A Brief on Scars: Good, Bad and Ugly Our skin acts as a barrier to the external environment. It pro- Common causes of disruption tects us from fluid loss, bacteria and other harmful organisms to the skin from entering our body and also from the harsh atmosphere surrounding us. It has several other functions like tempera-
Cns.iisc.ac.inInhibitory Contributions to Spatiotemporal Receptive-Field Structureand Direction Selectivity in Simple Cells of Cat Area 17 ADITYA MURTHY AND ALLEN L. HUMPHREYDepartment of Neurobiology, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania 15261 Murthy, Aditya and Allen L. Humphrey. Inhibitory contributions
spond well to motion in one direction across their receptive to spatiotemporal receptive-field structure and direction selectivity in fields and weakly or not at all to motion in the opposite simple cells of cat area 17. J. Neurophysiol. 81: 1212–1224, 1999.
direction. The mechanisms underlying this selectivity remain Intracortical inhibition contributes to direction selectivity in primary unresolved. However, among simple cells, important insights visual cortex, but how it acts has been unclear. We investigated thisproblem in simple cells of cat area 17 by taking advantage of the link have been gained through the study of spatiotemporal (S-T) between spatiotemporal (S-T) receptive-field structure and direction receptive-field structure. Many direction-selective simple cells selectivity. Most cells in layer 4 have S-T– oriented receptive fields in in cat area 17 have S-T– oriented receptive fields in which which gradients of response timing across the field confer a preferred response timing changes gradually across the field (Albrecht direction of motion. Linear summation of responses across the recep- and Geisler 1991; McLean and Palmer 1989; Movshon et al.
tive field, followed by a static nonlinear amplification, has been shown 1978; Reid et al. 1991; Saul and Humphrey 1992a). This previously to account for directional tuning in layer 4. We tested the organization confers directional tuning: a stimulus moving in a hypotheses that inhibition acts by altering S-T structure or the staticnonlinearity or both. Drifting and counterphasing sinewave gratings direction that successively activates receptive-field positions were used to measure direction selectivity and S-T structure, respec- with progressively shorter delays, or response phases, elicits a tively, in 17 layer 4 simple cells before and during iontophoresis of larger net excitatory response than a stimulus moving in the bicuculline methiodide (BMI), a GABA antagonist. S-T orientation opposite direction. In contrast, all nondirection-selective cells was quantified from fits to response temporal phase versus stimulus lack S-T– oriented receptive fields.
spatial phase data. Bicuculline reduced direction selectivity and S-T We recently showed (Humphrey and Saul 1998; Murthy et orientation in nearly all cells, and reductions in the two measures were al. 1998) that S-T structure is well correlated with directional well correlated (r 5 0.81) and reversible. Using conventional linear tuning in layer 4 of cat area 17. The degree to which cells are predictions based on response phase and amplitude, we found thatBMI-induced changes in S-T structure also accounted well for abso- S-T oriented accounts for over half of their directional tuning lute changes in the amplitude and phase of responses to gratings on average. We also showed that a linear-nonlinear, or expo- drifting in the preferred and nonpreferred direction. For each cell we nent, model (Albrecht and Geisler 1991; Heeger 1993) ac- also calculated an exponent used to estimate the static nonlinearity.
counts well for directional tuning in most layer 4 cells. The Bicuculline reduced the exponent in most cells, but the changes were model consists of two stages: a linear process in which S-T not correlated with reductions in direction selectivity. We conclude orientation confers a directional bias and a static nonlinear that GABA -mediated inhibition influences directional tuning in layer process that amplifies the bias to accentuate selectivity. The 4 primarily by sculpting S-T receptive-field structure. The source of nonlinearity may be a threshold or, equivalently, an exponen- the inhibition is likely to be other simple cells with certain spatiotem- tial amplification, either of which accentuates differences in poral relationships to their target. Despite reductions in the two response amplitude to optimal versus nonoptimal stimuli. The measures, most receptive fields maintained some directional tuningand S-T orientation during BMI. This suggests that their excitatory exponent model, however, does not account for directional inputs, arising from the lateral geniculate nucleus and within area 17, tuning in layer 6 because receptive fields there are weakly S-T are sufficient to create some S-T orientation and that inhibition ac- oriented and unrealistically large static nonlinearities are re- centuates it. Finally, BMI also reduced direction selectivity in 8 of 10 quired to account for their tuning (Murthy et al. 1998). Dy- simple cells tested in layer 6, but the reductions were not accompanied namic nonlinear interactions (Emerson and Citron 1992) likely by systematic changes in S-T structure. This reflects the fact that S-T predominate in layer 6.
orientation, as revealed by our first-order measures of the receptive Intracortical inhibition is important for direction selectivity, field, is weak there normally. Inhibition likely affects layer 6 cells via as evidenced by the fact that blocking GABA -mediated inhi- more complex, nonlinear interactions.
bition reduces selectivity in most simple cells (Sillito 1984).
How inhibition acts is not clear, however. One hypothesis isthat it creates or enhances S-T orientation. If so, then blocking inhibition should produce a reduction in S-T orientation that is The analysis of object motion in the visual world begins in correlated with a loss of directional tuning. An alternative, primary visual cortex (area 17) through the action of direction- though not mutually exclusive, hypothesis is that inhibition is selective neurons (Hubel and Wiesel 1962). These cells re- ‘‘flat,'' merely suppressing weak responses (Sato et al. 1995).
It might act by lowering membrane potentials relative to spikethreshold. This iceberg effect should enhance an initial direc- The costs of publication of this article were defrayed in part by the payment tional bias but not affect response timing. In terms of the of page charges. The article must therefore be hereby marked ‘‘advertisement''in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
exponent model, if this was the primary inhibition, then block- 0022-3077/99 $5.00 Copyright 1999 The American Physiological Society INHIBITION, S-T STRUCTURE, AND DIRECTION SELECTIVITY ing it should reduce direction selectivity and change the valueof the exponent but not alter S-T orientation.
To evaluate these two hypotheses, S-T structure and direc- tion selectivity were assessed in simple cells by measuringresponses to stationary counterphasing and drifting gratings,respectively. Bicuculline methiodide (BMI), a GABA antag- onist, was applied iontophoretically to reduce intracortical in-hibition. We observed that BMI reduced direction selectivity inmost cells. In layer 4, the effect was paralleled by a well-correlated reduction in S-T orientation. In layer 6, no system-atic changes in S-T structure were seen. We also calculated foreach layer 4 cell the value of an exponent that represents thestatic nonlinearity. Application of BMI reduced the exponentin most cells, but the reduction was not correlated with thechanges in directional tuning. Thus inhibition affects directionselectivity in layer 4 primarily by enhancing S-T orientation Example of a titration to assess the antagonistic effects of bicucul- and secondarily by accentuating the static nonlinearity. The line methiodide (BMI) on exogenously applied GABA. Plot shows 1st har- measures used here do not allow us to discern how inhibition monic amplitude in response to a grating drifting in the cell's preferreddirection. Horizontal bars mark the periods of drug application. Dashed line operates in layer 6.
(left) marks the control response during the 1st 2 min. Iontophoresis of GABAsuppressed the response within 5 min. Concurrent ejection of BMI antagonizedthe GABA effect and the control response returned. Cessation of BMI again led to GABA-induced suppression followed by a return to control-level re- sponsiveness (recovery) when the GABA ejection current was turned off.
