THE JOURNAL OF COMPARATIVE NEUROLOGY 374~354361 (1996)

Variation of Dorsal Horn Cell Dendritic Spread With Map Scale

PAUL B. BROWN, RONALD MILLECCHIA, JAMES L. CULBERSON,

Departments of Physiology (P.B.B., R.M., W.G.), Anatomy (J.L.C.), and Biology (D.C.-D.), West Virginia University, Morgantown, West Virginia 26506

WILBERT GLADFELTER, AND DOROTHY COVALT-DUNNING

ABSTRACT Cells in laminae 111, IV, and V of cat dorsal horn were injected with horseradish peroxidase

or neurobiotin. Dorsal views of the dendritic domains were constructed in order to measure their lengths, widths, areas, and lengthiwidth ratios in the horizontal plane (the plane of the somatotopic map). Dendritic domain width and area in the horizontal plane were negatively correlated with fractional distance between the medial and lateral edges of the dorsal horn. These results are consistent with the hypothesis that dendritic domain width varies with map scale, which is maximal in the medial dorsal horn. This is similar to the variation in widths of primary afferent bouton distributions.

The parallel variation of dorsal horn cell dendritic domain width and primary afferent bouton distribution width with map scale suggests that there is a causal relation between morphology and map scale in the dorsal horn representation of the hindlimb. This variation of adult morphology with map scale must reflect mechanisms responsible for the assembly of receptive fields. c 1996 Wiley-Liss, Inc.

Indexing terms: spinal cord, morphology, somatosensory, convergence, divergence

Mammalian somatosensory nuclei form precise spatial representations of the skin, from topographically (somato- topically) organized receptive fields (e.g., Brown and Fuchs, 1975; Kaas et al., 1979; Felleman et al., 1983; Koerber et al., 1993). Within the representations of the limbs, the areas devoted to highly innervated distal skin (foot and toes) are characterized by smaller receptive fields (RFs) and larger map scale (e.g., Mountcastle, 1957; Brown and Fuchs, 1975). These physiological properties are believed to be neural correlates of tactile acuity.

Theories of the development of somatotopic maps must account not only for the topographical orderliness, but also for the relations among peripheral innervation density, map scale, and RF size. Such accounts must predict quanti- tatively the connectivities underlying RF formation. This task is easiest for nuclei closest to the periphery he . , dorsal horn and dorsal column nuclei) because there is less circuitry intervening between the primary afferent fibers and the cells of interest.

However, it is difficult even in the first relay nuclei to describe exactly the circuitry underlying RFs. In the dorsal horn, many different sources contribute inputs which influ- ence the cells’ RF geometries, including monosynaptic input from primary afferent fibers, local interneuronal connections, descending connections, and presynaptic modu- lation. A logical starting point for elaboration of RF assem-

bly is with the monosynaptic connections between primary afferent axons and dorsal horn cells.

Much is now known about primary afferent connections to dorsal horn cells, especially the connections between low threshold cutaneous mechanoreceptive afferent fibers and cells in laminae I11 and IV. At least in the case of slowly adapting type 1 (SA1) axons, there are some direct connec- tions to SA1 responsive cells of laminae I11 and IV from all parts of the cells’ excitatory low-threshold RFs (Brown et al., 1973). Although primary afferent fibers have long- ranging connections in the spinal cord to regions in which they do not contribute to dorsal horn cell excitatory RFs (e.g., Merrill and Wall, 1972; Koerber and Brown, 1982; Wall and Shortland, 1991; Shortland and Wall, 1992; Wall and McMahon, 1994), the main projections of these axons generally fall within the dendritic domains of cells whose RFs do normally include the afferent axons’ RFs (Brown and Culberson, 1981; Koerber and Brown, 1980, 1982; Brown et al., 1991). In summary, although many other inputs may influence a cell’s RF, the best first approxima- tion of excitatory low threshold RF is the aggregate of RFs of primary afferent fibers monosynaptically connected to the cell.

Accepted May 28, 1996. Address reprint requests to Paul Brown, Physiology Department, Health

Sciences Center, West Virginia University, Morgantown, WV 26506.

