Context-Sensitive Binding by the Laminar Circuits of V1 and V2 Unified Model Perceptual Gro

Context-Sensitive Binding by the

Laminar Circuits of V1and V2:

A Uni?ed Model of Perceptual Grouping,

Attention,and Orientation Contrast

Rajeev D.S.Raizada and Stephen Grossberg

Department of Cognitive and Neural Systems

Boston University

677Beacon Street

Boston,MA02215

Phone:617-353-7858or-7857

Fax:617-353-7755

Email:rajeev@https://www.wendangku.net/doc/847357913.html,,steve@https://www.wendangku.net/doc/847357913.html,

Visual Cognition,forthcoming.

Originally submitted:July18,2000.Revised:November3,2000

Technical Report CAS/CNS TR-2000-008

Keywords:visual cortex,attention,grouping,orientation contrast,cortical layers,V1,V2,feedback, neural network

Abstract

A detailed neural model is presented of how the laminar circuits of visual cortical areas V1and V2implement context-sensitive binding processes such as perceptual grouping and attention.The model proposes how speci?c laminar circuits allow the responses of visual cortical neurons to be determined not only by the stimuli within their classical receptive?elds,but also to be strongly in?uenced by stimuli in the extra-classical surround.This context-sensitive visual processing can greatly enhance the analysis of visual scenes,especially those containing targets that are low con-trast,partially occluded,or crowded by distractors.We show how interactions of feedforward, feedback and horizontal circuitry can implement several types of contextual processing simultane-ously,using shared laminar circuits.In particular,we present computer simulations which suggest how top-down attention and preattentive perceptual grouping,two processes that are fundamental for visual binding,can interact,with attentional enhancement selectively propagating along group-ings of both real and illusory contours,thereby showing how attention can selectively enhance object representations.These simulations also illustrate how attention may have a stronger facilita-tory effect on low contrast than on high contrast stimuli,and how pop-out from orientation contrast may occur.The speci?c functional roles which the model proposes for the cortical layers allow sev-eral testable neurophysiological predictions to be made.The results presented here simulate only the boundary grouping system of adult cortical architecture.However,we also discuss how this model contributes to a larger neural theory of vision which suggests how intracortical and intercor-tical feedback help to stabilize development and learning within these cortical circuits.Although feedback plays a key role,fast feedforward processing is possible in response to unambiguous in-formation.Model circuits are capable of synchronizing quickly,but context-sensitive persistence of previous events can in?uence how synchrony develops.Although these results focus on how the interblob cortical processing stream controls boundary grouping and attention,related modeling of the blob cortical processing stream suggests how visible surfaces are formed,and modeling of the motion stream suggests how transient responses to scenic changes can control long-range apparent motion and also attract spatial attention.

1Introduction:Contextual effects and binding in visual cortex

This article continues the development of a neural model aimed at providing a uni?ed explana-tion of how the laminar circuits of visual cortical areas V1and V2interact with the LGN to control cortical development,learning,perceptual grouping,and attention(Grossberg,Mingolla,&Ross, 1997;Grossberg,1999a;Grossberg&Raizada,2000;Grossberg&Williamson,2000;Ross,Mingolla, &Grossberg,2000).In particular,the model has begun to clarify how preattentive and attentive perceptual mechanisms are intimately linked within the laminar circuits of visual cortex,notably how bottom-up,top-down,and horizontal cortical connections interact within the cortical layers. To this end,we quantitatively simulate a number of phenomena about visual contextual process-ing,contrast-sensitive grouping and attention to illustrate the model’s predictive power.In this regard,it has long been known that a neuron’s response to stimuli inside its classical receptive?eld (CRF)can be strongly in?uenced by stimuli outside in the surround(e.g.Blakemore&Tobin,1972; Maffei&Fiorentini,1976;Nelson&Frost,1978).Only more recently,however,has the functional importance of these contextual effects for real-world visual processing been widely appreciated (e.g.von der Heydt et al.,1984;Grossberg&Mingolla,1985;Allman et al.,1985;Gilbert&Wiesel, 1990;Knierim&Van Essen,1992;Grossberg,1994;Kapadia et al.,1995;Lamme,1998;Sillito et al., 1995;Polat et al.,1998;Sugita,1999).

a b

c

Figure1:Contrast-dependent perceptual grouping in primary visual cortex,showing how collinear?ankers have a net facilitatory effect at low stimulus contrasts,but then“crossover”into being net inhibitory at high contrasts.(a): Example stimuli of the sort used by Kapadia et al.(1998),consisting of three bars of equal contrast:a central target bar,and two collinear?ankers.The image shown here is an actual stimulus that was presented to the model network.

(b):Cross-section of V1layer2/3neural activity in the model,in response to low contrast bar stimuli.The solid line shows activity when the target and?ankers are presented together,with the responses to each of the bar corresponding to a“hump”of activity in the cross-section.Above-threshold layer2/3groupings form between the collinear bars,as shown by the regions of non-zero activity?lling the inter-bar spaces.The dotted line shows the neural response to the central target bar alone,presented without any?ankers.It can be seen that the target bar elicits more activity when the ?ankers are present,showing that the grouping has a net facilitatory effect at this low stimulus contrast.(c):Same set of V1layer2/3cross-sections as in(b),but now with all three bars presented at high contrast.Strong above-threshold collinear groupings form between the bars,but the net effect of the?ankers on the target is nonetheless inhibitory.By “net inhibitory”,we mean that the the central stimulus elicits a weaker response when the?ankers are present than when they are absent,with“net facilitatory”meaning the reverse.Thus,the net facilitatory effect of the?ankers in panel(b) can be seen by the fact that the solid with-?ankers activity trace is above the dotted without-?ankers line.In panel(c), the relative positions of these two lines have switched,indicating that the?ankers have now“crossed over”into being net inhibitory.

A particularly vivid example of a contextual effect is the collinear grouping of oriented stimuli, which enhances the detection of grouped targets(Polat&Sagi,1993;Kapadia et al.,1995)and of smooth contours(Field,Hayes,&Hess,1993;Kovacs&Julesz,1993),and which also gives rise to the percept of illusory contours(Kanizsa,1979)when the inducing stimuli also cause a brightness difference across the two sides of the collinear group(Grossberg&Mingolla,1985;Grossberg,1994). Psychophysical evidence suggests that grouping occurs without the need for top-down attention (Moore&Egeth,1997).Perceptual grouping mechanisms are particularly needed for detecting tar-gets which are surrounded by distractors or which are of low contrast.The relevance of contrast for such grouping is further illustrated by recent neurophysiological studies of cortical area V1which have shown that the contextual effects are contrast-dependent,with low-contrast targets being fa-cilitated by collinear?ankers,but high-contrast targets being depressed,as shown in Figure1(Polat et al.,1998;Kapadia,Westheimer,&Gilbert,1998).

Top-down attention can also be viewed as a form of contextual processing,in that it plays an impor-tant role when a target is surrounded by distractors,but may have much less effect when a target is presented on its own(Motter,1993;De Weerd et al.,1999).Attentional effects have been observed

throughout visual cortex,including many recent studies of attention in V1(e.g.Roelfsema et al., 1998;Watanabe et al.,1998;Somers et al.,1999;Ito&Gilbert,1999;Brefczynski&DeYoe,1999).Like collinear grouping,attention also has its greatest facilitatory effect when the target is low contrast, as illustrated in the study by De Weerd et al.(1999,data shown in Figure2c).Moreover,attention interacts in important ways with other contextual effects,in particular with grouping:attention can spread itself along visual groupings(He&Nakayama,1995;Davis&Driver,1997),and can propagate along both real and illusory contours(Roelfsema&Spekreijse,1999;Moore,Yantis,& Vaughan,1998),as illustrated in Figures4and5.We have elsewhere argued that top-down at-tention and related feedback pathways are mechanisms whereby the cortex can stabilize its initial development and subsequent learning(Grossberg,1980,1999a,1999b;Grossberg&Williamson, 2000).

Another important contextual effect is orientation contrast,in which an element whose orientation differs from that of its neighbors“pops out”from the background.Such effects have been observed psychophysically,and also neurophysiologically in V1(Nothdurft,1991;Knierim&Van Essen, 1992;Sillito et al.,1995;Nothdurft et al.,1999).Like grouping,this mechanism is particularly useful for picking out targets which are surrounded by distractors,as shown for example in Figure3. The process of visual binding is very closely related to these contextual processes;it too is needed most when the visual scene is cluttered with distractors.Indeed,when there is just a single visual element in a display the binding problem cannot arise.It is therefore unsurprising that many exper-imental studies of binding have concentrated on the effects discussed above,in particular grouping and attention(for recent reviews,see Gray(1999)on the relations between binding and Gestalt grouping,and Reynolds and Desimone(1999)and Treisman(1999)on binding and attention). Discussions of the neural processes underlying binding are often somewhat vague.In this article, we will attempt to address this problem by proposing speci?c and testable neurophysiological sub-strates for two visual processes which are fundamental for binding,namely attention and grouping. In particular,we suggest detailed laminar circuits in V1and V2for implementing these processes, and propose ways in which they can interact with each and with other visual contextual https://www.wendangku.net/doc/847357913.html,puter simulations from a neural network implementation of this architecture will be presented,demonstrating the viability of the proposed scheme and illustrating the details of its op-eration.Although the model proposed here concentrates primarily on the spatial aspects of bind-ing,rather than possible relations to temporal phenomena such as neural synchronization,it has elsewhere been shown that variants of this model are capable of rapidly synchronizing their emer-gent states during both perceptual grouping and attentional focusing;see Grossberg and Somers (1991)and Grossberg and Grunewald(1997).An important but often overlooked aspect of visual binding is also addressed by the present model:the question of how the information distributed across different cortical regions,and across different cortical layers within the same region,can be bound together.We suggest speci?c mechanisms of intercortical and intracortical feedback which allow the different layers and regions of cortex to mutually in?uence and even synchronize their visual processing.On the other hand,our results also illustrate how a very fast bottom-up sweep of information through the cortex can be suf?cient if the visual stimuli are suf?ciently unambiguous.