Details of surgical preparation are described elsewhere (Murthy et quency. Control responses usually reflected ;5 trials of each phase of al. 1998; Saul and Humphrey 1990). Briefly, adult cats were anesthe- the counterphase stimulus and 12–20 trials of each direction of the tized throughout the experiment using halothane in nitrous oxide and drifting stimulus. Trial duration was 4 – 6 s. The number of trials oxygen. A tracheostomy was performed, paralysis was induced using during BMI application was variable; it depended on the latency and gallamine triethiodide and D-tubocurarine chloride, and the animal strength of the BMI effect but was usually greater than the number of was ventilated artificially. Heart rate, mean arterial blood pressure, control trials.
and the cortical electroencephalogram were monitored continuously To test the effectiveness of BMI and determine approximate levels to assess physiological state. The halothane level was adjusted to of ejection current, we performed a standard titration procedure (Sil- maintain the dominant frequencies of the electroencephalogram ,4 lito 1977, 1984) on most cells. GABA was iontophoresed at a current Hz during all stages of the experiment. Lactated Ringer solution was that just suppressed the response to a grating moving in the preferred infused intravenously to maintain hydration. The corneas were cov- direction. While continuing to eject GABA, BMI ejection current was ered with contact lenses fitted with 3-mm artificial pupils.
activated and increased until the response returned to control levels.
One such titration is shown in Fig. 1, which plots a cell's responseamplitude over time. Here, 6 nA of GABA was sufficient to suppress Recording, visual stimulation, and iontophoresis the control responses. Within 3 min of simultaneously applying BMI Extracellular recordings of single neurons were made using mi- at 12 nA, the GABA effects were reversed. Cessation of both drugs cropipettes filled with 0.6 –2.0 M KCl (;35–20 MV). The recording led to recovery of the cell's firing rate to original values within 2 min.
electrode was glued to a three-barrel micropipette array, with its tip Having determined an effective BMI current by this procedure, we protruding from the array by ;20 mm (Havey and Caspari 1980). The then used it as a starting value in the visual tests.
array tip was broken to 5–7 mm, yielding an inner diameter of ;1.5 mm per barrel. Each barrel contained one of the following solutions: Data analysis bicuculline methiodide (BMI; 2.5 mM in 165 mM NaCl, pH 5 3);gamma-amino butyric acid (GABA; 0.5 M, pH 5 3); or sodium Action potentials were summed into peristimulus time histograms acetate (2.0 M) for current balancing, to which 4% Pontamine sky (PSTHs) to measure the average response per cycle of the periodic blue was added. Drug barrels were subject to constant retaining stimulus. First harmonic (6SE) amplitude and temporal phase were currents when not in use, using 218 nA for GABA and 210 to 215 obtained for each PSTH, with response phase expressed in cycles nA for BMI. Currents were controlled using a Neurophore iontophore- relative to the stimulus (Saul and Humphrey 1990).
From responses to moving gratings we computed a directional Receptive fields initially were mapped manually on a tangent index (DI) as DI 5 (PD 2 NPD)/(PD 1 NPD), where PD and NPD screen. All subsequent stimuli were presented monocularly at 57 cm are the response amplitudes in the preferred and nonpreferred direc- on a Tektronix 608 monitor driven by a Picasso image synthesizer tions of motion, respectively. Ratios of 0 and 1, respectively, reflect linked to an LSI-11/73 computer. Mean luminance was 15 cd/m2 and no and complete direction selectivity. An estimate of the standard Rayleigh-Michelson contrast was ;0.4. Simple cells were identified error of DI was obtained from separate DI measurements on individ- using standard criteria (Hubel and Wiesel 1962; Skottun et al. 1991).
ual sets of trials. Only cells with DIs .0.33 were considered selective Drifting sinewave gratings were used to determine each cell's and subsequently tested with BMI.
optimal stimulus orientation and spatial and temporal frequency Responses to counterphasing gratings were used to characterize S-T (Humphrey and Saul 1998). A set of randomly interleaved counter- receptive-field structure. We relied primarily on a recently developed phasing and drifting gratings then was presented before (control) and method described in detail by Murthy et al. (1998). Because the during iontophoresis of BMI and, when possible, after recovery from present results require understanding the method and its rationale, we the drug (postcontrol). The counterphasing grating was presented at summarize it here. In a strictly linear model of directional tuning, a eight spatial phases over one-half cycle of the stimulus spatial fre- stationary, counterphasing sinewave grating elicits predictable pat- A. MURTHY AND A. L. HUMPHREY terns of response amplitude and phase as a function of grating position near 0.25 and 0.75 cycles. Third, the phase functions were shifted in the simple-cell receptive field. For a completely direction-selective vertically to pass through the origin. The BMI and postcontrol data cell, as the spatial phase of the grating changes, response phase were shifted similarly to maintain spatial phase correspondence in the changes monotonically with a slope of 1.
In contrast, amplitude remains constant and an amplitude modula- tion (AM) ratio, defined as [1 2 (min amp/max amp)], is 0. For a Reconstructing recording sites directionally nonselective cell, response phase is constant (i.e.,slope 5 0) within each half of the grating cycle. However, amplitude Electrode penetrations were marked by extracellular deposits of varies sinusoidally with spatial phase and the AM ratio is 1. For a cell Pontamine dye. Animals were administered a lethal dose of Nembutal with intermediate directional tuning, amplitude also fluctuates sinu- and perfused with aldehydes. Brain sections were stained for Nissl soidally and the modulation ratio lies between 0 and 1, and the substance, electrode tracks were reconstructed (Murthy et al. 1998), response phase versus spatial phase data do not follow a straight line and cells' recording locations were assigned using the laminar criteria but are fit by an arctangent function (e.g., Fig. 4A). We used this fit to of O'Leary (1941; Humphrey et al. 1985).
derive a spatiotemporal index (STI) for each cell that reflects the slopeof the function at the spatial phase generating the maximum response.
The STI is a metric that summarizes the S-T orientation of thereceptive field. STI is 1 and 0, respectively, for receptive fields that are All comparisons of means were made using a paired t-test (Miller completely S-T oriented or unoriented. In a strictly linear system, S-T and Freund 1985). Pearson product-moment (r) or Spearman rank (r ) orientation determines directional tuning; thus STI and DI values are correlations were used for other comparisons.
equal (see Murthy et al. 1998 for derivation of the relationship).
In a linear system, S-T orientation and AM to counterphased gratings are related inversely. Hence either or both measures poten- tially predict direction selectivity. However, cells are subject to non-linearities that, in the context of direction selectivity, have been Results are based on 27 simple cells from 17 cats; 17 cells modeled as static nonlinearities (Albrecht and Geisler 1991; Heeger were in layer 4 and 10 were in layer 6. We first describe 1993). They accentuate differences in amplitude to optimal versus changes in direction selectivity and S-T receptive-field struc- nonoptimal stimuli and hence increase AM ratios beyond those due to ture produced by GABA blockade in three representative linear summation. Thus conventional linear predictions of direction cells. We next summarize how the blockade affected the pop- selectivity that use response amplitude (e.g., Reid et al. 1991) under- ulation and show that effects on S-T structure differed between estimate the linear contribution because of nonlinear amplitude dis- layers 4 and 6. We then show that inhibition contributes to tortion. In contrast, response phase is not affected by static nonlin- direction selectivity in layer 4 mainly by increasing S-T ori- earities, so the phase-based measure, STI, provides a better estimate of the linear contribution. We used the STI to quantify changes in thetemporal organization of the receptive field induced during blockadeof GABA -mediated inhibition and to estimate their linear contribu- Effect of BMI on direction selectivity and S-T structure in tion to changes in direction selectivity. As described in the RESULTS, individual cells we also employed the STI to evaluate the contribution of staticnonlinear processes to directional tuning.