O 1996 WILEY-LISS, INC.

DORSAL HORN CELL DENDRITE WIDTH AND MAP SCALE 355

If the main determinant of a cell’s excitatory low thresh- old RF is the ensemble of RFs of afferent fibers connecting directly to it, then it should be possible to predict the largest possible RFs of cells from the distributions of their den- drites and the RFs and projections of sensory axons. If afferent fibers from region x on the skin do not have projections which overlap the dendrites of cells in a region Y of the map, map region Y should not contain cells whose RFs include skin region x. Specificity of connections could still restrict cell RFs to smaller areas than the largest possible RFs permitted by overlap of axonal and dendritic domains.

In the dorsal horn, the somatotopic map is laid out in the horizontal (mediolateral, rostrocaudal) plane. The pre- and postsynaptic somatotopy of the different laminae are in register with each other, and the neutral axis of somatotopy is radial to the laminar boundaries (roughly dorsoventral except for the lateral dorsal horn, where the superficial laminae wrap around the deeper laminae: viz., Brown et al., 1991). For this reason, it has become customary for map organization to be analyzed in dorsal views of the dorsal horn, just as the somatotopy of the somatosensory cortex is analyzed in views of the flattened cortical surface. Thus, the most likely variations of connectivity which could influence map scale and RF size are variations in the horizontal plane. Aspects of connectivity which vary with depth in the dorsal horn are more likely to relate to submodality conver- gence properties and the destinations of tract cell axons, because different types of receptors project to different laminae, and the cells of origin of different tracts have different laminar distributions.

The mediolateral axis of the dorsal horn maps a distoprox- imal and embryologically ventrodorsal skin gradient, whereas the rostrocaudal axis maps the dermatomal trajec- tory. Because the dorsal horn map of the skin is some 300 times longer than it is wide, it is not surprising that the mediolateral gradient is the steeper of the two, with a large distoproximal fraction of the leg (ca. 15-20 cm) mapped onto a narrow neuropil (ca. 800 Fm). The simplest model of RF formation is passive assembly, in which cell dendrites sample afferent terminals available to them-that is, that portion of the somatotopically organized presynaptic neuro- pi1 which lies within the spans of their dendrites. It is reasonable to assume that relatively widespread axonal arbors afford greater opportunity for divergence, and rela- tively widespread dendritic trees afford greater opportunity for convergence. Clearly, in terms of opportunity to inter- sect the terminal domain of an axon, the mediolateral width of a dendritic tree is the most important variable to study. Total dendritic surface areas and details of branching patterns are irrelevant, as are the locations of cell bodies. Such morphological variation could provide a basis for differences in map scale and RF sizes in different parts of the map.

We have shown (Millecchia et al., 1991) that axons have wider terminal distributions in medial dorsal horn, where map scale is larger and RFs are smaller. The implications of variable axonal arbor width for the assembly of dorsal horn cell RFs can only be understood in the context of a similar study of dorsal horn cell dendrites. Therefore, we have undertaken a study of the dendritic distributions of dorsal horn cells as a function of their mediolateral position, for comparison with axonal variations. We have characterized and stained a sufficient number of lamina 111-IV cells to

verify that their dendrites do vary in width as a function of mediolateral position.

A preliminary report of these results has been published (Brown et al., 1993).

MATERIALS AND METHODS Subjects were adult cats (2-4 kg), of either sex. Animals

were first anesthetized with a mixture of xylazine (0.5 mg/kg) and ketamine (20 mgikg) intramuscularly. Tra- cheal, carotid, and external jugular cannulae were installed for artificial respiration, blood pressure monitoring, and infusion of liquids and drugs, respectively. Alpha chloralose was administered intravenously in a series of supplements as the original anesthetic wore off, to a total of 70 mg/kg. Surgical anesthesia was maintained with supplements of chloralose whenever nociceptive reflexes could be detected, the pupils were responsive to light, or the blood pressure responded to noxious stimuli. The fur of the hindlimb was clipped to a length of approximately 1 mm. A rectal probe provided feedback control for a temperature-controlled heating pad and heating lamps, used to maintain core temperature at 38°C.