1.1Doing different types of contextual processing at once:the preattentive/attentive

interface problem for cortex and for cortical models

The neurophysiological studies mentioned above provide compelling evidence that the processes of preattentive perceptual grouping and top-down visual attention coexist within the same cortical

a

Input to network

b Attentional feedback to target c

d Model simulation

Figure 2:Attention has a stronger facilitatory effect on low contrast stimuli than it does at high contrasts,as shown in the study by De Weerd et al.(1999)and the model’s simulation of it.(a):Example stimuli of the sort used by De Weerd et al.,consisting of a variable-contrast oriented grating surrounded by three distractor discs.The image shown here is an actual stimulus that was presented to the model network.(b):Attentional feedback directed in the model to the location of the target grating,implemented simply as a diffuse Gaussian of corticocortical feedback activity.(c):Data from the macaque study,reproduced with permission from De Weerd et al.(1999,Fig.3b).The solid line with circles shows the monkeys’orientation discrimination thresholds when the target and distractors were presented in an unlesioned visual quadrant,hence with intact top-down attention.Task performance was very good across all conditions,even when the target grating was very low contrast.The dashed line with squares shows that when the stimuli were presented in a visual quadrant from which V4had been lesioned,hence impairing top-down attention,task performance was still good at high grating contrasts,but degraded signi?cantly as the contrast reduced.Hence,top-down attention has more of an effect on low contrast stimuli.(d):Model simulation of the Deweerd et al.data.In the model,attention can simply be turned on and off,rather than by having to lesion any higher-level cortical areas.Indeed,these higher task-encoding areas,presumably in prefrontal and inferotemporal cortex,are not simulated in the present model,which considers only V1and V2.Thus,the Gaussian of attention is positioned over the target grating by specifying its coordinates in the simulation computer program,rather than by being steered by a simulated higher cortical area.The network’s “behavioral threshold”is simply operationalized as the reciprocal of V1layer 2/3oriented activity,since these are the cells which pass information about the grating’s orientation forward to higher areas.

a

Bar alone b

Iso-orientation c

Cross-orientation d

e

Model simulation

Figure 3:Orientation contrast in V1and in the neural model.(a-c):Stimuli of the sort used by Knierim and Van Essen (1992).The neural responses elicited by an isolated bar are recorded,then compared with responses when the same bar is embedded either in an iso-orientation or cross-orientation texture surround.The images shown here are the actual stimuli that were presented to the model network.(d):Neurophysiological data from macaque V1,adapted with permission from Knierim and Van Essen (1992,Fig.10).The icons along the -axis indicate that the stimuli presented were of the sorts shown in (a),(b)and (c)respectively.It can be seen that both sorts of texture surrounds have a suppressive effect on neural activity,compared to when the bar is presented on its own,but that the orthogonal surround produces less suppression.This is consistent with the perceptual effect that the bar seems to “pop-out”from the orthogonal background but not from the iso-orientation surround.(e):Model simulation of the orientation contrast effect.It can be seen here too that both kinds of surround have a net inhibitory effect,with the cross-orientation surround being less suppressive.

areas,namely V1and V2.

We wish to argue that the ability of the cerebral cortex to implement these different contextual processes all at once,within the same brain areas is a more non-trivial functional feat than is widely appreciated.In particular,although the individual tasks of implementing attention and perceptual groupings such as illusory contour grouping may,considered separately,be relatively tractable,

the task of performing both processes at once within the same cortical circuit raises the dif?cult problem of distinguishing the preattentive from the attentive,the external from the internal:the cortex must be able to tell the difference between activity that conveys information about objects in the environment as opposed to activity that has arisen purely as a result of top-down cortical processing.

For attention,this problem is as follows:top-down attention can enhance the?ring of cells which are already active,but if it were to produce above-threshold activity in the absence of any bottom-up retinal input,then the brain would be in danger of hallucinating:activity in V1and V2gets passed up to higher areas regardless of how it was caused,and these higher areas would have no means of telling the internally and externally created signals apart.It has,in fact,been elsewhere suggested how a breakdown in this process can lead to hallucinations,such as during the positive symptoms of schizophrenia(Grossberg,2000).

Four possible mechanisms would each seem to provide plausible solutions to this problem.How-ever,we will argue that they all fail,and that a more subtle solution utilizing the laminar architec-ture of cortex must be used instead.

Firstly,it seems that cortex could ensure that top-down attention on its own never produces above-threshold activity simply by keeping attentional feedback very weak.However,numerous physio-logical studies show that attention can exert extremely powerful effects in visual cortex,for example modulating the activity of MST cells by113%(Treue&Maunsell,1996).As well as being strongly facilitatory,attention can also be strongly suppressive,causing neurons to respond weakly even when their receptive?elds contain stimuli that would otherwise elicit optimal responses(Reynolds, Chelazzi,&Desimone,1999).

A second possibility,often adopted by other computational models,would be to make top-down feedback have a purely multiplicative effect on cortical?ring(e.g.Neumann&Sepp,1999),for instance by having feedback act exclusively on NMDA channels,which open only when the post-synaptic cell is active(e.g.Lumer,Edelman,&Tononi,1997).This would ensure that only already existing activity could be enhanced.However,this possibility fails to account for the fact mentioned above that attention can be inhibitory,as well as facilitatory.In particular,there is psychophysical and neurophysiological evidence that attention has a facilitatory on-center and suppressive off-surround form(Downing,1988;Caputo&Guerra,1998;Mounts,2000;Smith,Singh,&Greenlee, 2000;Vanduffel,Tootell,&Orban,2000).Moreover,there is evidence that corticocortical feedback axons act on both non-NMDA and NMDA channels(Cauller&Connors,1994).

The third possible way of solving the preattentive/attentive interface problem would by if cortex were to enforce the simple rule that only those cells whose CRFs contain visual stimuli should be allowed to be active.However,here the functional dif?culties of simultaneously implementing multiple types of contextual processing start to become apparent.This would-be rule is disobeyed by neurons that respond to Kanizsa-type illusory contours.Such neurons give above-threshold responses without having any visual stimuli within their CRFs,and are known to exist in V2(von der Heydt et al.,1984;Peterhans&von der Heydt,1989)and possibly also in V1(Nguyen&Lee, 1999).Responses to illusory contours induced by offset gratings have also been found in V1(Redies, Crook,&Creutzfeldt,1986;Grosof,Shapley,&Hawken,1993;Sheth,Sharma,Rao,&Sur,1996). Although the receptive?elds of neurons responding to such stimuli are not completely empty,since they contain line endings,they do not contain any stimuli which have the same orientation as the illusory contour itself.

Thus,cortex is faced with the problem of ensuring that top-down attention can have only a mod-

ulatory effect on bottom-up stimuli,even though groupings like an illusory contour can generate suprathreshold responses at positions that do not receive bottom-up inputs.A fourth possible so-lution,then,might be simply to ensure that attentional and perceptual grouping are kept?rmly separated in cortical processing.However,as mentioned above and illustrated in Figures4and 5,there exists neurophysiological and psychophysical evidence that attention actually propagates along both real and illusory contour groupings(He&Nakayama,1995;Moore et al.,1998;Roelf-sema et al.,1998;Roelfsema&Spekreijse,1999).Thus attention and grouping are intimately linked within the same neural circuitry.That is why we refer to this as an interface problem.How,then,are their different,even apparently contradictory,properties generated at an appropriately designed cortical interface?

The experiment by Moore et al.(1998)provided a particularly elegant demonstration that attention can?ow along illusory contours.They presented subjects with two“pacmen”stimuli which to-gether induced an illusory Kanizsa rectangle.That is,the two pacmen were separated by retinally unstimulated space,but were perceived as jointly forming a single object in virtue of the illusory contours that connected them.Moore et al.then cued attention to one end of the illusory bar by brie?y?ashing one of the pacmen,and found that the reaction time to a probe stimulus presented at the other end of the bar was improved,showing that the speed-enhancing effect of attention had spread from one side of the illusory contour to the other.In a control condition when the illusory contours were blocked,but all other aspects of the stimuli left the same,the reaction time advantage was now restricted only to the cued pacman inducer.Thus,attention was able to spread across the retinally unstimulated space separating the pacmen if,and only if,they were already joined by a preattentively formed illusory contour grouping.Because attention did not create any new group-ings,but merely enhanced ones which the inducers had already formed,its preattentive/attentive interface constraint remained unviolated.

Given,then,that the candidate solutions considered above to the preattentive/attentive interface problem all fail,how does the cortex succeed?The fact that cortex does indeed succeed in solving the problem is evidenced by the co-existence and mutual interaction within V1and V2of the two crucial contextual effects of top-down attention and preattentive perceptual grouping.We suggest that the question of how cortex integrates these diverse and seemingly con?icting tasks is one that must be addressed by any descriptively adequate computational model of contextual processing. We now present a neural model of V1and V2which proposes a speci?c solution to this problem using known properties of cortical laminar design,and which shows in computer simulations how the contextual effects of attention,perceptual grouping and orientation contrast can all be simul-taneously implemented.The model builds on and extends previous work presented in Grossberg et al.(1997),Grossberg(1999a)and Grossberg and Raizada(2000).

b Visual

input to network

c

Figure4:Demonstration of attention?owing along the neural representation of a visual boundary in V1of the neural model.A similar result was found neurophysiologically by Roelfsema et al.(1998)and Roelfsema and Spekreijse(1999).

(a):A diffuse Gaussian of top-down attention directed to the end of a line,(b),which was presented as visual input to the model network.(c):Cross-section of V1layer2/3activity elicited by the line visual stimulus with the attention directed to the leftmost end.Attention can enter layer2/3via two routes,both of which render the attentional enhancement subthreshold via a balance of excitation and inhibition.In one route,attentional feedback passes into layer6,is folded back up into the modulatory on-center off-surround layer64path,and then passes up into layer2/3.In the second route,attentional feeds back into in V1layer1,where it is collected by the apical dendrites of layer2/3pyramidal cells and also by the dendrites of inhibitory interneurons with their soma and axons in layer2/3but dendrites in layer1 (Lund&Wu,1997).It can be seen that attention enhances the end to which it is directed,but that this enhancement?ows along the length of the line beyond the range of the attentional Gaussian itself,gradually decaying over distance.This lateral?ow is carried by long-range horizontal axons from pyramidals in layer2/3.The slight dip in neural activity next to the maximally boosted region at the leftmost end is due to the off-surround layer64inhibition which attention also induces.

b Visual

input to network

c

Spatial position(pixels)

Figure5:(a):Demonstration of attention?owing along the neural representation of an illusory contour in V1of the neural model.A similar result was demonstrated psychophysically by Moore et al.(1998).(a):A diffuse Gaussian of top-down attention directed to the end of a dotted line,(b),which was presented as visual input to the model network.

(c):Cross-section of V1layer2/3activity elicited by the dotted line visual stimulus and the attention directed to the leftmost end.Note that regions of above-threshold layer2/3activity form between the segments of the dotted line, preattentively completing the neural representation of the boundary contour.As in Figure4,attention?ows along the neural representation of the contour,carried by the long-range layer2/3horizontal axons linking pyramidal cells which are?ring above-threshold.Note that without the preattentive completion of the layer2/3boundary representation, attention on its own would have been able only to provide a subthreshold prime to the neurons whose classical receptive ?elds fall on the gaps in the dotted line.The attentional enhancement extends well beyond the range of the top-down feedback Gaussian itself.

