THE LIGHT AND DARK OF VISUAL SIGNAL PROCESSING
Gloria Luo-Li
A thesis submitted in fulfilment of the requirements for the degree of Doctor of Philosophy
Faculty of Medicine and Health The University of Sydney 2020
ii Acknowledgements
Ten years ago, I was approached by a friend who advised me to extend my study in the field of medicine. I was excited about this idea and started exploring the journey.
In 2012, I met Dr Alan Freeman who subsequently supervised my part-time project.
Now, eight years later, my thesis is ready for submission! I knew that saying some general words of gratitude lack weight, however that’s the first and foremost way to express my appreciation. I am profoundly grateful to my supervisor Dr Alan Freeman for his unwavering support, insight and guidance and for his enthusiastic response to, and feedback on, every single step of the development of my project and extracurricular learning. Without the generous contributions of imparting his knowledge and confidence, this thesis would not be.
I sincerely thank my associate supervisor Professor David Alais for his constant assistance by providing his laboratory for some of the experiments. Professor Alais always gives advice and comments positively and confidently to dispel my doubts and encourage me to work things through free of stress. I thank him from the bottom of my heart for his valuable contribution to my first published paper, Chapter 5 in this thesis, and his admirable attitude towards our relationship during the past eight-year research journey.
To the American co-authors of my second published article – Distinguished Professors
Alonso and Zaidi, and Dr Mazade, I thank them for their generosity far above what I asked or expected, by allowing me to use the data from their animal studies. They also made many thoughtful contributions throughout the drafting of Chapter 6. I cannot find the right words to express my endless appreciation to Professor Alonso and his team members who sacrificed their valuable time to support me so well.
iii A special thank you goes to Dr Elaine Wong for her comments on this thesis, for her enthusiasm and generosity of sacrificing her own and family time working on my thesis, especially during the night-times after settling her newborn baby, for which I feel so thoroughly touched.
To all the administrative staff, undergraduate and postgraduate students who participated in the experiments, I cannot thank them enough for being a part of my research. Thank you to Professor Frank Lovicu, Dr Peter Knight, Associate Professor
Fazlul Huq, Dr Elizabeth Hegedus, Dr Jaimie Polson, Dr Darren Reed, Dr Helen
Ritchie, Dr David Mor, and Dr Aaron Camp, who gave encouraging and assertive comments on each of my annual progress review reports. Thank you to Ms Felicia Lim,
Dr Jin Huang, Dr Nastaran Hesam Shariati, Mr Gautham Jayachandran, Mrs Ann
Korabelnikoff, Mrs Helen Ananin, Ms Ruth Rinot, Dr Md Sheikh Anwar, Dr Zaynab
Al-Eisawi, and Dr Munira Al-Dossari, who gave me a helping hand whenever I needed it.
A special thank you to Dr Patrick Lam, General Practitioner, Senior Lecturer and researcher in the field of biomechanics and orthopaedics at UNSW. Dr Lam generously shared his incredible ideas from his reflection on his experience in research and teaching. I thank him for inspiring me, and the time he sacrificed for reading and commenting as I drafted my thesis.
Last, but not least, my heartfelt thanks to my family, especially my two teenage sons, for all your patience and loving encouragement throughout the years. To you, I dedicate this thesis.
iv Brief contents
Acknowledgements ...... iii
Brief contents ...... v
Detailed contents ...... vi
Summary ...... x
Chapter 1. Literature review ...... 1
Chapter 2. Aims ...... 29
Chapter 3. General methods ...... 33
Chapter 4. Pilot study ...... 45
Chapter 5. Responses to light and dark stationary stimuli ...... 59
Chapter 6. Responses to light and dark moving stimuli ...... 73
Chapter 7. Mechanisms underlying motion direction selectivity ...... 83
Chapter 8. Discussion ...... 111
References ...... 118
Publications ...... 128
v
Detailed contents
Acknowledgements ...... iii
Brief contents ...... v
Detailed contents ...... vi
Summary ...... x
Chapter 1. Literature review ...... 1
Visual pathways ...... 1
Overall structure of the visual pathways ...... 1
Retinal structure ...... 3
Major retinal signal processing pathways ...... 4
Centre-surround receptive fields ...... 5
Bipolar cell function ...... 8
Retinal ganglion cell function ...... 10
Lateral geniculate nucleus ...... 11
Cellular structure of the primary visual cortex ...... 12
Cortical receptive fields ...... 14
Conclusion ...... 15
Light/dark response asymmetries ...... 15
Motion sensitivity ...... 17
Neurophysiology of motion sensitivity ...... 17
Subcortical mechanisms ...... 18
Cortical mechanisms ...... 20
vi Mechanisms underlying direction selectivity ...... 26
Chapter 2. Aims ...... 29
Chapter 4 aims ...... 29
Chapter 5 aims ...... 29
Chapter 6 aims ...... 30
Chapter 7 aims ...... 30
Chapter 7 aims ...... 31
Chapter 7 aims ...... 30
Summary …………………………………………………………………………………………………………………….31
Chapter 3. General methods ...... 33
Psychophysics ...... 33
Subjects ...... 33
Equipment ...... 34
Visual stimulus ...... 34
Calibration ...... 37
Electrophysiology ...... 42
Animal surgery and preparation ...... 42
Electrophysiological recordings and data acquisition ...... 43
Visual stimulation and data analysis ...... 44
Chapter 4. Pilot study ...... 45
Abstract ...... 45
Introduction ...... 46
Experiment 1: Detection of light increments and decrements ...... 48
vii Methods ...... 48
Results ...... 49
Discussion ...... 50
Experiment 2: Detection of stimulus location ...... 51
Methods ...... 52
Results ...... 52
Discussion ...... 56
Psychometric function ...... 56
Reaction time ...... 56
Chapter 5. Responses to light and dark stationary stimuli ...... 59
Contributions of authors ...... 59
Chapter 6. Responses to light and dark moving stimuli ...... 73
Contributions of authors ...... 73
Chapter 7. Mechanisms underlying motion direction selectivity ...... 83
Abstract ...... 83
Introduction ...... 84
Methods ...... 85
Animal preparation ...... 85
Electrophysiology recordings and data collection ...... 85
Visual stimuli ...... 87
Results ...... 89
Response latency ………………………………………………………………………………………………………. 89
viii Sparse noise ...... 91
Response time course ...... 91
Receptive field map ...... 95
On-to-off direction ...... 98
Moving bar ...... 101
Comparing preferred and on-to-off direction ...... 107
Discussion ...... 108
A mechanism based on inhibition ...... 109
Chapter 8. Discussion ...... 111
Chapter 5 ...... 112
Chapter 6 ...... 112
Chapter 7 ...... 113
Biological advantage of parallel subcortical pathways ...... 113
Biological advantage of light/dark asymmetries ...... 114
Source of light/dark response asymmetries ...... 115
Methodology ...... 116
Contrast sensitivity……………………………………………………………………………………………116
References ...... 118
Publications ...... 128
Journal paper ...... 128
Abstracts and conference proceedings ...... 128
ix
Summary
Subcortical visual pathways are prominently divided into two types. On-centre neurons respond best to stimuli lighter than the background and off-centre neurons prefer darker stimuli. The convergence of these two pathways at the cortex results in a dramatic change in neuronal properties. Subcortical responses are basically photograph-like, but cortical responses are feature-based: cortical neurons encode features such as contour orientation, motion direction, and depth.
It has recently become clear that off-centre signals are not simply inversions of on- centre signals. Instead, signals in off-dominated cortical neurons are stronger, faster and more prevalent than their on-dominated counterparts. The aim of my thesis was to explore asymmetries of responses to light and dark, and to learn more about the influence of these asymmetries on feature processing. I used two methodologies, human psychophysics and the analysis of cortical neuronal data from a colleague’s laboratory. There are four results chapters.
Chapter 4 establishes and develops the psychophysical methods in two ways. First, it shows that the use of very low contrasts can capture response characteristics that are not appreciable at higher contrasts. Second, I developed an experimental design capable of simultaneously measuring response accuracy and reaction time. This dual approach adds weight to the experimental conclusions.
Chapter 5 psychophysically characterises the dominance of responses to stationary stimuli. I first show that the stimulus is invisible for contrast magnitudes less than about 0.01. This is assumed to result from the resting hyperpolarisation of simple cells.
I then demonstrate that responses to dark stimuli are more accurate and faster than
x are those to light stimuli. This result is modelled by assuming that off-dominant neurons in primary visual cortex have higher contrast sensitivity than do their on- dominant neighbours. Finally, I decomposed gratings into light and dark components and showed that the light component had to be delivered before the dark component to obtain optimal orientation discrimination. This result can be interpreted to mean that responses to dark and light need to arrive at the cortex at about the same time for optimal discrimination.
The focus in Chapter 6 is on moving rather than stationary stimuli. I delivered light and dark moving bars and asked my subjects to indicate motion direction.
Surprisingly, they were more accurate and faster for light than for dark bars. As a check on this result I analysed neuronal data from cat primary visual cortex; these data came from the laboratory of Jose Manuel Alonso. Neuronal responses to moving light bars had lower latency than responses to dark bars of the same contrast magnitude, provided that bar speed was low. Differences were not significant at high speed. The psychophysics and neuronal data are therefore in agreement. The speed advantage for light bars was interpreted in terms of differing contrast-response functions for on- and off-dominated cortical neurons.
