(Received 29 June 1972)

Transcription

1 J. Physiol. (1973), 228, pp With 9 text-figures Printed in Great Britain CONTRASTS IN SPATIAL ORGANIZATION OF RECEPTIVE FIELDS AT GENICULATE AND RETINAL LEVELS: CENTRE, SURROUND AND OUTER SURROUND BY P. HAMMOND From the Medical Research Council Group, Department of Communication, University of Keele, Keele, Staffordshire ST5 5BG (Received 29 June 1972) SUMMARY 1. The organization of receptive fields of retinal ganglion cells and A-laminae cells from the dorsal lateral geniculate nucleus (LGN) of the cat are compared under identical conditions. Some aspects of the geniculate data have been given elsewhere (Hammond, 1972b). 2. The receptive fields of geniculate cells consist of three zones - centre, antagonistic surround and synergistic outer surround - compared with only two for retinal cells. This result further supports the theory that the centre and surround of geniculate cell receptive fields derive from convergent, but discrete, retinal inputs. 3. The surrounds of geniculate receptive fields are known to be more powerfully antagonistic on their centres than is true of retinal cells. This relationship is re-examined. 4. Unlike geniculate fields, the locus of maximum sensitivity for the receptive field surround of retinal cells is not invariant either to stimulus geometry or adaptational state. 5. The latter result strongly suggests that the surround mechanism for retinal cells extends through the centre of the field. It establishes unequivocally that the overlap between receptive field centre and surround mechanisms, only marginal in geniculate, is very extensive indeed in retina. INTRODUCTION Binocularity apart, there has in recent years been a growing realization that the lateral geniculate nucleus does much more than simply provide a throughput between retina and cortex. Hubel & Wiesel (1961) have demonstrated differences in firing patterns and enhanced potency of receptive field surrounds for LGN neurones when compared with retina.

2 116 P. HAMMOND More recently Singer & Creutzfeldt have provided direct evidence that each geniculate cell is the focus of highly specific convergence from a number of discrete ganglion cells in the retina (Singer & Creutzfeldt, 197). Their basic conclusions, and independent substantiating evidence (Maffei & Fiorentini, 1971, 1972; Cleland, Dubin & Levick, 1971a, b; Hammond, 1971, 1972a, b; Nakayama, 1971) have been summarized elsewhere (Hammond, 1972b). The model proposed by Singer & Creutzfeldt for the organization of retinal input to the LGN is that each single cell in the nucleus receives field-centre input from one (or several) retinal fibres of similar type (all on-centre or all off-centre), and field surround input from a number of retinal fibres of complementary type. Maffei & Fiorentini (1972), whilst making it clear that the evidence to date is not conclusive, suggest forcefully in discussion that a model in which the fibre input to the geniculate surround is of the same type as that to the field centre (via an inhibitory interneurone) tallies better with the known properties of geniculate cells, a point already made independently by H. B. Barlow (personal communication). In either case the implications of convergence are basically the same. The purpose of this report is to tie up some of the remaining loose ends. First, evidence for convergence is entirely consistent with Hubel & Wiesel's (1961) observation that the influence of the receptive field surround is increased at the level of the geniculate; this property is re-examined here. Secondly, Singer & Creutzfeldt's model for convergence takes no account of the retinal receptive field surrounds. The contribution which retinal surrounds make to the spatial profile of a geniculate receptive field is considered in detail here, and a preliminary account of findings has already been published (Hammond, 1972a). Finally, the nature of convergence is important in determining the spatial distribution of the mechanisms of excitation and inhibition which underly the concentric organization of retinal and geniculate receptive fields, and these distributions are compared. METHODS Except as specified, surgical preparation, visual stimuli and recording techniques are detailed elsewhere (Hammond, 1972b). Essentially, animals were immobilized with Flaxedil whilst being maintained under light pentobarbitone anaesthesia. Rectal temperature and gas exchange were under continuous control, monitored in conjunction with carotid pressure and pulse and the surface cortical e.e.g. as criteria of anaesthetic adequacy throughout paralysis. Ventilation was maintained at 28 strokes/min to ca. 4x2 % end-tidal CO2. Pupils were dilated and nictitating membranes retracted pharmacologically. The corneae were protected with two-curve, neutral contact lenses. Supplementary trial lenses provided the necessary focal correction. Animals viewed monochromatic spot or annular stimuli of wave-length

3 RECEPTIVE FIELD ORGANIZATION 525 nm, flash-presented on a matt white screen 56-7 in. distant, which was diffusely lit with tungsten light at a luminance of 18 log cd/m2 (mid-mesopic, Hammond & James, 1971). All stimuli for each particular unit were of identical quantum flux per unit area (approximately 1 log unit more intense than a threshold spot of optimal size). The geniculate data are largely from the population of units described previously (Hammond, 1972b), together with a few further units from one additional preparation. These units were recorded with low-impedance, 4M-NaCl micropipettes (tip diameter ,um) from the A-laminae of the LGN in eleven cats ( kg), with subsequent histological confirmation of recording site. Activity in single optic tract fibres was also recorded from three of these preparations and from two additional animals, during nine vertical penetrations with high-impedance, 2-7 M-KC1 micropipettes between Horsley-Clarke co-ordinates A 95-A 1- and RL 7-- RL 9'. Histological confirmation was in this case unnecessary (Andrews & Hammond, 197a). 117 RESULTS Synopsis Results are selected from a total of one-hundred-and-nine units. Fiftyseven were from the lateral geniculate nucleus, and fifty-two from the optic tract posterior to the chiasm. The distribution of receptive field centres in the contralateral visual half-fields is shown in Fig. 1. All reported units were influenced by both cones (556 nm) and rods (52 nm). Specifically, optic tract fibres were all tonic in character (forty-eight on-centre, four off-centre; thirty-three driven by the contralateral and nineteen by the ipsilateral eye). The break-down for LGN units was as follows: forty-six on-centre, eleven off-centre; thirty-four cells from layer A (contralaterally driven), sixteen cells from layer AL (ipsilaterally driven), and seven optic radiation fibres (five contra- and two ipsilaterally driven); thirty-three tonic and fourteen phasic, five with intermediate characteristics and five unclassified. The basic procedures have been described in detail earlier (Hammond, 1972b). Comparisons are drawn between the activity of optic tract units and LGN cells (more specifically the tonic class), studied under identical conditions. Concentric suprathreshold spots, annuli or simultaneously presented spots and annuli are used for evaluation of both retinal and geniculate activity. Responses of on-centre retinal and geniculate units are compared in Figs. 2, 3 and 7 for these three situations, and the results are illustrated graphically in Figs. 4, 6 and 8. The findings are equally applicable to off-centre units, with the proviso that the off-centre sample is limited in number. Within these three basic classes of experiment the break-down was as follows: spot stimulation - twenty-one OT fibres (all tonic) and thirty-six LGN units (twenty-six tonic, ten phasic); annular stimulation - thirty-five