2Model neural network

The laminar architecture of the present model is constructed out of two fundamental building blocks:an on-center off-surround circuit running from layer6to layer4,and intrinsic horizon-

Connection in model Functional interpretation Selected references

V2layer6V1layer1Standard intercortical laminar feedback(Salin&Bullier,1995,p.110)Rockland and Virga(1989)

16(within a layer5pyr.)Corticocortical fdbk into6:Lay5pyr.,apic.dend.in1,axon in6.Lund and Boothe(1975,Fig.7),Gilbert and Wiesel(1979,cat)

V2(unknown layer)V1layer6Direct corticocortical feedback into V1layer6Gattass et al.(1997,Fig.4)

2/36Boundary groupings feedback into64on-center off-surround Blasdel et al.(1985,Fig.13),Kisvarday et al.(1989,Fig.7)

15Corticocortical fdbk into5:Lay5pyr.with apic.dend.in1Valverde(1985,Fig.24o),Peters and Sethares(1991,p.7)

2/35Part of indirect2/36path Lund and Boothe(1975,Fig.8),Callaway and Wiser(1996)

56Continuation of indirect routes into6,via5Blasdel et al.(1985,Fig.17),Kisvarday et al.(1989,Fig.7) Table1:All references are to macaque monkey unless otherwise noted.

tal connections in layer2/3which perform collinear integration and perceptual grouping.Each of these two sub-circuits has assigned to it a well-de?ned functional role,and is constructed from model neurons with empirically determined connectivity and physiological properties,as sum-marised in Table1.When these building blocks are connected together according to the known anatomy of V1and V2,as shown in Figure6,a cortical network is formed whose properties can be understood from the interactions of the functional sub-circuits,but whose behavior is much richer than that of any sub-circuit taken individually.

Attention in the model is mediated by a new mechanism that we call folded feedback(Grossberg, 1999a),whereby signals from higher cortical areas,and also the V1supragranular layers,pass down into V1layer6and are then“folded”back up into the feedforward stream by passing through the layer64on-center off-surround path(Figure6b),thus giving attention an on-center off-surround form,enhancing attended stimuli and suppressing those that are ignored.

A key prediction of the model is that the on-center of the64path is modulatory(or priming, or subthreshold),consistent with the?nding that layer4EPSPs elicited by layer6stimulation are much weaker than those caused by stimulation of LGN axons or of neighbouring layer4sites(Strat-ford et al.,1996),and also with the fact that binocular layer6neurons synapse onto monocular layer 4cells of both eye types without reducing these cells’monocularity(Callaway,1998,p.56).We sug-gest that the on-center excitation is inhibited down into being modulatory by the overlapping and broader off-surround.Thus,although the center excitation is weak,the suppressive effect of the off-surround inhibition can be strong.Because attentional excitation passes through the64path,it inherits this path’s properties:the attentional on-center is modulatory,able to enhance existing ac-tivity but only slightly to elevate neurons’baseline?ring rates in the absence of visual input(Luck,

Chelazzi,Hillyard,&Desimone,1997),but the off-surround can select strongly against unattended stimuli.The model would still be supported if weak suprathreshold excitatory responses in layer4 could be created by layer6stimulation,as long as these responses meet the crucial condition that they be too weak to cause suprathreshold groupings to occur within the horizontal connections of layer2/3.

Several routes exist through which feedback from higher cortex can reach V1layer6,as shown in Table1.Figure6b illustrates the route whereby feedback signals pass into layer1,where the majority of V2feedback axons terminate(Rockland&Virga,1989),and then stimulate the apical dendrites of layer5pyramidal cells whose axons send collaterals into layer6(Lund&Boothe,1975; Gilbert&Wiesel,1979),where the attentional signals are“folded”back up into the64on-center off-surround.Reversible deactivation studies of monkey V2have shown that feedback from V2to V1does indeed have an on-center off-surround form(Bullier,Hup′e,James,&Girard,1996),and moreover that the V1layer whose activation is most reduced by cutting off V2feedback is layer6 (Sandell&Schiller,1982).

We suggest that the mechanism of folded feedback is also used to help select the?nal layer2/3 grouping.If the visual information coming into the brain is unambiguous,then the correct group-ings could,in principle,form due to the?rst incoming wave of activation across layer2/3hori-zontal connections.However,in response to scenes or images with multiple grouping possibilities, the initial groupings that are formed in layer2/3may need to be pruned to select those that are correct.Like attentional signals from higher cortex,the groupings which start to form in layer2/3 also feed back into the64path(Figure6c),to enhance their own positions in layer4via the6

4on-center,and to suppress input to other groupings via the64off-surround.There exist direct layer2/36connections in macaque V1,as well as indirect routes via layer5(Table1).This com-petition between layer2/3groupings,via layer2/3642/3feedback,causes the strongest groupings to be selected,while it suppresses weaker groupings,ungrouped distractors,and noise. The interlaminar feedback also binds the cortical layers together into functional columns.

The fact that both attention and perceptual grouping share the properties of enhancing weak stim-uli,and of suppressing signals from nearby rival inputs,can thus be parsimoniously explained by the hypothesis that both processes share the64folded feedback path.This laminar architec-ture also resolves the preattentive-attentive interface problem described above,since despite their shared properties and coexistence side-by-side within V1and V2,attention and grouping behave quite differently in parts of visual space where there is no bottom-up visual stimulus.Above-threshold boundary groupings can form over regions with no bottom-up support,e.g.illusory con-tours.These groupings form in layer2/3.However,the only way top-down attentional signals can enter layer2/3is by?rst passing through a pathway in which a balance of overlapping excitation and inhibition damps down the attentional feedback into being subthreshold,or priming.Thus, attention can only modulate layer2/3,but cannot on its own cause above-threshold activation,and its internal/external problem is thereby resolved.

2/3 4 6

4

6

6

LGN

2/3

a b c

d

e

Figure6:Caption on following page.

Caption to Figure6.How known cortical connections join the layer64and layer2/3building blocks to form the entire V1/V2laminar model.Inhibitory interneurons are shown?lled-in black.(a):The LGN provides bottom-up activation to layer4via two routes.Firstly,it makes a strong connection directly into layer4.Secondly,LGN axons send collaterals into layer6,and thereby also activate layer4via the64on-center off-surround path.Thus,the combined effect of the bottom-up LGN pathways is to stimulate layer4via an on-center off-surround,which provides divisive contrast normalization(Grossberg,1973,1980;Heeger,1992)of layer4cell responses(see Appendix).(b):Folded feedback carries attentional signals from higher cortex into layer4of V1,via the modulatory64path.Corticocortical feedback axons tend preferentially to originate in layer6of the higher area and to terminate in the lower cortex’s layer 1(Salin&Bullier,1995,p.110),where they can excite the apical dendrites of layer5pyramidal cells whose axons send collaterals into layer6(the triangle in the?gure represents such a layer5pyramidal cell).Several other routes through which feedback can pass into V1layer6exist(see Table1for references).Having arrived in layer6,the feedback is then “folded”back up into the feedforward stream by passing through the64on-center off-surround path(Bullier et al., 1996).(c):Connecting the64on-center off-surround to the layer2/3grouping circuit:like-oriented layer4simple cells with opposite contrast polarities compete(not shown)before generating half-wave recti?ed outputs that converge onto layer2/3complex cells in the column above them.Like attentional signals from higher cortex,groupings which form within layer2/3also send activation into the folded feedback path,to enhance their own positions in layer4beneath them via the64on-center,and to suppress input to other groupings via the64off-surround.There exist direct layer2/36connections in macaque V1,as well as indirect routes via layer5(Table1).(d):Top-down corticogeniculate feedback from V1layer6to LGN also has an on-center off-surround anatomy,similar to the64path.The on-center feedback selectively enhances LGN cells that are consistent with the activation that they cause(Sillito et al.,1994),and the off-surround contributes to length-sensitive(endstopped)responses that facilitate grouping perpendicular to line ends.

(e):The entire V1/V2circuit:V2repeats the laminar pattern of V1circuitry,but at a larger spatial scale.In particular, the horizontal layer2/3connections have a longer range in V2,allowing above-threshold perceptual groupings between more widely spaced inducing stimuli to form(Amir,Harel,&Malach,1993).V1layer2/3projects up to V2layers6and 4,just as LGN projects to layers6an4of V1.Higher cortical areas send feedback into V2which ultimately reaches layer 6,just as V2feedback acts on layer6of V1(Sandell&Schiller,1982).Feedback paths from higher cortical areas straight into V1(not shown)can complement and enhance feedback from V2into V1.

In the earlier version of this model presented in Grossberg and Raizada(2000),the only pathway via which attention could enter layer2/3was the folded-feedback layer642/3circuit described above.Since the majority of feedback axons from higher cortical areas terminate in V1layer1 (Rockland&Virga,1989),we also discussed the possibility that attentional signals may modulate layer2/3more directly by stimulating the layer1apical dendrites of layer2/3pyramidals.Lund and Wu(1997)have shown that there exist inhibitory interneurons in layer2/3macaque V1which also have dendrites in layer1.Hence,we suggested that there may also exist a balance of excitation and inhibition keeping this direct attentional path into layer2/3modulatory,or subthreshold,just as the layer64off-surround overlaps with and balances64the on-center.Going beyond this earlier paper,we have now implemented these connections in the present simulations(see Equations20and21).As will be discussed in Section3.3below,the extended model incorporating these anatomical connections is still able to keep top-down feedback’s facilitatory effect within layer 2/3purely modulatory.

The notion of activity being subthreshold,or modulatory,is given a simple instantiation in the model’s equations:layer2/3activity below a?xed value,,produces no output from the cells in that layer.When the input activity starts to exceed,the output starts to climb from zero at the same rate,as described in Equation13.In both V1and V2,was?xed at0.2for all the simula-tions performed.Because this layer2/3signal function is continuous,and gives the output the same gain as the input,the behavior of the network changes continuously and predictably if is changed:smaller values would tend to allow stronger layer2/3bipole groupings to form,for

example allowing V1groupings to bridge over slightly larger visual gaps than https://www.wendangku.net/doc/847357913.html,rger values would tend slightly to weaken the groupings,and to mean that larger top-down attention signals would be required to in?uence the groupings that do form.

We also extend the network dynamics of the model presented in Grossberg and Raizada(2000)in another respect.In that study,we showed how the model could simulate the?nding by Polat et al. (1998)that neurons in cat V1responding to a low-contrast target Gabor stimulus are net facilitated by the presence of collinear?anking Gabor patches,but when the target is high-contrast,the ef-fect of the?ankers“crosses over”into being net inhibitory.The simulation in the Grossberg and Raizada(2000)paper showed that a long-range range V2grouping between the?anking elements fed back a subthreshold prime to the V1location of the central target Gabor,facilitating it by raising it above threshold at low-contrasts.However,shunting64inhibition from the?ankers also had a divisive effect on neural responses to the target,lowering their gain and causing the net effect of the?ankers to be suppressive at high-target contrasts.Thus,this simulation showed how contrast-dependent perceptual grouping can emerge as a result of network behavior,without needing to take into account possible differential effects of contrast on individual excitatory and inhibitory neurons.