Chapter 7, the final results chapter, analyses neuronal responses to explore motion direction selectivity. The hypothesis I tested, which derives from the results in Chapter
6, was that the preferred motion direction of a simple cell is from its on-subfield to its off-subfield. I analysed responses to flashed stationary spots of light and dark to obtain receptive field maps, and predicted preferred direction from the map. I then compared the prediction with measurements made with moving bars. The prediction matched the empirical result for one dataset, but the result was not statistically significant in a
xi second dataset. The evidence for the hypothesis is therefore weak. I discuss an alternative hypothesis at the end of the chapter.
xii
Chapter 1. Literature review
Visual pathways
Overall structure of the visual pathways
Visual input is transmitted from the retina to the lateral geniculate nucleus (or LGN) of the thalamus, and then to the primary visual cortex (V1) (Solomon & Lennie, 2007;
Figure 1.1). The retina is the innermost layer of the eye, which lies in front of the choroid and behind the vitreous body. The retina is described as being a multi-layered tissue containing a number of cell types. While the neural circuitry within the retina is rather complex, the flow of visual information from photoreceptors to the optic nerve is mostly through a series of three neurons: photoreceptor cell to bipolar cell to ganglion cell. Light entering the eye triggers the process of phototransduction within the photoreceptor layer of the retina, which is comprised of rods and cones. The output from these photoreceptor cells travel through the bipolar cells (as well as horizontal cells) until it reaches the retinal ganglion cells.
Retinal ganglion cell axons from all areas of the retina converge at the optic disc and exit the eye along the optic nerve. The retinal ganglion cells are generally divided into two groups, each projecting to one of two portions of the LGN. The majority of retinal ganglion cells project to dorsal LGN, providing information for visual perception. The second projection is to a small group of cells, the pretectum, lying adjacent to the LGN.
These cells provide the basis for other visual functions such as the pupillary light reflex. The initial component of the pupillary light reflex pathway is a bilateral projection from the retina to the pretectum, which in turn, projects to the Edinger-
Westphal nucleus that lies next to the nucleus of the oculomotor nerve in the midbrain.
1 Chapter 1. Literature review
The primate LGN is a six-layered nucleus of the thalamus. Interactions or synapses between the retinal ganglion cells and their corresponding cell layer within the LGN implements the flow of information between the retina and the LGN. This will be detailed in a later section, but in short, there are two paths (or channels) whereby LGN outputs can travel to V1. These channels are known as the parvocellular (P) and magnocellular (M) pathways. Each pathway carries specific qualities of visual information to different areas of V1 (Solomon & Lennie, 2007).
Figure 1.1. Primate visual pathway. The left panel outlines the visual pathway from retina to the primary visual cortex through the lateral geniculate nucleus (LGN). The red line represents the pathway commencing from temporal retina; the green line represents the projections from nasal retina.
The top right figure shows the basic structure of the retina. The axons of ganglion cells form the optic nerve, which carries visual signals from the retina. The lower right figure shows that axons of LGN
2 Visual pathways neurons project predominantly to layer 4 of the primary visual cortex (Figure from Box 1 in Solomon &
Lennie (2007).
Retinal structure
The retina measures 0.2 mm in thickness on average. There are five retinal neurons: photoreceptors, bipolar cells, horizontal cells, amacrine cells and retinal ganglion cells.
The cell bodies and their axons are layered in an alternating fashion, giving it a stratified appearance (Figure 1.2). The cell bodies of the retinal neurons are found in the inner and outer nuclear layers, as well as the ganglion cell layer, while the axons of the neurons and synapses between neurons occupy the inner and outer plexiform layers (Figure 1.2).
Figure 1.2. The retina is made of many types of neurons, which are arranged in layers. This figure is an adaptation of Netter’s drawings published in Felten et al. (2015).
3 Chapter 1. Literature review
The most distal layer of the retina is the pigment epithelium followed by the photoreceptor layer. Both types of photoreceptors consist of an outer and inner segment. The outer segment of the photoreceptor is partially embedded within the pigment epithelium. The pigment epithelium is an important retinal layer that maintains the photoreceptor neurons, and sheds the used portions of their outer segments, which are made of discs containing photosensitive pigments. The inner segment of the photoreceptors consists of the cell body and its synaptic terminal.
Within the outer plexiform layer of the retina, photoreceptor axons synapse with the bipolar cells and horizontal cells. The cell bodies of the bipolar, horizontal and amacrine cells are located in the inner nuclear layer. The processes of the bipolar and amacrine cells synapse with each other and with the retinal ganglion cell dendrites.
The cell bodies of the retinal ganglion cells make up a layer called the ganglion cell layer, and finally, the most proximal layer is the nerve fibre layer consisting of the axons emerging from the retinal ganglion cell soma. These retinal ganglion cell nerve fibres bundle up to form the optic nerve, and the visual information formed within the retina then travels to the remaining sections of the visual pathway (Purves et al., 2017).
Major retinal signal processing pathways
The processing of visual information signals begins within the outer segment of the photoreceptor cells. The light-sensitive pigments contained within the photoreceptor’s membranous discs absorb photons of light. The absorption of light triggers a cascade of chemical events (a process known as phototransduction), which converts electromagnetic stimuli into electrical signals. Electrical signals generated in photoreceptors are passed to bipolar cells located in the outer plexiform layer. The horizontal cells, which are also present in the outer plexiform layer, play a role in modulating signal transmission between photoreceptors and bipolar cells. In
4 Visual pathways particular, horizontal cells allow for lateral interactions between photoreceptors and bipolar cells, so as to increase the spatial contrast of an image and facilitate edge detection. Bipolar cell signals are transmitted to the retinal ganglion cells in the inner plexiform layer. Amacrine cells found in that layer modulate the signal transmission between bipolar and retinal ganglion cells. There are various subtypes of amacrine cells, with various image processing functions such as the signalling of contrast, colour, brightness and movement.
Centre-surround receptive fields
Bipolar and retinal ganglion cells have a rather distinctive receptive field arrangement
(Figure 1.3). A small circular patch of photoreceptors supplies a bipolar cell and a corresponding retinal ganglion cell. Hence the receptive fields of bipolar and retinal ganglion cells are roughly circular, similar to neurons of other sensory systems, but unique in their own way because the receptive fields are concentric, consisting of a central region which is surrounded by a ring (DeAngelis, Anzai, Ohzawa, & Freeman,
1995).
There are two types of bipolar cells, each responding differently to light applied on the centres of their receptive fields. They are called on-centre and off-centre bipolar cells.
An on-centre bipolar cell is one that becomes depolarised (or excited) in the presence of a light stimulus applied to the centre of its receptive field. However, when light is delivered to the surround of an on-centre bipolar cell, the cell becomes hyperpolarised
(or inhibited). The second type of bipolar cell is the off-centre cell, which behaves in the exact opposite way: light on the receptive field centre hyperpolarises (or inhibits) the bipolar cell, while light on the surround depolarises (or excites) the cell.
5 Chapter 1. Literature review
The retinal ganglion cells have two types of receptive fields, similar to bipolar cells.
The first type of receptive field has an on-centre with an off-surround. The second type consists of an off-centre/on-surround. Like bipolar cells, the receptive fields of retinal ganglion cells display centre-surround antagonism. There is a difference, however, between bipolar and retina ganglion cells. Photoreceptors, bipolar cells, horizontal cells, and most amacrine cells have graded postsynaptic signals which can be either depolarising or hyperpolarising; none of these retinal neurons generate action potentials. Retinal ganglion cells are the first cells along the visual pathway that generate action potentials. Therefore, on-centre and off-centre ganglion cells respond by depolarising or hyperpolarising and the magnitude of these changes modulates their firing rate. Rather, retinal ganglion cells respond by increasing or decreasing the frequency with which they discharge action potentials.
6 Visual pathways
Figure 1.3. Contour map of the receptive field of geniculate, simple and complex cell. Note that bipolar cells and retinal ganglion cells have similar circular receptive fields to those of LGN cells. Green and red represent excitatory and inhibitory subfields, respectively (DeAngelis, Ohzawa, & Freeman, 1995).
To illustrate how the concentric receptive fields responds to light, consider an on- centre retinal ganglion cell, which has an on-centre and off-surround (as with the LGN cell in Figure 1.3). When a small spot of light is directed in the central region of the retinal ganglion cell’s receptive fields, the firing rate of the retinal ganglion cell increases from resting levels (when no light is falling on the receptive fields).
Conversely, when the spot of light is presented within the region of the surround, the activity recorded from the on-centre retinal ganglion cell decreases. When the entire central region of the receptive field is illuminated, there is an optimum response from the retinal ganglion cell. The response polarity of bipolar cells depends on the type of
7 Chapter 1. Literature review glutamate receptor expressed (AMPA/KA or mGluR6) at the photoreceptor-bipolar cell synapse. The neurotransmitter is glutamate for both on and off-centre bipolar cells.
