Human experimental evidence of predictive influence in coherent motion integration 

Get Complete Project Material File(s) Now! »

Excitatory versus inhibitory horizontal connectivity

According to Kisvarday and Eysel (1993), in the cat, inhibitory lateral connections only come from GABA related basket cells. To the contrary of excitatory horizontal connectivity, the distribution of inhibitory interneurons collaterals does not present the orientation selective “patchy” aspect of excitatory lateral connectivity. It seems to be more uniform and to cover a smaller distance of about 2.5 mm (Albus et al., 1991; Kisvarday and Eysel, 1993; Kisvarday et al., 1994, 1997). In those regions, cells contact both pyramidal cells and interneurons, forming a relatively large inhibitory field, but also a des-inhibitory field reaching 4 to 5 mm.
In contrast to what has been reported for excitatory projections, Kisvarday and Eysel (1993) did not observe significant differences between the iso-oriented, cross-oriented or oblique preferred orientations of boutons contacted by inhibitory cells in area 17 and 18 of the cat. Investigating further, Kisvarday et al. (1994) highlighted some differences: 43% of inhibitory axonic endings reach cortical sites with similar orientations, 35% oblique sites and 22% contact cross-oriented domains, results that were later reproduced (Kisvarday et al., 1997). These authors concluded to a significant difference between excitatory and inhibitory long-distance projections: even though those two networks strongly overlap, the iso-orientation preference bias of the inhibitory network was found weaker than the excitatory one.

Role of feature geometry in response modulation

Whether facilitatory or suppressive, it was shown both in cat and monkey that maximal modulations are observed when the central and peripheral stimuli are iso-oriented and co-aligned (cat: Nelson and Frost, 1985; Chen et al., 2001, macaque: Knierim and Van Essen, 1992; macaque and human: Kapadia et al., 1995, 2000), strongly advocating for the implication of horizontal connectivity in those modulatory effects. It seems that suppressive modulation is the predominant type when central and peripheral stimuli are iso-oriented and co-aligned but the picture is not that simple and needs to be decomposed. Indeed, it has been shown that iso-oriented and co-aligned surrounding stimuli mostly have a suppressive effect when the contrast of the center stimulus is high. It appears that the contrast of the central stimulus controls the sign of the response modulation: at weak contrast, a similar stimulus presented in the surround can facilitate the responses but suppress it if the contrast of the stimulus in the center is high (Figure 2.2) (Toth et al., 1996; Polat et al., 1998; Sengpiel et al., 1997; Levitt and Lund, 1997a).

Overall center-surround contextual modulations

Surprisingly, the results of Sillito et al., (1995) showed that facilitatory modulations were also present for a center stimulation orthogonal to the optimal one, as long as the surrounding annulus grating was orthogonal to the center stimulus. Such cells were termed “orientation contrast” detectors. In the same line of evidence, the tuning of surround modulation in the macaque V1 depends on the stimulus orientation presented to the RF. Compared to the sole stimulation of the center, facilitation even emerges in many cells when both the RF and its surround are non-optimally stimulated, as long as the two stimuli presented are cross oriented (Shushruth et al., 2012), (Figure 2.3), highlighting a real adaptive facilitation based on the relative orientation between center and surround. According to these authors, tuned lateral inhibition (via the surround pathways) of untuned local recurrent connections causes maximal withdrawal of recurrent excitation at the feedforward-input orientation, resulting in the stimulus-dependent tuning of the surround.
Center-surround interactions can be considered as context-dependent modulations, relying on dynamic adaption to global statistical changes in the visual environment and the RF surround. They enhance the transmission of relevant information even when the center stimulation is not optimal for a given RF and suppress irrelevant redundant features of a visual scene, conferring to the RFs a status of constantly evolving entity.

Intracellular recordings: spatio-temporal subthreshold characterisation of the receptive field silent surround