Substantial neurophysiological evidence exists,however,showing that at high stimulus contrasts, inhibition starts to predominate over excitation as a combined result of several diverse factors: inhibitory interneurons have higher gains than excitatory pyramidal cells at high contrast(Mc-Cormick,Connors,Lighthall,&Prince,1985),inhibitory synapses depress less than excitatory synapses(Varela,Song,Turrigiano,&Nelson,1999),with synapses from inhibitory interneurons onto pyramidals in fact actively facilitating(Thomson,1997;Markram,Wang,&Tsodyks,1998; Reyes,Lujan,Rozov,Burnashev,Somogyi,&Sakmann,1998).This complex mixture of pre-and post-synaptic factors cannot be completely captured without greatly complicating the existing model; we approximate the total net effect on inhibition by passing the population inhibitory activity through a sigmoidal signal function,as shown in Eqs.16and17.This signal function starts off at low values,and then rapidly increases as higher contrast stimuli cause greater levels of inhibitory activation.Since the network pyramidals gradually saturate at increasing contrasts,the net effect is for inhibition to start to predominate(c.f.Grossberg,1970;Stemmler et al.,1995;Somers et al., 1998;Grossberg&Kelly,1999).By extending the previous model in this way,the model can capture an even wide-ranger of contrast-sensitive grouping effects.An example is the simulation of recent data from Kapadia et al.(1998)presented in Section3.4below.

3Results

The model presented here captures several aspects of visual contextual processing.The following simulations and explanations illustrate how the laminar architecture of cortex brings about this behavior.

3.1Attention has a greater effect on low-contrast stimuli

As was remarked in the introduction,top-down attention is needed most when a visual target is of low salience due to being surrounded by distractors,or being of low contrast.It would therefore be functionally advantageous for attention to provide a strong boost to low contrast targets,but to have a relatively weaker effect at high contrasts.This is exactly what was observed in the recent

behavioral study of macaque monkeys by De Weerd et al.(1999).The monkeys’task was to dis-criminate the orientation of a variable-contrast grating patch which was surrounded by distractors. Stimuli of this sort were presented to the model network,as shown in Figure2a.De Weerd et al. placed the stimuli in either an unlesioned visual quadrant,or ones in which lesions had been made to cortical areas V4or TEO,both of which are known to play important roles in visual attention. Their?nding,illustrated in Figure2c,was that the absence of these attentional regions severely impaired the monkeys’performance when the target grating was low contrast,but had relatively little effect when the target was high contrast.As can be seen from Figure2d,the model simu-lation produces very similar behavior.Here,the“behavioral threshold”of the network is simply operationalized as the reciprocal of the activity of the V1layer2/3cells which respond to verti-cal orientations.Because the model only simulates V1and V2,rather than higher areas such as prefrontal cortex which presumably control the behavioral responses made by the macaque,the network does not literally have a“behavioral threshold”.However,the layer2/3neurons which respond to vertical orientations are the cells which would pass forward information about the grat-ing’s orientation to higher areas.The greater the activity of these cells,the stronger and hence the more discriminable is the information which is passed forward.Since high discriminability would result in a low behavioral threshold,and vice versa,the simplest way of embodying this process in an equation is to take the reciprocal.Instead of having to lesion higher cortical areas,we are able simply to turn attention on and off in the model;attention is implemented as a diffuse Gaussian of unoriented cortical feedback directed to the target’s location.

In the model,attention aids discrimination by boosting the neural representation of the target through the layer64on-center,and also via the direct attentional projection into layer2/3. It also suppresses the distractors,which fall into attention’s layer64off-surround.However, these facts alone are not enough to explain why attention facilitates the lower contrast targets more than the high contrast ones.This behavior follows from two closely related network phenomena: shunting inhibition and neural saturation.High-contrast stimuli induce strong64on-center excitation at their own locations,but also bring with them divisive shunting inhibition from the overlapping64off-surround,thereby reducing their own contrast gain.Hence,lower contrast stimuli have higher gain and can therefore be boosted more by attention.Similarly,the simple fact of neural saturation means that cells which are?ring far below their maximal rate can be signi?-cantly boosted by attention,but cells which are pushed close to saturation by high contrast stimuli cannot.

3.2Orientation contrast

Another important contextual effect exhibited by the model network is orientation contrast,in which stimuli which are embedded in orthogonally oriented texture surrounds are seen to“pop-out”,whereas stimuli which are in iso-orientation surrounds do not(Nothdurft,1991;Knierim& Van Essen,1992;Sillito et al.,1995;Nothdurft et al.,1999).This perceptual effect is re?ected in the activity of V1neurons:although both iso-and cross-orientation surrounds have a net suppres-sive effect on the neural response to an isolated bar,the cross-orientation surround is signi?cantly less suppressive(Knierim&Van Essen,1992,data shown in Figure3d).Examples of the types of stimuli used by Knierim&Van Essen are shown in Figure3a-c.These images were in fact pre-sented as stimuli to the model network.As shown in Figure3e,model V1neurons exhibit the same qualitative pattern of behavior.The explanation for this is simply that in the model,layer6

4iso-orientation off-surround inhibition is stronger than the cross-orientation inhibition.The key question is:how did the inhibition come to be that way?The relative strengths of the iso-and cross-

orientation inhibitory projective?elds were not speci?ed by hand,but instead were self-organized in the developmental laminar model of Grossberg and Williamson(2000)which used the same lam-inar architecture as the present model,but without the corticocortical attentional connections.In the course of that model’s self-organizing development,the synapses tracked the statistics of visual inputs which were presented to the network.These inputs contained visual structure,in partic-ular straight edges,which caused iso-orientation correlations between neurons positioned along the length of the edge.The inhibitory synapses tracked these iso-orientation correlations,with the result that the the iso-orientation inhibition grew stronger than that for cross-orientations.

3.3Attention?ows along real and illusory contours

As remarked in introduction,the ability of attention to?ow along real and illusory contours places important constraints on visual cortical processing.Attention must be able to?ow along contour groupings which are already preattentively active,but cannot cause above-threshold activity on its own.The fact that attention does indeed?ow along groupings is no mere epiphenomenon,but is the key mechanism uniting spatial and object-based attention in early visual cortex.In particular, attention can thereby selectively enhance an entire object by propagating along its boundaries. Grossberg and Raizada(2000)simulated the study by Roelfsema et al.(1998),including the delayed time-course of attentional enhancement,caused by the time needed for attention to propagate along the representation of the curve.Here we show in more detail the spatial spread of attention along a real or illusory curve,illustrated in Figures4and5.In both cases,a Gaussian of diffuse attentional feedback directed to one end of a line stimulus causes excitation that does not just boost the directly attended location,but also spreads along some of the length of the line,even when the line is physically discontinuous but perceived as forming a collinear grouping(Figure5).

This lateral spread of attentional excitation is carried by the long-range horizontal connections in layer2/3of V1and V2.As described in Section3above,there exist two routes by which attention can get into layer2/3.The main route is that attention passes into layer6,is then folded back up into the layer64on-center off-surround path,where the balance between the on-center excita-tion and the overlapping off-surround inhibition ensures that the attentional enhancement which can then feed on into layer2/3is purely subthreshold.The second route is the direct attentional connection into layer2/3,illustrated in Figure6e and described in Equations20and21.The model layer2/3contains inhibitory interneurons as well as excitatory pyramidal cells,in order to control the formation of groupings through layer2/3horizontal connections,and the attentional feed-back synapses onto both of them,again providing a balance of suppressive and facilitatory forces which ensures that attentional enhancement remains subthreshold.Although this subthreshold signal would on its own be unable fully to activate layer2/3,it can nonetheless boost preatten-tively formed collinear groupings which form along the line stimuli,and,in the case of the dotted line,bridge over the gaps of retinally unstimulated space.Note that in both cases,the attentional enhancement gradually declines with distance from the attention focus,due to decay of neural ac-tivity.The rate of fall-off with distance is smaller in V2than in V1,due to the longer-range layer 2/3horizontal connections found in the higher area(Amir et al.,1993).

3.4Contrast-sensitive grouping and inhibition

Contextual effects can be either facilitatory or inhibitory,depending on stimulus contrast.In par-ticular,the effect of collinear?ankers on a target can“crossover”from being net excitatory at low

contrasts to being net suppressive at high-contrasts,either when the central target alone varies in contrast(Polat et al.,1998),or when the target and?ankers all vary in contrast together.As discussed in Section3above,several pre-and post-synaptic factors may contribute to the predom-inance of inhibition over excitation at higher stimulus contrasts,although network-level effects alone can be suf?cient to account for the Polat et al.(1998)data,as shown in our previous paper (Grossberg&Raizada,2000).

Figure1shows a crossover effect using three bars of equal contrast,as demonstrated experimen-tally by Kapadia et al.(1998).The?anking bars exert both excitatory and inhibitory effects on the central target.At low contrasts,the layer64inhibitory sigmoidal signal function still takes low values,and inhibition is weaker than the collinear layer2/3excitation,giving a net facilitatory effect(Figure1b).At higher stimulus contrasts,the total amount of inhibition starts to fall into the rapidly growing section of the sigmoidal inhibitory signal function,allowing inhibition from the?ankers to overwhelm the excitation that they also supply,making their net effect suppressive (Figure1c).

4Discussion

The neural model presented here shows how visual cortex can implement several types of contex-tual processing at once,and also allow them to interact.In doing so,it builds upon and extends the simulations presented in Grossberg et al.(1997),Grossberg and Raizada(2000),and Grossberg and Williamson(2000).Moreover,the model proposes speci?c functional roles for known laminar circuits to carry out the contextual processing,and suggests how attention and perceptual grouping can interact within this laminar circuitry to solve the preattentive/attentive interface problem.

As far as we are aware,no other existing model meets the challenge of this problem by attempting to emulate cortex’s ability to perform attention and perceptual grouping simultaneously.Whereas the functional importance of top-down attention is clear,the formation of illusory contours may at ?rst sight appear to be an almost epiphenomenal consequence of the seemingly more fundamental process of collinear facilitation.However,illusory contours can perform a crucial task which mere facilitation cannot:they can actively close incomplete boundaries,a process that requires that cells with unstimulated CRFs can nonetheless become active.This boundary closure can guide surface reconstruction,complete boundaries over visual gaps caused by the blind-spot and retinal veins, and also provide enhanced information for the recognition of partially occluded objects(Grossberg, 1994).Several other models of collinear grouping in V1produce facilitation but not illusory con-tours,and hence are unable to capture this important aspect of cortical processing(Stemmler et al., 1995;Li,1998;Somers et al.,1998;Yen&Finkel,1998).Those models which do implement illu-sory contours either leave out any consideration top-down cortical feedback(Williams&Jacobs, 1997;Heitger,von der Heydt,Peterhans,Rosenthaler,&Kubler,1998),fail to capture the on-center off-surround form of attention by treating top-down feedback as having a purely excitatory mul-tiplicative effect(Neumann&Sepp,1999),or treat“reentrant”feedback signals from higher areas “as if they were signals from real contours in the periphery entering via”(Finkel&Edelman, 1989,p.3197),thereby creating the risk of perceptual hallucinations.Conversely,many models of top-down feedback in visual processing do not implement perceptual grouping(e.g.Harth,Un-nikrishnan,&Pandya,1987;Mumford,1992;Olshausen,Anderson,&Van Essen,1993;Ullman, 1995;Tsotsos,Culhane,Wai,Lai,Davis,&Nu?o,1995;Usher&Niebur,1996;Rao&Ballard,1999), therefore leaving untouched what we suggest are crucial design constraints which shape the func-tional laminar architecture of cortex.