If the entire surround is illuminated, there is an opposite effect on the retinal ganglion cell, that is, the retinal ganglion cell is maximally inhibited. When both centre and surround are illuminated, the retinal ganglion cell’s activity rises just above resting levels. This example demonstrates that uniform illumination across the entire receptive fields is not as efficient in increasing the activity of a retinal ganglion cell compared to well-defined spot of light, or a bar of light that passes through the centre of the on-centre retinal ganglion cell’s receptive field. This property accounts for why retinal ganglion cells are sensitive to differences in illumination levels across the receptive fields (luminance contrast).
Bipolar cell function
The responses of on-centre and off-centre bipolar cells vary according to the changes in light intensity. Photoreceptors hyperpolarise in the presence of light; the degree of hyperpolarisation is dependent on light intensity. Hyperpolarisation of photoreceptors decreases the release of neurotransmitter, which may cause depolarisation, or hyperpolarisation of the bipolar cell, depending upon the type of neurotransmitter released at the synapse between photoreceptor and bipolar cell
(Purves et al., 2017). When light is presented on the centre of an on-centre bipolar cell, the photoreceptor becomes hyperpolarised, causing depolarisation of the bipolar cell because of the sign-inverting synapse between them. Excitatory signals from the on- centre bipolar cell causes on-centre ganglion cells to increase their action potential discharge rate (Figure 1.4).
8 Visual pathways
Consider the effect of light shining within the centre of the receptive field of an off- centre bipolar neuron. Again, the first change is hyperpolarisation of the photoreceptor cell, with a subsequent hyperpolarisation of the off-centre bipolar cell due to the excitatory nature of the synapse between the photoreceptor and the off- centre bipolar cell. There is subsequent hyperpolarisation of the off-centre ganglion cell, which decreases its action potential frequency (Figure 1.4).
Figure 1.4. Light stimuli depolarise the on-centre bipolar cell and at the same time hyperpolarise the off-centre bipolar cell, subsequently resulting in vigorous firing of action potentials from the on-centre ganglion cell, and the decline in the action potential firing rate from the off-centre ganglion cell, respectively (Kandel, Schwartz, & Jessell, 2000).
The centre-surround structure of the bipolar cell’s receptive field is transmitted to the ganglion cell via synapses located in the inner plexiform layer, as shown in Figure 1.4.
Some synapses connect on-centre bipolar cells to on-centre ganglion cells, while others connect off-centre bipolar cells to off-centre ganglion cells. The accentuation of
9 Chapter 1. Literature review contrasts by the centre-surround receptive fields of the bipolar cells is thereby preserved and passed on to the ganglion cells, and ultimately to the visual cortex
(Purves et al., 2017). Vision depends on our ability to distinguish contrasts between objects and the backgrounds behind them. The initiation of parallel pathways for visual processing beginning in the retina is one of the mechanisms that make this discrimination possible.
Retinal ganglion cell function
The retinal ganglion cells are important for shape recognition and detecting the movement of objects. In primates, these stimulus properties are processed by two major types of retinal ganglion cells: parvocellular (P) or midget cells, and magnocellular (M) or parasol cells.
Midget ganglion cells, as their name suggests, are small cells that vastly outnumber their parasol ganglion cells counterparts in the retina by about eight times. The axons of midget ganglion cells travel through the optic nerve and tract and are destined for the parvocellular layers of the LGN (Dacey & Petersen, 1992). Midget cells have small dendritic fields and cell bodies. Within the retina, they receive inputs from one to a few cone bipolar cells, which in turn are connected to single cones. These physical properties are the reasons why midget cells are sensitive to colour and have small centre-surround receptive fields. Functionally, the midget cell produces sustained responses to stimuli that are centred in its receptive field and weaker responses when there is stimulus movement. Midget cells are best suited for responding to changes in colour (Dacey & Petersen, 1992).
Compared to midget cells, parasol ganglion cells have larger dendritic trees and cell bodies. They make up about 10% of the retinal ganglion cell population, and travel
10 Visual pathways within the optic nerve and tract to the LGN where they terminate in the magnocellular layers of the LGN. Within the retina, parasol cells receive synaptic inputs from a relatively large numbers of bipolar cells, which are in turn connected to numerous rods and cones. Hence, parasol cells do not possess colour-sensitive properties. They do, however, have large concentric receptive fields and conduct neural signals faster. This enables parasol cells to convey motion signals and to be more sensitive to low-contrast stimuli.
Lateral geniculate nucleus
The LGN is a nucleus located within the ventral aspect of thalamus that receives the majority of optic tract fibres. Information that the LGN received from the retinal ganglion cells is then passed on to the primary visual cortex, from which the LGN also receives significant feedback. The LGN is a laminated structure with six principal layers of cells. The largest LGN cells make up the innermost two layers (i.e. layers 1 and 2, known as the magnocellular layers), while smaller cells form the outermost four layers (i.e. layers 3, 4, 5, and 6, known as the parvocellular layers) (Purves et al., 2017).
There are also thin layers of the smallest cells (called koniocellular neurons) that are interposed between the six principal layers. The optic tract fibres from each eye synapse in different layers of the LGN. As a result, each LGN neuron responds to stimulation of one eye only (Hubel & Wiesel, 1961; Reid & Alonso, 1995), as shown in
Figure 1.5.
The functional properties of LGN neurons are similar to those of retinal ganglion cells by way of association. Midget retinal ganglion cells project to the parvocellular layers of the LGN and synapse with the cells located there (pLGN cells). Since pLGN cells process inputs from midget retinal ganglion cells, their properties are similar: pLGN
11 Chapter 1. Literature review cells are colour-sensitive and have small centre-surround receptive fields, hence they are best suited for detecting fine detail (Martinez et al., 2005). Parasol cells synapse with the neurons of the magnocellular layers (mLGN cells). Like parasol cells, mLGN cells have relatively large receptive fields and are colour-insensitive. They respond well to movement of visual stimuli (Purves et al., 2017). The third group of LGN neurons are the koniocellular cells (kLGN cells). Koniocellular cells receive inputs from midget retinal ganglion cells of the retina. kLGN cells show sensitivity particularly to information derived from short-wavelength-sensitive cones (Martin,
White, Goodchild, Wilder, & Sefton, 1997). The axons of these different types of LGN neurons terminate in different layers or sublayers of the primary visual cortex (V1).
Cellular structure of the primary visual cortex
Like the other parts of the neocortex, the visual cortex is a horizontally stratified structure consisting of six layers, labelled layers I to VI (Figure 1.1). The type of neurons distinguishes each layer from the next, and each layer is specialised in either receiving or sending neural information. Layer IV, for example, contains numerous stellate cells, described as being small-sized neurons with dendrites that radiate out around the cell body, and which specialises largely in receiving connections from the
LGN. In the primary visual cortex, layer IV is divided into three sublayers designated
IVA, IVB, and IVC. Layer IVC is further subdivided into IVCα and IVCβ. Output axons from the LGN transmit information from the eye along various pathways that project mainly into layer IVC.
12 Visual pathways
Figure 1.5. The magnocellular and parvocellular pathways in LGN. Midget ganglion cells synapse with small parvocellular LGN cells, while parasol ganglion cells synapse with large magnocellular cells of the LGN. These form the parvocellular (P) and magnocellular (M) pathways, respectively. They play an important role in the visual perception of colour, motion, and other fine details of stimuli (Kandel et al., 2000).
In addition to the horizontal stratification of the visual cortex, the cortex is also divided into vertical columns in which all the neurons respond to the same characteristic (for
13 Chapter 1. Literature review example, colour, contrast, ocular dominance, movement and orientation) of a given portion in the visual field. The columns thus form functional units that run perpendicular to the surface of the cortex.
Cortical receptive fields
Unlike the neurons of the retinal ganglion layer or the LGN, the receptive fields of the neurons of the primary visual cortex are not circular, but rather rectangular in shape
(Jones & Palmer, 1987; Martinez et al., 2005) (Figure 1.3). These rectangular receptive fields have on-subfields that respond actively to light, flanked by off-subfields that respond to darkness (Hubel & Wiesel, 1959).
A beam of light, which is not oriented precisely parallel to the boundary between on- and off-subfields, is simply not effective for the neuron. Thus, these types of neurons respond particularly well to stimuli that are oriented in a specific direction, and they are called simple cells (Hubel & Wiesel, 1962). The simple cell receptive fields are thought to be the result of the convergence of several adjacent circular receptive fields of cells in LGN. Since most output neurons from the LGN project to layer IV of the primary visual cortex, the cells in layer IV are mainly simple cells (Tanaka, 1983). Other cells in the primary visual cortex outside of layer IV have complex receptive fields: they are orientation selective but relatively insensitive to stimulus location (Hubel &
Wiesel, 1962).
Inputs from many simple cells converge on a complex cell (Hubel & Wiesel, 1961;
Hubel & Wiesel, 1965). These simple cells convey information about the same orientation, from overlapping receptive fields spread across the whole receptive field of the complex cell. Therefore, complex cells have large receptive fields, which are lacking in clear subfields of excitatory and inhibitory regions (Martinez et al., 2005)
14 Visual pathways
(Figure 1.3). Simple and complex cells are direction selective. They are excited by a moving edge that displays a specific orientation and moving in a specific direction of motion.
Conclusion
I conclude that visual information is processed in various stages from the retinal ganglion cells, to LGN cells, and to the simple and complex cells of the visual cortex.
