Influence of stimulus length on directional bias of complex cells in cat striate cortex

Neuroscience Vol. 18, No. 1, pp. 25 30, 1986 Printed in Great Britain

0306-4522/86 $3.00 + 0.00 Pergamon Press Ltd © 1986 IBRO

I N F L U E N C E OF STIMULUS L E N G T H ON D I R E C T I O N A L BIAS OF COMPLEX CELLS IN CAT STRIATE C O R T E X P. HAMMOND and G. S. V. MOUAT Department of" Communication and Neuroscience, University of Keele, Keele, Staffordshire ST5 5BG, U.K. Abstract--Visual cortical cells respond optimally to an oriented bar moving either in one unique direction or in directions 180° apart. Length-dependence of this direction selectivity was investigated in the striate cortex of lightly anaesthetized cats. Approaching half of all complex cells showed some lability in their direction selectivity. The incidence was highest in standard and intermediate (length-summating) complex cells, less amongst special complex cells (those with only localized summation) and least amongst end-stopped cells, especially those of the special category. By contrast, direction selectivity of simple cells was length-independent. No correlation between a cell's overt selectivity (i.e. bidirectional, directionbiased or direction-selective) and its lability with bar length or polarity of contrast (light/dark) was evident. Moreover, since individual neurons amongst a population of complex cells could exhibit either increase, decrease, or no change of direction selectivity with length, it is unlikely that length p e r se can be coded by direction-selectivity.

Gilbert 4 segregated complex cells of the cat's striate cortex into length-summating (standard) and nonlength-summating (special) classes. Recently, we have re-examined this classificatory scheme and have had occasion to modify it. 2~8 We found that 39% of cells were special, responding optimally to contours of optimal orientation but much shorter than the minimum response field estimate of receptive field length; 43% were standard, responding optimally to bars whose length equalled or exceeded the length of the minimum response field (by its nature, the minimum response field method significantly underestimates receptive field length). However, 18% of complex cells could not be typed in this way: they had length summation characteristics intermediate between those of special and standard complex cells; possibly they represent a continuum in the length summation domain, but there is compelling evidence to suggest that they constitute a distinct third group of cells. Indeed, rigid application of the yardstick that the ratio of length summation to receptive field length must equal or exceed unity for standard complex cells raises the proportion of these unclassifiable cells from 18% to as much as 34%. Peterhans e t al. 2~ have recently described direction selectivity of simple cells in the cat's striate cortex, applying a quantitative measure which compares the strength of response peaks in a cell's preferred and opposite directions:

We have therefore adopted a slightly modified comparison, which takes account of these parameters and which we have applied primarily to complex cells. Thus, a cell's direction selectivity is defined by the ratio: (preferred - non-preferred)/ (preferred - resting discharge). [Note that it is not necessary to make allowance for resting discharge in the numerator, i.e. (preferred - resting discharge) ( n o n - p r e f e r r e d - resting discharge) since the resting discharge terms cancel out.] The ratio is zero in bidirectional cells, between zero and unity in direction-biased cells; unity in directionselective cells; and greater than unity in directionselective cells exhibiting null-suppression. A reversal of preferred direction yields a negative ratio. In the course of the previously published measurements of complex cells' length summation, 2'8 we noted that the direction bias of a significant proportion of cells varies with stimulus length. These results are reported here, together with data from a few simple cells for comparison. EXPERIMENTAL

PROCEDURES

Recordings were made from lightly anaesthetized, paralyzed, cats) ,6,~-~3 Except that animals recovered between periodic recording sessions, 5 these techniques were conventional. On a prior occasion, animals had been implanted under surgical anaesthesia with a closed chamber and a peg cemented to the skull for painless head restraint) After induction with nitrous oxide/oxygen/halothane (Fluothane, I.C.I.), animals were anaesthetized with 72.5%:27.5% nitrous oxide:oxygen plus 0.25q).6% halothane during recording. Gallamine triethiodide (Flaxedil, May and Baker) was given intravenously; mydriasis (1% atropine sulphate; Minims, Smith and Nephew), retraction of nictitating membranes (10% phenylephrine hydrochloride; Minims, Smith