In our previous paper(Grossberg&Raizada,2000),we presented simulations of the earlier version of this model,which differed from the present one only in lacking the direct attentional connec-tions into layer2/3,and the layer64inhibitory signal function.Three types of behavior were simulated in the earlier paper:attention protecting a target from the suppressive effect of?ankers (Reynolds et al.,1999),the time-course of attention?ow along a curve(Roelfsema et al.,1998),and contrast-sensitive perceptual grouping of Gabor patches(Polat et al.,1998).As can be seen from the simulations presented in the current paper,these properties still hold in the extended version of the model,although here they are applied to different,but related,sets of stimuli.Thus,the new extensions to the model maintain and extend its previous qualitative patterns of behavior,although the exact quantitative behavior is not identical,due to the addition of the new circuitry.

These modeling results also bear upon other issues concerning cortical coding.For example,in response to unambiguous visual information,a boundary grouping can start to form very rapidly in response to a feedforward sweep of signal from layer4to layer2/3.Thus the existence of corti-cal feedback does not preclude fast cortical processing(Thorpe,Fize,&Marlot,1996).Intracortical feedback is predicted to become increasingly important when multiple groupings of the image or scene are possible.Even here,the model’s selection of a?nal grouping can often converge within one or at most a few feedback cycles between layers42/36 4.Intercortical feedback may be needed when attention must select some cue combinations over others,based on higher-order constraints.The model shows how very high-order constraints can,in principle,modulate even low-order feature detectors by propagating across multiple cortical regions via their layers6,with-out ever fully activating their groupings in layer2/3.An open experimental question concerns whether and how such a propagating priming effect is attenuated as a function of the number of cortical regions that are traversed.It has also been simulated how these grouping and attentional circuits may rapidly synchronize,even generating fast synchronizing oscillations under some con-ditions(Grossberg&Somers,1991;Grossberg&Grunewald,1997)

All of these statements require quali?cation,however.For example,the context-dependent persis-tence of previously grouped images may interfere with the synchrony of subsequent groupings,as illustrated by the model of Francis,Grossberg,and Mingolla(1994).Also,the fact that attention-induced increases in?ring rate can propagate along perceptual groupings(see Figures4and5), thereby selectively enhancing object representations,shows that synchronous activation of an ob-ject by attention is not necessary in all cases(Roelfsema et al.,1998).Finally,one needs to emphasize that all the explanations and simulations presented above,and those in earlier articles about this evolving cortical model,concern only processing of visual boundaries within the interblob stream of visual cortex,as opposed to the processing of surface brightness and color within the blob stream. Boundary groupings within the interblob stream are predicted,in the absence of surface featural in-formation,to be invisible,or amodal.Hence,all of the results in this series of articles strictly concern only the salience of boundary groupings,not the perception of the surfaces which these boundaries enclose.Visibility is predicted to be a property of surface representations within the blob stream, with these surfaces arising due to the?lling-in of brightness and color within closed boundary groupings formed in the interblob stream(Grossberg,1994).Whereas contour salience and visibil-ity often covary,this is not always the case:for example,Glass patterns(Glass,1969)contain highly salient concentric contour groupings,but do not induce any brightness differences that would cause bright Ehrenstein-like circular surfaces to be visible.Another limitation of the present model is that it does not describe how transient responses to changing or moving stimuli can rapidly attract vi-sual attention.One major pathway for this mechanism is likely to be the“where”dorsal cortical stream.Recent models of motion processing clarify the key role of these transient responses(Chey, Grossberg,&Mingolla,1997,1998;Baloch,Grossberg,Mingolla,&Nogueira,1999),and also how

they can attract visual attention(Grossberg,1998).

Several studies providing important data on grouping and also attention in V1have recently been carried out by Charles Gilbert and colleagues.In particular,Kapadia et al.(1998)used oriented line stimuli of the sort shown in Figure1a to investigate the spatial arrangement of contextual facilita-tion and inhibition induced by?anking lines which were of the same orientation as the target.They found that the?ankers were facilitatory when they were placed to be approximately collinear with the target line,but were inhibitory when they were located to its sides.In the present model,stimuli induce a pool of layer64off-surround inhibition around them which extends in all directions, as shown in Figure7,and also induce a more strongly anisotropic region of layer2/3facilitation, oriented primarily collinearly with the stimulus itself(Figure8).These regions of facilitation and inhibition spatially overlap.However,the collinear excitation at the ends of an oriented line can be strong enough to overwhelm the inhibition which is also generated there,giving a net facilitatory effect,especially at low stimulus contrasts.Thus,we suggest that the existence of a net excitatory effect at locations collinear with a line ending does not imply that the inhibitory off-surround is restricted to being present only by the line’s sides.In fact,the existence of off-surround inhibition at a line ending can be very useful functionally,for example in generating end-cuts(Grossberg& Mingolla,1985).

Ito and Gilbert(1999)examined the interaction of top-down attention and collinear facilitation in V1 of macaque monkeys which were performing a brightness comparison task.Although this study is pioneering in investigating the interaction of these visual processes,we have not simulated their neural data here since their results were not consistent across the two monkeys from which record-ings were made.In one monkey,focal attention directed to a target line was found to increase the facilitatory effect upon that line of a collinear?anker.In the other monkey,the opposite effect was found.Several factors might contribute to this discrepancy.As remarked by Ito and Gilbert them-selves,the monkeys had undergone different amounts of training.Another possibility is that the requirements of the behavioral task were not well-suited for probing the most commonly needed functions of attention and grouping:as we argued in Section1,these processes are needed most when a visual target is weak or low-contrast,and hence hard to detect.In such circumstances,one would expect both processes,and their interaction,to be facilitatory.However,in the Ito&Gilbert study,the target was bright and easily detectable,and the task was to discriminate its brightness as accurately as possible.Thus,a simple net facilitation of neural activity could actually hinder the monkey’s brightness judgment.This con?ict between the speci?c task demands and the most com-mon ecological uses of attention and grouping may partially account for the differences between the monkeys.In the present model,a possible mechanism which might underlie such a difference would be the width of the attentional focus directed at the target line.If the focus is narrow,it will enhance the target,but attention’s off-surround will actively suppress the collinear?anker, and attention will tend to reduce the?anker’s facilitatory effect.If the focus is slightly wider,the attentional on-center will fall on both target and?anker,and the facilitatory effect of the collinear grouping will be enhanced.

Because the present model assigns speci?c functional roles to many aspects of cortical laminar cir-cuitry,many testable predictions can be derived from it.Several such predictions are presented in the conclusion of Grossberg and Raizada(2000).The simulations presented here extend and broaden the scope of the model,and also generate new predictions over and above those already presented.Perhaps the most directly testable of these concern the spread of attention along illusory as well as real contour groupings(see Figures4and5).We suggest that there should exist measur-able neurophysiological correlates of such?ow,in particular in layer2/3of V2and possibly also of V1.This could be tested by replicating the Roelfsema et al.(1998)study,but having the mon-

古代晋灵公不君、齐晋鞌之战原文及译文

晋灵公不君（宣公二年） 原文： 晋灵公不君。厚敛以雕墙。从台上弹人，而观其辟丸也。宰夫胹熊蹯不熟，杀之，寘诸畚，使妇人载以过朝。赵盾、士季见其手，问其故而患之。将谏，士季曰：“谏而不入，则莫之继也。会请先，不入，则子继之。”三进及溜，而后视之，曰：“吾知所过矣，将改之。”稽首而对曰：“人谁无过？过而能改，善莫大焉。诗曰：‘靡不有初，鲜克有终。’夫如是，则能补过者鲜矣。君能有终，则社稷之固也，岂惟群臣赖之。又曰：‘衮职有阙，惟仲山甫补之。’能补过也。君能补过，衮不废矣。” 犹不改。宣子骤谏，公患之，使鉏麑贼之。晨往，寝门辟矣，盛服将朝。尚早，坐而假寐。麑退，叹而言曰：“不忘恭敬，民之主也。贼民之主，不忠；弃君之命，不信。有一于此，不如死也！”触槐而死。 秋九月，晋侯饮赵盾酒，伏甲将攻之。其右提弥明知之，趋登曰：“臣侍君宴，过三爵，非礼也。”遂扶以下。公嗾夫獒焉。明搏而杀之。盾曰：“弃人用犬，虽猛何为！”斗且出。提弥明死之。 初，宣子田于首山，舍于翳桑。见灵辄饿，问其病。曰：“不食三日矣！”食之，舍其半。问之，曰：“宦三年矣，未知母之存否。今近焉，请以遗之。”使尽之，而为之箪食与肉，寘诸橐以与之。既而与为公介，倒戟以御公徒，而免之。问何故，对曰：“翳桑之饿人也。”问其名居，不告而退。——遂自亡也。 乙丑，赵穿①攻灵公于桃园。宣子未出山而复。大史书曰：“赵盾弑其君。”以示于朝。宣子曰：“不然。”对曰：“子为正卿，亡不越竟，反不讨贼，非子而谁？”宣子曰：“呜呼!‘我之怀矣，自诒伊戚。’其我之谓矣。” 孔子曰：“董狐，古之良史也，书法不隐。赵宣子，古之良大夫也，为法受恶。惜也，越竞乃免。” 译文： 晋灵公不行君王之道。他向人民收取沉重的税赋以雕饰宫墙。他从高台上用弹弓弹人，然后观赏他们躲避弹丸的样子。他的厨子做熊掌，没有炖熟，晋灵公就把他杀了，把他的尸体装在草筐中，让宫女用车载着经过朝廷。赵盾和士季看到露出来的手臂，询问原由后感到很忧虑。他们准备向晋灵公进谏，士季说：“如果您去进谏而君王不听，那就没有人能够再接着进谏了。还请让我先来吧，不行的话，您再接着来。”士季往前走了三回，行了三回礼，一直到屋檐下，晋灵公才抬头看他。晋灵公说：“我知道我的过错了，我会改过的。”士季叩头回答道：“谁能没有过错呢？有过错而能改掉，这就是最大的善事了。《诗经》说：‘没有人向善没有一个开始的，但却很少有坚持到底的。’如果是这样，那么能弥补过失的人是很少的。您如能坚持向善，那么江山就稳固了，不只是大臣们有所依靠啊。

The way常见用法

The way 的用法 Ⅰ常见用法： 1)the way+ that 2)the way + in which（最为正式的用法） 3)the way + 省略（最为自然的用法） 举例：I like the way in which he talks. I like the way that he talks. I like the way he talks. Ⅱ习惯用法： 在当代美国英语中，the way用作为副词的对格，“the way+ 从句”实际上相当于一个状语从句来修饰整个句子。 1)The way =as I am talking to you just the way I’d talk to my own child. He did not do it the way his friends did. Most fruits are naturally sweet and we can eat them just the way they are—all we have to do is to clean and peel them. 2）The way= according to the way/ judging from the way The way you answer the question, you are an excellent student. The way most people look at you, you’d think trash man is a monster. 3）The way =how/ how much No one can imagine the way he missed her. 4）The way =because