4 118 P. HAMMOND OT fibres (all tonic) and twenty-five LGN units (nineteen tonic, six phasic); simultaneous spot/annular stimulation - nine OT fibres (all tonic) and fifteen LGN units (all tonic). Centre-surround organization Responses from tonic on-centre optic tract and geniculate units to spot or annular stimulation are typified by the pair of units illustrated in Figs. 2 and 3, which have been selected for similarity of field-centre size. Fe-.. o.*~*. 5 to~~~~~~~~~s ~~~~~~~~~~ -- a. so ~~~~~~ * ~~~~~~~15 ~~~ * ~~~~~~~~5 ~~~~ Fig. 1. Location of receptive field centres of geniculate and retinal units in the contralateral visual half-field, with respect to the plotted areae centrales. Data for the two eyes are superimposed, with torsional correction for an assumed mean position angle of 25 for each optic disk (Bishop, Kozak & Vakkur, 1962). The majority of units were from the right geniculate and optic tract, with receptive fields in the left half-fields as illustrated. Receptive fields of a few units from the left nucleus and tract are mirrorimaged from their correct location in the right half-fields (- principal A-laminae LGN cells, and optic radiation fibres; optic tract fibres. The overlapping pairs of symbols each show the location for an LGN cell and one of its constituent fibre inputs, recorded simultaneously from the nucleus). 35

5 RECEPTIVE FIELD ORGANIZATION 119 Stimuli were identical even in absolute flux per unit area, and minimum thresholds were similar in each case. The established receptive field organization for retinal and geniculate units is superficially similar, centre and concentric annular surround (but see also the following section). With stimuli suitably restricted to either the centre or the surround, the response from the surround is always of opposite polarity to that from the centre. For on-centre units it consists essentially of an on-discharge accompanied by off-inhibition from the centre, and on-inhibition followed by an off discharge from the surround, and conversely for off-centre units. But a number of contrasts emerge. The initial transient of the response from LGN cells is more discrete, and 75 OT fibre osu IJ114 na122,14 LGN cell fi - U-td2lt_ 1 =1 M"llll~~1h _I1 4' -. L la 1*5 1L.X I11u 1 :S 1' us E. 2 D I H1l II I I -II I I - I *..11 "NM 1,iif-". I - Fig. 2. Comparison between tonic on-centre OT fibre and LGN cell responses: effect of varying stimulus spot diameter. Stimuli were concentrically presented spots of wave-length 525 nm, all of identical quantum flux per unit area (at an intensity level approximately 1 log unit above that of a threshold spot of optimum diameter). Flashes (4 msec every 1-4 sec, indicated by the step-function below each column of records) were presented against a uniform achromatic adapting field with a luminance of -18 log cd/m2. Spike records: positive up, retouched; horizontal bar calibration, 1 sec. PST8: each for twenty consecutive responses accumulated on a storage CRT; 1 data bins at bin-widths 1 msec; vertical bar calibration, 1 spikes/bin; horizontal bar calibrations, stimulus and 1 sec time marks. Optic tract fibre (OT 48-1): ipsilateral, field centre 22 eccentric. LGN cell (LGN 46.9): layer A contralateral, field centre 13 eccentric.

6 12 P. HAMMOND its peak impulse frequency is much higher, than for optic tract fibres (compare columns of PSTs in Fig. 2). The clear pause between the initial transient and the maintained phase of the discharge is a common feature of LGN cell responses; this is frequently absent and rarely as obvious for optic tract fibres. The surrounds ofgeniculate fields show much more powerful antagonism towards their centres than is true of retinal fields (Hubel & Wiesel, 1961). This difference is apparent even from casual inspection of the raw data and in V OT fibre II,1 ''I 11,.'II,, I, 1, I II II, 1.I 1S.F I~~~~~~~~~~~~~~~~~~~~~ LGN cell r I. 1 - it- RI I 11 -I r It --1 I I-. - I. V E A2 C M E r_ C rnINI11 IUttijfill1o _ 4 _ 5 _ 6,..1 _ I I II *1 I I'l' t TN j--a.at~a 1 II1 M 1. II II I. -1[iI~ [. 1r 1r I 1_ II 1 _ B1sllrlr-lrlll-l-llla ii-. Fig. 3. Comparison between tonic on-centre OT fibre and LGNcellresponses: effect of varying mean diameter of annular stimuli. All annuli were of wavelength 525 nm, flash-presented for 4 msec at intervals of 1-4 sec (as indicated by the step-function at the foot of each column of records) against a uniform adapting field at 418 log cd/m2. Quantum flux per unit area was maintained constant (approximately 1 log unit above that for a threshold spot of optimal size). S.F. = spontaneous firing. Spike record: positive up, retouched; horizontal bar calibration, 1 sec. PST8: each for twenty consecutive responses accumulated on a storage CRT; 1 data bins at binwidth 1 msec; vertical bar calibration, 1 spikesibin; horizontal bar calibrations, stimulus and 1 sec time marks. Same units as in Fig. 2. PSTs illustrated in Figs. 2 and 3. The effects of annular stimulation (oninhibition and the ensuing off discharge from the surrounds of on-centre retinal units) are weak when compared with their effect on geniculate fields. The attenuation of the response to spots larger-than-optimal is much less marked for optic tract fibres than for LGN cells (Fig. 2). Field-centre stimulation results in qualitatively similar responses from both retinal and geniculate units. The difference shows up in response to