As previously developed, the cortical area activated by a localized stimulus does not pertain to the feedforward retinotopic imprint of this latter but spread over larger distances via horizontal connections. Intrinsic and voltage sensitive dye imaging reveal the existence of a large depolarising field encompassing several times the size of the spiking response studied extracellularly. Thus, the RF definition, especially its spatial extent, is function of the technique used to characterise it and of the type of stimuli used (Frégnac and Bringuier, 1996; Walker et al., 2000). Different experimental techniques provide different types of information and characterisations of RFs. Contrary to optical imaging techniques and electrophysiological extracellular recordings, intracellular recordings give access to the subthreshold synaptic activity (synaptic bombardment inputs that a given cell receives) as well as the output spiking responses at the single cell level.
The Minimum Discharge Field (MDF) represents the region of space eliciting spiking responses. In accordance with the findings of Das and Gilbert (1995) showing large subthreshold depolarising field with intrinsic optical imaging, Bringuier et al. (1999) showed using intracellular recordings in vivo in the primary visual cortex of the cat that the spatial extent of the depolarising field (DF) is much larger than the MDF and the hyperpolarising field (HF) displays a sparser spatial profile. These authors showed that the spatial sensitivity profile of the DF envelope is co-centered with the MDF (Figure 2.4A). Taken together, the topographical combination of the MDF, the DF and the HF covers on average 5 ± 2.4° of visual angle, an area 4 times larger than the MDF classically defined by extracellular studies. In that study, significant subthreshold depolarizing responses could be evoked by annular flashed gratings surrounding the MDF up to 11.3° without stimulation of this latter, indicating the existence of excitatory sensitivity to large surrounding stimuli in the “far periphery” of the classical MDF, much further than originally thought from extracellular studies. The strength of the depolarizing response decreases linearly regarding eccentricity from the MDF center (Figure 2.4A). Moreover, the slope of that linear decrease depends on the spatial summation provided by the test stimulus itself: it is comparable between two-dimensional sparse noise and flashed bars (Figure 2.4B and C). However, the spatial attenuation of depolarising responses was found weaker for annular sinusoidal gratings of larger eccentricity, i.e. the slope of response attenuation over space was much lower when spatial summation was increased (Figure 2.4D). Moreover, the modulations of responses to annular gratings observed in that study were selective to orientation, indicating a cortical origin of interaction mechanisms.
Another source of information on the origin of those center-surround modulations is the onset latency of the responses. Indeed a strong correlation was observed in the study of Bringuier et al., (1999) between the relative eccentricity of the flashed stimuli from the MDF center and the latencies of the evoked depolarizing responses (Figure 2.4 E-G). Taking into account the magnification factor between the visual field and the cortex, the authors converted the visual field distance between stimuli positions into cortical distance. The speed of horizontal propagation of activity was directly derived from the inverse of the slopes obtained from the latency basins distributions. They found a speed of horizontal activity propagation ranging from 0.1 to 0.5 m/s (Figure 2.4 H), a speed matching the velocity of activity spread confirmed by optical imaging techniques (Das and Gilbert, 1995; Grinvald et al., 1994; Shoham et al., 1999; Per Roland, 2002; Jancke et al., 2004; Chavane et al., 2011). Although the propagation of activity responsible for center-surround modulations could come from higher-order cortical areas via top-down feedback connections, it is unlikely. Indeed, the speed of horizontal propagation directly estimated from those spatio-temporal characterizations of receptive fields is consistent with in vitro intracellular measurements of horizontal action potential propagation speed in rat and cat visual cortex (Salin and Prince, 1996, Chervin et al., 1988; Komatsu et al., 1988; Hirsch and Gilbert, 1991; Nowak and Bullier, 1998). Feedback connections from higher areas which displays larger MDFs than V1 could explain large synaptic integration fields in the primary visual cortex but there is evidence against that interpretation. Indeed, by opposition to constant speed propagation of cortical wave of activity by horizontal connections, feedback connections do not account for the linear spatio-temporal dependency in visual onset latencies observed here, because of their much higher conduction speed velocity (superior to HC propagation speed by a factor of 10: Girard et al., 2001; Angelucci and Bullier, 2003; Briggs and Usrey, 2007). Only two milliseconds are needed to convey feedback information from MT to V1. In contradiction with common belief, late visual cortex responses do not necessarily only signal reverberation between cortical areas since the slowest signal propagation, in the order of several tens of milliseconds, may result from unmyelinated propagation along long distance horizontal connections intrinsic to V1.

Stimulus-induced cooperativity is necessary for the anisotropic spread of lateral activity