2013年安徽省中考语文试题及答案

2013年安徽省初中毕业学业语文考试卷 安徽省阜阳市颍泉区邵营中学王海军 一、语文积累与综合运用 (35分) 1.默写古诗文中的名句名篇。(10分） (1)补写出下列名句中的上句或下句。(任选其中 ．．．．6．句．) ①择其善者而从之，。（《论语》） ②，若出其里。(曹操《观沧海》) ③苔痕上阶绿，。(刘禹锡《陋室铭》) ④无可奈何花落去，。(晏殊《淀溪沙》) ⑤，处江湖之远则忧其君。(范仲淹《岳阳楼记》） ⑥，柳暗花明又一村。(陆游《游山西村》) ⑦夕阳西下，。(马致远《天净沙·秋思》) ⑧，宫阙万间都做了土。(张养浩《山坡羊·潼关怀古》 (2)默写白居易《钱塘湖春行》的后四句 ．．．。 2、阅读下面的文字，完成(1)-(4)题。(9分) 古往今来，人们都在编织着梦想与希jì。如果说梦想是天边的星晨，永恒地照亮匆匆的人生，那么，现实就是脚下的土地，真实地记录着行走的足迹。人，既不能生活在华而不实的梦想里，也不要沉nì于喧嚣纷扰的现实中。在人生的道路上，我们要一边种植梦想，，随时采撷，收获希望。 (1)根据拼音写出相应的汉字，给加点的字注音。 希jì( ) 沉nì( ) 喧嚣．( ) (2)文中有错别字的一个词是“”，这个词的正确写法是“”。 ①“华而不实”中，“华”的意思是。“采颇”的意思是。 ②根据文意，仿照文中画线的句子，在横线处填上一句恰当的话。 3、运用你课外阅读积累的知识，完成(1)~(2)题。(4分) (1)下面一段文字节选自《骆驼祥子》，读一读，指出A、B两处的人物是谁。 A心里有数儿。他晓得祥子是把好手，即使不拉他的车，他也还愿意祥子在厂子里。有祥子在这儿，先不提别的，院子与门口永远扫得干干净净。B更喜欢这个傻大个儿，她说什么，祥子老用心听着，不和她争辩；别的车夫，因为受尽苦楚，说话总是横着来；她一点不怕他们，可是也不愿多搭理他们；她的话，所以，都留给祥子听。 (2)安徒生笔下，卖火柴的小女孩借着火柴的亮光，看见了温暖的、喷香的烤鹅、美丽的圣诞树、和蔼的。 4、九年级开展以“文明出行”为主题的综合实践活动。活动中有一些问题，请你参与解决。（12分） (1)下面是同学们在活动中整理的《闯红灯原因统计表》，请你用简洁的语言表述其中是信息。(不．得出现数 ．．．．字．)

初中阶段主要句型的用法

初中阶段主要句型的用法 1. It's time to (for) .... 表“时间到了；该干…的时间了”之意。 It's time to go home. It's time for school. 注意：to的后面接动词短语，而for的后面接名词。 2. It's bad (good) for ... 表示“对…有害（有益）的”含义。 Please don't smoke. It's bad for your health. Please take more exercise. It's good for your health. 3. be late for (school) 是“上学（迟到）”之意。for后面还可以接meeting或class。He was late for school this morning. Don't be late for class, please. 4. had better do / not do sth. 表示“最好做/不做某事”之意。 You had better put on the coat when you go out. It's cold outside. We had better stop to have a rest. 注意：用had better时，后面一定要直接跟动词原形，决不能加to do sth. 否定一定用在第二个动词之前，是动词不定式的否定式。 5. be (feel) afraid of ... 表示“恐怕”，“害怕…”之意。 He is afraid of snakes. Mary feels afraid of going out alone. 6. enjoy doing sth. 是“喜欢（爱好）做某事”之意，doing是动名词作动词enjoy 的宾语。enjoy有欣赏之意。 Are you enjoy living in Beijing? Do you enjoy listening to music? 7. stop doing sth. 停止做某事 stop to do sth. 停下来做某事 It's time for class. Stop talking, please. When she saw me, she stopped to talk with me. 8. Let (make) sb. do 让（使）某人做某事。 Let's go to school. Father made his son clean the room again. 注意：当make sb. do sth. 用于被动语态时，第二个动词前面就一定要加to 了。如： His son was made to clean the room again. he was made to work twelve hours a day in the past. 9. like to do / like doing sth. 是“喜欢做某事”之意。 like to do sth. 是表示比较具体的喜欢做某事；like doing sth. 是表示宠统的喜欢做某事。如： I like to swim in the swimming pool. 我喜欢在游泳池中游泳。（喜欢的具体的地方游泳） I like swimming. 我喜欢游泳（只讲喜欢这项运动） 10. ask (tell) sb. to do. (not to do) sth. 请（让）某人（不）做某事 Jim asked his mother to help him with his lessons. When did you tell him not to shout loudly?

The way的用法及其含义(二)

The way的用法及其含义（二） 二、the way在句中的语法作用 the way在句中可以作主语、宾语或表语： 1.作主语 The way you are doing it is completely crazy.你这个干法简直发疯。 The way she puts on that accent really irritates me. 她故意操那种口音的样子实在令我恼火。The way she behaved towards him was utterly ruthless. 她对待他真是无情至极。 Words are important, but the way a person stands, folds his or her arms or moves his or her hands can also give us information about his or her feelings. 言语固然重要,但人的站姿,抱臂的方式和手势也回告诉我们他(她)的情感。 2.作宾语 I hate the way she stared at me.我讨厌她盯我看的样子。 We like the way that her hair hangs down.我们喜欢她的头发笔直地垂下来。 You could tell she was foreign by the way she was dressed. 从她的穿著就可以看出她是外国人。 She could not hide her amusement at the way he was dancing. 她见他跳舞的姿势，忍俊不禁。 3.作表语 This is the way the accident happened.这就是事故如何发生的。 Believe it or not, that's the way it is. 信不信由你, 反正事情就是这样。 That's the way I look at it, too. 我也是这么想。 That was the way minority nationalities were treated in old China. 那就是少数民族在旧中

2013年安徽省中考物理试卷及答案.docx

2 0 1 3 年安徽省初中毕业学业考试 物理 题号一二三四总分 得分 注意事项： 1．本卷共四大题 23 小题，满分 90 分。物理与化学的考试时间共 120 分钟。 2．本卷试题中 g 值均取 10N／kg。 一、填空题（每空 2 分，共 26 分；将答案直接写在横线上，不必写出解题过程）1．小林坐在行驶的火车上，以他乘坐的火车为参照物，他是（选填“运动” 或“静止”）的。 2．商场里的保洁员在拖过地后，常用电扇对着地面吹，这样做的目的是通过加快空气流动 来加快（填物态变化的名称），使地面尽快变干， 3．小明在一次测量体温时，体温计的示数如图所示，则他的体温为℃。 ℃ 35 36 37 38 39 4041 42 第 3 题图 4．古诗“花气袭人知骤暖”的意思是，从花的芳香气味变浓可以知道周边的气温突然升高。 从物理学的角度分析，这是因为温度越高，。 5．小强在立定跳远起跳时，用力向后蹬地，就能获得向前的力，这是因为物体间力的作用 是的。离开地面后，由于，他在空中还能继续向前运动。 6．如图，能说明发电机工作原理的是图（选填“甲”或“乙”）。 图甲图乙 第 6 题图

7．如图，静止在花朵上的一种叫“全碳气凝胶”的固体材料，它是我国科学家研制的迄今 为止世界上最轻的材料。一块体积为100cm3的“全碳气凝胶”的质量只有0.016g ，则它的 密度为kg/m3。 第 7 题图第 8 题图第 9 题图 8．我国设计的“中华月球车”今年将登陆月球，图为该车的模型。物体在月球表面附近受 到的重力仅为在地球表面附近所受重力的1/6 ，月球车的质量为 120kg，它在月球表面附近 受到的重力为N。若它在月球表面上方竖直下落4m，重力对它做的功为J。 9． 2013 年 5 月，象征“和平、友爱和快乐”的巨型大黄鸭在香港与游客见面，图为大黄鸭 漂浮在海面上的情景。大黄鸭的质量约为 600kg，海水的密度取1.0103 kg / m 3，它浸在水 中的体积约为m3。 L2 10．如图，小灯泡L1和 L2均标有“，0.3A ”字样，AB两端的电压U A B ＝。开关 S 断开时，通过L2的电流为A。闭合开关 S，两只小L1S 灯泡消耗的总电功率为W 。 U 第 10 题图 二、选择题（每小题 3 分，共 21 分；每小题给出的四个选 项中，只有一个选项符合题意，请把符合题意的选项序号填入下列表格内，答案 ．．．．．．． 未填入表格内的不能得分） 题号11121314151617 答案 11．下列各图描述的实验中，用来说明声音的传播需要介质的是 A．发声的音叉激B．音叉发出的声音C．钢尺伸出桌边的长度D．抽取玻璃罩内的空气，起水花越响，乒乓球被弹开变短，振动时声音音调变听到罩内的音乐声减小 的越远高

如何翻译古文

如何翻译古文 学习古代汉语，需要经常把古文译成现代汉语。因为古文今译的过程是加深理解和全面运用古汉语知识解决实际问题的过程，也是综合考察古代汉语水平的过程。学习古代汉语，应该重视古文翻译的训练。 古文翻译的要求一般归纳为信、达、雅三项。“信”是指译文要准确地反映原作的含义，避免曲解原文内容。“达”是指译文应该通顺、晓畅，符合现代汉语语法规范。“信”和“达”是紧密相关的。脱离了“信”而求“达”，不能称为翻译；只求“信”而不顾“达”，也不是好的译文。因此“信”和“达”是文言文翻译的基本要求。“雅”是指译文不仅准确、通顺，而且生动、优美，能再现原作的风格神韵。这是很高的要求，在目前学习阶段，我们只要能做到“信”和“达”就可以了。 做好古文翻译，重要的问题是准确地理解古文，这是翻译的基础。但翻译方法也很重要。这里主要谈谈翻译方法方面的问题。 一、直译和意译 直译和意译是古文今译的两大类型，也是两种不同的今译方法。 1．关于直译。所谓直译，是指紧扣原文，按原文的字词和句子进行对等翻译的今译方法。它要求忠实于原文，一丝不苟，确切表达原意，保持原文的本来面貌。例如： 原文：樊迟请学稼，子曰：“吾不如老农。”请学为圃。子曰：“吾不如老圃。”（《论语?子路》） 译文：樊迟请求学种庄稼。孔子道：“我不如老农民。”又请求学种菜蔬。孔子道：“我不如老菜农。”（杨伯峻《论语译注》） 原文：齐宣王问曰：“汤放桀，武王伐纣，有诸？”（《孟子?梁惠王下》） 译文：齐宣王问道：“商汤流放夏桀，武王讨伐殷纣，真有这回事吗？（杨伯峻《孟子译注》） 上面两段译文紧扣原文，字词落实，句法结构基本上与原文对等，属于直译。 但对直译又不能作简单化理解。由于古今汉语在文字、词汇、语法等方面的差异，今译时对原文作一些适当的调整，是必要的，并不破坏直译。例如： 原文：逐之，三周华不注。（《齐晋鞌之战》） 译文：〔晋军〕追赶齐军，围着华不注山绕了三圈。

