Divisive normalization processors in the early visual system of the Drosophila brain

Divisive normalization is a model of canonical computation of brain circuits. We demonstrate that two cascaded divisive normalization processors (DNPs), carrying out intensity/contrast gain control and elementary motion detection, respectively, can model the robust motion detection realized by the early visual system of the fruit fly. We first introduce a model of elementary motion detection and rewrite its underlying phase-based motion detection algorithm as a feedforward divisive normalization processor. We then cascade the DNP modeling the photoreceptor/amacrine cell layer with the motion detection DNP. We extensively evaluate the DNP for motion detection in dynamic environments where light intensity varies by orders of magnitude. The results are compared to other bio-inspired motion detectors as well as state-of-the-art optic flow algorithms under natural conditions. Our results demonstrate the potential of DNPs as canonical building blocks modeling the analog processing of early visual systems. The model highlights analog processing for accurately detecting visual motion, in both vertebrates and invertebrates. The results presented here shed new light on employing DNP-based algorithms in computer vision.

Is predictive coding falsifiable?

Predictive-coding has justifiably become a highly influential theory in Neuroscience. However, the possibility of its unfalsifiability has been raised. We argue that if predictive-coding were unfalsifiable, it would be a problem, but there are patterns of behavioural and neuroimaging data that would stand against predictive-coding. Contra (vanilla) predictive patterns are those in which the more expected stimulus generates the largest evoked-response. However, basic formulations of predictive-coding mandate that an expected stimulus should generate little, if any, prediction error and thus little, if any, evoked-response. It has, though, been argued that contra (vanilla) predictive patterns can be obtained if precision is higher for expected stimuli. Certainly, using precision, one can increase the amplitude of an evoked-response, turning a predictive into a contra (vanilla) predictive pattern. We demonstrate that, while this is true, it does not present an absolute barrier to falsification. This is because increasing precision also reduces latency and increases the frequency of the response. These properties can be used to determine whether precision-weighting in predictive-coding justifiably explains a contra (vanilla) predictive pattern, ensuring that predictive-coding is falsifiable.

Effects of α-lactalbumin on strength, fatigue and psychological parameters: a randomised double-blind cross-over study

PURPOSE

The neurotransmitter serotonin has a strong effect on behaviour and motor control. Regarding motor control, serotonin contributes to the development of fatigue and is also involved in the ability of motor neurones to operate across a large range of forces (gain control). The consumption of tryptophan-rich supplements (such as α-lactalbumin) is of interest because this amino acid is the only precursor for brain serotonin synthesis. Therefore, the purpose of this study was to determine the effects of α-lactalbumin supplementation on neuromuscular performance.

METHODS

Using a randomised double-blind cross-over design, 16 healthy participants performed plantar flexor and handgrip maximal voluntary contractions, a 30-s submaximal handgrip contraction, and a plantar flexor fatigue protocol before and 90 min after consuming either 40 g of α-lactalbumin, an isonitrogenous beverage (Zein) or an isocaloric beverage (corn-starch). Sleepiness, mood, and cognition were assessed to evaluate any psychological effects.

RESULTS

α-Lactalbumin decreased force steadiness by 25% during the sustained submaximal handgrip contraction (p < 0.01) and induced greater fatigue (15% reduction in total torque-time integral, p = 0.01) during the fatigue protocol. These effects were not observed for the other control beverages. No effects were found for maximal or explosive strength, or psychological measurements.

CONCLUSIONS

40 g of α-lactalbumin increased handgrip force variability and reduced performance during fatiguing muscle contractions but did not influence brief maximal contractions or psychological parameters in healthy individuals. These findings support the hypothesis that the consumption of α-lactalbumin can increase motor neurone input-output gain and exacerbate central fatigue during sustained maximal exercise.