Figure 2A shows average responses of a layer 4 cell to one In addition to the STI, we used phase and amplitude measures to cycle of a sinewave grating drifting at 2 Hz during control, make conventional linear predictions of directional tuning for com- BMI, and postcontrol conditions. During control trials, the cell parison with our STI-based measures and to examine relationships was highly direction selective (DI 5 0.92), discharging vigor- between S-T structure and direction selectivity that require informa- ously in the preferred direction of motion (Fig. 2A, bottom) and tion about amplitude. We used a superposition method similar to thatof Jagadeesh et al. (1997). It derives from the fact that the sum of two weakly in the nonpreferred direction (top). Within 4 min of counterphasing gratings in spatial and temporal quadrature constitutes iontophoresing BMI, selectivity was abolished (DI 5 0). To a drifting grating. Assuming linearity, the responses to the counter- facilitate comparison, control responses (z z z) are superimposed phasing gratings equal those to the drifting grating. We identified pairs on the BMI data. Interestingly, the loss of direction selectivity of gratings in spatial quadrature from the eight spatial phases tested.
in this cell reflected both an increase in response amplitude to First harmonic response amplitude and phase at each spatial phase the nonpreferred direction and a decrease in amplitude to the were expressed as a vector in polar coordinates. Temporal quadrature preferred direction. Additionally, there were shifts in response was simulated by translating the response phase of one grating in each timing that were most visible in the preferred direction: the pair by a quarter cycle. The paired responses were summed vectorially response during BMI was delayed by about a quarter cycle to give predicted responses to a grating drifting in each of two relative to the control response. The drug effect was reversible directions. A mean predicted amplitude and phase1 was calculatedfrom the four quadrature pairs. Predicted amplitudes to each direction as evident in the postcontrol trials taken within 3 min of of motion also were used to derive a predicted DI.
terminating BMI. We show later that BMI-induced changes in For ease in viewing, the counterphase data from control trials were amplitude and timing to moving gratings reflect changes in the normalized in three ways. First, response phase was plotted so as to amplitude and timing structure of the receptive field.
increase with increasing spatial phase, thereby normalizing for pre- The cell's S-T structure during the three conditions is shown ferred direction of motion. Second, the response amplitude and phase in Fig. 3. The PSTHs illustrate responses to a 2-Hz counter- functions were shifted equally horizontally so that amplitude peaked phasing grating presented at different spatial phases in thereceptive field. During control trials (Fig. 3A), the receptive 1 Response phases also were normalized to compensate for the difference field displayed clear S-T orientation, as evidenced by a gradual between the spatial phases of the quadrature pairs. Taking spatial phases of 0 shift in response timing with increasing spatial phase. To and 0.25 as the first or reference pair, we subtracted 0.0625 cycles from the further illustrate this, mean phase values are plotted against second quadrature pair to simulate spatial alignment. Similarly, 0.125 and0.1875 cycles were subtracted from the third and fourth pairs, respectively.
spatial phase in Fig. 4A. An arctangent function fit to the INHIBITION, S-T STRUCTURE, AND DIRECTION SELECTIVITY Responses of 3 cells (A–C) to drifting gratings during control, BMI, and postcontrol conditions. For each condition, each pair of peristimulus time histograms (PSTHs) shows average discharge profiles to 1 cycle of the grating moving in the preferred(bottom) and nonpreferred (top) direction. For the BMI conditions, superimposed control responses (z) help to illustrate thedrug-induced changes in amplitude and/or timing. Directional index (DI) values are indicated (right) in each pair of PSTHs. Forresponses in A–C, the grating drifted at 2, 4, and 3 Hz, respectively; firing rate scaling (vertical bars) is 50, 22, and 26 impulses(ips), respectively. See text for details.
response phase data yielded an S-T orientation index (STI) as increases in response phase, and all were statistically sig- nificant (P , 0.05).
Figure 3B illustrates the changes in S-T structure during Figures 3B and 4C also reveal that the BMI-associated iontophoresis of BMI. Control responses are shown superim- response did not lag the control discharge uniformly across posed on the corresponding BMI profiles. Each pair of PSTHs spatial phase. Phase lags were greatest at zero spatial phase and is normalized to equate maximum firing rates to better illustrate progressively less up to ;0.44 cycles. Because of symmetry the relative changes in timing induced by the drug. Reduction (Movshon et al. 1978; Reid et al. 1991), the same timings were of inhibition produced clear timing changes at all spatial phas- duplicated in the second half cycle. The changes resulted in es; responses were elicited later than in control trials. These much more uniform timing within each half of the grating shifts are defined as phase lags. In Fig. 4C the lags are plotted cycle, and the receptive field became essentially S-T unori- Responses of the cell in Fig. 2A to a stationary grating undergoing sinusoidal luminance modulation at 2 Hz. Two cycles of stimulation are shown, with the second response in each PSTH being a duplicate of the first. A: control responses at 16 spatialphases spanning a grating cycle. First half-cycle was tested; symmetry allowed responses to be duplicated to complete the secondhalf-cycle. Receptive field was spatiotemporally (S-T) oriented. B: responses during application of BMI. To clarify the changes inresponse timing, BMI (—) and control (- - -) responses are superimposed, and each pair of histograms is normalized to equatemaximum firing rates. BMI produced clear changes in timing at all spatial phases, resulting in a loss of S-T orientation. C:postcontrol responses show the reestablishment of S-T orientation.
A. MURTHY AND A. L. HUMPHREY A and B: response phase and amplitude plotted as a function of stimulus spatial phase for thecontrol responses in Fig. 3A. Response phase in-creased monotonically with spatial phase. An arctan-gent fit to the phase data (— in A) yielded a spatio-temporal index (STI) of 0.62. C and D: responsesduring iontophoresis of BMI (E); control responses(F) are shown for comparison. Response phase at allpositions was significantly delayed by BMI, com-pared with controls, and the STI was reduced to 0.16.
Like response phase, amplitude changed systemati-cally, and the amplitude ratio increased from 0.67 to0.90. E and F: during the postcontrol run, S-T ori-entation and response amplitude and its modulationreturned to approximate control values. All error barsin this and the following figures indicate 61 SE.
ented (STI 5 0.16). The loss of S-T orientation would be 0.75– 0.88 cycles). However, as before, the phase shifts re- expected in a linear model of directional tuning.
sulted in more uniform timing across the receptive field.
In such a model, amplitude profiles vary systematically with The decrease in this cell's direction selectivity also was direction selectivity (see METHODS) (Murthy et al. 1998). For a accompanied by increases in response amplitude to the coun- cell with a DI of 1.0, amplitude should be constant (i.e., terphasing grating (Fig. 5D), although the AM ratio changed unmodulated) as the position of a counterphased grating little, from 0.89 to 0.81. The superposition analysis predicted a changes. As DI decreases, the degree of modulation should reduction in DI from 0.31 to 0.19. After cessation of the increase. Figure 4D shows that BMI altered the AM ratio, block, timings, S-T orientation and amplitudes re- increasing it to 0.90 from a control value of 0.67. Overall, then, turned to approximate control values (not illustrated).
the changes in phase and amplitude during GABA blockade Figure 2C illustrates the effect of reducing inhibition on a simple cell in layer 6. Like the previous example, the are consistent with a linear spatiotemporal model of direction response to each direction of motion increased during BMI selectivity. This was supported by the conventional linear but the relative change in the nonpreferred direction was predictions using the superposition method: BMI reduced pre- greater, reducing DI from 0.97 to 0.65. Strong directional dicted DI from 0.47 to 0.08.
tuning returned in the postcontrol trials. Figure 6 illustrates After cessation of the GABA antagonist, the control pattern the cell's S-T structure. Unlike the layer 4 cells, this recep- of response timings was reinstated (Figs. 3C and 4E), S-T tive field lacked prominent S-T orientation during control orientation was again clearly discernable (STI 5 0.70), and the trials (STI 5 0.14; Fig. 6, A and C). BMI induced slight AM ratio decreased to 0.75, its approximate control value. The phase leads at some positions but the timing shifts did not conventional linear prediction of direction selectivity (0.40) significantly change S-T orientation (Fig. 6, B and C). This also returned to a near-control value.
is not surprising given the initially low S-T orientation. The The aforementioned cell was one of the most striking ex- reduction in DI was accompanied by an increase in the AM amples of the effects of reducing inhibition. Another simple ratio, from about 0.67 to 0.81 (Fig. 6D). The superposition cell in layer 4, the direction selectivity of which was reduced analysis predicted a reduction in DI from 0.37 to 0.17. Thus but not abolished by BMI, is shown in Fig. 2B. This result was for this and two other layer 6 cells (not illustrated), changes the more typical one. The cell was highly direction selective in S-T structure correctly predicted the reduction in direc- (DI 5 0.93) during control trials. Within 2 min of applying tion selectivity. However, unlike layer 4 cells, the changes BMI, responses to both directions of motion increased by about largely reflected alterations in the AM ratio rather than in the same amount. However, the relative increase in response to S-T orientation. Further, in other layer 6 cells (see following the nonpreferred direction was greater, resulting in a 53% text) the minor changes in S-T structure induced by GABAA reduction in DI. The effect of BMI was reversible, as seen in blockade predicted an increase in DI but a decrease was the postcontrol data. Unlike the previous cell, these changes seen. Overall, changes in S-T structure accounted poorly for were not accompanied by any significant shift in response changes in directional tuning in this and other layer 6 cells.
phase to the drifting grating.