(preferred - non-preferred)/preferred. This measure has its merit but also its limitations because, expressed in this way, direction selectivities cannot exceed i00% --i.e. no account is taken of resting discharge levels, or of null suppression whereby a stimulus moving in the direction opposite to that preferred by a cell may actually depress its firing below the resting level. 25

26

P. HAMMONDand G. S. V. MOUAT

and Nephew), corneal protection with unpowered contact lenses, 5 mm diameter artificial pupils and supplementary lenses ensured optical correction for a viewing distance of 57 cm. Artificial ventilation to 3.8-4.0% end-tidal carbon dioxide was maintained via a cuffed endotracheal tube. Carbon dioxide, electroencephalogram, electrocardiogram and heart rate, rectal temperature and neuronal firing rates were monitored throughout. Four-M NaCl-filled micropipettes were used to isolate single cells. Vertical electrode tracks at precisely defined spacings were made on consecutive occasions through either cerebral hemisphere (3-6.5 mm behind the interaural plane, <2 mm from the midline). Thus, recordings were safely within area 17, from the lower contralateral quadrant of the visual field, within 9° of the area centralis. Raw data were stored on tape cassettes. Waveform and polarity of neural impulses, and dot raster displays where each dot represented one impulse and each row one stimulus cycle, were monitored from storage oscilloscope displays. Spike timings were stored to 1 ms accuracy on computer disc: directional tuning curves (directions in 10° steps), averaged response histograms (50 ms/bin), and averaged spike firing (impulses/s) were derived on-line. Analyses of length summation and the derived plots of directional bias were made off-line. Stimulus orientation and width were optimized. Since directionality may vary with stimulus velocity, z° optimal velocity was used in all cases except where video-recorded stimuli were presented. In the latter case, velocity was fixed at 4°/s, which represents a reasonable compromise for striate cells. Ocular dominance, together with classical tests for distinguishing complex from simple cells t6 (including sensitivity to moving randomly textured flelds6'l°A3), and preliminary tests of length summation and end-inhibition, were assessed. Thereafter the non-dominant eye was occluded throughout. Light or dark bars were presented on a CRT display (Hewlett-Packard 1304A) at 57cm, and were swept backand-forth at optimal velocity against uniform or stationary, randomly textured backgrounds (256 x 256 elements subtending 10 x 10° square, generated at 50 frames/s; more recently 512 x 512 elements at 100 frames/s). 8 The presence of such a background has no influence, providing it remains stationary, t Average luminance was 1.1 log cd/m 2 and light and dark bars were, respectively, 0.3 log units brighter or 0.6 log units darker than the background. Animals either viewed this display directly, or video recordings of the same display presented on a monochrome TV monitor, centred on each cell's receptive field. When preparing the video recordings, the CRT display was masked by a 10° diameter circular window. Batches of stimuli were presented in pre-determined pseudorandom order. Quantitative derivations of directional preference and tuning for oriented light or dark bars moving against stationary textured or uniform backgrounds were made manually or under computer-control.6'7'~3 Standard (lengthsummating) cells,4 especially, are more finely tuned for long than for short bars: thus, to achieve the most reliable fix on preferred orientation, essential for unambiguous assessment of length summation and directional bias, tuning was always assessed with long bars exceeding the receptive field length; in end-stopped cells the longest bar consistent with a reliable response was used. 2'8 Receptive fields were mapped as rectangular "minimum response fields". 3 The extreme locations of response as a long bar of optimal orientation entered and left the receptive field defined its lateral margins. Its ends were defined with a short bar whose trajectory was systematically altered until a response was barely audible. Length summation was assessed quantitatively with bars whose orientation was constant but whose length was varied according to a pseudorandom sequence. In the case of the directly viewed CRT-display, a short bar of unvarying

length and of optimal velocity was presented alternately with a longer bar in batches of 10 or 16 trials, as a control for response variability. With video-recorded stimuli, bars of two different lengths were presented alternately (10 presentations of each at a fixed velocity of 4°/s). Each stimulus moved back-and-forth (1 s in each direction), followed by a 1 s pause. Resting discharge levels were measured over the same number of cycles, with only stationary texture or a uniform background present. Each complex cell was classified as standard (length-summating), special (optimal response to short contours) or as intermediate (length summative properties intermediate between those of standard and special complex cells), according to Gilbert's criteria4 modified in light of our own experience. 2'8 Length summation data were replotted as length vs directional bias functions for each cell. Directional bias was defined as (P - N)/(P - S), where P and N were the response strengths (in impulses/s, without correction for resting discharge) in the preferred and null directions of motion, respectively, and S was the resting discharge level of the cell. Thus, a ratio of 1 implies that a cell is direction selective, with no response in the null direction; 0 that it is bidirectional (equal responses in either direction of motion across the receptive field). Ratios > 1 imply direction selectivity together with null suppression; direction-biased cells yield ratios between 0 and I; a negative value implies a reversal of bias. RESULTS