The second study contributing to our working hypothesis investigated more specifically the requirements for the anisotropic spread of lateral activity. To better understand the functional role of lateral interactions, Chavane et al. (2011) combined voltage sensitive dye imaging (done in the lab of Amiram Grinvald, Weizmann) and intracellular recordings (done at UNIC) in area 17 and 18 of the anaesthetized and paralysed cat in vivo. VSD imaging showed that locally oriented stimuli evoked an orientation-selective activity restricted to the cortical feedforward input of the stimulus. Beyond that feedforward imprint of approximately the size of an hypercolumn, the laterally activated area gradually lost its orientation selectivity with a space constant of about 1 mm. Intracellular recordings showed that this loss of orientation preference comes from the orientation preference diversity of converging synaptic input originating from outside the classical RF. However, increasing the stimulus size provoked an extension of the anisotropic spread of cortical activity beyond the feedforward imprint, suggesting that stimulus-induced cooperativity enhances the long-range spread of iso-preference.
First, the authors presented local oriented sinusoidal luminance gratings through a circular aperture whose size was adjusted to the average RF dimensions. Maps evoked by local stimuli were compared to maps of cortical activation obtained from full-field stimulation. Stimuli were presented at four different orientations and normalized by a “blank” stimulus (Figure 3.4 A). The white contours on that figure represent the domain within which each pixel activation was significantly higher, on a trial by trial basis, than the spontaneous level. The time sequence shows an initial local activation at a latency of about 40 ms after stimulus onset before a gradual spread over most of the imaged cortical surface. The speed of this long-range lateral spread was estimated by latency regarding eccentricity and reached a velocity of about 0.09 m/s (a value compatible with that inferred from synaptic echoes: Bringuier et al, 1999). To distinguish feedforward and laterally activated areas, retinotopic mapping was obtained by presenting local stimuli in adjacent visual positions. Local magnification factors were calculated by comparing the cortical distance between the centers of gravity of each visual stimulus cortical activation and the visual distance between the centers of the corresponding stimuli in visual space. These local magnification factors were used to approximate the extent of the retinotopic representation (Figure 3.4 C, black ellipse). The superposition of the stimulus retinotopic representation on the area covered by the spread allow to see that the lateral spread indeed covers large cortical territory extending far beyond the retinotopic activation and covering various orientation domains. (Figure 3.4 D).


Table of contents :