Neural mechanism of experience-dependent sensory gain control in C. elegans

Animals' sensory systems adjust their responsiveness to environmental stimuli that vary greatly in their intensity. Here we report the neural mechanism of experience-dependent sensory adjustment, especially gain control, in the ASH nociceptive neurons in Caenorhabditis elegans. Using calcium imaging under gradual changes in stimulus intensity, we find that the ASH neurons of naive animals respond to concentration increases in a repulsive odor 2-nonanone regardless of the magnitude of the concentration increase. However, after preexposure to the odor, the ASH neurons exhibit significantly weak responses to a small gradual increase in odor concentration while their responses to a large gradual increase remain strong. Thus, preexposure changes the slope of stimulus-response relationships (i.e., gain control). Behavioral analysis suggests that this gain control contributes to the preexposure-dependent enhancement of odor avoidance behavior. Mathematical analysis reveals that the ASH response consists of fast and slow components, and that the fast component is specifically suppressed by preexposure for the gain control. In addition, genetic analysis suggests that G protein signaling may be required for the regulation of fast component. We propose how prior experience dynamically and specifically modulates stimulus-response relationships in sensory neurons, eventually leading to adaptive modulation of behavior.

Long-range connections enrich cortical computations

The cerebral cortex can perform powerful computations, including those involved in higher cognitive functions. Cortical processing for such computations is executed by local circuits and is further enriched by long-range connectivity. This connectivity is activated under specific conditions and modulates local processing, providing flexibility in the computational performance of the cortex. For instance, long-range connectivity in the primary visual cortex exerts facilitatory impacts when the cortex is silent but suppressive impacts when the cortex is strongly sensory-stimulated. These dual impacts can be captured by a divisive gain control model. Recent methodological advances such as optogenetics, anatomical tracing, and two-photon microscopy have enabled neuroscientists to probe the circuit and synaptic bases of long-range connectivity in detail. Here, I review a series of evidence indicating essential roles of long-range connectivity in visual and hierarchical processing involving numerous cortical areas. I also describe an overview of the challenges encountered in investigating underlying synaptic mechanisms and highlight recent technical approaches that may overcome these difficulties and provide new insights into synaptic mechanisms for cortical processing involving long-range connectivity.

Dynamic biosonar adjustment strategies in deep-diving Risso's dolphins driven partly by prey evasion

Toothed whales have evolved flexible biosonar systems to find, track and capture prey in diverse habitats. Delphinids, phocoenids and iniids adjust inter-click intervals and source levels gradually while approaching prey. In contrast, deep-diving beaked and sperm whales maintain relatively constant inter-click intervals and apparent output levels during the approach followed by a rapid transition into the foraging buzz, presumably to maintain a long-range acoustic scene in a multi-target environment. However, it remains unknown whether this rapid biosonar adjustment strategy is shared by delphinids foraging in deep waters. To test this, we investigated biosonar adjustments of a deep-diving delphinid, the Risso's dolphin (Grampus griseus). We analyzed inter-click interval and apparent output level adjustments recorded from sound recording tags to quantify in situ sensory adjustment during prey capture attempts. Risso's dolphins did not follow typical (20logR) biosonar adjustment patterns seen in shallow-water species, but instead maintained stable repetition rates and output levels up to the foraging buzz. Our results suggest that maintaining a long-range acoustic scene to exploit complex, multi-target prey layers is a common strategy amongst deep-diving toothed whales. Risso's dolphins transitioned rapidly into the foraging buzz just like beaked whales during most foraging attempts, but employed a more gradual biosonar adjustment in a subset (19%) of prey approaches. These were characterized by higher speeds and minimum specific acceleration, indicating higher prey capture efforts associated with evasive prey. Thus, tracking and capturing evasive prey using biosonar may require a more gradual switch between multi-target echolocation and single-target tracking.

Binocular vision adaptively suppresses delayed monocular signals

A neutral density filter placed before one eye will produce a dichoptic imbalance in luminance, which attenuates responses to visual stimuli and lags neural signals from retina to cortex in the filtered eye. When stimuli are presented to both the filtered and unfiltered eye (i.e., binocularly), neural responses show little attenuation and no lag compared with their baseline counterpart. This suggests that binocular visual mechanisms must suppress the attenuated and delayed input from the filtered eye; however, the mechanisms involved remain unclear. Here, we used a Steady-State Visual Evoked Potential (SSVEP) technique to measure neural responses to monocularly and binocularly presented stimuli while observers wore an ND filter in front of their dominant eye. These data were well-described by a binocular summation model, which received the sinusoidal contrast modulation of the stimulus as input. We incorporated the influence of the ND filter with an impulse response function, which adjusted the input magnitude and phase in a biophysically plausible manner. The model captured the increase in attenuation and lag of neural signals for stimuli presented to the filtered eye as a function of filter strength, while also generating the filter phase-invariant responses from binocular presentation for EEG and psychophysical data. These results clarify how binocular visual mechanisms-specifically interocular suppression-can suppress the delayed and attenuated signals from the filtered eye and maintain normal neural signals under imbalanced luminance conditions.