Figure 5, B and C, shows that the BMI-induced changes in the cell's direction selectivity reflected changes in S-T orien- EFFECT OF BMI ON DIRECTION SELECTIVITY.
Figure 7 plots the tation: STI decreased from 0.51 to 0.18. Unlike the previous DI under control versus BMI conditions for each cell. Most cell, timing changes were associated with phase leads not lags, (89%) cells lie significantly below the line of unity slope, and they did not occur at all positions. Significant shifts oc- indicating that the drug reduced their direction selectivity.
curred only between spatial phases of 0.25 and 0.38 cycles (and However, the strength of the effect varied widely, from slight INHIBITION, S-T STRUCTURE, AND DIRECTION SELECTIVITY Effect of BMI on responses of the cell in Fig. 2B to a grating counterphasing at 4 Hz. Conventions are as in Figs. 3 and 4. A: receptive field was moderately S-T oriented during control trials. B: BMI produced a leftward shift (i.e., phase lead) inresponse timing at spatial phases 0.25– 0.38 cycles (and 0.75– 0.88 cycles). C: response phase vs. spatial phase for control and BMIconditions. BMI application reduced S-T orientation. D: response amplitude versus spatial phase for the 2 conditions shows thatBMI increased amplitude.
reductions to complete loss. For 37% of the cells, DI was direction selectivity. Mean control DI was 0.80 6 ;0.05 reduced to ,0.33, our criterion for selectivity, but most of both for layers 4 and 6. During BMI application the absolute these cells maintained a directional bias. Interestingly, two value of each mean dropped to 0.41 6 0.06 and 0.51 6 0.06, cells reversed their preferred direction: one was patently selec- respectively. Both values were significantly lower than tive and the other was biased for direction.
normal (P , 0.05) but not different from each other There was no laminar difference in the effect of BMI on (P . 0.1).
Effect of BMI on responses of the layer 6 cell in Fig. 2C to a grating counterphasing at 3 Hz. A: this direction-selective receptive field was weakly S-T oriented in control trials. B: BMI produced slight phase leads at most spatial phases, relative tocontrol responses. C: response phase vs. spatial phase data reveal no significant change in S-T orientation induced by BMI. D: BMIincreased response amplitude at most spatial phases. Conventions are as in Figs. 3 and 4.
A. MURTHY AND A. L. HUMPHREY A similar analysis comparing conventional linear predictionsagainst actual direction selectivity is shown in Fig. 9B. Thechange in DI was correlated moderately with the change inpredicted DI, although the relationship was more variable (r 50.67) than between DI and STI.
We previously showed that the S-T structure of most layer 4 receptive fields accounts for a substantial fraction of theirdirectional tuning (Murthy et al. 1998), although linearpredictions nearly always underestimate actual tuning (Al-brecht and Geisler 1991; DeAngelis et al. 1993b; Reid et al.
1991). Here we asked whether S-T structure accounted for asimilar fraction of direction selectivity in control and BMIconditions despite the drug-induced changes. Figure 9Cplots the proportion of DI attributable to STI for each cell inthe two conditions. Values for 9 of the 16 cells fell on ornear the line of unity slope, indicating that S-T orientationcontributed to a similar proportion of directional tuning inboth conditions. For four additional cells (with ordinate Comparison of the directional indexes of single cells during control values of 0) STI accounted for a moderate to high proportion and BMI conditions. BMI reduced direction selectivity in most cells, but the of DI in control trials. During BMI, their STIs dropped to effect varied widely across the sample. Two cells below the zero line reversed their preferred direction of motion during drug application.
5 0.04), as did most of their DIs (mean 5 0.16), again revealing a strong dependence of directional EFFECT OF BMI ON SPATIOTEMPORAL STRUCTURE.
ences between layers 4 and 6 were observed in the action ofBMI on S-T structure. Figure 8A plots the S-T orientation ofeach cell during control and BMI conditions. During controltrials, layer 4 cells displayed a wide range of STI values, from0.16 to 0.83 (mean 5 0.41 6 0.04). During blockade ofinhibition, STI was significantly reduced in 12 of the 16 cells(mean STI 5 0.17 6 0.03; P , 0.05). Similar to the effect onDI, however, the change in STI varied among cells; a fewbecame completely S-T unoriented but most continued to dis-play an obvious spatiotemporal bias.
As noted above, direction-selective cells in layer 6 normally exhibit little or no S-T orientation. In the present sample,control STIs ranged from 0 to 0.22 (mean 5 0.09 6 0.02).
Reducing GABA -mediated inhibition did not systematically affect STI in this layer (mean STI 5 0.13 6 0.03). It wasunaltered in three cells, increased slightly in four cells, andreversed slightly in one cell.
Figure 8B summarizes the effect of BMI on predicted direc- tion selectivity, obtained using the superposition method. Al-though inclusion of amplitude in these conventional predic-tions changed the distributions relative to Fig. 8A, similartrends were observed. In layer 4, BMI caused changes in S-Tstructure in most (81%) cells that predicted a reduction in DI,from 0.33 6 0.04 to 0.21 6 0.04 (P , 0.05), on average. Inlayer 6, predicted DI was not consistently affected (controlmean 5 0.14 6 0.04; BMI mean 5 0.18 6 0.04). Because theaction of BMI on S-T structure was clearest in layer 4, wefocus on this layer for the rest of the RESULTS.
SELECTIVITY IN LAYER 4.
Figure 9 shows that the changes in direction selectivity during blockade of GABA mediated in- A: comparison of S-T orientation indexes during control and BMI hibition largely were accounted for by the alterations in S-T conditions. Most layer 4 cells (F; n 5 16) lie below the line of unity slope, receptive-field structure. Figure 9A plots the percent change in revealing that BMI reduced their STIs. In contrast, the effect of BMI on STI STI versus percent change in DI induced by BMI. Most cells in layer 6 cells (E; n 5 8) was weak and variable, with no net change across lie in the third quadrant, confirming that a reduced DI was the sample. B: conventional linear predictions of direction selectivity for thesame cells during control and BMI conditions reveals similar trends. Three almost always accompanied by a lowered STI. For these cells, cells were omitted from both scatterplots due to unreliable counterphase reductions in the two measures were well correlated (r 5 0.81).
responses during 1 condition.
INHIBITION, S-T STRUCTURE, AND DIRECTION SELECTIVITY counted for by the predictions was roughly similar in thetwo conditions for about three-fourths of the cells.
Taken together, these results indicate that GABA -mediated inhibition affects directional tuning in layer 4 largely throughchanges in spatiotemporal receptive-field structure, particularlyby increasing S-T orientation.
Because the BMI effects on direction selectivity in layer 4 largely could be accounted for by changes in S-Tstructure, we wondered whether alterations in response ampli-tude to drifting gratings could be predicted similarly from thecounterphase data. Using the superposition method, we com-puted a predicted response amplitude for each direction ofmotion in the control and BMI conditions. Subtraction of theBMI predicted amplitudes from control predicted amplitudesyielded the predicted amplitude change for each direction dueto the reduced inhibition. Likewise, the measured amplitudes todrifting gratings in the two conditions were subtracted to yieldthe observed amplitude change due to BMI.