7 RECEPTIVE FIELD ORGANIZATION 121 large-spot or diffuse illumination. In this case on-centre retinal units invariably respond with nett excitation; one never sees suppression of firing at stimulus onset and seldom more than a suspicion of an increase in firing frequency at stimulus offset. By contrast, for LGN units the surround frequently predominates over the centre; diffuse illumination may evoke a response of opposite polarity to that for small spots. Thus, for on-centre LGN units, firing may be depressed or even abolished during illumination, to be followed by an off discharge. Fig. 4 makes the graphical comparison for the same pair of units as in Fig. 2. Both units were studied under identical conditions and it can be seen that they possessed receptive field centres of approximately the same size. The slope of the curve for supra-optimal stimuli, where response was inversely related to stimulus diameter, was significantly steeper for geniculate than for retinal units. Most convincing was the same effect observed for two unit-pairs recorded quasi-intracellularly from the LGN. In each case the record was from a tonic geniculate cell, together with e.p.s.p.s from one of its constituent retinal fibre inputs. All e.p.s.p.s were of identical amplitude and time course, and the receptive field centres of both units were exactly superimposed. Every geniculate cell spike took off from an e.p.s.p. and the shape of its rising phase in combination with the e.p.s.p. was extremely stable. The retinal fibre involved in each case thus provided the sole excitatory input to the geniculate field centre (cf. Cleland et al a). Conversely, not every e.p.s.p. was the source of a cell spike; the 'input/ output' ratio approached unity for small-spot stimulation, but increased significantly for larger stimuli. A comparison of the effectiveness of surround antagonism between retinal and geniculate units is made in Fig. 5. In the case of optic tract units, a reduction in response to 6 % of optimal was the largest effect that could be systematically produced by diffuse illumination. Thus for comparison the diameter of stimulus spot required to reduce the response to this value was plotted as a function of receptive field centre diameter. The size of the field centre was given by the stimulus diameter which evoked the greatest discharge. Both values were interpolated from a smooth curve drawn through the data points, such as shown in Fig. 4. It is clear from Fig. 5 that the lines of best fit to each of the two sets of data tend to approximate to one another for units with larger field centres, but deviate significantly for the smaller fields. Two conclusions can be drawn from this result. First, geniculate surrounds are more potent than retinal surrounds. Secondly, the difference is appreciably more marked for small field-centre units. Possible implications are considered in Discussion.

8 122 P. HAMMOND 1 A ~ ) f 4 2 IL l l lillii Spot diameter (degrees, log scale) 1 _ B Spot diameter (degrees, log scale) Fig. 4. Influence of stimulus area on the responsiveness of tonic on-centre units: comparison between OT fibre (A) and LGN cell (B). Stimulus wavelength 525 nm, at constant quantum flux (1 log unit above that for a threshold spot of optimal size), presented against an achromatic adapting field with a luminance of -18 log cd/m2. OT fibre 48/1 ipsilateral, field centre 22 eccentric; mean spike count equivalent to 1%, 112 spikes/ sec. LGN cell 46/9: layer A contralateral, field centre 13' eccentric; mean spike count equivalent to 1 %, 114 spikes/sec. Each plotted point is the mean for twenty consecutive responses.

9 RECEPTIVE FIELD ORGANIZATION 123 Retino-geniculate convergence: the outer surround Classically the receptive fields of LGN cells are concentrically organized, with a centre and an annular antagonistic surround not unlike fields of ganglion cells in the retina. However, if one accepts the convergence model proposed by Singer & Creutzfeldt (197), then by virtue of the contribution due to the receptive field surrounds of the fibre inputs from the retina, the expectancy is that geniculate cell receptive fields will possess an additional annular zone (illustrated by the diagram of Fig. 9). 4 U 3~~~~~~~~~~~~~~~~~~~~ *~~~~.~~~~~~~~~~~~~~~~~~ 15 ~ ~ ~~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~1 C~~~~~~~~~~~~~~~~~~C 4) ~~~~~~~~~~~~~~~~~~ -75 U_~ ~ ~ ~ ~ E * Spot diameter for 6% optimum response (degrees, log scale) Fig. 5. Comparison of the potency of surround antagonism fortonic on-centre OT fibres and LGN cells. Relationship between receptive field centre size (spot diameter evoking optimal response) and the diameter of a larger-thanoptimal spot required to reduce the response to 6% of maxmu, both values interpolated from graphical plots such as illustrated in Fig. 4. All stimuli of wave-length 525 nm and of identical quantum flux per unit area (approximately 1 log unit more intense than a threshold spot of optimal size). [] OT fibres;fl LGN cells. This zone, hereafter termed the outer surround, will predictably be of the same polarity as the receptive field centre, i.e. excitatory in on-centre fields, inhibitory in off-centre fields. The majority of geniculate receptive fields have circular symmetry. Assuming this tenet it is equally clear from Fig. 9 that the most potent stimulus for either the surround or the predicted outer surround of a geniculate field will be an annulus, concentric with the field centre.