On the Subspace Invariance of Population Responses

In cat visual cortex, the response of a neural population to the linear combination of two sinusoidal gratings (a plaid) can be well approximated by a weighted sum of the population responses to the individual gratings - a property we refer to as subspace invariance. We tested subspace invariance in mouse primary visual cortex by measuring the angle between the population response to a plaid and the plane spanned by the population responses to its individual components. We found robust violations of subspace invariance arising from a strong, negative correlation between the responses of neurons to individual gratings and their responses to the plaid. Contrast invariance, a special case of subspace invariance, also failed. The responses of some neurons decreased with increasing contrast, while others increased. Altogether the data show that subspace and contrast invariance do not hold in mouse primary visual cortex. These findings rule out some models of population coding, including vector averaging, some versions of normalization and temporal multiplexing.

Development of contrast normalization mechanisms during childhood and adolescence

Contrast sensitivity is regulated by neural mechanisms that flexibly adjust responsiveness to optimize stimulus encoding across different environments. Here we studied the developmental status of gain control mechanisms in school-age children (5-17years) and adults using a visual masking paradigm. A variable contrast, spatially random 2-D noise test pattern was masked by the presence of a superimposed independent noise pattern presented at 0, 12 and 40% contrast. Frequency-tagged steady state visual evoked potentials were used to separately record responses to the test (5.14Hz) and the mask (7.2Hz). By incrementally increasing the test contrast we measured contrast response functions for each mask contrast. The unmasked contrast response functions were largely similar in shape across age, but peak amplitude was higher in the children. Masking shifted the contrast response function rightward on the contrast axis in both the adults and older children, elevating contrast thresholds by a similar factor across age. However, in younger children, masking resulted in a change in the slope of the contrast response function. These findings suggest that immaturity in the contrast normalization process persists until approximately 11years of age.

Subliminal stimulation and somatosensory signal detection

Only a small fraction of sensory signals is consciously perceived. The brain's perceptual systems may include mechanisms of feedforward inhibition that protect the cortex from subliminal noise, thus reserving cortical capacity and conscious awareness for significant stimuli. Here we provide a new view of these mechanisms based on signal detection theory, and gain control. We demonstrated that subliminal somatosensory stimulation decreased sensitivity for the detection of a subsequent somatosensory input, largely due to increased false alarm rates. By delivering the subliminal somatosensory stimulus and the to-be-detected somatosensory stimulus to different digits of the same hand, we show that this effect spreads across the sensory surface. In addition, subliminal somatosensory stimulation tended to produce an increased probability of responding "yes", whether the somatosensory stimulus was present or not. Our results suggest that subliminal stimuli temporarily reduce input gain, avoiding excessive responses to further small inputs. This gain control may be automatic, and may precede discriminative classification of inputs into signals or noise. Crucially, we found that subliminal inputs influenced false alarm rates only on blocks where the to-be-detected stimuli were present, and not on pre-test control blocks where they were absent. Participants appeared to adjust their perceptual criterion according to a statistical distribution of stimuli in the current context, with the presence of supraliminal stimuli having an important role in the criterion-setting process. These findings clarify the cognitive mechanisms that reserve conscious perception for salient and important signals.

Cholinergic and serotonergic modulation of visual information processing in monkey V1

The brain dynamically changes its input-output relationship depending on the behavioral state and context in order to optimize information processing. At the molecular level, cholinergic/monoaminergic transmitters have been extensively studied as key players for the state/context-dependent modulation of brain function. In this paper, we review how cortical visual information processing in the primary visual cortex (V1) of macaque monkey, which has a highly differentiated laminar structure, is optimized by serotonergic and cholinergic systems by examining anatomical and in vivo electrophysiological aspects to highlight their similarities and distinctions. We show that these two systems have a similar layer bias for axonal fiber innervation and receptor distribution. The common target sites are the geniculorecipient layers and geniculocortical fibers, where the appropriate gain control is established through a geniculocortical signal transformation. Both systems exert activity-dependent response gain control across layers, but in a manner consistent with the receptor subtype. The serotonergic receptors 5-HT1B and 5HT2A modulate the contrast-response curve in a manner consistent with bi-directional response gain control, where the sign (facilitation/suppression) is switched according to the firing rate and is complementary to the other. On the other hand, cholinergic nicotinic/muscarinic receptors exert mono-directional response gain control without a sign reversal. Nicotinic receptors increase the response magnitude in a multiplicative manner, while muscarinic receptors exert both suppressive and facilitative effects. We discuss the implications of the two neuromodulator systems in hierarchical visual signal processing in V1 on the basis of the developed laminar structure.