Relationship between BMI-induced changes in DI and S-T recep- Figure 10, A and B, plots the predicted versus observed tive-field structure for layer 4 cells. A and B: plots of the percent change in DI amplitude change for the preferred and nonpreferred directions vs. percent change in STI and predicted DI, respectively, induced by BMI. For of motion, respectively. For most cells, BMI increased re- most cells, the reduction in DI was accompanied by a proportional reduction sponse amplitudes to drifting gratings for each direction. For in STI and predicted DI. One cell reversed preferred direction during BMI andis represented as a 2130% change in DI. For another cell, linear predictions both directions, most points in the sample fell on or near the incorrectly indicated a slight reversal in preferred direction, which is repre- line of unity slope, indicating that the amplitude changes were sented in B as 2110%. Correlation coefficient and slope are indicated by r and well accounted for by the predictions. Interestingly, BMI m, respectively. C and D: plots of the fraction of DI accounted for by STI and caused a decrease in amplitude in three cells (Figs. 2A and predicted DI, respectively, in control and BMI conditions. See text for de-scription.
10A), which were predicted by the changes in S-T structure. Ingeneral, BMI had a similar effect on response amplitudes in tuning on S-T orientation. For the three other cells, the individual cells to drifting and counterphasing gratings, in- fraction differed significantly in the two conditions. A sim- creasing amplitudes to both stimuli in most cells, and decreas- ilar analysis, done using conventional linear predictions, is ing it to both in the three cells. The stronger responses during shown in Fig. 9D. Again, the fraction of selectivity ac- BMI were expected given its action in reducing inhibition. The Relationship between the BMI-induced changes in response amplitude and phase to driftinggratings and that predicted from changes in S-T struc-ture. A and B: change in amplitude— either an in-crease or decrease in mean firing rate—for the pre-ferred direction, was well accounted for by the linearpredictions. A weaker relationship held for the non-preferred direction. C and D: change in responsephase— either a phase lead (negative cycle) or lag(positive cycle)–to each direction was well accountedfor by the linear predictions. Four cells for whichcontrol amplitudes were ,2 ips are omitted from Band D.
A. MURTHY AND A. L. HUMPHREY weaker responses in a few cells were surprising, and mayreflect the action of BMI on complex neural networks (e.g.,disinhibition of inhibitory neurons feeding back on the cellbeing studied).
The reduction in direction selectivity often was accompanied by a shift in the phase of response to the driftinggrating (e.g., Fig. 2A). To assess whether these timing shiftsalso could be explained by changes in S-T structure, we per-formed the superposition analysis as above but focused onresponse phase. Figure 10, C and D, plots the change inpredicted versus observed phase for the preferred and nonpre-ferred direction, respectively. For most layer 4 cells, BMIinduced a response phase lag to drifting gratings; other cellsunderwent a slight phase lead. Importantly, most of these shiftswere well predicted by the concomitant changes in S-T struc-ture.
Although the BMI-induced changes in timings and ampli- tudes across the receptive field underlie the shifts in timing tomoving stimuli, the causal relationships are not obvious fromsimple inspection of the static plots. Clearly, the response to adrifting grating reflects the convolution of the stimulus profilewith the amplitude and temporal structure of the receptivefield. For layer 4 simple cells, the superposition method cap-tures essential aspects of these S-T interactions, revealingcausal relationships between changes in S-T structure andchanges to moving gratings.
DOES BMI ALSO AFFECT DIRECTION SELECTIVITY IN LAYER 4 VIA CHANGES IN THE STATIC NONLINEARITY? Effect of BMI on the static nonlinearity, n , and its relationship structure correlated with DI in control and BMI conditions, to directional tuning. A: plot of n values in control vs. BMI conditions for 13 cells in which the exponent could be reliably estimated. BMI reduced the linear predictions underestimated DI in most cells, indicating exponent in nearly all cells. B: plot of the percent changes in n that nonlinear processes also operate in both conditions. In induced by BMI. Reductions in the 2 measures were not correlated. r , layer 4 these processes can be modeled as a static nonlinear- Spearman rank correlation coefficient.
ity—an exponent—that follows linear summation (Albrechtand Geisler 1991; Heeger 1993; Murthy et al. 1998). Here we and BMI conditions, respectively. The reduction reflected the fact asked whether reduction of inhibition altered the exponent and, that although BMI increased the AM ratio in most cells, the if so, whether the change contributed systematically to de- decrease in STI was disproportionately greater so that the STI- creased directional tuning.
predicted modulation more closely approximated that observed The nonlinearity was evaluated using a procedure developed during BMI. Figure 11B plots the percent change in n by Murthy et al. (1998). To summarize, the exponent can be for these cells. Although both measures were reduced by BMI in estimated using the amplitude and STI data. One compares the most cells, the changes were not correlated. We note that express- modulation in amplitude to a counterphased grating with the ing change in n in terms of percentage is not ideal, given its modulation expected from the STI value. As described in exponential nature. As alternatives, we analyzed change in terms METHODS, the STI fully predicts the AM ratio and the two of absolute reduction in n and DI and by normalizing for measures are inversely related if summation is strictly linear.
concomitant changes in STI. Neither procedure improved the If, however, a static nonlinearity follows the linear stage, then correlation. We conclude that although reducing GABA -medi- the modulation predicted by the STI will underestimate the ated inhibition altered both S-T structure and the static nonlinear- actual modulation. This is because the nonlinearity accentuates ity, the structural change was the more sensitive predictor of the differences in amplitude to the null and optimal spatial phases reduction in direction selectivity.
of the grating, increasing modulation. The nonlinearity can beestimated by calculating an exponent, n , which is required to raise the STI-predicted AM ratio to that observed. We previ-ously showed (Murthy et al. 1998) that n Numerous studies have shown that iontophoretic application another exponent, n , that is required to precisely match of bicuculline reduces direction selectivity (Sato et al. 1995; linearly predicted and actual direction selectivity to drifting Sillito 1975, 1977; Sillito et al. 1980; Tsumoto et al. 1979; gratings. Thus n is a reasonable estimate of the static non- Wollman and Palmer 1993), thus demonstrating an important linearity. Here we compared n in control and BMI condi- role for intracortical inhibition in generating this perceptually tions for each layer 4 cell.
important (Pasternak et al. 1985) response property. However, Figure 11A shows that BMI application reduced the value of translating these results into an understanding of how the in most cells; geometric means were 1.7 and 1.1 in control inhibition operates has proved to be more difficult. Our study INHIBITION, S-T STRUCTURE, AND DIRECTION SELECTIVITY reveals that, at least for layer 4 simple cells, inhibition sculpts 1997). The firing rates could be accounted for by applying a the spatiotemporal structure of the receptive field, accentuating simple threshold to the membrane potentials followed by a S-T orientation so as to produce greater directional tuning. In linear gain.
terms of the linear-static nonlinear exponent model, inhibition These results suggest that the reduction in n appears to operate primarily at the linear stage of spatiotem- to changes in membrane potential relative to spike threshold.
Given that inhibition acts to keep a cell's membrane potential Here we discuss the results in light of the spatiotemporal below threshold (Berman et al. 1992; Ferster and Jagadeesh model. We then consider the heterogeneity in the effect of BMI 1992), reducing inhibition by BMI should increase the propor- on direction selectivity and S-T structure, and its implications tion of time the potential rises above threshold. This will lead for excitatory mechanisms. Finally, we illustrate a connectional to increased firing rates in the preferred and nonpreferred scheme that accounts for our observations. We focus on layer directions, a change that we observed in most cells (Fig. 10).
4 because BMI effects there can be interpreted in the context of However, because the increase in the nonpreferred direction is the exponent model, although we briefly consider the layer 6 proportionately greater, due to its smaller control response, direction selectivity will decrease. The BMI-induced reductionin n thus may reflect an increase in membrane potential Understanding BMI effects in the context of the LN model relative to spike threshold.
