1977. The distribution of pulvinar terminals inareas 17 and 18 of the monkey. Brain Res 137:343–350.

Ogren MP, Hendrickson AE. 1979. The morphology and distribution ofstriate cortex terminals in the inferior and lateral subdivisions of themacaca monkey pulvinar. J Comp Neurol 188:179–199.

Olshausen BA, Anderson CH, Van Essen DC. 1993. A neurobiological modelof visual attention and invariant pattern recognition based on dynamicrouting of information. J Neurosci 13:4700–4719.

Olszewski J. 1952. The thalamus of the Macaca mulatta: an atlas for usewith the stereotoxic Instrument. Basel: Karger.

Petersen SE, Robinson DL, Keys W. 1985. Pulvinar nuclei of the behavingrhesus monkey: visual responses and their modulations. J Neuro-physiol 54:867–886.

Petersen SE, Robinson DL, Morris JD. 1987. Contributions of the pulvinarto visual spatial attention. Neuropsychologia 25:97–105.

Rafal RD, Posner MI. 1987. Deficits in human visual spatial attentionfollowing thalamic lesions. Proc Natl Acad Sci USA 84:7349–7353.

Rezak M, Benevento LA. 1979. A comparison of the organization of theprojections of the dorsal lateral geniculate nucleus, the inferior pulvi-

AXON ANALYSIS OF PULVINOCORTICAL CONNECTIONS 249

nar and adjacent lateral pulvinar to primary visual cortex (area 17) inthe macaque monkey. Brain Res 167:19–40.

Robinson DL, Cowie RJ. 1997. The primate pulvinar: structural, functional,and behavioral components of visual salience. In: Steriade M, Jones EG,McCormick DA, editors. Thalamus: organization and function, vol. II.Amsterdam: Elsevier. p 53–92.

Robinson DL, Petersen SE. 1992. The pulvinar and visual salience. TINS15:127–132.

Rockland KS. 1985. A reticular pattern of intrinsic connections in primateV2 (area 18). J Comp Neurol 235:467–478.

Rockland KS. 1992. Configuration, in serial reconstruction, of individualaxons projecting from area V2 to V4 in the macaque monkey. CerebCortex 2:353–374.

Rockland KS. 1995. The morphology of individual axons projecting fromarea V2 to MT in the macaque. J Comp Neurol 355:15–26.

Rockland KS. 1996. Two types of corticopulvinar terminations: round (type2) and elongate (type 1). J Comp Neurol 368:57–87.

Rockland KS, Virga A. 1990. Organization of individual cortical axonsprojecting from area V1 (area 17) to V2 (area 18) in the macaquemonkey. Vis Neurosci 4:11–28.

Shinoda Y, Kakei S, Futami T, Wannier T. 1993. Thalamocortical organiza-tion in cerebello-thalamo-cortical systems. Cereb Cortex 3:421–29.

Shipp S, Zeki S. 1989. The organization of connections between areas V5and V1 in macaque monkey visual cortex. Eur J Neurosci 1:309–332.

Stepniewska I, Kaas JH. 1997. Architectonic subdivisions of the inferiorpulvinar in New World and Old World monkeys. Vis Neurosci 14:1043–60.

Steriade M, Jones EG, McCormick DA, editors. 1997. Thalamus, vol. I.Organization and function. Amsterdam: Elsevier.

Trojanowski JQ, Jacobson S. 1976. Areal and laminar distribution of somepulvinar cortical efferents in rhesus monkey. J Comp Neurol 169:371–392.

Trojanowski JQ, Jacobson S. 1977. The morphology and laminar distribu-tion of cortico-pulvinar neurons in the rhesus monkey. Exp Brain Res28:51–62.

Ungerleider LG, Ganz L, Pribram KH. 1977. Size constancy in rhesusmonkeys: effects of pulvinar, prestriate, and inferotemporal lesions.Exp Brain Res 27:251–269.

Ungerleider LG, Galkin TW, Mishkin M. 1983. Visuotopic organization ofprojections from striate cortex to inferior and lateral pulvinar in rhesusmonkey. J Comp Neurol 217:137–157.

Ungerleider LG, Desimone R, Galkin TW, Mishkin M. 1984. Subcorticalprojections of area MT in the macaque. J Comp Neurol 223:368–386.

Van Essen DC. 1985. Functional organization of primate visual cortex. In:Jones EG, Peters A, editors. Cerebral cortex, vol. 3. New York: PlenumPress. p 259–329.

Wong-Riley MTT. 1977. Connections between the pulvinar nucleus and theprestriate cortex in the squirrel monkey as revealed by peroxidasehistochemistry and autoradiography. Brain Res 134:249–267.

Yoshioka T, Levitt JB, Lund JS. 1992. Intrinsic lattice connections ofmacaque monkey visual cortical area V4. J Neurosci 12:2785–2802.

Yoshioka T, Blasdel GG, Levitt JB, Lund JS. 1996. Relation betweenpatterns of intrinsic lateral connectivity, ocular dominance, and cyto-chrome oxidase-reactive regions in macaque striate cortex. CerebCortex 6:297–310.

Zeki SM. 1978. The third visual complex of rhesus monkey prestriatecortex. J Physiol 277:273–290.

250 K.S. ROCKLAND ET AL.