Synopsis The results are a subset o f a large sample o f simple and complex cells recorded from superficial and deep layers o f area 17 o f 15 adult cats o f either sex (mean weight 3.4 kg; range 2.5-4.2 kg). Length s u m m a t i o n and directional bias were assessed in 81 complex cells, o f which 31 were e n d - s t o p p e d to some d e g r e e - - 3 4 special (18 end-stopped); 35 standard (10 endstopped); 12 intermediate in properties (3 ends t o p p e d ) - - i n c l u d i n g 56 cells whose length s u m m a t i o n behaviour alone has been reported previously. 2'8 D a t a for two o f the e n d - s t o p p e d complex c e l l s - - o n e special, one intermediate in c h a r a c t e r i s t i c s - - s h o w e d extreme variability and were discarded from the outset. C o m p a r a b l e data for 5 simple cells (3 end-stopped) are included for comparison. Table 1 summarizes these data for each functional class.

Variation o f direction selectivity with stimulus length Figure 1 illustrates examples o f three complex cells showing, respectively, no change, decreasing and increasing direction selectivity with increase in bar length. The c o m m o n picture was that seen in Fig. I(A)-(B), especially a m o n g s t cells totally selective for direction and lacking a resting discharge. However, changes in direction selectivity with length could be quite dramatic, the cell o f Fig. I ( C ) - ( D ) being strongly biased for direction for short bars but approaching bidirectionality for tong bars, whereas the cell o f Fig. I(E)-(F) was direction biased for short bars and direction selective for long ones. O f 79 complex cells, direction selectivity altered systematically with length in 44% (35 cells) and was un-

Directional bias in complex cells i

,

,

,

,

,

,

i

~

,

,#~1

r

i

r

27

,

i

i

i

i

i

i

i

i

I~

i: ,

4~n3

°~

~

~'~

2"

'

~

~

!

~

ooo

,~ ~

~,,

~

i

i

x o ~ ................., ............:___._~

o

g

0

[.z.1 . .

r.z.l ¢~ ,,-% o.~ e-.

¢~

J

x o L

o

>~

~

~

g

~

.~

>~

~

1.4_

~

2

o ~, ~ ~,~-~

0

~

;

L.--J

[

1

~

1

1

1

1

1

1

1

1

1

1

1

1

1

~ . ~

~ g~

/ o o m

Ji

,"

~

~ ~ ~ - ' ~ .=

m~

i

L o -o ii ..................................

I .............

~ ......................

÷ .... c o

u c o

o

o

E

o

nm m

-

m

i

morn

S

g

~~

.~ ~ ~

t--q

i

i

i

i

i

~

i

(o~s/s~ind~)

~

i

~

i

asuods~a

i

I

i

I

u~ N

i

i

i

i

X#~^~og~as

i

i

0

.~,

i

•~ ' ~

~ o ~ 0

~

~

~ .~ .-~

0

~r-~

i ~.~ ~-0

.~

c

xo®

u .

~<

i

on~

'.~ on

•v ~ o [

cm

9

i

L__J

I~J

i

i

~

L

i

i

i

o'~-,,

~=h

I’.

2x Table

HAMMOW

1. Direction

and

selectivity:

Number of cells

G

S. V.

dependence Direction

I2 h IX

I.3 4 I7

Special Special-H All special

lb I7 33

7 I4 21

Y 3 I2

Y 2 II 79

4 I 5 44

5 I 6 35

2 3 5

2 3 3

0 0 0

Suffix -H implies end-stopping

length

vs length

No change/Change

‘5 IO 35

Simple cells Simple Simple-H All simple

on stimulus selectivity

Complex cells Standard Standard-H All standard

Intermediate Intermediate-H All intermediate Totals:

MOIIAT

x

i

2

?