10 124 P. HAMMOND The stimuli chosen were therefore a series of narrow annuli, all of the same format: outer diameter/inner diameter = n/(n -1), where n is in degrees, variable in.5 steps from 1.5 to 1. Each annulus was identified by its mean diameter (n -.5), and the flux per unit area 1 8 A 1- -o 6 a) WC Q. (A 4 2 _ 1 r- - Il I I I I I IifIIIIII Annulus mean diameter (degrees, log scale) B\ 8 I- 1- ọ - as to va 6 F- 4 e 13 MA 2. N Annulus mean diameter (degrees, log scale) Fig. 6. For legend see facing page.

11 RECEPTIVE FIELD ORGANIZATION 125 was maintained constant within any given series of trials. Thus although the total flux of a stimulus varied with its dimensions (as in the case of concentric spot stimuli), each annulus similarly provided half-a-degree of receptive field coverage between inner and outer circumferences. A typical result is illustrated graphically in Fig. 6B for a tonic on-centre LGN cell. The equivalent spontaneous firing level is given by the dashed line. Three receptive field zones were clearly indicated by the data for this unit. Small annuli falling partially or entirely within the receptive field centre evoked nett excitation; larger annuli evoked suppression from the inhibitory surround; still larger annuli again evoked nett excitation. In the latter case, the term disinhibition is more appropriate than excitation, for the peak in mean firing rate to large annuli did not invariably rise above the spontaneous firing level. Disinhibition, although a comparatively weak phenomenon, was seen for almost all tonic and phasic geniculate cells investigated. There were two exceptions; a phasic cell for which no discharge at all could be evoked by any stimulus outside the receptive field centre, and a tonic cell with an unusually large field centre (2.5). In the case of the tonic cell it is possible that the largest annulus which could be generated by the projection system (mean diameter 9.5) was still too small to influence the outer surround. The result, whilst easier to demonstrate in on-centre units, has also been confirmed for each of several offcentre units. Spot stimuli on their own do not show up the influence of the outer surround conclusively, but careful inspection of the graphed results to spot stimulation reveals a slight, but consistent, inflexion in the curve (e.g. the small displacement of the data point at 7 in Fig. 4B). The disinhibitory component shows up as a longer-latency component of the geniculate cell response, with a minimum latency of the order of 8 msec, as can be seen from the histograms for annuli with mean diameters of 5 and 6 in the right-hand column of Fig. 3 (arrowed). Thus Fig. 6. Response variation as a function of annulus mean diameter: comparison between OT fibre (A) and LGN cell (B) with approximately the same field centre size and eccentricity. Stimulus wave-length 525 nm, at constant flux (1 log unit above that for a threshold spot of optimal size), presented against an achromatic adapting field with a luminance of 418 log cd/m2. Mean spike counts are expressed as percentage maximum for comparison (U increasing diameters; U decreasing diameters). Equivalent spontaneous firing level is indicated by dashed lines. Each point is the mean from twenty consecutive responses. OTfibre 58/2: contralateral, field centre 14 eccentric; mean spike count equivalent to 1 %, 19 spikes/sec. LGN cell 46/7: layer A contralateral, field centre 11 eccentric; mean spike count equivalent to 1 %, 4 spikes/sec.

12 126 P. HAMMOND the disinhibitory peak tended to show up even more clearly than in the graph data of Fig. 6B for annular stimuli, if the initial transient of the discharge was excluded, by suitably gating the spike count. It follows that disinhibition in phasic cells was often difficult to observe with this technique. The longer latency components of the discharge from phasic cells were absent or represented comparatively weakly, and in any case their surrounds were generally less potent than for tonic cells (Hammond, 1972 b). By contrast, the identical experiment performed on tonic on-centre fibres of the optic tract repeatedly failed to show any suggestion of a disinhibitory outer surround (Fig. 6A). Increase in annulus mean diameter beyond the field centre evoked progressively weakening inhibition. Firing frequency simply approximated to the spontaneous level beyond the limits of the surround. No evidence of elevation of firing above the spontaneous level, or even of a peak in firing, was seen in any instance. Thus the phenomenon is not comparable to the periphery effect observed by McIlwain (1964). It must be concluded that geniculate cell receptive fields consist of three zones - centre, antagonistic surround, and synergistic outer surround - compared with only two for retinal fields. The result has recently been confirmed independently by Maffei & Fiorentini (1972), and is a strong argument in favour of retino-geniculate convergence. It should also be noted here that the negative slope of the function as annulus coverage passes progressively from centre to surround tends to be steeper for geniculate than for retinal units. This is apparent only from an over-all inspection of the data. It is not shown up by the particular pair of units illustrated in Fig. 6, but is clear in the comparative data of the following section (see Fig. 8). The stimulus geometry meant that the smallest annulus of the series was confined to the receptive field centre only in the case of large field centre units, so that a quantitative comparison cannot be made. However, the result clearly implies that the centre/surround boundary is sharpened at the level of the geniculate, and further substantiates the convergence hypothesis. Spatial distribution of centre and surround mechanisms The results of experiments with simultaneously presented, concentric spot and annular stimuli in geniculate have already been detailed elsewhere (Hammond, 1972b). In this section the results are contrasted with results from comparable experiments on optic tract units (Figs. 7 and 8). These experiments throw some light on the spatial distribution of excitatory and inhibitory mechanisms which are related to the centresurround organization of receptive fields at the two levels. But it is also worth drawing attention to the enormous qualitative differences in centre-