At every stage, each cell is able to process more complex detail than the level below it, such that the cells at the higher stages are most capable in breaking down a visual image into features, which are the building blocks of object recognition.
Light/dark response asymmetries
Early studies of the subcortical visual system assumed that the responses of on- and off-centre neurons were of equal magnitude and opposite in sign. Some of the first indications of on/off asymmetry came from psychophysics. Krauskopf (1980) delivered both incremental and decremental light pulses to human subjects and found that their sensitivity was higher in the latter case. Bowen, Pokorny, and Smith (1989) stimulated their subjects with mirror-image sawtooth temporal waveforms and found higher contrast sensitivity for rapid-off than for rapid-on waveforms. Visual evoked potentials supported these findings: Zemon, Gordon, and Welch (1988) presented both positive- and negative-contrast checkerboard stimuli to human subjects and found higher contrast gain for negative contrasts. These results all indicate stronger visual responses to stimuli darker than the background compared with responses to light stimuli.
15 Chapter 1. Literature review
Cortical electrophysiology provided further evidence for this asymmetry. Jin et al.
(2008) found that on- and off-centre geniculate afferents tend to segregate in cat primary visual cortex and that the cortical region representing the central area of retina is dominated by off-centre inputs (Jin et al., 2008). Yeh, Xing, and Shapley (2009) recorded single-unit activity in monkey primary visual cortex and found that dark- dominated neurons substantially outnumbered light-dominated neurons in layer 2 and 3 but the numbers were similar in the input layer 4C: see Figure 1.6. They therefore concluded that dark dominance arises cortically. Xing, Yeh, and Shapley (2010) refined this result by finding a small preference for dark-dominant cells in layer 4Cb, the layer that receives parvocellular input.
Figure 1.6. The frequency distributions of on/off ratio showed interlaminar differences. The figure shows the percentages of off-dominant neurons and light-dominant neurons in different layers of V1.
A large number of neurons in layer 2/3 showed stronger responses to dark than light, while in layer 4c and 4a/b the difference was of weaker significance (Yeh et al., 2009).
14 Light/dark response asymmetries
Responses to darks have an advantage over those to lights not only in response magnitude but also in timing. Jin, Wang, Lashgari, Swadlow, and Alonso (2011) found that inputs to cat primary visual cortex from off-centre geniculate neurons arrive 3-6 ms before on-centre inputs (Figure 1.7). Further, off-dominated cortical cells responded about 3 ms before their on-dominated neighbours (Komban et al., 2014).
The speed advantage for dark stimuli has been confirmed psychophysically. Komban,
Alonso, and Zaidi (2011) asked subjects to indicate the number of dark or light squares presented against a patterned background. Reaction time was lower for dark stimuli.
Komban et al. (2014) presented dark and light stimuli with a variable delay between them. Subjects responded faster, by 6-14 ms, when dark stimuli were presented first.
Figure 1.7. Distributions of response latency and peak time in cat LGN. The upper graph shows off- centre geniculate cells have shorter response latency (~4ms) than on-cells. The response peak in off-
15 Chapter 1. Literature review centre cells is ~3 ms faster than for on-cells, as shown in the lower graph (Jin, Wang, Lashgari, et al.,
2011).
Another type of on/off asymmetry has recently been reported, this time in the contrast-response function. Kremkow et al. (2014) presented spots of light or dark against a grey background and recorded from both LGN and cortical neurons in cats.
Dark spots produced a response that grew almost linearly with spot luminance but for light spots, the contrast-response function was compressive, saturating at higher luminances. It should be noted, however, that both luminance and response axes were normalised, making it difficult to compare sensitivities for small luminance changes.
It has recently been shown that on/off asymmetries play a role in the spatial organisation of primary visual cortex. Using long horizontal penetrations of cortex,
Kremkow, Jin, Wang, and Alonso (2016) showed that off-dominance did not vary across ocular dominance columns, and that off- and on-dominance were negatively corelated when ocular dominance was constant. They concluded that the axis along which on/off dominance varies is spatially orthogonal to the axis with varying ocular dominance. They also found that simple cell off-subfields varied smoothly in visual field location across any penetration. The on-subfield, however, tended to vary in position around the off-subfield. Lee, Huang, and Fitzpatrick (2016), who recorded from tree shrew cortex using calcium imaging, corroborated this last result.
The origin of on/off asymmetries is controversial. It has been suggested that the asymmetry originates at the synapse between photoreceptors and bipolar cells (Jin,
Wang, Swadlow, & Alonso, 2011). The synapse with on- and off-bipolars is metabotropic and ionotropic, respectively, and the former is known to be slower. On- centre retinal ganglion cells in the primate, however, have been shown to respond faster than their off-centre neighbours (Chichilnisky & Kalmar, 2002), casting doubt
16 Light/dark response asymmetries on a retinal origin. A different retinal source for the asymmetry is ganglion cell density: off-cells are slightly more numerous than on-cells (Wässle, Boycott, & Illing, 1981), giving them a potential advantage. The finding by Yeh et al. (2009) of dark-dominance in monkey cortical layers 2 and 3, but not in the input layer, is suggestive of a cortical origin for the asymmetry. But this cannot be true in the cat because geniculocortical inputs are dark-dominated in that species (Jin et al., 2008). Given these mixed results, the source of on/off asymmetries remains to be determined.
Motion sensitivity
Neurophysiology of motion sensitivity
All visual neurons respond to a changing pattern of light on the retina, including that produced by a moving stimulus. Some neurons have a property, motion direction selectivity, that allows them to distinguish motion from dynamic stationary stimuli.
This selectivity means that they respond best to motion in a specific direction and less to other directions. Direction selectivity was first described in the cat’s primary visual cortex (Hubel & Wiesel, 1959) as illustrated in Figure 1.8. Since then direction selectivity has been found in a number of other species, including rabbit (Barlow, Hill,
& Levick, 1964), monkey (Hubel & Wiesel, 1968), mouse (Yoshida et al., 2001) and fly
(Single & Borst, 1998). The origin of direction selectivity, however, differs between species. Direction selective neurons are found in the retina of flies, rabbits and mice.
By contrast, the direction selectivity seen in cat and monkey cortex originates cortically. The following describes mechanisms underlying direction selectivity, first subcortically and then in cortex.
17 Chapter 1. Literature review
Subcortical mechanisms
Rabbit and mouse
Barlow et al. (1964) were the first to describe direction-selective retinal ganglion cells in the rabbit. Barlow and Levick (1965) performed a variety of experiments with the aim of finding the mechanisms underlying the direction selectivity. In their key experiment, cells were stimulated with two bars of light at different receptive field locations. The response when the bars were presented sequentially in the antipreferred direction (that is, the opposite to the preferred direction) was less than the sum of the responses when the bars were presented singly. Barlow et al. (1964) therefore reasoned that the antipreferred response was less than the preferred because of inhibition. Their model included inhibition in the antipreferred direction, with a delay to increase its effectiveness.
18 Motion sensitivity
Figure 1.8. Responses from a neuron in cat primary visual cortex. A bar of light was moved in a variety of directions across the receptive field (left column). The neuron responds best to a specific orientation and motion direction (right column) (Hubel & Wiesel, 1959).
The subsequent search for the origin of retinal direction selectivity focussed on the starburst amacrine cell. This cell has a radially symmetric dendritic tree that receives bipolar cell input over the whole tree (Vaney & Taylor, 2002). Only the distal parts of the tree contact ganglion cells. Immuno-ablation of mouse starburst amacrines abolishes direction selectivity in ganglion cells, suggesting that the amacrines are responsible for the selectivity (Yoshida et al., 2001). Borg-Graham and Grzywacz
(1992) proposed that distal dendrites respond better to motion away from the cell body than towards it. Selective connections of the amacrine to a ganglion cell could then generate direction selectivity.
More recent work has provided support for this idea. Euler, Detwiler, and Denk (2002) used two-photon calcium imaging to show that the release site on an amacrine cell dendrite prefers centrifugal motion over centripetal motion. Briggman, Helmstaedter, and Denk (2011) showed that ganglion cells receive inhibition primarily from amacrine cell dendrites whose centrifugal direction matches the antipreferred direction of the ganglion cell. How does direction selectivity arise in an amacrine cell dendrite? One suggestion is that dendritic cable properties augment summation in the centrifugal direction (Vlasits et al., 2016) but there are other possibilities (Mauss, Vlasits, Borst, &
Feller, 2017). Another model for amacrine cell direction selectivity relies on the spatial and temporal offsets of on- and off- bipolar cell inputs (Mauss et al., 2017). These alternatives continue to be tested. The end result is that direction-selective ganglion cells respond to both onset and offset of light in addition to their direction selectivity.
19 Chapter 1. Literature review
Fly
As in rabbits and mice, the anatomical origin of motion direction selectivity in flies has been found. The fly optic lobe has five layers: photoreceptors, lamina, medulla, lobula and lobula plate. Neurons can be divided into on- and off-varieties, and direction selectivity arises independently for each contrast polarity. For on-channels, medulla neurons Mi1 and Tm3, which are not direction selective (Behnia, Clark, Carter,
Clandinin, & Desplan, 2014), synapse onto T4 neurons in the lobula plate, which have been shown to be direction selective (Joesch, Schnell, Raghu, Reiff, & Borst, 2010).