IO 5

3 X

7

4 0 4

3

-l

I 4

0 7

22

1;

(hypercomplexity)

changed in 56% (44 cells). Amongst complex cells, increasing direction selectivity with increase in length was more prevalent (28%; 22 cells) than decrease in selectivity (16%; 13 cells). By contrast, direction selectivity was independent of bar length in all five of

-_

.-_

Fig. 2. Length summation and direction selectivity in an end-stopped, direction-selective simple cell. Bar width 0.3’. uniform background; all other conventions as in Fig. 1.

the simple cells studied for comparison (Fig. 2), whether they were inherently direction-selective or bidirectional. There was no correspondence between a cell’s overt direction selectivity and its susceptibility to change of directionality with stimulus length. Thus bidirectional or direction-biased cells were no more (or less) likely to show fluctuations of directional selectivity with length than direction-selective cells-even when the latter exhibited null suppression. Moreover. cells in all directionality groups (whether bidirectional, direction-biased or direction-selective) were equally likely to show length-dependent increases or dccreases in direction selectivity. Length summation and direction selectivity vs length were examined for both light and dark bars in 14 cells. In all but two cases the length summation and dependence of direction selectivity on length were similar for stimuli of either contrast (see also Refs 8, II and 12). Complex cell class dependency The distribution of length dependency in direction selectivity for all classes of complex cells, including end-stopped members of each class, is shown in Table 1. With the following exceptions, there were no clearcut distinctions between the different functional groups of cells. Direction selectivity of standard and intermediate complex cells was more likely to be affected by stimulus length (49 and 55% of cells, respectively) than was the case for special complex cells (36%). This was especially noteworthy if one distinguishes between end-stopped and non-endstopped cells in each group. Direction selectivities of only a small minority of end-stopped special complex cells (18%) were susceptible to changes in stimulus length; in fact this was true to a lesser extent for end-stopped members of all classes of complex cells. In common with Gilbert,4 we noted a much higher incidence of end-stopping amongst special than amongst other classes of complex cells.

Directional

bia end-stopped cells. but actually the reverse in :I pro-

Our

results

show

that.

regardless

ap-

hall’ of all complex cells show some degree

proaching

of dependency of direction length.

of class.

However.

bctwc~

each cell’s overt

bidircctronal.

selectivity

on stimulus

there was no obvious direction

direction-biased

correlation

selectivit)

or direction-selcctivc).

to change of directional

or it\ susceptibility

(vk.

bias. and

bar Icngth. Coupled with the fact that individual from

all

complex-cell

classes

becominp mot-c. others lengrh.

it is itnplausihlc ach

were affected, some

Icss. direction-selective

encoded bq a neuron’s For

cells wirh

that length per .S(J can he direction

measurcmcnts

selectivity.

to bc reliable.

it is ob\i-

ou\ly csscntial for them to be made at precisely each cell’\ optimal slight11 >iclti

orientation.

non-optimal

an cntircly

tirstl!

Measurements

cwcntaCons difcrent

might This

result.

at

could aribc.

hecause long bar\ are more likely

s~mult;lncou> flanking

coverage

of

wcn

conccivablq

discharge

to provide centre

and

regions of the rcceptivc ticld than arc short

bar\ M hen presented other than at the cell’s optimal orxnlalion.

leading

raponsc:

\ccondl!

10

prefcri cd directions Phirdlv.

prcmaturc

a

decrement of’

preferred

beca~~se

and

in sonic cells are not

Ihe orientation

selectivity

non-

I80 apal-1.”

of length

sun-

mating cells (simple ccll~ and standard complex cells). Airpens

with

incrcasc in bar length:’ “.”

responxc

~iicread

level resulting

from

thus

an

increase in

Icngth ma) hc partly

offset by the greater response

dccrcmcnr ussoci:lU~

with mismatch

for lo11g than foi short

bars. This

i-c\ult in coniplcu ccllz in ditti-rent hcing JitYcrcntiall> thar aharpnes5 partly

in orientation

factor

also

might

cortical

I~iiinac

aKectcd, since thcrc is cvidcncc

of tuning

by Iani~na.““’ ‘_ For

ror orientation

is dictawd

this reason all our mea-

\urcmcnth v,ccl-cmade at precisely the preferred oricntatIon ~(II- each cell. and ail assessments

oforientatlon

tuning \\cre made with long. rather than short, \Vhat of the special

GISC

of‘

barb.

end-stopped cells and

rhe comparat~vc 5rabili(> of their direction selectivity’! Thcrc

i\ indepcndcnt evidence which bears on this

tinding.