General Introduction
1.1 The construction of a global percept
1.1.1 Psychophysical rules underlying perceptual binding and information extraction
1.2 Receptive fields along the visual pathway
1.2.1 Simple and complex cells of the primary visual cortex
1.2.2 A columnar organization
1.2.3 A hierarchical processing of information
1.2.4 Questioning the hierarchical model
1.3 Center-surround interactions: physiological definition, spatial scale, connectivity’s origins across species and link with perception
1.3.1 Contrast-dependent spatial summation of the Receptive field
1.3.2 Thalamo-cortical contribution of macaque LGN inputs to V1 center-surround modulations …..
1.3.3 Aggregate receptive field and intra V1 lateral connectivity’s contribution to macaque V1 near and far surround modulation
1.3.4 Lateral connectivity, collinear facilitation and link with perception
1.3.5 Differences and similarities between cat and monkey’s lateral connectivity: implication for perception
1.4 Connectivity types and canonical circuits involved in lateral processing
1.4.1 Thalamo-cortical feedforward connectivity’s contribution to V1 center-surround modulations 53
1.4.2 Feedback connectivity
1.4.3 A laminar dependency?
1.4.4 Local recurrent connectivity
1.4.5 Horizontal connectivity
1.5 Role of V1 lateral connectivity in low level perception
Part I – Lateral connectivity and the propagation of network belief
I-1. Background
I-1.1 Anatomy
I-1.1.1 Iso-orientation bias in excitatory horizontal connectivity?
I-1.1.2 Spatial spread of horizontal connections
I-1.1.3 Iso-orientation binding across the retinotopic map
I-1.1.4 Excitatory versus inhibitory horizontal connectivity
I-1.2 Functional role
I-1.2.1 Role of feature geometry in response modulation
I-1.2.2 Overall center-surround contextual modulations
I-1.2.3 The depolarizing field hypothesis
I-1.2.4 Intracellular recordings: spatio-temporal subthreshold characterisation of the receptive field silent surround
I-2. Working Hypothesis
I-2.1 Spread of horizontal activity in V1
I-2.1.1 Lateral and feedforward activity interact in a spatio-temporal coherent way to shape the overall propagation of cortical activity
I-2.1.2 Stimulus-induced cooperativity is necessary for the anisotropic spread of lateral activity ……
I-2.1.3 The existence of a synaptic dynamic association field favouring the integration of iso-aligned elements composing a centripetal flow
I-2.2 Exploring in depth the dynamic association field and its implication in the propagation of a prediction travelling through the V1 network
I-2.3 Filling-in and predictive responses
I-3. Visual stimuli design
I-3.1 Geometric design of the stimulation and definition of a common cellulo-centric referential
I-3.2 General Spatio-Temporal design of apparent motion (AM) sequences
I-3.3 Contrast conditions
I-3.4 Probing filling-in or predictive responses
I-3.5 Probing the implication of horizontal connectivity
I-3.6 Manipulating the spatial coherence of the flows
I-3.7 Manipulating the spatial and temporal coherence of the flows
I-3.8 Probing the existence of local versus global motion flow detectors
I-4. Material and methods
I-4.1 Animal breeding
I-4.2 Surgical procedure and animal preparation
I-4.3 Intracellular recordings
I-4.4 On-line characterization of receptive fields
I-4.5 Histology
I-4.6 – Part I Assessing cells responses linear prediction
I-5. Results
I-5.1 Assessing visual response significance
I-5.2 Input summation during centripetal apparent motion
I-5.3 Spatio-temporal coherence is necessary for binding lateral and horizontal waves
I-5.4 Apparent motion speed needs to match the cortical horizontal propagation speed
I-5.5 Predictive/filling-in responses
I-5.6 Reversal of V1 neuron axial sensitivity as a function of the retinal flow speed
I-6. Discussion
I-6.1. General context of the findings
I-6.2. Summary of the main findings
I-6.3 Gain control during centripetal apparent motion
I-6.4 Non–linear Center–Surround spatio–temporal integration during apparent motion: the roles of spatial synergy and temporal coherence and their implication in the predictive coding hypothe
I-6.5 Experimental evidence of prediction influence throughout the cortical hierarchy
I-6.6 Human experimental evidence of predictive influence in coherent motion integration
I-6.7 Filling-in as a prediction
I-6.8 V1’s latency advance as a neuronal correlate of psychophysical speed overestimation of co-aligned elements apparent motion
I-6.9 Functional shift of the synaptic integration field anisotropy as a function of the speed of the global motion flow
I-6.10 Reconfiguration of retinal flow integration by lateral connectivity during oculomotor exploration
I-6.11 Binding in-phase versus rolling wave
I-6.12 Future work and bottleneck
Part II – Lateral connectivity and hallucinatory-like states in V1
II-1. Background
II-1.1 On the origins of Geometric hallucinations in the brain
II-1.2 Beyond Mackay’s after-effect: sequential and simultaneous contrast adaptation and spatial opponency
II-1.3 Theoretical models of selection and stabilisation of spatio-temporal stable states intrinsic to V1’s functional architecture
II-2. Working Hypothesis
II-2.1 Geometric pattern formation results from a repulsive shift in orientation within and between hypercolumns: link between experimental data and theoretical models
II-2.1.1 Neural Stripes of activity in V1 results in a smooth spatial gradient of orientation preference shifts
II-2.1.2 Psychophysical evidence of repulsive shift in orientation
II-2.1.3 Electrophysiological evidence of repulsive shift in orientation
II-2.1.4 Lateral connectivity as a self-organizing engine for the spatial propagation of fast repulsive orientation shift adaptation
II-2.1.5 Model of coupled hypercolumns under global repulsive shift
II-2.1.6 Formation of static vs dynamic hallucinatory planforms
II-2.2 Geometric physical inducers constrain orientation domains exploration: towards a directly inducible and temporally stable percept
II-2.2.1 Experimental evidence of smooth spontaneous orientation transition between neighbouring hypercolumns
II-2.2.2 Biasing ongoing activity’s orientation exploration by the physical presentation of a geometric planform inducer combined with a perturbation
II-2.2.3 Oscillatory activity accompanies dynamic, propagating percepts: implications of 1/f α filtered noise in pattern formation and propagation
II-2.3 The cat as a model of geometric hallucinations in the brain
II-3. Visual stimuli design
II-3.1 Probing the geometric nature of hallucinatory-like propagating waves of activity and maximizing their detectability at the single cell level
II-3.2 Probing the need of a synergistic interaction between geometric inducer and 1/fα noise in the emergence of hallucinatory-like waves of activity
II-3.3 Geometry’s position and luminance equalization across conditions
II-3.4 Temporal display of the stimuli
II-3.5 Generation of 1/fα spatiotemporally filtered noise
II-3.5.1 Spatial filtering of the noise
II-3.5.2 Temporal filtering of the noise
II-4. Results
II-4.1 Detecting oscillatory activity
II-4.2 Stimulus-locked analysis
II-4.3 Sustained nature of the oscillations and autocorrelation measures
II-4.4 Comparison between the Power spectral content of stimulus locked and stimulus-induced response components
II-4.4.1 Average cross spectrum between each pair of trials and dephasing
II-4.4.2 Signal, Noise, total signal power and Signal to Noise Ratio evaluation
II-4.5 Population analysis
II-4.5.1 Power spectral content of the trials
II-4.5.2 Dephasing of the cross spectrum average and power spectral content of stimulus-locked oscillations
II-4.5.3 Signal, noise, and total signal power spectral estimation
II-4.5.4 Signal to noise ratio and information rate
II-4.6 Supplementary figures
II-5. Discussion
II-5.1. From subjectivity to an Archetype
II-5.2. Lateral diffusion of repulsive shift in orientation between neighbouring hypercolumns as a potential mechanism of neural stripe formation on V1
II-5.3. Multiscale evidence of repulsive shift in orientation as the substrate of our model
II-5.4. Potential implications and parametrization
II-5.5. Experimental framework: subthreshold detectability of hallucinatory-like propagating waves of activity by intracellular recordings in the cat V1
II-5.6. Neural stripe formation and propagation refractory period as a signature of an intrinsic clock of the primary visual cortex
II-5.7. Future work and potential implications
Bibliography ..


Related Posts