Blockade of dorsolateral pontine 5HT1A receptors destabilizes the respiratory rhythm in C57BL6/J wild-type mice

The neurotransmitter serotonin (5HT) acting via 5HT1a receptors (5HT1aR) is a potent determinant of respiratory rhythm variability. Here, we address the 5HT1aR-dependent control of respiratory rhythm variability in C57BL6/J mice. Using the in situ perfused preparation, we compared the effects of systemic versus focal blockade of 5HT1aRs. Blocking 5HT1aRs in the Kölliker-Fuse nucleus (KFn) increased the occurrence of spontaneous apneas and accounted for the systemic effects of 5HT1aR antagonists. Further, 5HT1aRs of the KFn stabilized the respiratory rhythm's response to arterial chemoreflex perturbations; reducing the recovering time, e.g., the latency to return to the baseline pattern. Together, these results suggest that the KFn regulates both intrinsic and sensory determinants of respiratory rhythm variability.

Dynamics of Cochlear Nonlinearity

Dynamic aspects of cochlear mechanical compression were studied by recording basilar membrane (BM) vibrations evoked by tone pairs ("beat stimuli") in the 11-19 kHz region of the gerbil cochlea. The frequencies of the stimulus components were varied to produce a range of "beat rates" at or near the characteristic frequency (CF) of the BM site under study, and the amplitudes of the components were balanced to produce near perfect periodic cancellations, visible as sharp notches in the envelope of the BM response. We found a compressive relation between instantaneous stimulus intensity and BM response magnitude that was strongest at low beat rates (e.g., 10-100 Hz). At higher beat rates, the amount of compression reduced progressively (i.e. the responses became linearized), and the rising and falling flanks of the response envelope showed increasing amounts of hysteresis; the rising flank becoming steeper than the falling flank. This hysteresis indicates that cochlear mechanical compression is not instantaneous, and is suggestive of a gain control mechanism having finite attack and release times. In gain control terms, the linearization that occurs at higher beat rates occurs because the instantaneous gain becomes smoothened, or low-pass filtered, with respect to the magnitude fluctuations in the stimulus. In terms of peripheral processing, the linearization corresponds to an enhanced coding, or decompression, of rapid amplitude modulations. These findings are relevant both to those who wish to understand the underlying mechanisms and those who need a realistic model of nonlinear processing by the auditory periphery.

Crowding, grouping, and gain control in schizophrenia

Visual paradigms are versatile tools to investigate the pathophysiology of schizophrenia. Contextual modulation refers to a class of paradigms where a target is flanked by neighbouring elements, which either deteriorate or facilitate target perception. It is often proposed that contextual modulation is weakened in schizophrenia compared to controls, with facilitating contexts being less facilitating and deteriorating contexts being less deteriorating. However, results are mixed. In addition, facilitating and deteriorating effects are usually determined in different paradigms, making comparisons difficult. Here, we used a crowding paradigm in which both facilitation and deterioration effects can be determined all together. We found a main effect of group, i.e., patients performed worse in all conditions compared to controls. However, when we discounted for this main effect, facilitation and deterioration were well comparable to controls. Our results indicate that contextual modulation can be intact in schizophrenia patients.

Conditioned frequency-dependent hearing sensitivity reduction in the bottlenose dolphin (Tursiops truncatus)

The frequency specificity of conditioned dampening of hearing, when a loud sound is preceded by a warning sound, was investigated in a bottlenose dolphin. The loud sounds were 5 s tones of 16, 22.5 or 32 kHz, sound pressure level of 165 dB root mean square (RMS) re. 1 µPa. Hearing sensitivity was tested at the same three frequencies. Hearing sensitivity was measured using pip-train test stimuli and auditory evoked potential recording. The test sound stimuli served also as warning sounds. The durations of the warning sounds were varied randomly to avoid locking a conditioning effect to the timing immediately before the loud sound. Hearing thresholds before the loud sound increased, relative to the baseline, at test frequencies equal to or higher than the loud sound frequency. The highest threshold increase appeared at test frequencies of 0.5 octaves above the loud sound frequencies.