Spinal cord segments L3-S3 were exposed and covered with mineral oil. The segment chosen for recording was stabilized on a platform. A bilateral pneumothorax was performed, and the animal was paralyzed with gallamine triethiodide (initial dose 20 mg i.v., supplemented with 10 mg every half hour). End-expiratory COs was maintained at 3.5-4.0% with a respirator.

In earlier experiments, horseradish peroxidase (HRP) was used to stain cells; in later experiments, we switched to neurobiotin (Vector). When HRP was used, microelectrodes were filled with 30% HRP (Sigma Type VI) in 0.1 M phosphate buffer, pH 7.6, and 0.2 M or 0.5 M KCl. These electrodes were broken back to 20-40 MR.

When extracellular single-unit potentials were encoun- tered, their low-threshold excitatory cutaneous mechanore- ceptive fields were mapped with small probes, outlining the area within which light touch stimulation evoked increased discharge rates, if they were judged to be postsynaptic (RFs larger than those of primary afferent fibers). Upon penetra- tion of a unit with a mechanoreceptive field on the leg, it was determined to be a dorsal horn cell if postsynaptic potentials were evident and the electrode tip was 800-1,500 km below the surface of the cord. The low threshold excitatory cutaneous RF was mapped by using gentle manual stimulation. The outline of the spike discharge zone usually was within 1-2 mm of the outline obtained by extracellular mapping, although this zone sometimes fell within a larger region from which inhibitory postsynaptic potentials (IPSPs) or excitatory postsynaptic potentials (EPSPs) could be evoked. If the cell had a resting potential of more than -40 mV, it was injected ionophoretically with HRP or neurobiotin. Each injection lasted 1 minute, using 375-msec pulses recurring every 500 msec (injected charge 20 nA.min sustained plus 15 nA.min intermittent). If after an injection the resting and action potentials were not seriously degraded, an additional injection was made.

At least 20 minutes after injection of the last cell, the animal was perfused for histological processing. When HRP was used, the cat was perfused with 1 liter 0.9% NaCl solution with 0.1 g procaine hydrochloride and 0.1 g NaNOz at 38T, followed by 2 liters of 0.75% glutaraldehyde and

356 P.B. BROWN ET AL.

Fig. 1. Photomicrographs to illustrate filling of distal dendrites with reaction product. A Dendritic branches near their rostra1 limit (500 pm from cell body). B: Cell body and proximal dendrites. C: Dendritic fragments 500 pm caudal to cell body. Scale bar = 50 pm.

0.5% paraformaldehyde in 0.1 M phosphate buffer (pH 7.41, at approximately 24°C.

Tissue blocks were soaked overnight in sucrose-saturated 0.1 M phosphate buffer (pH 7.4) at 4°C. Serial transverse sections were cut at 50 pm on a freezing microtome, and HRP-injected tissue was processed with the cobalt-en- hanced diaminobenzidine (DAB) method (Adams, 1977) and counterstained with Neutral Red. All sections were mounted and examined for filled profiles.

When neurobiotin was used electrodes were filled with 2% neurobiotin (Vector) in 0.1 M KCl. Electrodes were broken back to 25-40 MfZ ( < 1 pm). All other aspects of cell recording and iontophoretic injection were the same as for HRP injections. Animals were perfused with 1 liter 0.9% NaCl followed by 1 liter 4% paraformaldehyde in pH 7.4 phosphate buffer. Tissue sections were permeabilized with

0.4% Triton X-100 in phosphate buffer for 2-4 hours, incubated in avidin-biotin-HRP complex (ABC reagent, Vector) for 12-16 hours, and then rinsed and stained with the cobalt-enhanced DAB procedure in which the initial concentrations of 1% HzOz and 5.5% DAB were doubled one at a time in alternation every 15 minutes until the sections were a light charcoal gray. Visual inspection of wet sections was used to confirm high-contrast staining of dendrites, in order to determine the endpoint.