unit10bythetimeigotoutsidethebushadalreadyleft期末复习

期末复习提纲（十） Un it 10 By the time I got outside, the bus had already left. 重点词组与句型： 1. take a shower 2.get in the shower 3.get up 4.get outside 5.get to school 6.1 eave sth.+ 地点 7.get back to sp. 8.start teach ing 9.wake up late 10.be late for 11.go off 12.wait for https://www.wendangku.net/doc/847357913.html,e out 14.have some breakfast 15.run off 16.get there 17.on time https://www.wendangku.net/doc/847357913.html,e by 19.give sb. a ride 20.bri ng sth. to sp. 21.break dow n 22.show up 23. get dressed 24.stay up all night 25. invite sb. to do sth. / invite sb. to sp. 26. by the time 27.happa n to sb. 28. a costume party 29.thousa nds of 30.h un dreds of * (31.flee from sp. 32. April Fool ' s DOyB.stop doing sth. 34.as …as one can 35. fool sb. 36.ask sb. to do sth. 37.both…and 38.an hour later 39. get married 40.play a joke on sb) 40. By the time I got up, my brother had already gotte n in the shower. 41. By the time she got up, her brother had already gone into the bathroom. 42. By the time she went outside, the bus had already gone. 43. By the time she got to class, the teacher had already started teach ing. 44. Whe n she got to school, she realized she had left her backpack at home. 45. By the time I got back to school, the bell had rung. 46. My alarm clock didn ' t go off. 47.1 took a quick shower, and had some breakfast, and the n ran off to the bus stop. * (48.Have you ever forgotte n to bring your homework to school? 49.Wells was so convincing that hun dreds of people believed the story. 50.She was thrilled, because she wan ted to get married. 语法： 过去完成时。 练习： 一、根据句子意思和所给首字母提示，用适当单词完成句子。 1. Bad luck. I missed the early bus because I o ______ this morni ng. 2. Perhaps he ' ll dress himself up as a clown in the c _________ party. 3. You must be e ________ a fter swimmi ng for along time. Come and have a rest.

(完整版)the的用法

定冠词the的用法： 定冠词the与指示代词this ,that同源,有“那(这)个”的意思,但较弱,可以和一个名词连用,来表示某个或某些特定的人或东西. (1)特指双方都明白的人或物 Take the medicine.把药吃了. (2)上文提到过的人或事 He bought a house.他买了幢房子. I've been to the house.我去过那幢房子. (3)指世界上独一无二的事物 the sun ,the sky ,the moon, the earth (4)单数名词连用表示一类事物 the dollar 美元 the fox 狐狸 或与形容词或分词连用,表示一类人 the rich 富人 the living 生者 (5)用在序数词和形容词最高级,及形容词等前面 Where do you live?你住在哪? I live on the second floor.我住在二楼. That's the very thing I've been looking for.那正是我要找的东西. (6)与复数名词连用,指整个群体 They are the teachers of this school.(指全体教师) They are teachers of this school.(指部分教师) (7)表示所有,相当于物主代词,用在表示身体部位的名词前 She caught me by the arm.她抓住了我的手臂. (8)用在某些有普通名词构成的国家名称,机关团体,阶级等专有名词前 the People's Republic of China 中华人民共和国 the United States 美国 (9)用在表示乐器的名词前 She plays the piano.她会弹钢琴. (10)用在姓氏的复数名词之前,表示一家人 the Greens 格林一家人(或格林夫妇) (11)用在惯用语中 in the day, in the morning... the day before yesterday, the next morning... in the sky... in the dark... in the end... on the whole, by the way...

2013年安徽省中考英语试题(word版-含标准答案)

２０13年安徽省中考英语试卷 第一部分听力（共五大题,满分３0分) I.关键词语选择（共５小题；每小题1分,满分5分) 你将听到五个句子。请在每小题所给的Ａ、Ｂ、C 三个选项中选出一个你所听到的单词或短语。每个句子读两遍。 1.Ａ. traｄｅ B.ｔrｉp C. tｒain ２． A. warn B．wash C.ｗａtch ３. A. ｓofｔ B. ｓaｆe Ｃ．sａme ４. Ａ. ticｋet B. bａskeｔＣ. pｒｅsｅnｔ' ５. A. ｋeep ｏff B．gｅt oｆfＣ．turn off Ⅱ.短对话理解（共10小题；每小题１分,满分10分）你将听到十段对话，每段对话后有一个小题。请在每小题所给的Ａ、Ｂ、C三个选项中选出一个最佳选项。每段对话读两遍。 6. Wｈat is the ｍａn's job? 7.Wｈat is the man ｇoinｇto ｄo? ８.Whichａnimａｌｉｓｍｅnｔｉoｎeｄiｎtｈe ｃonvｅrsaｔｉon? 9．Wｈat spｏrt wｉll tｈe ｍaｎdo this Saturdaｙ? １0. Ｗhaｔis the mａn looking ｆor？ 1１. Whｅre can the wｏman parｋthe ｃａｒ? A. Ｂy the road. B. In the buiｌdiｎg．Ｃ. Beｈindｔhe builｄiｎｇ. 12．How doｅｓthｅman ｇoｔｏｗorｋthese days？ A. Bｙcar. B. By taxi.C．Bｙｂike. １３. Whａｔdｏｅs the mａｎmｅan? Ａ．He likes hｉs job. B. Ｈe hates hｉs jｏb. C.Ｈｅoftｅn ｌoses jｏbs. 1４.Hoｗmuｃh wilｌｔhe man pay?

齐晋鞌之战原文和译文

鞌之战选自《左传》又名《鞍之战》原文：楚癸酉，师陈于鞌(1)。邴夏御侯，逢丑父为右②。晋解张御克，郑丘缓为右(3)。侯日：“余姑翦灭此而朝食(4)”。不介马而驰之⑤。克伤于矢，流血及屦2 未尽∧6)，曰：“余病矣(7)！”张侯曰：“自始合(8)，而矢贯余手及肘(9)，余折以御，左轮朱殷(10)，岂敢言病吾子忍之！”缓曰：“自始合，苟有险，余必下推车，子岂＿识之(11)然子病矣！”张侯曰：“师之耳目，在吾旗鼓，进退从之。此车一人殿之(12)，可以集事(13)，若之何其以病败君之大事也擐甲执兵(14)，固即死也(15)；病未及死，吾子勉之(16)！”左并辔(17) ，右援拐鼓(18)。马逸不能止(19)，师从之，师败绩。逐之，三周华不注(20) 韩厥梦子舆谓己曰：“旦辟左右！”故中御而从齐侯。邴夏曰：“射其御者，君子也。”公曰：“谓之君子而射之，非礼也。”射其左，越于车下；射其右，毙于车中。綦毋张丧车，从韩厥，曰：“请寓乘。”从左右，皆肘之，使立于后。韩厥俛，定其右。逢丑父与公易位。将及华泉，骖絓于木而止。丑父寝于轏中，蛇出于其下，以肱击之，伤而匿之，故不能推车而及。韩厥执絷马前，再拜稽首，奉觞加璧以进，曰：“寡君使群臣为鲁、卫请，曰：‘无令舆师陷入君地。’下臣不幸，属当戎行，无所逃隐。且惧奔辟而忝两君，臣辱戎士，敢告不敏，摄官承乏。” 丑父使公下，如华泉取饮。郑周父御佐车，宛茷为右，载齐侯以免。韩厥献丑父，郤献子将戮之。呼曰：“自今无有代其君任患者，有一于此，将为戮乎”郤子曰：“人不难以死免其君，我戮之不祥。赦之，以劝事君者。”乃免之。译文1：在癸酉这天，双方的军队在鞌这个地方摆开了阵势。齐国一方是邴夏为齐侯赶车，逢丑父当车右。晋军一方是解张为主帅郤克赶车，郑丘缓当车右。齐侯说:“我姑且消灭了这些人再吃早饭。”不给马披甲就冲向了晋军。郤克被箭射伤，血流到了鞋上，但是仍不停止擂鼓继续指挥战斗。他说:“我受重伤了。”解张说:“从一开始接战，一只箭就射穿了我的手和肘，左边的车轮都被我的血染成了黑红色，我哪敢说受伤您忍着点吧！”郑丘缓说:“从一开始接战，如果遇到道路不平的地方，我必定（冒着生命危险）下去推车，您难道了解这些吗不过，您真是受重伤了。”daier 解张说:“军队的耳朵和眼睛，都集中在我们的战旗和鼓声，前进后退都要听从它。这辆车上还有一个人镇守住它，战事就可以成功。为什么为了伤痛而败坏国君的大事呢身披盔甲，手执武器，本来就是去走向死亡，伤痛还没到死的地步，您还是尽力而为吧。”一边说，一边用左手把右手的缰绳攥在一起，用空出的右手抓过郤克手中的鼓棰就擂起鼓来。（由于一手控马，）马飞快奔跑而不能停止，晋军队伍跟着指挥车冲上去，把齐军打得打败。晋军随即追赶齐军，三次围绕着华不注山奔跑。韩厥梦见他去世的父亲对他说:“明天早晨作战时要避开战车左边和右边的位置。”因此韩厥就站在中间担任赶车的来追赶齐侯的战车。邴夏说:“射那个赶车的，他是个君子。”齐侯说: “称他为君子却又去射他，这不合于礼。”daier 于是射车左，车左中箭掉下了车。又射右边的，车右也中箭倒在了车里。（晋军的）将军綦毋张损坏了自己的战车，跟在韩厥的车后说: “请允许我搭乗你的战车。”他上车后，无论是站在车的左边，还是站在车的右边，韩厥都用肘推他，让他站在自己身后——战车的中间。韩厥又低下头安定了一下受伤倒在车中的那位自己的车右。于是逢丑父和齐侯（乘韩厥低头之机）互相调换了位置。将要到达华泉时，齐侯战车的骖马被树木绊住而不能继续逃跑而停了下来。（头天晚上）逢丑父睡在栈车里，有一条蛇从他身子底下爬出来，他用小臂去打蛇，小臂受伤，但他（为了能当车右）隐瞒了这件事。由于这样，他不能用臂推车前进，因而被韩厥追上了。韩厥拿着拴马绳走到齐侯的马前，两次下拜并行稽首礼，捧着一杯酒并加上一块玉璧给齐侯送上去，