13 RECEPTIVE FIELD ORGANIZATION 127 surround organization at retina and geniculate, highlighted by Fig. 7. The spike records and PSTs compared are for exactly the same pair of units as in Figs. 2 and 3. A 4.5 annulus totally abolished the response of the LGN cell to the test spot and also evoked an off-discharge. For the OT fibre, field-centre stimulation powerfully dominated the response to all spot/ annulus pairs; suppression of the response to the test spot alone was slight compared with the LGN cell (Fig. 8). OT fibre Test spot Dal-;; - -_-I-R1 - -a alone 1*5 LGN cell _1_mW MT- 111_11 I L_ I 2-5 bo 3.5 i-t-szl 1,45 WI L 6-5 I L f tltsl L tl c -*4-5 CL C ma 9.5 HrIllh-in-l-h.1 11 pe1 s. ]I -1 I Fig. 7. Comparison between retinal and geniculate receptive field surrounds: responses to annuli presented simultaneously with a centred test spot. Spot and annuli of identical wave-length (525 nm) and quantum flux per unit area, presented against an achromatic adapting field with a luminance of 18 log cd/m2. Test spot diameter: 1 for OT fibre;.75 for LGN cell. Same units and additional detail as in Figs. 2 and 3. Apart from these obvious differences in centre-surround balance for retinal and geniculate fields, the comparison favours mechanisms with entirely different spatial distributions in retina and geniculate, although the receptive fields themselves possess concentric organization in common. The result for these experiments in geniculate was that local adaptation of the field centre by the test spot, whilst significantly decreasing the diameter of the centre-surround boundary (the annulus mean diameter for the point of cross-over from nett excitation to nett inhibition in Fig. 8), 5 PHY 228

14 128 P. HAMMOND was without influence on the size of annulus needed to optimally activate the surround. It was concluded that the locus of maximum sensitivity for the receptive field surround must be a concentric circle, rather than a point at the geometric centre of the receptive field, basically without overlap between the centre and surround mechanisms (Hammond, 1972b). Annulus mean diameter (degrees, log scale) r n 8A lb. i M *o 61 (" 2 18 OQ 16 B E W C ~1 Annulus mean diameter (degrees, log scale) I,,,,.,. 1 11H _,,,,,,,, O. <, 2 \ ;' /a a / Fig. 8. Response variation as a function of annulus mean diameter, whenpresented simultaneously with a centred test spot. The optic tract and geniculate units compared are the same pair as in Figs. 2 and 7. The spot and annular stimuli were of identical wave-length (525 nm) and quantum flux per unit area, presented simultaneously for 4 msec each 1-4 see against a white adapting field of luminance -18 log cd/m2. Mean spike counts for twenty consecutive responses to each stimulus pair are expressed as a percentage of the response to the test spot alone, for annual of increasing (U) and decreasing (E]) diameter. Tedt spot diameter: 1 in A;.75 in B. Mean pike count equivalent to 1% (response to teat spot alone): 121 spikes/sec in A; 76 spikes/sec in B. The experiments were more exacting in the case of the optic tract, because it was crucial to hold each unit long enough to obtain comprehensive data for both annular and combined spot/annulus presentations. This rarely presented problems in recording from LGN, for cells could

15 RECEPTIVE FIELD ORGANIZATION 129 routinely be held for many hours. But since the complete experiment took anything up to an hour, OT fibres were lost all too frequently during collection of data. It was essential to show that there was not undue hysteresis between the beginning and end of a run, and that the level of spontaneous activity remained stable throughout. As a precaution, annuli were therefore presented in order of increasing diameter, with a return run for intermediate annual in reverse order. Spontaneous firing levels were checked at regular intervals. Equally essential was the presence of a clearcut response minimum in subsequent graph plots, in order that these minima could be satisfactorily compared for the annulus-alone and spot/ annulus conditions. In contrast to geniculate, the result of introducing a central test spot over retinal fields was that the response minimum for concentric annular stimulation was markedly shifted towards the centre of the field, in addition to a reduction in diameter of the effective centre-surround boundary. In the particular case of the unit illustrated by Fig. 8A the response minimum to annular stimulation alone was at an annulus mean diameter of approximately 5.5. This figure was reduced to just over 3.5 by the addition of a test spot. Thus the mechanism of the retinal surround must actually extend through the field centre, or at least the overlap between centre and surround must be very substantial indeed. This conclusion, together with the nature of local field-adaptation involved, is considered in greater detail in Discussion. DISCUSSION The two major issues for discussion are: first, the precise nature of convergence from retina on to cells in the A-laminae of the dorsal division of LGN; secondly, differences in spatial distribution of the centre and surround mechanisms for receptive fields in retina and geniculate, related to this convergence. In addition, some comments are made about the possible alternatives for local adaptation by spatially selective stimulation within receptive fields, together with speculations as to the site of this adaptation. Retino-geniculate and geniculo-cortical convergence The exploded diagram of Fig. 9 presents a model for retino-geniculate convergence, proposed by Singer & Creutzfeldt (197). The model has been elaborated to take account of the contributions made by receptive field surrounds of the several retinal fibre inputs to each geniculate cell. All the available evidence which is summarized in the introduction and elsewhere (Hammond, 1972b), together with that provided by this paper, is entirely consistent with the model. Of course the diagram is not intended to be to scale and is deliberately simplified in several respects. For instance, the 5-2

16 13 P. HAMMOND number of discrete retinal fibre inputs to the geniculate field centre remains an open question. It is variously described as ranging from a single fibre (Singer & Creutzfeldt, 197; and see Results section, this paper) to one-to-several fibres (Cleland et al a). Again, the symmetry of inputs is unlikely to be as perfect as in the diagram, and it is equally unlikely that the respective fibre inputs to the geniculate surround are present without multiple overlap with one another (the geniculate sur- -round is not a single retinal field-centre diameter in width). However, the,- _N + + /~ ~ ~ ~ ~ ' + \ ( J/ +) \+ / Fig. 9. Exploded diagram of a model for convergence of retinal fibre inputs on to a single LGN cell. The resultant geniculate receptive field (lower), consisting of centre, surround and outer surround, takes account of the receptive field centres and surrounds of the constituent retinal inputs (top), spatially juxtaposed on the retina as in the composite diagram (centre). The model is equally applicable to off-centre geniculate cells. For further details see text. region of overlap between the inputs to geniculate centre and surround appears to be comparatively narrow (Hammond, 1972b) and such overlap as there may be has not been included in the diagram. A further point is that the retinal surrounds are comparatively much larger than illustrated