Selectivity arises in T4’s dendrites but the mechanism is uncertain (Mauss et al., 2017).
A similar circuit exists for the off-neuron T5. T4 and T5, which have relatively small receptive fields then converge onto the wide-field tangential cell.
Cortical mechanisms
There are subcortical direction-selective neurons in the cat (Cleland & Levick, 1974) but they project almost exclusively to the brainstem (Fukuda & Stone, 1974). Instead, cortical direction selectivity in carnivores and primates arises in visual cortex itself.
This direction selectivity was first described more than fifty years ago (Hubel &
Wiesel, 1962), but the mechanisms that underlie it are still unknown. In what follows
I will describe some of the characteristics of cortical direction selectivity and some of the hypotheses about its mechanism.
The direction selectivity of a cortical neuron can be well described with a plot of response magnitude versus stimulus direction. Two types of plots are commonly used. A Cartesian plot, illustrated in Figure 7.11 shows direction and response on the horizontal and vertical axes, respectively. This plot has two peaks, with the higher
20 Motion sensitivity peak at the preferred direction. Polar plots represent stimulus direction by the angle from the horizontal axis and response by radial distance.
A convenient metric, the direction selectivity index, can be calculated from plots such as these. Two forms have typically been used for the index
��� = (1.1)
(Peterson, Li, & Freeman, 2004) and
��� = (1.2)
(DeAngelis, Ohzawa, & Freeman, 1993a), where � and � are the responses in the preferred and antipreferred (that is, opposite) directions, respectively. Form 2 can be used where the response refers to the increment above resting level and has the advantage that it lies between 0 and 1, values which indicate absent and maximum direction selectivity, respectively. Figure 1.9A (Gizzi, Katz, Schumer, and Movshon
(1990) shows the frequency histogram of the direction selectivity index for 116 simple cells in cat. The histograms are biased towards the right side indicating that direction- selective neurons are common. Figure 1.10 is the measurement of the distribution of direction selectivity index for simple cells by Peterson et al. (2004). Direction-selective cells are not only prevalent but also largely indifferent to contrast sign. Yamane,
Maske, and Bishop (1985) measured direction selectivities for moving light and dark bars and found that 94% of sampled cells had the same preferred direction for the two signs. Another group (Casanova, Nordmann, Ohzawa, & Freeman, 1992) found a similar figure, 90%.
A simple way to implement this model is to assume that a cortical neuron linearly sums its subcortical inputs and that there is a temporal delay between the two sides of
21 Chapter 1. Literature review the cortical receptive field. Reid, Soodak, and Shapley (1987) tested this idea on cat simple cells. They placed stationary contrast-reversing gratings at a range of locations across the receptive field and recorded the amplitude and temporal phase of the resulting responses. Direction-selective cells showed a progressive delay across their receptive field whereas unselective cells did not. The authors then attempted to predict responses to moving gratings by adding suitably delayed responses to stationary gratings. The prediction was only partially successful in that it accounted for only about half of the measured direction selectivity. Reid et al. concluded that linear mechanisms can account for only part of the observed direction selectivity. Further, while they found a progressive delay of responses across the receptive field, they did not provide a mechanism for its source.
22 Motion sensitivity
Figure 1.9. Directionality index of striate neurons for optimally oriented gratings. The directionality index of cat’s striate neurons is given by 1 − � ⁄� where the response in the preferred direction is represented by � , and � is the response in the nonpreferred direction. A is the distribution for 116 simple cells, B is the distribution for 129 complex cells, whereas C is the distribution for 107 cells from the lateral suprasylvian cortex (LS) cells (Gizzi et al., 1990).
Reid et al. (1987) then used the measured responses to stationary stimuli to predict response amplitudes for stimulus moving in the preferred and antipreferred directions. They found that nonlinearities facilitated the preferred direction response by 11% but suppressed the antipreferred response by a much larger margin, 42%. They concluded that nonlinear mechanisms have a much greater suppressive than faciliatory effects. This conclusion agreed with previous work (Goodwin, Henry, &
Bishop, 1975).
Figure 1.10. Measurements of the distribution of direction selectivity index for cat simple cell responses to drifting gratings. The open triangles and filled triangles represented the distribution of direction selectivity index predicted from the linear and nonlinear receptive field respectively. The vertical dashed line indicates the criterion for categorising neurons as direction-selective (Peterson et al., 2004).
While Reid et al. (1987) showed a temporal delay that varies across the receptive field, two groups (Adelson & Bergen, 1984; Watson & Ahumada, 1985) provided a new way of visualising it. Instead of showing the receptive field as a map in two spatial dimensions, they used a space-time map: the spatial dimension gave distance
23 Chapter 1. Literature review perpendicular to the preferred orientation and the temporal axis indicated time after the start of a stimulus. In a neuron that is not direction-selective, this map can be described by the product of spatial and temporal functions. Direction-selective neurons differ in that the plot is slanted, where the slope gives preferred velocity.
McLean and Palmer (1989) used this technique to predict both the preferred direction and optional velocity in their sample of cat cortical simple cells. DeAngelis, Ohzawa, and Freeman (1993b), who also recorded from simple cells, used the spatiotemporal receptive field to demonstrate a diversity of direction selectivities (Figure 1.11). They measured the direction selectivity index conventionally, using drifting gratings, and from the (slanted) spatiotemporal receptive field. The preferred direction matched in
95% of cases, but the index predicted from the receptive field was only about half of that obtained with gratings. DeAngelis et. al. showed that this discrepancy was due to an output nonlinearity, such as the action potential threshold, in the responses to gratings, whereas the index calculated from the receptive field assumed linearity.
24 Motion sensitivity
Figure 1.11. Simple cell spatiotemporal receptive-field profiles (contour maps). Data for three simple cells (A-C), cell A is from adult cat; cell B is from 8-week-old kitten, cell C is from 4-week-old kitten
(DeAngelis et al., 1993b).
Jagadeesh, Wheat, Kontsevich, Tyler, and Ferster (1997), who recorded intracellularly from simple cells in cat cortex, confirmed the findings of Reid et al. (1987). They showed two ways in which nonlinearities play a role in direction selectivity. First, membrane potential in response to a stationary contrast-reversing does not necessarily have a sinusoidal time course like that of the stimulus: the neuronal time course can be more like a sawtooth. Second, the direction selectivity index measured from
25 Chapter 1. Literature review membrane potential is about half that obtained with action potential rates. The shortfall in the index predicted from assumptions of linearity are therefore due to thresholding of membrane potential.
Ferster’s laboratory (Priebe & Ferster, 2005; Priebe, Lampl, & Ferster, 2010) also recorded intracellularly from cat cortical simple cells stimulated with moving gratings.
Priebe and Ferster (2005) measured excitatory and inhibitory synaptic conductance by injecting currents of varying strength through their recording electrode. They found that both conductances varied cyclically with the stimulus and that the amplitude of the variation was greater in the preferred direction than the antipreferred. Inhibitory conductance, however, was out of phase with excitatory conductance. They were able to predict the neuron’s spike response to movement in both directions by assuming that excitatory and inhibitory synaptic potentials sum linearly and are then thresholded to obtain impulse rate.
Mechanisms underlying direction selectivity
Neural models of direction selectivity assume that the visual image is sampled at two or more locations in visual space, and that the signal at one location is delayed relative to other location. These signals are then combined in visual cortex to yield a selective response. Models differ in two ways: 1. The source of the delay; 2. Attribution of the selectivity to an enhancement of the response in the preferred direction, response suppression in other directions, or a combination of these effects. I will now describe some of the published models.
Mastronarde (1987) described two types of relay cell, lagged and nonlagged, in the lateral geniculate nucleus. Responses to step-onset stimuli were delayed by 40-80 ms in lagged relative to nonlagged cells. Humphrey and Saul (2002) suggested that these
26 Motion sensitivity cells, which project to visual cortex, could together contribute to direction selectivity.
This idea, however, is hard to reconcile with the finding that the cortex produces robust responses to stimuli which change every 40 ms or less (DeAngelis et al., 1993b).
Another model features the Meynert cell, which has a dendritic field asymmetrically located relative to the cell body. Livingstone (1998) suggested that the delay in signals travelling through the dendritic tree to the soma could result in direction selectivity.
The main problem with this proposal is that Meynert cells reside in cortical layer 5 and
6 whereas direction-selective cells are prevalent in layer 4.
More recent work has found a new candidate for the delay underlying direction selectivity. On-centre inputs to cortex are delayed relative to off-centre inputs
(Komban et al., 2014). Further, on- and off-centre inputs to a cortical neuron are spatially offset (Reid & Alonso, 1995; Tanaka, 1983) providing an extra condition for direction selectivity. Hesam Shariati and Freeman (2012) built a computational model on these findings, using a latency difference between on- and off-inputs of about 4 ms.
The model showed robust direction selectivity and replicated a variety of empirical results on motion sensitivity. The preferred direction, however, was always from the off-subfield to the on-subfield, a consistency that is not found in the laboratory. In short, there is currently no consensus on the underlying mechanism for direction selectivity.
27
28
Chapter 2. Aims
The studies reviewed in Chapter 1 show clear asymmetries between responses to visual stimuli lighter and darker than the background. They also leave a number of questions unanswered; I have attempted to answer several of these questions in this thesis. I now describe the questions, indicate how I have tackled them, and state the aims of each of the results chapters that follow.