“the way+从句”结构的意义及用法

“tｈｅwａｙ+从句”结构的意义及用法 首先让我们来看下面这个句子: Ｒead the foｌloｗｉnｇpassagｅａnd talkａbout ｉt wi ｔh your classmaｔes．Try tｏｔeｌl whaｔyｏu thiｎk ｏf Tom aｎd ｏｆｔhe ｗay the chiｌｄrenｔrｅated ｈim. 在这个句子中，ｔhe waｙ是先行词,后面是省略了关系副词that或in whｉch的定语从句。 下面我们将叙述“the ｗaｙ+从句”结构的用法。 1.tｈe wａy之后，引导定语从句的关系词是tｈat而不是hoｗ,因此，<<现代英语惯用法词典＞＞中所给出的下面两个句子是错误的：Ｔhｉs is ｔｈeｗayｈowｉtｈａppened. This is the way how he ａlwａｙs treats me. 2．在正式语体中,thaｔ可被in which所代替;在非正式语体中，ｔhat则往往省略。由此我们得到theｗay后接定语从句时的三种模式：1) tｈe ｗaｙ＋ｔhａt-从句２）the way +ｉn whiｃh－从句3) the ｗay +从句 例如：Ｔｈe way(in whiｃh ，thａt) ｔhｅｓｅｃomｒade ｓloｏｋａｔｐrobｌems is wrong．这些同志看问题的方法

不对。 Ｔｈｅｗａy(that ,ｉn ｗhiｃh)yoｕ’ｒe doinｇit is comple ｔeｌy crａzy.你这么个干法,简直发疯。 Wｅａdmiｒｅd him for thｅｗay ｉｎｗhｉch he fａｃeｓdiｆficultieｓ． Waｌlａｃe ａｎd Ｄarwiｎｇreed ｏn the way ｉｎwhi ｃh ｄｉfｆｅrｅnt forms ｏf life had ｂegｕn.华莱士和达尔文对不同类型的生物是如何起源的持相同的观点。 Thｉs is ｔｈe ｗaｙ（tｈａｔ) hｅdｉd it. I lｉkeｄｔhe waｙ(ｔhat) ｓｈeｏｒganized ｔhe ｍeetｉng． 3.theｗay(tｈat)有时可以与hｏw(作“如何”解）通用。例如： That’s tｈe ｗaｙ(tｈaｔ) shｅspoke. = Tｈat’s how shｅspoke.

2013年安徽省中考真题

2013年安徽省初中毕业学业考试 数 学 本试卷共八大题，计23小题，满分150分，考试时间120分钟. 一、选择题（本大题共10小题，每小题4分，满分40分）每小题都给出代号为A 、B 、C 、D 的四个选项，其中只有一个是正确的，请把正确选项的代号写在题后的括号内，每一小题，选对得4分，不选，错选或选出的代号超过一个的（不论是否写在括号内）一律0分 1．-2的倒数是（ ） （A ）- 21 （B ）2 1 （C ） 2 （D ）—2 2．用科学记数法表示537万正确的是（ ） （A ）537×104 （B ）5.37×105 （C ）5.37×106 （D ）0.537×107 3．图中所示的几何体为圆台，其主（正）视图正确的是（ ） 4．下列运算正确的是（ ） （A ）2x +3y =5xy （B ）5m 2·m 3=5m 5 （C ）()2 22 a b a b -=- （D ）m 2·m 3=m 6 5．已知不等式组? ??≥+>-0103x x 其解集在数轴上表示正确的是（ ） 6．如图，AB ∥CD ，∠A +∠E =75°，则∠C 为（ ） （A ）60° （B ）65° （C ）75° （D ）80°

7．目前我国已建立了比较完善的经济困难学生资助体系，某校去年上半年发放给每个经济困难学生389元，今年上半年发放了438元.设每半年．．．发放的资助金额的平均增长率为x ，则下面列出的方程中正确的是（ ） （A ）4382(1)x +=389 （B ）3892(1)x +=438 （C ）389(12)x +=438 （D ）438(12)x +=389 8．如图，随机闭合开关K 1、K 2、K 3中的两个，则能让两盏灯泡同时发光的概率为（ ） （A ）61 （B ）31 （C ）21 （D ）3 2 9．图1所示矩形ABCD 中，BC =x ，CD =y ，y 与x 满足的反比例函数关系如图2所示，等腰 直角三角形AEF 的斜边EF 过C 点，M 为EF 的中点，则下列结论正确的是（ ） （A ）当x =3时，EC ＜EM （B ）当y =9时，EC ＞EM （C ）当x 增大时，EC ·CF 的值增大. （D ）当y 增大时，BE ·DF 的值不变. 10．如图，点P 是等边三角形ABC 外接圆⊙O 上的点，在以下判断中，不正确．．．的是（ ） （A ）当弦PB 最长时，APC △是等腰三角形 （B ）当APC △是等腰三角形时，PO ⊥AC （C ）当PO ⊥AC 时，∠ACP =30° （D ）当∠ACP =30°，PBC △是直角三角形

《鞌之战》阅读答案(附翻译)原文及翻译

《鞌之战》阅读答案（附翻译）原文及翻 译 鞌之战[1] 选自《左传成公二年（即公元前589年）》 【原文】 癸酉，师陈于鞌[2]。邴夏御齐侯[3]，逢丑父为右[4]。晋解张御郤克，郑丘缓为右[5]。齐侯曰：余姑翦灭此而朝食[6]。不介马而驰之[7]。郤克伤于矢，流血及屦，未绝鼓音[8]，曰：余病[9]矣！张侯[10]曰：自始合，而矢贯余手及肘[11]，余折以御，左轮朱殷[12]，岂敢言病。吾子[13]忍之！缓曰：自始合，苟有险[14]，余必下推车，子岂识之[15]？然子病矣！张侯曰：师之耳目，在吾旗鼓，进退从之[16]。此车一人殿之[17]，可以集事[18]，若之何其以病败君之大事也[19]？擐甲执兵，固即死也[20]。病未及死，吾子勉之[21]！左并辔[22]，右援枹而鼓[23]，马逸不能止[24]，师从之。齐师败绩[25]。逐之，三周华不注[26]。 【注释】 [1]鞌之战：春秋时期的著名战役之一。战争的实质是齐、晋争霸。由于齐侯骄傲轻敌，而晋军同仇敌忾、士气旺盛，战役以齐败晋胜而告终。鞌：通鞍，齐国地名，在今山东济南西北。 [2]癸酉：成公二年的六月十七日。师，指齐晋两国军队。陈，

列阵，摆开阵势。 [3]邴夏：齐国大夫。御，动词，驾车。御齐侯，给齐侯驾车。齐侯，齐国国君，指齐顷公。 [4]逢丑父：齐国大夫。右：车右。 [5]解张、郑丘缓：都是晋臣，郑丘是复姓。郤（x ）克，晋国大夫，是这次战争中晋军的主帅。又称郤献子、郤子等。 [6]姑：副词，姑且。翦灭：消灭，灭掉。朝食：早饭。这里是吃早饭的意思。这句话是成语灭此朝食的出处。 [7]不介马：不给马披甲。介：甲。这里用作动词，披甲。驰之：驱马追击敌人。之：代词，指晋军。 [8] 未绝鼓音：鼓声不断。古代车战，主帅居中，亲掌旗鼓，指挥军队。兵以鼓进，击鼓是进军的号令。 [9] 病：负伤。 [10]张侯，即解张。张是字，侯是名，人名、字连用，先字后名。 [11]合：交战。贯：穿。肘：胳膊。 [12]朱：大红色。殷：深红色、黑红色。 [13]吾子：您，尊敬。比说子更亲切。 [14]苟：连词，表示假设。险：险阻，指难走的路。 [15]识：知道。之，代词，代苟有险，余必下推车这件事，可不译。 [16]师之耳目：军队的耳、目（指注意力）。在吾旗鼓：在我们

(完整版)time的用法总结

一．time的短语 from time to time 有时 on time 准时, in time 及时; all the time 始终,一直; at the same time 同时， ahead of time提前 at no time 绝不 some time一段时间 sometime“在某一时候”。可用来指过去或将来 sometimes (at times, from time to time) “有时，不时” at a time （a time） at one time （once）

at times ( sometimes) in no time （immediately）立刻,马上; have a good/nice time (enjoy oneself) “过的愉快 for the time being “暂时” Many a time/many times 多次 take one’s time从容 kill time消磨时间 【活学活用】选出与画线部分意思相同或相近的选项 1. Jim comes to visit us from time to time. That’s always the happiest time for the family. A.on time B. sometime C. at times D. some times

2．At no time _____study though __ ___great progress. A. should we give up; we have made B. shouldn’t we give up; we have made C. we should give up; we have made D. we shouldn’t give up; have we made 3.---When shall we visit the Science Museum?” ---_________ next week.” A.Sometime B. Sometimes C. Some time D. Sometimes 答案：1. C2.A 3A 二．time相关从属连词高考常考点 1. every time / each time每次 Every time I call on him, he is out.

way 用法

表示“方式”、“方法”，注意以下用法： 1.表示用某种方法或按某种方式，通常用介词in(此介词有时可省略)。如： Do it (in) your own way. 按你自己的方法做吧。 Please do not talk (in) that way. 请不要那样说。 2.表示做某事的方式或方法，其后可接不定式或of doing sth。 如： It’s the best way of studying [to study] English. 这是学习英语的最好方法。 There are different ways to do [of doing] it. 做这事有不同的办法。 3.其后通常可直接跟一个定语从句(不用任何引导词)，也可跟由that 或in which 引导的定语从句，但是其后的从句不能由how 来引导。如： 我不喜欢他说话的态度。 正：I don’t like the way he spoke. 正：I don’t like the way that he spoke. 正：I don’t like the way in which he spoke. 误：I don’t like the way how he spoke. 4.注意以下各句the way 的用法： That’s the way (=how) he spoke. 那就是他说话的方式。 Nobody else loves you the way(=as) I do. 没有人像我这样爱你。 The way (=According as) you are studying now, you won’tmake much progress. 根据你现在学习情况来看，你不会有多大的进步。 2007年陕西省高考英语中有这样一道单项填空题： ——I think he is taking an active part insocial work. ——I agree with you_____. A、in a way B、on the way C、by the way D、in the way 此题答案选A。要想弄清为什么选A，而不选其他几项，则要弄清选项中含way的四个短语的不同意义和用法，下面我们就对此作一归纳和小结。 一、in a way的用法 表示：在一定程度上，从某方面说。如： In a way he was right.在某种程度上他是对的。注：in a way也可说成in one way。 二、on the way的用法 1、表示：即将来(去)，就要来(去)。如： Spring is on the way.春天快到了。 I'd better be on my way soon.我最好还是快点儿走。 Radio forecasts said a sixth-grade wind was on the way.无线电预报说将有六级大风。 2、表示：在路上，在行进中。如： He stopped for breakfast on the way.他中途停下吃早点。 We had some good laughs on the way.我们在路上好好笑了一阵子。 3、表示：(婴儿)尚未出生。如： She has two children with another one on the way.她有两个孩子，现在还怀着一个。 She's got five children，and another one is on the way.她已经有5个孩子了，另一个又快生了。 三、by the way的用法