17 RECEPTIVE FIELD ORGANIZATION 131 (the assumption that the subtense between inner and outer diameters of the retinal surrounds is equivalent to one field-centre diameter is made so as to simplify the composite diagram at the centre of the Figure). Finally, no account is taken of the scale-changes in retinal receptive fields which are known to be related to adaptation level (Andrews & Hammond, 197b). Their consequences for the organization of geniculate cell receptive fields have already been considered elsewhere (Hammond, 1972b) - videlicet invariance of receptive field centre size and a modest increase in diameter of the surround. Unlike retina, in geniculate the changeover from phototopic to scotopic conditions is not accompanied by a reduction in sensitivity of the receptive field surround relative to that of the centre (P. Hammond, in preparation); even in total dark-adaptation the geniculate surround is still functional (Maffei & Fiorentini, 1971). However, one obvious prediction is that the outer surround of the geniculate receptive field should be eliminated by dark adaptation to levels where the retinal surrounds cease to be influential (Barlow, Fitzhugh & Kuffler, 1957) and this expectation is currently under investigation. Before passing on to the second issue for discussion it is worth considering some of the consequences of reorganization of retinal information in LGN. There is now a whole gamut of evidence which favours two parallel projections from retina to cortex, remaining essentially separate in passage through the geniculate. These two parallel systems have been variously termed X and Y cells (Enroth-Cugell & Robson, 1966), sustained and transient (Cleland et al b), or tonic and phasic (Fukada, 1971; Fukada & Saito, 1972). An alternative parallel slow- and fast-conducting axonal classification has also been made (Stone & Freeman, 1971; Stone & Hoffmann, 1971), a distinction not reflected by histological examination of fibre diameter (Stone & Holldnder, 1971). The crucial information linking the slow-conducting system with tonic neuronal activity, and the fast system with phasic activity, has been provided by Fukada (1971) and Fukada & Saito (1972). For tonic (slow) and phasic (fast) systems the geniculate receptive field surrounds are stronger than those in the retina. At both levels there is also a general tendency for cells with smaller field centres to have more potent surrounds (Wiesel, 196; Hubel & Wiesel, 1961). One corollary of this is that tonic cells, which over-all tend to have smaller field centres (Fukada, 1971; Hammond, 1972b), do in fact possess stronger surrounds than phasic cells. Equally, in geniculate the outer surround is more clearly defined for tonic neurones. The enhanced surround potency for geniculate receptive fields is consistent with sharpening-up of spatial tuning in LGN, and is significantly more marked for units with small field centres (Fig. 5). Further, Fukada (1971) makes the important observation that tonic cells

18 132 P. HAMMOND tend to occur more frequently towards the centre of the visual field. Phasic cells predominate increasingly towards the periphery and are found in isolation in the extreme periphery. Such factors tie in rather nicely with a more recent suggestion that spatial information is processed by the tonic system and temporal information by the phasic system (Fukada & Saito, 1971). Surprisingly the relationship between receptive field centre size and eccentricity in the visual field is rather weak; there does appear to be a positive correlation, but the scatter is high (unpublished observations). If spatial information is primarily processed by the tonic system, what are the implications of retino-geniculate convergence and in particular the geniculate outer surround for the receptive fields of striate cortical neurones? Hubel & Wiesel (1962) proposed a hierarchical organization for cortex, in which simple cells project serially on to complex cells. Hoffmann & Stone (1971) challenge this proposition, since they have recently demonstrated that the phasic (fast) fibre system projects to complex cells and the tonic (slow) system to simple and hypercomplex cells. The response latency of simple cortical cells to electrical stimulation of the optic chiasm or radiations is greater than for complex cells. However, the latency differences between slow and fast systems in response to electrical stimulation are trivial compared with latencies to natural stimulation. Before accepting that their results are inconsistent with a hierarchical organization of cells, it is therefore essential to know the difference in response latencies of tonic and phasic retinal elements to natural visual stimulation. It may well be that the magnitude of these latter differences is sufficient to preserve the hierarchical order. If one accepts the basic notion that simple cortical cells receive projections from tonic LGN neurones, then their receptive field profile should reflect the geniculate outer surround. A common example for illustration is a simple cortical field with a discharge centre and inhibitory side-bands. Hubel & Wiesel (1962) proposed a model embodying geniculo-cortical projection from a single row of on-centre geniculate cells. This might alone be sufficient to account for the subliminal flanks of excitation recently observed by Bishop, Henry & Smith (1971), which lie adjacent to each inhibitory side-band. In this case the subliminal excitatory flanks would be formed by the geniculate outer surrounds. There are, however, a number of factors which are inconsistent with the model for geniculo-cortical convergence proposed by Hubel & Wiesel (see also Discussion in Hammond, 1971). First, cortical cells are less responsive to diffuse illumination than LGN cells, and their inhibitory side-bands are even more powerfully influential than the surrounds of geniculate fields. If projection were from a single row of geniculate cells, one would except the relative weighting factors between geniculate centres and surrounds