The neuronal basis of on-line visual control in smooth pursuit eye movements

Smooth pursuit eye movements allow us to maintain the image of a moving target on the fovea. Smooth pursuit consists of separate phases such as initiation and steady-state. These two phases are supported by different visual-motor mechanisms in cortical areas including the middle temporal (MT), the medial superior temporal (MST) areas and the frontal eye field (FEF). Retinal motion signals are responsible for beginning the process of pursuit initiation, whereas extraretinal signals play a role in maintaining tracking speed. Smooth pursuit often requires on-line gain adjustments during tracking in response to a sudden change in target motion. For example, a brief sinusoidal perturbation of target motion induces a corresponding perturbation of eye motion. Interestingly, the perturbation ocular response is enhanced when baseline pursuit velocity is higher, even though the stimulus frequency and amplitude are constant. This on-line gain control mechanism is not simply due to visually driven activity of cortical neurons. Visual and pursuit signals are primarily processed in cortical MT/MST and the magnitude of perturbation responses could be regulated by the internal gain parameter in FEF. Furthermore, the magnitude and the gain slope of perturbation responses are altered by smooth pursuit adaptation using repeated trials of a step-ramp tracking with two different velocities (double-velocity paradigm). Therefore, smooth pursuit adaptation, which is attributed to the cerebellar plasticity mechanism, could affect the on-line gain control mechanism.

The tilt illusion: phenomenology and functional implications

The perceived orientation of a line or grating is affected by the orientation structure of the surrounding image: the tilt illusion. Here, I offer a selective review of the literature on the tilt illusion, focusing on functional aspects. The review explores the merits of mechanistic accounts of the tilt illusion based upon sensory gain control in which neuronal responses are normalized by the pooled activity of other units. The role of inhibition between orientation-selective neurons is discussed, and it is argued that their associated disinhibition must also be taken into account in order to model the full angular dependence of the tilt illusion on surround orientation. Parallels are drawn with adaptation as modulation by the temporal rather than spatial context within which an image fragment is processed. The chromatic selectivity of the tilt illusion and the extent of its dependence on the visibility of the surround are used to infer characteristics of the neuronal normalization pools and the loci in the cortical processing hierarchy at which gain control operates. Finally, recent evidence is discussed as to the possible clinical relevance of the tilt illusion as a biomarker for schizophrenia.

Motion-based prediction explains the role of tracking in motion extrapolation

During normal viewing, the continuous stream of visual input is regularly interrupted, for instance by blinks of the eye. Despite these frequents blanks (that is the transient absence of a raw sensory source), the visual system is most often able to maintain a continuous representation of motion. For instance, it maintains the movement of the eye such as to stabilize the image of an object. This ability suggests the existence of a generic neural mechanism of motion extrapolation to deal with fragmented inputs. In this paper, we have modeled how the visual system may extrapolate the trajectory of an object during a blank using motion-based prediction. This implies that using a prior on the coherency of motion, the system may integrate previous motion information even in the absence of a stimulus. In order to compare with experimental results, we simulated tracking velocity responses. We found that the response of the motion integration process to a blanked trajectory pauses at the onset of the blank, but that it quickly recovers the information on the trajectory after reappearance. This is compatible with behavioral and neural observations on motion extrapolation. To understand these mechanisms, we have recorded the response of the model to a noisy stimulus. Crucially, we found that motion-based prediction acted at the global level as a gain control mechanism and that we could switch from a smooth regime to a binary tracking behavior where the dot is tracked or lost. Our results imply that a local prior implementing motion-based prediction is sufficient to explain a large range of neural and behavioral results at a more global level. We show that the tracking behavior deteriorates for sensory noise levels higher than a certain value, where motion coherency and predictability fail to hold longer. In particular, we found that motion-based prediction leads to the emergence of a tracking behavior only when enough information from the trajectory has been accumulated. Then, during tracking, trajectory estimation is robust to blanks even in the presence of relatively high levels of noise. Moreover, we found that tracking is necessary for motion extrapolation, this calls for further experimental work exploring the role of noise in motion extrapolation.