The mediolateral distributions of stained dendritic seg- ments of each cell were measured in each transverse section using a camera lucida and a nomogram, adjusting the nomogram so that 0.0 fell at the medial edge of laminae I11 and IV, and 1.0 fell at the lateral edge. Each dorsal view of a dendritic domain is the combination of mediolateral extents for all transverse sections, regardless of density of dendritic

DORSAL HORN CELL DENDRITE WIDTH AND MAP SCALE 357

/-

A

Fig. 2. Camera lucida drawings of two neurons to compare mediolat- era1 dendritic span at different mediolateral positions. Cells were partially reconstructed by using camera lucida by superimposing and linking stained fragments from five sequential 50-km sections centered on the cell body. A Lamina V cell in medial dorsal horn, with wide dendritic tree. B: Lamina IV cell in lateral dorsal horn, with narrow dendritic tree. Scale bar = 200 km.

branches or the branching pattern. Cells that were encoun- tered on the right side of the spinal cord are displayed as left-right mirror images on left dorsal views. Each neuron’s dendritic field was mapped by starting at the cell body and working rostrally until no more dendritic fragments were seen, and then working caudally from the cell body in the same fashion. In most animals, there was only one cell injected on each side or there were large gaps between pairs of cells injected on the same side. In order to avoid overlap of dendritic trees of two cells within the same dorsal horn,

A L6

L7

S1

5 3

B - L6

- L7

-S1

-s2

-53

Fig. 3. Dorsal views of areas occupied by dorsal horn cell dendritic domains. In this and other dorsal views, laminae 111 and IV of the left dorsal horn are represented with rostral toward the top, caudal toward the bottom, lateral to the left, and medial to the right. The outermost outline represents the edges of the left dorsal horn; smaller outlines within it represent the areas occupied by dorsal horn cell dendritic domains. The dotted line extending the length of the dorsal view is the outline of the foot representation (Koerber et al., 1993). A Dorsal view with mediolateral and rostrocaudal scale the same. B: Dorsal view with the mediolateral scale expanded for greater clarity. Mediolateral and rostrocaudal dimensions of segments are proportional to average dimensions, averaged across many animals in this and earlier studies.

such cells were only included in the sample if, by tracing dendritic fragments from section to section, the dendritic domains of the two cells were clearly separated (e.g., separated by a wide gap mediolaterally within any section). Cells were not included for analysis if the staining faded with distance from the cell body (except for decreased optical density due to decreased dendritic diameter).

Figure 1 illustrates well-stained dendritic fragments at the rostral (Fig. 1A) and caudal (Fig. IC) tips of the dendritic field of a representative injected cell. Figure 1A,C

358 P.B. BROWN ET AL.

\ i

- L6

- L7

- s l

-

-s2

-s3

was photographed from sections 0.5 mm rostral and caudal to the cell body (Fig. 1B).

In order to analyze the effect of map location on the morphology of dendrites, we performed the same analyses on this sample of dorsal horn neurons as were performed for primary afferent bouton distributions (Millecchia et al., 1991). We examined the relation between dendritic domain length (rostrocaudal extent), width (widest mediolateral extent on a single section), area (of dorsal view outline), and lengthiwidth ratio as a function of mediolateral position (location of center of gravity of dendritic outline) within the dorsal horn. We use similar illustrations in this report for ease of comparison with the results of the previous study. The Pearson correlation coefficient r was calculated, from which a t-value could be derived for testing statistical significance. The criterion for statistical significance was P < .05.

RESULTS Sixteen cells, in 11 animals, were stained adequately for

analysis in laminae 111, IV, and V: two of these were in L6, six in L7, seven in S1, and one in S2. Neurons with somata in the medial dorsal horn tended to have larger mediolat- era1 extents than those in lateral dorsal horn. Two cells were selected for camera lucida reconstruction from serial transverse sections, to illustrate typical mediolateral spread of dendrites of cells in medial and lateral dorsal horn (Fig. 2). The photomicrographs of Figure 1 illustrate the appear- ance of individual sections from the cell of Figure 2B.