Whereas the reductions in n help to account for height- Two aspects of our results clearly show that inhibition ened amplitudes and decreased direction selectivity, the reduc- contributes to direction selectivity by altering S-T structure.
tions were not correlated with changes in DI, unlike the First, BMI-induced reductions in selectivity nearly always changes in S-T orientation. This indicates that inhibition af- were accompanied by reductions in S-T orientation, and fects direction selectivity primarily by accentuating S-T orien- changes in the two parameters were well correlated. Second, tation. Inhibition secondarily affects the static nonlinearity, conventional linear predictions, based on response timing and probably by lowering membrane potentials relative to spike amplitude, accounted well for absolute changes in the ampli- threshold so as to suppress responses in the nonpreferred tude and phase (Fig. 10) of responses to drifting gratings. The direction (Movshon et al. 1978). An additional possibility, success of these predictions confirms other evidence (Albrecht which our data cannot address, is that inhibition also alters the and Geisler 1991; DeAngelis et al. 1993a,b; Humphrey and gain of feedforward and recurrent excitation (Suarez et al.
Saul 1998; Jagadeesh et al. 1997; McLean et al. 1994; Reid et al. 1991) for linear spatiotemporal summation as a critical The ability to dissociate the influence of BMI on linear mechanism of directional tuning in simple cells. Also, it ex- versus static nonlinear mechanisms rests on the assumption tends those population studies by showing how S-T structure that the nonlinearity does not affect response phase and hence and direction selectivity covary in single cells.
does not alter S-T orientation. This assumption is reasonable Although changes in structure account for much of the given our measure of timing: fundamental response phase. For reduction in DI, the STI and conventional linear predictions example, membrane potential fluctuations in response to sine- underestimate directional tuning in control and BMI conditions wave gratings are not always sinusoidal (cf. Figs. 7 and 8 in (Fig. 9, C and D). This is expected because, in both conditions, Jagadeesh et al. 1997). For a waveform that deviates from a nonlinearities exist that amplify directional biases produced by sinusoid, simple DC shifts in the waveform relative to spike linear spatiotemporal summation. Although this helps to ex- threshold might alter the resulting discharge profile and the plain discrepancies between predicted and observed DI, we timing of some spikes. However, the phase of the fundamental still must ask why the absolute changes in amplitude to drifting response would be affected little, as would be the S-T orien- gratings were well predicted by the linear model (Fig. 10, A tation. Only temporal shifts of the whole profile would signif- and B) given the static nonlinearity. The answer may lie in the icantly alter fundamental phase. Additionally, if a static non- fact that, for most cells, BMI reduced n , our measure of the linearity did affect response phase and S-T orientation, then nonlinearity. Thus any change in amplitude to drifting gratings BMI-induced reductions in STI should correlate with the re- should have been well predicted by the linear model. In con- ductions in n , but they did not (r 5 20.12). Thus it is trast to amplitude, the excellent predictions of response phase unlikely that changes in the static nonlinearity contributed to changes to drifting gratings are expected because response the reductions in S-T orientation. Those reductions likely re- phase is not influenced by static nonlinearities.
flect changes in spatiotemporal interactions among cells' in- Interpreting the weaker static nonlinearity (n application requires knowing what the nonlinearity reflectsbiologically. We have modeled it as an exponent but it may Heterogeneity in the effect of BMI on direction selectivity reflect a spike threshold or threshold plus amplification. Inpractice, exponents and thresholds produce similar effects: the The action of BMI on direction selectivity varied widely accentuation of differences in cell discharge rates to optimal among cells, from small reductions to complete loss. These versus nonoptimal stimuli (Albrecht and Geisler 1991; Tol- results conflict with those of Sillito et al. (1980), who reported hurst and Heeger 1997). Carandini and Ferster (1998) recently that BMI eliminated direction selectivity in all simple cells provided support for threshold and amplification as processes studied. However, our results are in general agreement with underlying the static nonlinearity in simple cells. They mea- those of Tsumoto et al. (1979), Wollman and Palmer (1993), sured the modulation of membrane potentials and discharge and Sato et al. (1995), who also reported a wide range of BMI rates to drifting gratings. Directional tuning of the spikes was effects on direction selectivity. Numerous observations indi- always greater than that of the potentials (Jagadeesh et al.
cate that the heterogeneity here was not due to methodological A. MURTHY AND A. L. HUMPHREY problems. First, potential variations between electrodes in re- manner predicted by the changing phase relationships between liably ejecting BMI partly were controlled for by performing lagged and nonlagged cells as a function of temporal frequency titrations to assess the drug's ability to antagonize exogenously (Saul and Humphrey 1992b). Further, Ferster et al. (1996) applied GABA. Second, differential effects of BMI were ob- reported that cortical cooling, designed to suppress intracorti- served even in single penetrations. For example, in one track cal interactions, did not reduce directional tuning in layer 4 four cells were tested; BMI minimally affected direction se- cells, as measured from membrane potentials in response to lectivity in the first cell but virtually eliminated it in the last drifting gratings. These studies thus indicate that geniculocor- cell. Third, BMI significantly increased the visually evoked tical inputs play an important excitatory role in constructing firing rates of nearly all cells relatively independent of the direction-selective receptive fields.
effect on direction selectivity, indicating that it effectively The geniculate inputs likely contribute to S-T structure reduced some level of inhibition. Fourth, the strengths of BMI and directional tuning in layer 4 both by their direct con- ejection currents were often less (;30 vs. 50 nA) for cells nections to simple cells (Bullier and Henry 1979; Ferster the direction selectivity of which was abolished than for cells and Lindstrom 1983; Martin and Whitteridge 1984) and by showing little effect. Fifth, for a number cells, after collecting indirect connections via other cortical cells, some of which the main set of data we continued to iontophorese BMI for up are inhibitory. In this regard our results and those of others to 60 min and raised the current intensity as high as 200 nA to (Sillito 1984) appear to conflict with the conclusion of achieve maximal block. This nearly always resulted in oscil- Ferster et al. (1996) that, at least at the membrane potential latory, bursty discharges unlinked to the visual stimulus fol- level, inhibition is not necessary to produce directional lowed by a silencing of activity. At no time did these prolonged tuning. However, the discrepancy may be less than it ap- ejections produce any reduction in directional tuning beyond pears. The average DI in our layer 4 cells during BMI that observed at lower currents. Sixth, most inhibitory synapses ejection was ;0.4. This is on the high end of the DIs are located on or near the soma (Somogyi 1989) and the measured from membrane potential fluctuations (Jagadeesh concentrations of BMI used should have effectively blocked et al. 1997). Perhaps the residual selectivity in our cells their action. This is particularly the case in layer 4, where most reflects geniculate-based directional biases of the membrane cells are relatively small and compact (Martin and Whitteridge potentials. We would expect our DIs to be higher than those 1984). Taken together, these results indicate that genuine dif- observed intracellularly because spike thresholds still affect ferences exist among cells in the contribution of GABA - the BMI data, accentuating directional biases. Nevertheless, mediated inhibition to direction selectivity and, likewise, to we also found that inhibition accentuates S-T orientation S-T structure.
and simple changes in spike threshold do not account for We did not attempt to manipulate GABA -mediated inhibi- this. Therefore we predict that the removal of inhibition by tion. However, Baumfalk and Albus (1988) reported that ion- cortical cooling should produce changes in S-T orientation tophoresis of phaclophen, a GABA antagonist, rarely altered that are observable at the membrane potential level. Unfor- direction selectivity. In addition, intracellular blockade of both tunately, no data on S-T structure during cooling exist to test chloride (GABA ) and potassium (GABA ) channels causes a this prediction. It remains to be determined whether the BMI reduction but not an elimination of direction selectivity (Nel- and cooling results are compatible with a common interpre- son et al. 1994). These, and the results given here, indicate that the directional tuning and S-T orientation remaining after Figure 12 provides a simple illustration, compatible with the blockade of inhibition reflects excitatory inputs onto simple present and previous (Saul and Humphrey 1990, 1992a) find- ings, of how an S-T well-oriented receptive field may beproduced by inputs from other simple cells having certain Sources of inhibitory and excitatory inputs to direction spatiotemporal relationships. Simulated responses to a coun- selective cells terphasing grating are shown for an excitatory (A) and inhibi-tory (B) simple cell and their target (C). Only the first half of Here we consider the nature and sources of inputs to direc- the grating spatial cycle is illustrated. Although not shown, tion-selective cells in layer 4. The inhibitory inputs are clearly responses in Fig. 12A are produced by rectified inputs from two cortical in origin and most likely arise from other simple cells LGN-like units—lagged and nonlagged— having relative spa- with receptive fields that are spatially and temporally offset tial and temporal offsets of 0.1 and 0.15 cycles, respectively.