19 RECEPTIVE FIELD ORGANIZATION 133 to be more-or-less preserved at cortex. Secondly, Hubel & Wiesel's model is not consistent with the observations that the spatial organization of cortical receptive fields is invariant to adaptational change (Hammond, 1971), unlike retinal fields (Andrews & Hammond, 197b), and that geniculate fields are intermediate in character (Hammond, 1972b). Third, there is the evidence for retino-geniculate convergence (Singer & Creutzfeldt, 197; Hammond, 1971, 1972a, b; Cleland et al. 1971a, b; Maffei & Fiorentini, 1971, 1972). It is an eminently reasonable assumption that a principle which is good for convergence and spatial sharpening at one level in the visual system is likely to obtain also at higher levels. Admittedly, this latter point is only circumstantial evidence. But the experimental data, together with those findings inconsistent with Hubel & Wiesel's model, are all fitted by a model in which the discharge centre of a simple cortical cell receptive field receives input from one row of geniculate cells, whereas the inhibitory side-bands receive principal projections from neighbouring rows of geniculate cells to either side. As in the case of retino-geniculate convergence, these might be of the same type (on-centre or off-centre) as those projecting to the discharge centre, or of opposite type via an inhibitory synapse or interneurone. In this model the subliminal flanks of excitation noted by Bishop and others (1971) would be primarily accounted for by the surrounds of those geniculate cells projecting to the inhibitory side-bands. Of course the prediction from this is that the outer surrounds from these same inputs would lead to yet further subliminal bands of inhibition, lying beyond the subliminal flanks of excitation, i.e. a receptive field somewhat resembling a grating pattern, with no less than seven separate, parallel zones. The consequence of such convergence would be a further stage of sharpening in the visual process, resulting in a cell relatively insensitive to luminance and optimally tuned for the processing of spatial information. Spatial distribution of receptive field mechanisms in retina and geniculate All recent work on the nature of retino-geniculate convergence supports the composite structure for a geniculate cell receptive field, summarized by Fig. 9. The locus of maximum sensitivity for the geniculate surround is a concentric circle, and there is basically no spatial overlap between centre and surround mechanisms (Hammond, 1972b) (this, of course, excepts the weak contribution which the surrounds of the retinal fibre inputs make to the centre and surround of the geniculate field, as opposed to its outer surround. In any case, it is clear from Fig. 9 that the spatial overlap of the retinal fields is such that these subsidiary contributions will act largely in the same sense as the principal inputs to the geniculate field centre and surround).

20 134 P. HAMMOND The issue is by no means as clear-cut for retinal ganglion cells, because in a spectrally non-opponent system (as in the cat) it is difficult to find a method which suitably discriminates between the mechanisms of the receptive field centre and surround. In retinae which do contain spectrally opponent cells, it is possible to achieve such a discrimination by differential wave-length stimulation and adaptation. But even here there -is controversy. A number of authors provide evidence to show that both mechanisms have maximal sensitivity at the centre of the receptive field (Hubel & Wiesel, 196; Wolbarsht, Wagner & MacNichol, 1961; Hammond, 1968), but in a class of opponent cell which possesses a centresurround organization nothing like as stereotyped as that for ganglion cells in the cat retina. Michael (1968b) provides the neatest evidence from a class of blue-green opponent cells in ground squirrel retina; the centres receive information from green-sensitive cones, and the surrounds from blue-sensitive cones which also extend through the field centre. Against this is Michael's evidence for an entirely separate class of opponent cells, also from ground squirrel retina, where green-sensitive cones are restricted to the field centres, and blue-sensitive cones to the surrounds (i.e. exactly comparable to the spatial distribution in non-opponent cat geniculate fields). Spekreijse, Wagner & Wolbarsht (1972) reach a similar conclusion for spectrally opponent cells in goldfish retina. In cat, Rodieck & Stone (1965) present a variety of indirect evidence in favour of the notion that the surround mechanism in retinal ganglion cells does extend through the field centre. One of their strongest lines of argument is that when the retinal surround disappears during darkadaptation (Barlow et al. 1957) the sensitivity of the field centre increases. This result conflicts with a comparable experiment performed by Spekreijse et al. (1972) on goldfish, in which selective spectral adaptation of the surround was unaccompanied by any change in sensitivity of the field centre (or vice versa). Species difference is the most plausible explanation. The present result is an additional pointer that the mechanism of the retinal surround passes through the receptive field centre in cat. This result and the contrasting result for cat geniculate (Hammond, 1972b) are of necessity mutually dependent on one another. In geniculate, since the response minimum for annular stimulation is not reduced by adding a central test spot, one starts from the premise that the surround mechanism does not pass through the field centre and that there is a narrow but significant boundary of overlap between centre and surround mechanisms. The test spot was always significantly smaller than the field centre, yet the effective diameter of the centre-surround boundary was reduced. This is evidence, first for selective adaptation of the centre mechanism (which one would anyway predict if the surround and centre mechanisms were

21 RECEPTIVE FIELD ORGANIZATION 135 spatially discrete), and second that the adaptation is not primarily an effect resulting from local adaptation of receptors (although this must also occur). It suggests that the adaptation must act central to the site at which the receptor inputs associated with the field-centre mechanism converge. It follows that the sensitivity of the entire central mechanism must be lowered, and not simply that part of its distribution which lies within the area of the test spot. In the case of retinal ganglion cells even should the surround mechanism extend through the field centre, one still expects selective adaptation of the centre mechanism because of its much tighter spatial distribution (in ground squirrel, Michael (1968a) has independently demonstrated that central adaptation in spectrally non-opponent, contrast-sensitive retinal units is selective for the central mechanism in that it leads to an increase in the responsiveness of the surround). So once again a reduction in the effective centre-surround boundary occurs. However, in contrast to geniculate, this change is accompanied by a reduction in diameter of the annulus most effective in evoking a response from the surround. The experiment cannot resolve whether or not the sensitivity distribution of the surround mechanism has its maximum actually at the centre of the field. If one advances this as a theory one cannot escape the follow-up that even a very small test spot will to some extent 'dent' the centre of the distribution. But the result does suggest a very substantial overlap between centre and surround mechanisms, of an entirely different order of magnitude to that seen in geniculate fields. The support of the Medical Research Council is gratefully acknowledged. I wish to thank Dennis Andrews and Janusz Kulikowski for their helpful comments on the manuscript, particularly those relating to retinal adaptation, and express my appreciation for technical assistance given by Brian Whitehouse and by Mrs J. E. Clarke. REFERENCES ANDREWS, D. P. & HAMMOND, P. (197a). Mesopic increment threshold spectral sensitivity of single optic tract fibres in the cat: cone-rod interaction. J. Physiol. 29, ANDREWS, D. P. & HAMMOND, P. (197b). Suprathreshold spectral properties of single optic tract fibres in cat, under mesopic adaptation: cone-rod interaction. J. Physiol. 29, BARLOW, H. B., FITZHUGH, R. & KUFFLER, S. W. (1957). Change of organization in the receptive fields of the cat's retina during dark-adaptation. J. Physiol. 137, BISHOP, P. O., KozAx, W. & VAKXUR, G. J. (1962). Some quantitative aspects of the cat's eye: axis and plane of reference, visual field co-ordinates and optics. J. Physiol. 163, BISHOP, P. O., HENRY, G. H. & SMITH, C. J. (1971). Binocular interaction fields of single units in the cat striate cortex. J. Physiol. 216,