The dendritic domains of all the cells are illustrated in Figure 3, where they are combined in a single dorsal view of the relevant segments. Figure 3A is a dorsal view with a 1:l aspect ratio; note that all the dendritic trees are longer than they are wide. Figure 3B is a view with the mediolateral scale expanded for improved clarity. It should be kept in mind that the expanded mediolateral scale distorts the apparent lengthiwidth ratios of dendritic trees. In these and other dorsal views, a left dorsal view is shown in schematic form, with rostral to the top, caudal to the bottom, lateral (left) to the left, and medial (right) to the right. In this figure, average lengths and widths of seg- ments are used, computed from lengths and widths of all the segments in our database. In other dorsal views, average or normalized dimensions will be used as indicated. The combined data of Figure 3 do not compensate for interanimal variations, which we have demonstrated to consist mostly of a rostrocaudal shift of the map from one animal to another, referred to as pre- and postfixation (Koerber et al., 1993).

Each cell’s dendritic domain rostrocaudal length was calculated as number of sections spanned by stained den- drites times average section thickness (average segment length divided by number of sections in that segment). Lengths ranged from 437 to 1,059 (mean 2 standard devia- tion, 723 2 200) pm. There was no significant correlation of length with mediolateral location (r = -0.09, P = .7).

Figure 4 illustrates the relation between dendritic do- main mediolateral width and location in the somatotopic map, in a dorsal view with normalized segment lengths and

Fig. 4. Dorsal view with average segmental dimensions mediolater- ally, normalized dimensions rostrocaudally. Lines represent the widths of dendritic domains, centered at the locations of dendritic domain centers.

DORSAL HORN CELL DENDRITE WIDTH AND MAP SCALE 359

800

700

600

500

5 E 400 300

200

100

0

r = -0.62

p = 0.008

0

0

0 O 0

0

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 ML

Fig. 5. The relationship between dendritic domain width and mediolateral position. Medial edge of dorsal horn = 0.0, lateral edge = 1.0.

average segment widths. Each dendritic domain width is plotted as a mediolaterally oriented line segment centered on the dendritic domain center, with a width equal to the width of the dendritic domain on that section where the dendritic fragments spanned the greatest mediolateral ex- tent. The nolrmalized widths, calculated as fractional medio- lateral width times average dorsal horn width, ranged from 249 to 578 (384 +- 101) pm. Figure 5 plots width as a function of mediolateral position. There was a very signifi- cant correlation (r = -0.62, P = .008).

We also examined the relation between dendritic domain area and location in the somatotopic map, in a dorsal view with normallized segment lengths and widths. The areas ranged from 0.05 to 0.34 (0.14 & 0.08) mmz. As illustrated in Figure 6, there was a significant correlation of area with mediolateral location (r = -0.55, P = ,025).

Cells’ lengkhiwidth ratios were calculated as number of microns rostrocaudal length (based on number of sections spanned) over number of microns width (greatest mediolat- era1 extent on a single section), which ranged from 1.0 to 4.2 (2.0 +- 0.8). The correlation between lengthiwidth ratio and mediolateral position was not significant (r = 0.46, P = .07).

DISCUSSION Technical considerations

The cells examined in this study were unidentified dorsal horn cells, stained without regard to mechanoreceptor convergence except that they had to receive some low- threshold cutaneous mechanoreceptive input. The destina- tions of their axons were not identified by antidromic stimulation of ascending tracts. Thus, the cell types exam-

ined were probably quite heterogeneous. The diversity of cell types and the small sample might cause us to miss correlations, e.g., between mediolateral position and length/ width ratio. Interanimal variations, especially rostrocaudal shifting of the map from one animal to another, may also contribute variability which would obscure correlations. None of these sources of variability could introduce spuri- ous correlations.

In spite of the heterogeneous population, small sample size, and interanimal variation, it was still possible to detect significant correlations of dendritic tree width and area with mediolateral position. The correlation of areas may simply reflect the correlation of widths with mediolateral position. The correlation of width with position is particu- larly striking in view of reported differences in the dendritic geometries of cells of origin of different ascending spinal tracts (Brown et al., 1976, 1977; Brown and Fyffe, 1981).