Chapter 4 aims
Light/dark asymmetries have been studied psychophysically, but the methods used
(Bowen et al., 1989; Komban et al., 2011; Zemon et al., 1988) have been much less direct than those used in electrophysiological studies. Further, while asymmetries in sensitivity and timing have each been measured psychophysically, they have not been measured in the same experiments and using the same methodology. Another problem is response nonlinearity: Kremkow et al. (2014) found a difference in the contrast-response function of on- and off-centre neurons. High-contrast results might therefore be difficult to interpret because of saturation, adaptation and after-effects.
I therefore set out to develop improved psychophysical methods for measuring light/dark asymmetries. I used simple stimuli (gratings), measured response accuracy and reaction time simultaneously, and used very low contrasts. My aim was: to develop a direct, comprehensive, low-contrast experiment design for the psychophysical measurement of light/dark asymmetries.
Chapter 5 aims
This chapter tackles three questions that remain unresolved in previous work. First, contrast-response functions have been measured in, and compared between, on- and
29 Chapter 2. Aims off-centre neurons (Kremkow et al., 2014). While light and dark response thresholds have been measured psychophysically, there appears to be no recordings of contrast- response functions by this method. Second, while contrast-response functions have been fitted with curves such as the Naka-Rushton function (Kremkow et al., 2014), a more insightful approach requires a model for the mechanisms underlying the contrast-response function. Third, it is known that cortical orientation selectivity depends on the convergence of on- and off- inputs from the lateral geniculate nucleus onto cortical neurons (Jin, Wang, Swadlow, et al., 2011). How does the known delay in on-inputs affect selectivity? My aims in this chapter were therefore: to psychophysically measure contrast-response functions for stimuli lighter and darker than the background; develop a mechanistic model capable of explaining the results; psychophysically measure orientation discrimination when the light and dark components of the stimulus are presented asynchronously.
Chapter 6 aims
This chapter addresses a major gap in the literature. Light/dark asymmetries have been studied with stationary time-varying stimuli but not with moving stimuli. Do moving dark stimuli evoke stronger, faster responses than light stimuli? I studied this question both psychophysically and by analysing data collected from cat cortex in a colleague’s laboratory. My aim was: to describe light/dark asymmetries in responses to moving stimuli and to compare the results between human psychophysics and neuronal recordings in cat primary visual cortex.
Chapter 7 aims
The last results chapter is a new approach to an old problem: what is the neuronal mechanism underlying cortical selectivity for motion direction? This is an important
30 Chapter aims issue because cortical processing of motion signals starts with direction-selective neurons in primary visual cortex. The approach was prompted by a finding in the previous chapter, namely that responses to moving light bars are faster than responses to moving darks. My hypothesis here, therefore, was that the preferred direction for a simple cell was from its on- to its off-subfield. I tested this hypothesis on data collected from cat primary visual cortex by colleagues at the State University of New York. My aim was to: test hypotheses for the neuronal basis of motion direction selectivity in primary visual cortex.
Summary
My thesis is that asymmetries between responses to light and dark stimuli play an important role in mammalian vision. In particular, differences between responses to light and dark are important in:
• stimulus detection
• orientation discrimination
• motion direction discrimination
• the mechanism underlying direction discrimination
31
32
Chapter 3. General methods
This chapter describes the methods used in the experimental work, except where otherwise stated.
Psychophysics
Subjects
Ten human subjects (6-7 women, 3-4 men) took part in the experiments; they were aged 18 to 52. The experiments were performed on adult humans. Subjects were given a clear statement of the tasks they needed to perform but were unaware of the purpose of the experiment. A copy of the written consent form was provided for those who were eligible for the experiments, and they signed their agreement to participate. The experimental procedure was approved by the University of Sydney Human Research
Ethics Committee.
Each subject underwent vision tests to ensure they had normal, or corrected-to- normal, visual acuity and good stereo-acuity. A general vision test prior to the experiment was based on a standard visual acuity test (Snellen chart), stereotest (a
Titmus fly stereotest tool and a pair of polarizing glasses, Stereo Optical Co, Inc.), and near vision test (11966R, Richmond Products, Inc.) to test both eyes. Subjects were required to have an acuity of 6/6 or better for each eye and for both eyes together. The stereotest measured the subject’s stereo-threshold: a threshold of 60 seconds of arc or less was required. The requirement on near vision was that the subject be able to read
N5 text.
33 Chapter 3. General methods
Equipment
Visual stimuli were presented on a 13-inch cathode ray tube (CRT) computer monitor subtending 13° in width and 10° in height and displaying 1024 x 76 pixels. The spatial resolution of the monitor was 77 pixels/deg and the video frame rate was 75 Hz. The mean luminance of the raster was 42 cd/m2 and stimuli were monochromatic (� =
0.340, � = 0.329). The distance between the computer monitor and the eyes was 1140 mm. Subjects viewed the stimuli directly. Experiments were carried out in a quiet and darkened laboratory to ensure the highest performance of the subject with the lowest visual and auditory interference. Subjects responded to stimuli by pressing specified keys on a RT button box (Li, Liang, Kleiner, & Lu, 2010). Matlab (MathWorks,
Inc.) and the Psychophysics Toolbox (Brainard, 1997; Kleiner, Brainard, & Pelli, 2007) were used for stimulus delivery and data collection.
Visual stimulus
As shown in Figure 3.1A, a fixation square was located at the centre of the raster. The inner dimensions of the box were 2.5° x 2.5°, and the border width was 0.25°. All stimuli were shown within the square; Figure 3.1B shows a typical stimulus. The test stimulus was presented between 1 and 2 s from the start of the trial; Figure 3.2A shows stimulus timing, this duration was randomly sampled from a uniform probability density function. If the subject did not respond within 1 s of the test onset, an auditory prompt was presented to indicate a further second for the response (Figure 3.2B).
Reaction times for prompted responses were not included in the analysis.
34 Psychophysics
Figure 3.1. (A) CRT monitor and viewing area. Stimuli were presented within the fusion square, at centre. (B) A typical stimulus.
35 Chapter 3. General methods
Figure 3.2. Stimulus timing. (A) Unprompted response. Pre-stimulus intervals were set between 1-2 s and were sampled from a uniform probability density. If the subject pressed a key within one second after test onset, an unprompted response was recorded. (B) Prompted response. If the subject did not respond in the required time, an auditory prompt was given. An extra second was then given for the response.
36 Psychophysics
Calibration
Spatial settings
Measurement of the stimulus spatial settings was carried out by means of ruler placed on the monitor screen. The spatial dimensions of stimuli were measured and were within 2% of their settings.
Temporal settings
The resolution of the monitor’s setting was calibrated by using a two-channel digital storage oscilloscope (Tektronix, Inc., 70MHz, 1 GS/S). A photosensor was used to measure light levels on the monitor. All timing parameters could then be measured by counting frames on the oscilloscope (Figure 3.3A). Stimulus durations matched software settings. Reaction time was calibrated by connecting the photosensor to one oscilloscope channel and the RT button box to the other channel. Reaction time was computed from the number of frames between stimulus onset and button press. This value was compared with the time computed by the software delivering the stimulus and recording the button press; the oscilloscope and software measurements differed by a constant amount less than half the video frame duration.
37 Chapter 3. General methods
Figure 3.3. (A) Video-frame timing on the CRT monitor. The voltage recorded by a photosensor placed at the middle of the screen with no stimulus. Three video frames are shown. (B) Gamma function (left) and correction (right).
Contrast settings
Background luminance of the display monitor was measured using a spectrophotometer (Photo Research PR-650). To correct for luminance nonlinearities,
I modelled uncorrected luminance as a power function of monitor input:
� = �min + (�max − �min)� (3.1)
38 Psychophysics where
� = luminance
�min = luminance with zero input
�max = luminance with maximum input
� = monitor drive, that is, monitor input
� = constant
The variable � takes values between 0 and 1. It is convenient to normalise � by its maximum:
� = �min + (1 − �min)� (3.2) where
� = �/�max
�min = �min/�max
The left side of Figure 3.3B shows this gamma function. This nonlinearity was corrected by applying the inverse function:
⁄ � = ((� − � )⁄(1 − � )) (3.3)
The correction was implemented with a software look-up table.
Contrast is defined as
� = (� − � )⁄� (3.4) where
�stim = stimulus luminance
39 Chapter 3. General methods
�back = background luminance
The gamma-corrected plot of measured contrast versus contrast setting is shown in
Figure 3.3B. The fitted straight line accounts for 100% of the variance.
Small contrasts
I used very low contrasts in my experimental work, contrasts potentially close to the quantisation level of the video card digital-to-analog converter (DAC). It was therefore important to measure this quantisation level. To do this I alternated the luminance between a fixed setting and background, and measured contrast using the photometer and equation 3.4. The number of alternations was 12 and an average taken to obtain contrast. The result is shown by the circles in Figure 3.4A. The circles show a series of quantisation steps, and to measure the width of the steps I fitted a model as follows. Assume that contrast setting is converted to measured contrast in a series of processes as shown in Figure 3.4B. Contrast setting is converted to luminance using
Equation 3.4.