from their targets (Maex and Orban 1996; Pollen and Ronner Profiles in Fig. 12B reflect two other LGN inputs with similar 1981). This conclusion is supported by the observation that relative offsets. The cortical receptive fields that result are each reducing inhibition changed the temporal structure of layer 4 slightly S-T oriented (STIs 5 0.23) and share the same pre- receptive fields. Complex cells lack spatiotemporally modu- ferred direction of motion. Linear summation of these two lated responses, which are necessary to produce this effect.
units (C) produces a receptive field with greater S-T orientation Excitatory inputs to cortical cells arise from the LGN and (STI 5 0.64) and hence stronger directional tuning. An arc- cortex. We previously showed (Saul and Humphrey 1990) that tangent fit to the response phase versus spatial phase data is lagged and nonlagged LGN cells (Mastronarde 1987) convey shown in Fig. 12D (F). Removing the inhibitory input to cell to cortex the range of timings needed to construct S-T oriented C would expose the excitatory structure, resulting in systematic receptive fields. The unique timing signatures of the two af- shifts in the cell's response phase and a reduction in S-T ferent groups are observed readily in simple-cell receptive orientation (E), similar to that observed experimentally (e.g., fields and can account for the progression of timings across Fig. 5C). Note that receptive fields receiving direct LGN input these fields (Saul and Humphrey 1992a). Also direction selec- may be more or less S-T oriented than shown here, depending tivity in many simple cells varies with temporal frequency in a on the range of response timings among the inputs.
INHIBITION, S-T STRUCTURE, AND DIRECTION SELECTIVITY of the timings that produce S-T orientation. Clearly, in thecat, the necessary range of timing delays is present in theLGN relay cells (Saul and Humphrey 1990). Their existenceprecludes the need to create timing delays in cortex usingpolysynaptic circuits, N-methyl-D-aspartate receptors (Maexand Orban 1996) and/or GABA receptors (Suarez et al.
Directional mechanisms in layer 6 Unlike layer 4, direction-selective cells in layer 6 display weak first-order S-T orientation, and even the addition of staticnonlinearities does not account for their directional tuning(Murthy et al. 1998). Similarly, BMI had no consistent effecton the cells' first-order S-T structure despite reducing directionselectivity. These results indicate that dynamic nonlinear in-teractions predominate in layer 6. Such interactions are detect-able using two bars flashed sequentially across the receptivefield (Emerson and Citron 1992). The second-order S-T ori-ented structures revealed by this indicate that directional tuningreflects nonlinear facilitatory and suppressive interactions inthe preferred and nonpreferred directions. An obvious predic-tion is that BMI should lessen the suppression and reducesecond-order S-T orientation.
We thank P. Baker for computer programming, M. Kieler for electronics support, and J. Feidler and A. Saul for helpful discussions. We are particularlygrateful to Dr. Kaiqi Sun for instructing us in the manufacture and use of theiontophoresis arrays and for participating in the early experiments.
This research was supported by National Eye Institute Grant EY-06459 and a Core Grant for Vision Research (EY-08098) to the Eye and Ear Institute ofPittsburgh.
Present address of A. Murthy: Dept. of Psychology, 301 Wilson Hall, Vanderbilt University, Nashville, TN 37240.
Address for reprint requests: A. L. Humphrey, Dept. of Neurobiology, University of Pittsburgh, School of Medicine, E1440 Biomedical ScienceTower, Pittsburgh, PA 15261.
Illustration of a connectional scheme that would produce a highly S-T-oriented receptive field. Simulated responses of an excitatory (A) and Received 1 July 1998; accepted in final form 24 November 1998.
inhibitory (B) simple cell and their target (C) to a counterphase grating at 8spatial phases covering one-half of the grating cycle. Responses in A reflect therectified output of 2 ON center, LGN-like units (not illustrated) the spatial phases of which, relative to 0 phase of the grating, are 0.0 and 0.1 cycles, ALBRECHT, D. G. AND GEISLER, W. S. Motion selectivity and the contrast- respectively, and the temporal phases of which are 0.05 and 0.20 cycles, response function of simple cells in the visual cortex. Vis. Neurosci. 7: respectively. B: spatial and temporal phases of 2 other ON center inputs are 0.25 531–546, 1991.
and 0.25 cycles, respectively, for 1 input and 0.35 and 0.05 cycles for the other.
BAUMFALK, U. AND ALBUS, K. Phacolofen antagonizes baclofen-induced sup- Spatiotemporal offsets of the afferents produce slight S-T orientation in the pression of visually evoked responses in the cat's striate cortex. Brain Res. recipient receptive fields. Linear summation of the rectified outputs of the 2 463: 398 – 402, 1988.
cortical cells produces a third receptive field (C) with greater S-T orientation.
BERMAN, N. J., DOUGLAS, R. J. AND MARTIN, K.A.C. GABA-mediated inhibi- STI values are indicated for each receptive field. D: arctangent fit to the phase tion in the neural networks of visual cortex. In: Progress in Brain Research. data in A (E) and C (F), to illustrate how removal of inhibition would change GABA in the Retina and Central Nervous System, edited by R. R. Mize, timing and reduce S-T orientation in cell C.
R. E. Marc, and A. M. Sillito. New York: Elsevier, 1992, vol. 90, p.
This illustration does not address obvious complexities such as spatially opponent inhibition, excitatory and inhib- BULLIER, J. AND HENRY, G. H. Laminar distribution of first-order neurons and afferent terminals in cat striate cortex. J. Neurophysiol. 42: 1271–1281, itory feedback at all of the illustrated stages, and the large numbers of neurons that must interact. These and other CARANDINI, M. AND FERSTER, D. The iceberg effect and orientation tuning in spatiotemporal interactions have been modeled by others cat V1. Invest. Ophthalmol. Vis. Sci. 39: S239, 1998.
(Maex and Orban 1996; Suarez et al. 1995). Maex and DEANGELIS, G. C., OHZAWA, I., AND FREEMAN, R. D. Spatiotemporal organi- Orban (1996) have shown that the S-T interactions can zation of simple-cell receptive fields in the cat's striate cortex. I. Generalcharacteristics and postnatal development. J. Neurophysiol. 69: 1091–1117, account for many stimulus-dependent behaviors of direc- tion-selective cells. Here we illustrate excitation and inhi- DEANGELIS, G. C., OHZAWA, I., AND FREEMAN, R. D. Spatiotemporal bition as sharing the same preferred direction, but in prin- organization of simple-cell receptive fields in the cat's striate cortex. II.
ciple cross- and nondirectional inhibition could interact with Linearity of temporal and spatial summation. J. Neurophysiol. 69: 1118 –1135, 1993b.
excitatory inputs to produce S-T– oriented structure. A key EMERSON, R. C. AND CITRON, M. C. Linear and nonlinear mechanisms of difference between our model (Saul and Humphrey 1992a) motion selectivity in simple cells of the cat's striate cortex. In: Non- and that of Maex and Orban (1996), however, is the source linear Vision Determinants of Neural Receptive Fields, Function and Net- A. MURTHY AND A. L. HUMPHREY works, edited by R. B. Pinter and B. Nabet. Boca Raton: CRC, 1992, p.