The sample was limited mostly to L7 and S1. We restricted our sample to a region in which the mediolateral somato- topic gradient (distance of RF center from tips of toes as a function of mediolateral cell position) was relatively uni- form in order to minimize the effect of variable somatotopic gradient on the analysis. If data were combined from areas with very different somatotopic gradients, such a small sample might have led to apparent reversals of somatotopic gradient, and certainly would have led to greater variance of the RF position as a function of ML position.

Morphological correlations with map scale Although subjective impressions of dendritic domain

width variations with mediolateral position have been reported before, for cells injected with HRP (Proshansky and Egger, 1977) and for tract cells whose proximal den-

360

0.5

0.4

0.3 Q : 0.2

0.1

0

P.B. BROWN ET AL.

r = -0.55

p = 0.025

0

0

0

%* O O O O 0

0

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 ML

Fig. 6. Relation between dendritic domain area and mediolateral position.

drites were labeled retrogradely with HRP from their termination sites (Enevoldson and Gordon, 1989a,b), the present study is the first quantitative analysis of the relations between measurable anatomical variables and mediolateral location of dorsal horn cells. Previous observ- ers have not proposed any functional significance for the relationship between dendritic tree width and mediolateral position. The most obvious physiological correlate of den- dritic (and axonal: Millecchia et al., 1991) domain width in the hindlimb representation is map scale, which is maximal in medial dorsal horn and minimal in lateral dorsal horn. In the medial dorsal horn, where foot and toes are mapped, the representation of a unit area of skin is spread across a wider span than in the lateral dorsal horn, where more proximal parts of the hindlimb are mapped. In medial dorsal horn, where map scale is large, anatomical distributions of axon terminals and dorsal horn cell dendrites need not be as restricted mediolaterally as in lateral dorsal horn, where map scale is small.

Assembly of dorsal horn cell receptive fields We have presented arguments in the opening paragraphs

that dendritic width might affect the degree of spatial convergence upon dorsal horn cells, just as the widths of afferent axons’ terminal fields might affect the degree of divergence of their representations in the map. If there are some monosynaptic connections from all parts of dorsal horn cells’ RFs, as preliminary studies have indicated, then there should be a very strong correlation between RF locations and geometries, and the local variations of mor- phology of primary afferent terminal fields and dorsal horn cell dendritic trees. Our results indicate that the widths of both axon terminal fields and dorsal horn cell dendritic trees are correlated with map scale, which implies that

afferent terminal fields and cell dendritic trees have widths which co-vary. A larger sample of both populations would be necessary to test this hypothesis, but if it proved to be correct, the functional implications could be quite impor- tant. One possible consequence would be that the assembly of dorsal horn cell RFs is based on a competitive mecha- nism, in which convergence and divergence of connections between afferent fibers and cells are independent of map location. In this case, RF sizes would be a direct function of map scale, or more appropriately, both map scale and RF sizes are determined by some common factor such as peripheral innervation density.

ACKNOWLEDGMENTS This research was supported by USPHS grant NS27511.

We are indebted to Mark Jacquin for his advice on the use of neurobiotin and to Louis Ritz for comments on the manu- script. Avonell Painter performed the histological process- ing.

LITERATURE CITED Adams, J.C. (1977) Technical considerations on the use of horseradish

peroxidase as a neuronal marker. Neuroscience 2: 141-145. Brown, P.B., and J.L. Culberson (1981) Somatotopic organization of hind-

limb cutaneous dorsal root projections to cat dorsal horn. J. Neuro- physiol. 45:137-143.

Brown, P.B., and J.L. Fuchs 11975) Somatotopic representation of hindlimb skin in cat dorsal horn. J. Neurophysiol. 38:l-9.

Brown, A.G., and R.E.W. Fyffe (1981) Form and function of spinal neurones with axons ascending the dorsal columns in the cat. J. Physiol. ILond.) 321:31-47.