Thus

inhiblrion It

Orhan

c’r ~1’~~” noted

sclectivtl!

thus

bc espcctcd to influence

non-5pcclficall)

for

;IIMI Icport end-/one inhibition 111 Ihc prcfcrrcd slanca

Morcobcr.

direction

direction

all orientations

dll-ccrlonh of mo\cmcnt. However. a h!pci-complex 0211‘5axi

end-zone

in end-stopped cells.

14;~s pan-dircc~ional

might

that

and

the same au1hoi-s

to be maximal

orientation. of motion

along

operating onl) in sonic

home non-end-stopped

in-

direction-

hl;t\e.i ccllx may in f:lct be end-stopped in the null dit-cctIc)n

Icngh

01‘ motion

amimatlon

alone. These must

fac,lctorsimply

be greater

pi-ct’crrcd than in the preferred

in

direction

the

Ihal noti-

in snmc

portion would

of

non-end-stopped

cclls~ results

which

Indeed predict Icnpth-depcndcnt \ ariation

01‘

directional bias. Other factors which might habe induced apparent length

dependency in

dircchon

driving cells inlo saluration. of

the stimulus-~trcngth

w

because stimulus

either

&zcti\lry

Include

hcyond the linear phase r-csponx

contrast

rclalionsh~p.

\\:I$ \ct loo

because of length summation /MY SC’.or

htgh.

;I ccmhinatwn

of both. In the data presented. nei1hcr explanation particularly

hkcl!.

tirstlq

because \+c r:irely

cells beyond 100 impulse\ s. which their potcnU stimuli

restricted

heca~~sc

in contrast

(~32 Eu-

Proccdurc\).

Notably. tivit;”

helol\

much

maximal tiring rates: secondly

were suitahl)

perimcnlal

\\‘;I?,

15

drove

Ihe grcatcst \artarion

Lvith lcnglh

occurred

elicited comparativclq

for

01’ dirc&on short

~~lcc-

bars.

~+cah response\.

uhtch

irathcr than

for bars of optimal length. The pi-incipal cxplanatlon for this 12. ol‘cour~c. Oiar genuine changes strength

wrc

summation

grcarcr 1%a4

l’or shorr

in intei-pretntion

Caution

for

rhan

Ii iicx.

\\hidi

Icngrh

Ionpr

1,. how\cr.

since an) nicasurc of direction that \bhich \+c ;idopM.

111rcsponsc

ham-\. lor

\~intuli.

ncccswr!,

selccri\it!.

15 inhercnll!

hcrc’

including

LI~I-CII;I~IL’ in

cases of Lveak rcsponsc. The direction

\clccrlvit;< in-

dcx is susceptible to frank mislntcrprctallon

lilr

short

bars whose Icngth\ arc subthrc\hcrld l’or ;I ccl1 /.ero response

neazisaril~

dtctalcs

an

especially in direction-sele~ti\~

iii&\

of rcsponscs in rhc pt-cfcrrcd dircstton. (e.g. at 0.35 and 0.5 the index to unit) direction-selecti\c

dircctlnn

I(AJ.

hc)\\s\cr \llght

there l-~~ng. 111 rcalil!.

With

thi\ pro\i\o.

an!

all

truly

tzcIl\ h~clded an ~n\ai-i;~n~ unit)

index (or greater. II‘ null c.p. Figs

Hut

in I’ip. I .A). ~ninii‘dt;t~cl~ r;ti\c5

without

change of sctcctl\it!.

(~1‘ ICIO.

cell\. c~cn the wcakc\t

(R)

sclcctl\it?

supprc\\ton \\crc pre\ent).

and 71).

I 131.In

LTII\

u how

dccrc;i\ed \+1111bar icngth.

the

index apprcachcd ;I m;~\;~rn~l \aluc of unit\ t’oi- ,\horl bars whose howeher.