O'LEARY, J. L. Structure of the area striata of the cat. J. Comp. Neurol. 75: 131–164, 1941.
FERSTER, D., CHUNG, S., AND WHEAT, H. Orientation selectivity of thalamic PASTERNAK, T., SCHUMER, R. A., GIZZI, M. S., AND MOVSHON, J. A. Abolition input to simple cells of cat visual cortex. Nature 380: 249 –252, 1996.
of visual cortical direction selectivity affects visual behavior in cats. Exp. FERSTER, D. AND JAGADEESH, B. EPSP-IPSP interactions in cat visual cortex Brain Res. 61: 214 –217, 1985.
studied with in vivo whole-cell patch recording. J. Neurosci. 12: 1262–1274, POLLEN, D. A. AND RONNER S. F. Phase relationships between adjacent simple cells in the visual cortex. Science 212: 1409 –1411, 1981.
FERSTER, D. AND LINDSTROM, S. An intracellular analysis of geniculo- REID, R. C., SOODAK, R. E., AND SHAPLEY, R. M. Directional selectivity and cortical connectivity in area 17 of the cat. J. Physiol. (Lond.) 342: 181–215, spatiotemporal structure of receptive fields of simple cells in cat striate cortex. J. Neurophysiol. 66: 505–529, 1991.
HAVEY, D. C. AND CASPARI, D. M. A simple technique for constructing ‘‘piggy-back'' multibarrel microelectrodes. Electroencephalogr. Clin. Neu- ATO, H., KATSUYAMA, N., TAMURA, H., HATA, Y., AND TSUMOTO, T. Mech- rophysiol. 48: 249 –251, 1980.
anisms underlying direction selectivity of neurons in the primary visual cortex of the macaque. J. Neurophysiol. 74: 1382–1394, 1995.
EEGER, D. J. Modeling simple-cell direction selectivity with normalized, half-squared, linear operators. J. Neurophysiol. 70: 1885–1898, 1993.
SAUL, A. B. AND HUMPHREY, A. L. Spatial and temporal response properties of HUBEL, D. H. AND WIESEL, T. N. Receptive fields, binocular interaction and lagged and nonlagged cells in the cat lateral geniculate nucleus. J. Neuro- functional architecture in the cat's visual cortex. J. Physiol. (Lond.) 160: physiol. 64: 206 –224, 1990.
106 –154, 1962.
SAUL, A. B. AND HUMPHREY, A. L. Evidence of input from lagged cells in the HUMPHREY, A. L. AND SAUL, A. B. Strobe rearing reduces direction selectivity lateral geniculate nucleus to simple cells in cortical area 17 of the cat.
in area 17 by altering spatiotemporal receptive-field structure. J. Neuro- J. Neurophysiol. 68: 1190 –1207, 1992a.
physiol. 80: 2991–3004, 1998.
SAUL, A. B. AND HUMPHREY, A. L. Temporal frequency tuning of direction HUMPHREY, A. L., SUR, M., UHLRICH, D. J., AND SHERMAN, S. M. Projection selectivity in cat visual cortex. Vis. Neurosci. 8: 365–372, 1992b.
patterns of individual X- and Y-cell axons from the lateral geniculate SILLITO, A. M. The contribution of inhibitory mechanisms to the receptive field nucleus to cortical area 17 in the cat. J. Comp. Neurol. 233: 159 –189, 1985.
properties of neurones in the striate cortex of the cat. J. Physiol. (Lond.) 250: JAGADEESH, B., WHEAT, H. S., KONTSEVICH, L. L., TYLER, C. W., AND FERSTER, 305–329, 1975.
D. Direction selectivity of synaptic potentials in simple cells of cat visual SILLITO, A. M. Inhibitory processes underlying the directional specificity of cortex. J. Neurophysiol. 78: 2772–2789, 1997.
simple, complex, and hypercomplex cells in the cat's visual cortex.
MAEX, R. AND ORBAN, G. A. Model circuit of spiking neurons generating J. Physiol. (Lond.) 271: 775–785, 1977.
directional selectivity in simple cells. J. Neurophysiol. 75: 1515–1545, SILLITO, A. M. Functional considerations of the operation of GABAergic inhibitory processes in the visual cortex. In: Cerebral Cortex, edited by MARTIN, K.A.C. AND WHITTERIDGE, D. Form, function, and intracortical pro- E. G. Jones and A. Peters. New York: Plenum Press, 1984, vol. 2, p. 91–117.
jections of spiny neurones in the striate cortex of the cat. J. Physiol. (Lond.) SILLITO, A. M., KEMP, J. A., MILSON, J. A., AND BERARDI, N. A re-evaluation 353: 463–504, 1984.
of the mechanisms underlying simple cell orientation selectivity. Brain Res. MASTRONARDE, D. N. Two classes of single-input X-cells in cat lateral genic- 194: 517–520, 1980.
ulate nucleus. I. Receptive-field properties and classification of cells. J. Neu- SKOTTUN, B. C., DEVALOIS, R. L., GROSOF, D. H., MOVSHON, J. A., ALBRECHT, rophysiol. 57: 381– 413, 1987.
D. G., AND BONDS, A. B. Classifying simple and complex cells on the basis MCLEAN, J. AND PALMER, L. A. Contribution of linear spatiotemporal receptive of response modulation. Vision Res. 31: 1079 –1086, 1991.
field structure to velocity selectivity of simple cells in area 17 of cat. Vision SOMOGYI, P. Synaptic organization of GABAergic neurons and GABA Res. 29: 675– 679, 1989.
ceptors in the lateral geniculate nucleus and visual cortex. In: Neural MCLEAN, J., RAAB, S., AND PALMER, L. A. Contribution of linear mechanisms Mechanisms of Visual Perception, edited by D.M.K. Lam and C. D. Gilbert.
to the specification of local motion by simple cells in areas 17 and 18 of the Houston: Gulf, 1989, p. 35– 62.
cat. Vis. Neurosci. 11: 271–294, 1994.
SUAREZ, H., KOCH, C., AND DOUGLAS, R. Modeling direction selectivity of MILLER, I. AND FREUND, J. E. Probability and Statistics for Engineers. New simple cells in striate visual cortex within the framework of the canonical Delhi: Prentice-Hall, 1985.
microcircuit. J. Neurosci. 15: 6700 – 6719, 1995.
MOVSHON, J. A., THOMPSON, I. D., AND TOLHURST, D. J. Spatial summation in TOLHURST, D. J. AND HEEGER, D. J. Comparison of contrast-normalization and the receptive fields of simple cells in the cat's striate cortex. J. Physiol. threshold models of responses of simple cells in cat striate cortex. Vis. (Lond.) 283: 53–77, 1978.
Neurosci. 14: 293–309, 1997.
MURTHY, A. N., HUMPHREY, A. L., SAUL, A. B., AND FEIDLER, J. C. Laminar TSUMOTO, T., ECKART, W. AND CREUTZFELDT, O. D. Modification of orientation differences in the spatiotemporal structure of simple cell receptive fields in sensitivity of cat visual cortex neurons by removal of GABA-mediated cat area 17. Vis. Neurosci 15: 239 –256, 1998.
inhibition. Exp. Brain Res. 34: 351–363, 1979.
NELSON, S., TOTH, L., SHETH, B., AND SUR, M. Orientation selectivity of WOLLMAN, D. E. AND PALMER, L. A. The effects of GABA blockade on the cortical neurons during intracellular blockade of inhibition. Science 265: spatiotemporal receptive field structure of neurons in cat striate cortex.
774 –777, 1994.
Invest. Ophthal. Vis. Sci. 34: S908, 1993.
MICROMEDEX® HEALTHCARE SERIES User Guide Copyright © 1974-2008 Thomson Healthcare. All rights reserved. This manual, as well as the data and software implementation described in it, is furnished under license and may be used or copied only in accordance with the terms of such license. The content of this manual is furnished for informational use only, is subject to change without notice, and should not be construed as a commitment on the part of Micromedex.