Brown, P.B., H. Moraff, and D.N. Tapper 11973) Functional organization of the cat’s dorsal horn: Spontaneous activity and central cell response to single impulses in single Type I fibers. J. Neurophysiol. 36.427-839.

DORSAL HORN CELL DENDRITE WIDTH AND MAP SCALE 361

Brown, A.G., C.R. House, P.K. Rose, and P.J. Snow (1976) The morphology of spinocervical tract neurones in the cat. J. Physiol. (Lond.) 260:719- 738.

Brown, A.G., P.K. Rose, and P.J. Snow (1977) The morphology of spinocervi- cal tract neurones revealed by intracellular injection of horseradish peroxidase. J. Physiol. (Lond.) 270:747-764.

Brown, P.B., W.E. Gladfelter, J.L. Culberson, D. Covalt-Dunning, R.V. Sonty, L.M. Pubols, and R.J. Millecchia (1991) Somatotopic organization of single primary afferent axon projections to cat spinal cord dorsal horn. J. Neurosci. I1:29%309.

Brown, P.B., R.J. Millecchia, W.E. Gladfelter, J.L. Culberson, and D. Covalt-Dunning (1993) Variation of dendritic morphology with map location in hindlimb dorsal horn. SOC. Neurosci. Abstr. 19:327.

Enevoldson, T.P., and G. Gordon (1989a) Postsynaptic dorsal column neurons in the cat: a study with retrograde transport of horseradish peroxidase. Exp. Brain. Res. 755511-620.

Enevoldson, T.P., and G. Gordon (1989b) Spinocervical neurons and dorsal horn neurons projecting to the dorsal column nuclei through the dorsolateral fascicle: a retrograde HRP study in the cat. Exp. Brain. Res. 75:621-630.

Felleman, D.J., J.T. Wall, C.G. Cusick, and J.H. Kaas (1983) The representa- tion of the body surface in S-I of cats. J. Neurosci. 3:1648-1669.

Kaas, J.H., R.J. Nelson, M. Sur, C.S. Lin, and M.M. Merzenich (1979) Multiple representations of the body within the primary somatosensory cortex of primates. Science 204r521-523.

Koerber, H.R., and P.B. Brown (1980) Projections of two hindlimb cutane- ous nerves to cat dorsal horn. J. Neurophysiol. 44259-269.

Koerber, H.R., and P.B. Brown (1982) Somatotopic organization of hindlimb cutaneous nerve projections to cat dorsal horn. J. Neurophysiol. 48:481- 489.

Koerber, H.R., G. Hobbs, and P.B. Brown (1993) Precision and variability of hindlimb representation in cat dorsal horn and implications for tactile localization. J. Neurophysiol. 70:2489-2501.

Merrill, E.G., and P.D. Wall (1972) Factors forming the edge of a receptive field. The presence of relatively ineffective afferents. J. Physiol. (Lond.) 226:825-846.

Millecchia, R.J., L.M. Pubols, R.V. Sonty, J.L. Culberson, W.E. Gladfelter, and P.B. Brown (1991) Influence of map scale on primary afferent terminal field geometry in cat dorsal horn. J. Neurophysiol. 66:696-704.

Mountcastle, V.B. (1957) Modality and topographic properties of single neurons of cat's somatic sensory cortex. J. Neurophysiol. 20:408-434.

Proshansky, E., and M.D. Egger (1977) Dendritic spread of dorsal horn neurons in cats. Exp. Brain. Res. 28:153-166.

Shortland, P., and P.D. Wall (1992) Long range afferents in the spinal cord. 11. Arborizations that penetrate gray matter. Philos. Trans. R. Soc. Lond. [Biol.] 337:445455.

Wall, P.D., and S.B. McMahon (1994) Long range afferents in rat spinal cord. 111. Failure of impulse transmission in axons and relief of the failure after rhizotomy of dorsal roots. Philos. Trans. R. Soc. Lond. [Biol.] 343:211-223.

Wall, P.D., and P. Shortland (1991) Long-range afl'erents in the rat spinal cord. I. Numbers, distances and conduction velocities. Philos. Trans. R. Soc. Lond. [Biol. I 334235-93.