�set = �back(1 + �set)
But the background luminance was set equal to half the maximum luminance so that
� ⁄� = 0.5(1 + � )
That is
�set = 0.5(1 + �set) (3.5)
The video card framestore entry is obtained by applying the inverse gamma function
(Equation 3.3) to this value:
⁄ � = (� − � )⁄(1 − � )) (3.6)
40 Psychophysics and then digitising it:
� = ceil((� − 1)� )⁄(� − 1) (3.7) where ceil (�) is the least integer greater than � and
� = number of DAC levels
Monitor luminance is given by the gamma function (Equation 3.2)
� = � + (1 − � )� and contrast by the inverse of Equation 3.5:
� = 2� − 1 (3.8)
The DAC used had 10 bits, so the model was fitted to the data in Figure 3.4A by using
� = 2 . The model fitted well, � = 0.95, and showed that the quantisation step at background luminance was 0.004. This step is well below the important contrast differences described in the experimental work.
41 Chapter 3. General methods
Figure 3.4. (A) Contrast was determined by alternating contrast between a fixed setting and zero, and measuring contrast using the photometer. The circles show a series of quantisation steps that were well fitted by the model. (B) The model used to determine the quantisation steps. The flow diagram shows a series of processes for the conversion of set contrast to measured contrast.
Electrophysiology
Animal surgery and preparation
All procedures for the animal experiments were performed in accordance with the guidelines of the U.S. Department of Agriculture and approved by the Institutional
42 Electrophysiology
Animal Care and Use Committee at the State University of New York, State College of
Optometry.
Four adult male cats (6-12 months) were injected intramuscularly with acepromazine
(0.2 mg/kg) and initially anesthetized with ketamine (10 mg/kg). All vital signs were monitored – electrocardiogram (EKG), expired CO , electroencephalogram (EEG), pulse oximetry, blood pressure – and maintained by continuous infusions of propofol
(5–6 mg/kg/hr) and sufentanil (10–20 ng/kg/hr) for anesthesia through the two inserted intravenous catheters. Vecuronium bromide (0.2 mg/kg/hr) was given to paralyse the muscle, and saline (1-3 ml/hr) added to keep hydration. The cat’s pupils were dilated with 1% atropine sulfate, and nictitating membranes retracted with 2% neosynephrine. Contact lenses with a 3 mm pupil diameter focused the eyes on the computer screen placed 57 cm from the animal.
Electrophysiological recordings and data acquisition
Two linear 32-channel multielectrode arrays (inter-electrode distance = 0.1 mm,
Neuronexus) were inserted horizontally in layer 4 of primary visual cortex to record multi-unit neuronal activity: see Figure 7.1A. The multielectrode arrays were introduced with a small angle (<5 degrees) parallel to the cortical surface, and the inserted layer was estimated from the cortical depth, local field potentials and receptive fields recorded. A computer running Omniplex (Plexon, Inc) was used to collect the amplified signals of the stimuli (spike recordings) which were filtered between 250 Hz and 8 kHz. A 32-channel vertical array was introduced in the opposite hemisphere.
43 Chapter 3. General methods
Visual stimulation and data analysis
Visual stimuli were generated in Matlab (Mathworks, Inc) using Psychtoolbox extensions and presented on a calibrated LCD gamma-corrected monitor (120 Hz, 120 cd/m2). Gamma correction was obtained by measuring the input voltage and output luminance of the monitor and using Psychtoolbox to correct the gamma nonlinerarity.
Sparse noise was used to map the receptive fields of all the recorded channels. This technique presents brief light or dark squares at random locations across the visual field. The stimulus is continued until both contrasts have been presented at each location a specific number of times. The other type of stimulus used were light (240 cd/m2) and dark (0.27 cd/m2) moving bars (2.1 deg width) to determine the preferred direction/orientation for each cortical site. The stimuli moved on a midgrey background (120 cd/m2) from one side of the computer screen to the other. The speeds of the moving bars were set at 5, 10, 15, 17.3, 30, 60 deg/s. The orientation of the moving bar was one of 16 values evenly distributed across 360°. Each dark and light oriented bar was delivered four times, and the responses to dark and light for each motion orientation were smoothed impulse rate using a Gaussian weighting function
(SD of 1 bin).
44
Chapter 4. Pilot study
Abstract
Aims. 1. The visual contrast-response function in human subjects typically has small gradient at low contrast. The mechanism underlying this low sensitivity is unclear and I therefore aimed to measure the contrast-response function at low contrast. 2.
Recently published work shows that light decrements are detected faster than increments. I intended to see whether this observation held at low contrast. Method.
Two experiments were performed. In experiment 1, I used a raised-cosine Gabor as test stimulus to ensure that each presentation was a luminance increment or decrement but not both. Test stimuli had low contrast and random timing. The subject’s task was to indicate whether the test stimulus was a luminance increment or decrement. In experiment 2, I used a half Gabor located on the left or right side of the viewing area and asked subjects to indicate whether the stimulus was on the left or right side.
Response correctness and reaction time were recorded for both experiments. Results.
In both experiments the proportion correct was inconsistent with signal detection theory: there was a small range of contrasts, centred on zero, in which contrast sensitivity was close to zero. The vertical location of this “dead zone” differed between subjects, however, resulting in variability between subjects. Experiment 2 reduced this variability. Reaction times varied between contrast increments and decrements, but the difference was not significant. Conclusion. Low-contrast measurements revealed a “dead zone” in the psychometric function, but the reaction time difference between light and dark stimuli was not significant. Further changes in the experimental design, which might reveal this difference, are discussed.
45 Chapter 4. Pilot study
Introduction
One of the major goals of neuroscience is to show the connection between human perception and neuronal activity. One example of this linking process is the contrast- response function. Psychophysical and electrophysiological evidence suggest the same form for this function. Nachmias and Sansbury (1974) measured the detectability of a grating presented on a blank background, as shown by the open circles in Figure
4.1A. The result is a sigmoidal function of contrast. Dean (1981), who measured simple cell responses in cat primary visual cortex, showed that responses increased linearly with contrast but the intercept with the contrast axis was positive. As shown in Figure 4.1B, there was a contrast threshold below which the response was zero. Low sensitivity at low contrast is commonly seen both psychophysically and electrophysiologically, but the underlying mechanism is unclear (Foley & Legge,
1981). I therefore aimed to measure this part of the contrast-response function in more detail.
Another way of linking behaviour and neuronal activity is through processing speed.
Over the last few decades it has become apparent that the visual responses to stimuli darker than the background are not simple inversions of lighter stimuli.
Psychophysical (Bowen et al., 1989; Dannemiller & Stephens, 2001; Komban et al.,
2014; Komban et al., 2011), EEG (Zemon et al., 1988) and neuronal (Jin, Wang,
Swadlow, et al., 2011; Komban, 2014; Liu & Yao, 2014) evidence has shown that responses to darks are faster and stronger than responses to lights of the same magnitude. I aimed to show the timing difference between responses to lights and darks by measuring reaction time in human subjects.
46 Introduction
Figure 4.1. Previously published contrast-response functions. (A) Human psychophysics: the proportion of correct responses as a function of the contrast difference to be discriminated. Empty symbols represent lower contrast of zero; filled symbols represent lower contrast of 0.0079. Data for two subjects are shown (Nachmias & Sansbury, 1974). (B) Measurements on a simple cell in cat cortex; response amplitude is plotted both as depth of modulation (solid circles) and as increase in mean discharge rate (open circles) (Dean, 1981). The data in both A and B show a low-sensitivity zone near zero contrast.
47 Chapter 4. Pilot study
This chapter develops the methodology for measuring contrast sensitivity for low contrast stimuli to pursue the two aims stated in the Abstract: 1. The visual contrast- response function in human subjects typically has small gradient at low contrast. The mechanism underlying this low sensitivity is unclear and I therefore aimed to measure the contrast-response function at low contrast. 2. Recently published work shows that light decrements are detected faster than increments. I intended to see whether this observation held at low contrast. During the experiments, subjects were asked to respond to a low-contrast Gabor stimulus to indicate whether the stimulus was an increment or decrement (in Experiment 1), or whether it was on the left or right side
(in Experiment 2). I recorded the proportion of correct responses and reaction times for both experiments. My aims were only partially accomplished and will be fully developed in the following chapter.
Experiment 1: Detection of light increments and decrements
Methods
The methods applicable to all experiments have been described in the General Methods chapter. The methods specific to this chapter are as follows. The stimulus was a raised-cosine Gabor, as shown in Figure 4.2:
�(�, �) − � � + � = 0.5 × �(1 + cos(2���))exp (−0.5 × ) � �stim
where �(�, �) = luminance , � = background luminance , � = contrast; � = horizontal location (deg), � = vertical location (deg), �stim = 0.45 deg, �= 3 cycles/deg.
The stimulus time course is shown in Figure 3.2A. The pre-stimulus interval was set between 1-2 s and was sampled from a uniform probability density. If the subject
48 Experiment pressed a key within one second after the test stimulus, a response was recorded. If the subject missed the stimulus (presumably due to very low contrast), a sound was delivered to prompt the subject to respond, as shown in Figure 3.2B. The subject was given 1 s in which to respond after the prompt. The number of trials for each experimental condition was 1000.
Figure 4.2. On each trial, a Gabor patch was presented on the centre of the viewing area, with either positive or negative contrast. The subjects’ task was to indicate whether there was an increment or decrement in light.