22 136 P. HAMMOND CLELAND, B. G., DuBIN, M. W. & LEVICK, W. R. (1971a). Simultaneous recording of input and output oflateral geniculate neurones. Nature, New Biol. 231, CLELAND, B. G., DurBiN, M. W. & LEVICK, W. R. (1971b). Sustained and transient neurones in the cat's retina and lateral geniculate nucleus. J. Physiol. 217, ENROTH-CUGELL, CHRISTINA & RoBsoN, J. G. (1966). The contrast sensitivity of retinal ganglion cells of the cat. J. Physiol. 187, FUKADA, Y. (1971). Receptive field organization of cat optic nerve fibers with special reference to conduction velocity. Vision Res. 11, FUKADA, Y. & SAITO, H-A. (1971). The relationship between response characteristics to flicker stimulation and receptive field organization in the cat's optic nerve fibers. Vision Res. 11, FUKADA, Y. & SAITO, H-E. (1972). Phasic and tonic cells in the cat's lateral geniculate nucleus. Tohoku J. exp. Med. 16, HAMMoND, P. (1968). Spectral properties of dark-adapted retinal ganglion cells in the plaice (Pleuronectes platessa, L.). J. Physiol. 195, HAMMoND, P. (1971). Chromatic sensitivity and spatial organization of cat visual cortical cells: cone-rod interaction. J. Physiol. 213, HAMMOND, P. (1972a). Spatial organization of receptive fields of LGN neurones. J. Physiol. 222, 53-54P. HAMMOND, P. (1972b). Chromatic sensitivity and spatial organization of LGN neurone receptive fields in cat: cone-rod interaction. J. Physiol. 225, HAmmOND, P. & JAMEs, C. R. (1971). The Purkinje shift in cat: extent of the mesopic range. J. Physiol. 216, HOFFMANN, K.-P. & STONE, J. (1971). Conduction velocity of afferents to cat visual cortex: a correlation with cortical receptive field properties. Brain Res. 32, HUBEL, D. H. & WIESEL, T. N. (196). Receptive fields of optic nerve fibres in the spider monkey. J. Physiol. 154, Hu-BEL, D. H. & WIESEL, T. N. (1961). Integrative action of the cat's lateral geniculate body. J. Physiol. 155, HUBEL, D. H. & WIESEL, T. N. (1962). Receptive fields, binocular interaction and functional architecture in the cat's visual cortex. J. Physiol. 16, MAFFEI, L. & FIoRENTiNi, ADRIANA (1971). How the retinal channels build up the geniculate receptive fields. Proc. Int. Union Physiol. Sci. 9, 358 (XXV International Congress, Munich, July). MAFFEI, L. & FIORENTINI, ADRIANA (1972). Retinogeniculate convergence and analysis of contrast. J. Neurophysiol. 35, McILwMN, J. T. (1964). Receptive fields of optic tract axons and lateral geniculate cells: peripheral extent and barbiturate sensitivity. J. Neurophysiol. 27, MICHAEL, C. R. (1968 a). Receptive fields of single optic nerve fibres in a mammal with an all-cone retina. I: Contrast-sensitive units. J. Neurophysiol. 31, MICRAEL, C. R. (1968 b). Receptive fields of single optic nerve fibres in a mammal with an all-cone retina. M: Opponent colour units. J. Neurophysiol. 31, NAK.AyA, K. (1971). Local adaptation in cat LGN cells: evidence for a surround antagonism. Vision Res. 11, RODIECK, R. W. & STONE, J. (1965). Analysis of receptive fields of cat retinal ganglion cells. J. Neurophysiol. 28, SINGER, W. & CREUTZFELDT,. D. (197). Reciprocal lateral inhibition of on- and off-centre neurones in the lateral geniculate body of the cat. Expl Brain Res. 1,

23 RECEPTIVE FIELD ORGANIZATION 137 SPEKREIJSE, H., WAGNER, H. G. & WoLBARs=T, M. L. (1972). Spectral and spatial coding of ganglion cell responses in goldfish retina. J. Neurophy8iol. 35, STONE, J. & FREEMAN, R. B. Jr. (1971). Conduction velocity groups in the cat's optic nerve classified according to their retinal origin. Expl Brain Re8. 13, STONE, J. & HOFFMANN, K.-P. (1971). Conduction velocity as a parameter in the organization of the afferent relay in the cat's lateral geniculate nucleus. Brain Re8. 32, STONE, J. & HOLLXNDER, H. (1971). Optic nerve axon diameters measured in the cat retina: some functional considerations. Expl Brain Re8. 13, WIESEL, T. N. (196). Receptive fields of ganglion cells in the cat's retina. J. Phyeiol. 153, WOLBARSHT, M. L., WAGNER, H. G. & MAcNICHOL, E. F. Jr. (1961). Receptive fields of retinal ganglion cells: extent and spectral sensitivity. In The Vieual System: Neurophysiology and Psychophysic8, ed. JUNGm & KORwNHUBER, pp Berlin: Springer-Verlag.