I\ host

Icngth cwcctlcd

dirccrion

\electi\lt\

for e*atiipk,

tlic-ccl1

the index \\;Is Iowc\t for ~OI-L

thl-c4~olti. :ncrc.~\cd

(\I

I,ig

lwi \.

ccl15

iVi,h \\llll

I (I’)

IcnL$h. ;IIIC~

(f:).

30 2. Ahmed B. and Hammond P. (1984) Length summation characteristics of complex cells in cat striate cortex, hi)\4 t-otwxi is the “special” vs. “standard” classification? J. Physiol.. Land. 353, 24P. 3. Barlow H. B.. Blakemore C. and Pettipew .I. D. (1967) The neural mechanism of binocular depth dtscrmunatrorn J. Physiol., Lond. 193, 327 -342. 4. Gilbert C. D. (1977) Laminar diflerences in receptive field properties of cells in cat primary visual cortex. .I. l’hv.ricJ Lond. 268, 391421. 5. Hammond P. (1978) Inadequacy of nitrous oxide/oxygen mixtures for maintaining anaesthesia in cats: satisfactory alternatives. Pain 5, 143-l 5 I. 6. Hammond P. (1978) Directional tuning of complex cells in area 17 of the feline visual cortex. J. Phy.~iol.. Lomb. 285, 47949 1. 7. Hammond P. (1981) Simultaneous determination of directional tuning of complex cells in cat striate cortex for har and for texture motion. Exp. Bruin Res. 41, 364369. 8. Hammond P. and Ahmed B. (1985) Length summation of complex cells in cat striate cortex: a reappraisal of the “special/standard” classification. Neuroscience 15, 639-649. P. and Andrews D. P. (1978) Orientation tuning of cells in areas 17 and 18 of the cat’s visual cortex. E-X/J. 9. Hammond Brain Res. 31, 341.-3.51. P. and MacKay D. M. (1977) Differential responsiveness of simple and complex cells in cat striate cortex 10. Hammond to visual texture. Exp. Brain Res. 30, 275-296. Il. Hammond P. and MacKay D. M. (1983) Influence of luminance gradient reversal on simple cells in feline striate cortex. J. Physioi., Land. 337, 69-87. 12. Hammond P. and MacKay D. M. (1985) Influence of luminance gradient reversal on complex cells in feline striate cortex. J. Physiol., Lond. 359, 315--329. 13. Hammond P. and Smith A. T. (1983) Directional tuning interactions between moving oriented and textured stimuli in complex cells of feline striate cortex. J. Physiol., Lond. 342, 3549. 14. Henry G. H., Bishop P. 0. and Dreher B. (1974) Orientation, axis and direction as stimulus parameters for striate cells Vision Res. 14, 767-778. 15. Henry G. H., Dreher B. and Bishop P. 0. (1974) Orientation specificity of cells in cat striate cortex. J. Neuroph_v.Col. 37, 13941409. 16. Hubel D. H. and Wiesel T. N. (1962) Receptive fields, binocular interaction and functional architecture in the cat’s visual cortex. J. Physiol.. Lond. 160, 106-l 54. 17. Leventhal A. G. and Hirsch H. V. B. (1978) Receptive properties of neurones in different laminae of visual cortex of the cat. J. NeurophyCol. 41, 948-962. 18. Orban G. A., Kato H. and Bishop P. 0. (1979) End-zone region in receptive fields of hypercomplex and other striate neurones in the cat. J. Neurophysiol. 42, 818-832. 19. Orban G. A., Kato H. and Bishop P. 0. (1979) Dimensions and properties of end-zone inhibitory areas in recepttve fields of hypercomplex cells in cat striate cortex. J. Neurophysiol. 42, 833-849. 20. Orban G. A., Kennedy H. and Maes H. (1981) Response to movement of neurones in areas 17 and 18 of the cat: direction selectivity. J. NeurophJ,sio/. 45, 1059.--1073. 21. Peterhans E., Bishop P. 0. and Camarda R. M. (1985)Direction selectivity of simple cells in cat striate cortex to moving light bars, I. Relation to stationary flashing bar and moving edge responses. Exp. Bruin Res. 57, _512--522. (Acwpwd

I? Nowmber

1985)