Results
As shown in Figure 4.2, the test stimulus was a raised–cosine with a Gaussian envelope. Test stimulus contrast was sampled from a Gaussian probability density with zero mean. The subject’s task was to signal whether the test stimulus was a light increment or decrement. Figure 4.3A is the probability of choosing “increment” as a function of stimulus contrast for a single subject. The data were binned by selecting as far as possible equal numbers of samples per bin. As expected, the probability of choosing “increment” increases with contrast.
49 Chapter 4. Pilot study
There was an extra feature - a notable plateau at the centre of the sigmoidal curve. This feature also appeared for two other subjects in figure 4.3B, C. Their general trend also were sigmoidal curves, but the plateau varied vertically across subjects. Figure 4.3D indicates psychometric function for all subjects and illustrates the variability.
Discussion
The psychometric functions in Figure 4.3 are consistent across subjects in that the probability of choosing "increment" increased more or less monotonically with contrast. Further, there was a region around zero contrast in which sensitivity did not change. Subjects differed, however, in the mean probability across the contrast range.
Each subject therefore has a bias towards responding either "increment" or
"decrement". I sought to eliminate this bias in the next experiment.
50 Experiment
Figure 4.3. Experiment 1 result. (A), (B) & (C) were the psychometric functions for different subjects.
(D) Psychometric function for all subjects, with one line per subject. The vertical axis is the probability of choosing “increment”. The plateau varies vertically across subjects.
Experiment 2: Detection of stimulus location
Experiment 2 was designed to remove subject bias. The stimulus was randomly presented on one or the other side of the viewing area (Figure 4.4) and subjects were required to indicate the side on which the stimulus appeared. Subjects may still have a bias towards one side, but the randomisation removes the bias by averaging correct
51 Chapter 4. Pilot study and incorrect choices of the preferred side. The experiment also measured reaction times.
Figure 4.4. Stimuli in Experiment 2. On each trial, a half-Gabor was presented on either the left or right side of the viewing area, with either positive or negative contrast. The subject’s task was to indicate the side on which the stimulus appeared regardless of its contrast.
Methods
As shown in Figure 4.4, the test stimulus was a raised cosine with a Gaussian envelope.
Only the left or right half side was shown, with other conditions the same as in
Experiment 1. Test stimulus contrast was sampled from a Gaussian probability density with zero mean. The subject was asked to press one of the two keys to indicate the side and to respond as soon as they saw the stimulus. Reaction time was recorded as the interval between the start of the stimulus and the start of the response.
Results
Psychometric functions
Figure 4.5A is the proportion of correct responses plotted as a function of stimulus contrast for a single subject. The binning process was the same as in Experiment 1.
52 Experiment
Figure 4.5B is the proportion of trials in which the same subject chose the lighter side.
The advantage of plotting the data in this way is that it makes the psychometric function nearly monotonic as in Experiment 1. The data in Figure 4.5C (another subject) and 4.5D (all subjects) have a similar form. The plateau is now at much the same height for all subjects, showing that the new experiment design has removed the bias. An analysis of variance (factors: subject, contrast and its third and fifth powers) showed that the proportion of correct responses did not differ significantly between subjects ( � = 0.95, �(5, 51) = 0.22, � = 0.97). The � value is the fraction of the variance for which the model accounts, and therefore indicates the quality of the model.
53 Chapter 4. Pilot study
Figure 4.5. Psychometric functions, Experiment 2. (A) The proportion of correct responses as a function of stimulus contrast for a single subject. (B) The psychometric function of the same subject as in (A) showing the probability that the subject chooses the lighter side on a trial. (C) Psychometric function for another subject. (D) Psychometric functions for all subjects, with one line per subject. All subjects have a plateau in sensitivity at a vertical value of about 0.5.
Chronometric function
Figure 4.6A is a scatter plot of reaction time for a single subject with each trial shown by a circle. For both signs of contrast, reaction time reduced as absolute contrast increased: the scatter distribution declined on either side. Reaction time is binned and plotted as a function of contrast magnitude in Figure 4.6B; the subject is the same as in
A. Figure 4.6C (another subject) and Figure 4.6D (all subjects) both show that reaction time reduced as contrast magnitude increased. I used an analysis of variance test to
54 Experiment compare reaction times for contrast decrements and increments: there was no difference at the 5% significance level (factors: subject, contrast polarity, contrast magnitude and its square: � = 0.99, �(1,50) = 1.5, � = 0.30).
Figure 4.6. Chronometric functions, Experiment 2. (A) Reaction time for a single subject: red and blue circles represented responses to contrast increments and decrements respectively. (B), (C) and (D) were mean reaction time for (B) a single subject, (C) all subjects, and (D) pooled data.
55 Chapter 4. Pilot study
Discussion
Psychometric function
As shown in Figure 4.5, the revised design in Experiment 2 produced psychometric functions that were more consistent across subjects. Individual biases were removed by asking subjects to choose between stimulus locations rather than contrast polarities.
The revised psychometric functions show a clear plateau at zero contrast. This indicates that the subject has no information about the stimulus for a range of contrasts around zero. This observation is consistent with the resting hyperpolarisation found in simple cells in monkey primary visual cortex (Tan, Chen, Scholl, Seidemann, &
Priebe, 2014). Further, simple cells in cat primary visual cortex have a clear contrast threshold (Dean, 1981). Psychophysically, the low-gradient portion of the psychometric functions shown in Figure 4.1 could also stem from the same neuronal source.
Reaction time
Previous psychophysics (Komban et al., 2014) has shown that human responses to stimuli darker than the background are faster than responses to lighter stimuli. I expected to see the same result in Figure 4.6D. The reaction times here, however, did not differ significantly between lights and darks. A possible reason for this lack of significance can be seen in Figure 4.4. The stimuli I used had an edge at the midline.
Eye movements occur even when the eyes are fixated, and such movements during the stimulus, which lasted for 0.2s, could have shifted the fixation point back and forth across the edge. These movements could, in turn, have confused responses to lighter
56 Experiment and darker stimuli. This possible confound is addressed in the next chapter by removing the stimulus edge.
57
58
Chapter 5. Responses to light and dark stationary stimuli
Contributions of authors
This chapter reproduces a paper I published as part of my PhD. My contributions to the paper were as follows. All authors designed the experiments. I performed the experiments. Alan Freeman and I analysed the results. All authors contributed to writing the manuscript.
59 Chapter 5. Responses to light and dark stationary stimuli
Journal of Vision (2016) 16(15):18, 1–13 1
Orientation discrimination requires coactivation of on- and off-dominated visual channels School of Medical Sciences, The University of Sydney, Gloria Luo-Li Sydney, Australia # $ School of Psychology, The University of Sydney, Sydney, David Alais Australia # $ School of Medical Sciences, The University of Sydney, Alan W. Freeman Sydney, Australia # $
Orientation sensitivity depends on the cortical convergence of on- and off-center subcortical neurons. Introduction Off-center inputs are faster and stronger than their on- center counterparts: How does this asymmetry affect Attempts to link perception with neural function orientation discrimination? We tackled this question have a long history. Landmark studies in the late 1960s psychophysically with grating stimuli that either used classical psychophysical approaches such as increased or decreased luminance. The gratings were of adaptation and masking to study the spatial tuning and low contrast in order to avoid the complicating orientation selectivity of human vision (Blakemore & influences of nonlinearities such as response saturation, masking, and aftereffects. Gratings were presented in Campbell, 1969; Campbell & Kulikowski, 1966), either of two locations, and subjects indicated the revealing results that closely matched emerging findings perceived location. Stimuli were randomly timed, and about neurons in primary visual cortex (Hubel & response correctness and reaction time were recorded. Wiesel, 1968). In a similar vein, we were motivated to We found the following: (a) Contrast sensitivity was look for perceptual consequences of recent neuro- insignificant for a range of contrasts around zero. (b) physiological studies showing an asymmetry between Outside this range, contrast sensitivity for contrast responses to light increments and decrements of equal decrements exceeded that for increments by an average magnitude (Komban et al., 2014; C. I. Yeh, Xing, & of 15%. (c) Reaction times for contrast decrements were Shapley, 2009). As Phillips and Wilson (1984) have up to 45 ms less than for increments. (d) These findings pointed out, experiments using suprathreshold stimuli are reproduced by a signal-detection model which require careful interpretation. We have therefore taken incorporates recent physiological findings: Neurons in a different approach, using low-contrast stimuli to primary visual cortex are hyperpolarized at rest; these avoid nonlinearities such as response saturation, neurons respond more to darks than to lights; and off- masking, and afterimages. dominated cortical neurons have shorter latencies than The asymmetry between responses to light incre- their on-dominated neighbors. (e) We tested orientation ments and decrements has been shown psychophysi- discrimination by splitting a grating into two components, one containing the light bars and the other cally (Bowen, Pokorny, & Smith, 1989; Dannemiller & the dark, and presenting the two components Stephens, 2001; Komban et al., 2014; Komban, Alonso, asynchronously. Discrimination was optimal when light & Zaidi, 2011; Krauskopf, 1980; Lu & Sperling, 2012), bars preceded dark bars, consistent with coactivation of in visual evoked potentials (Zemon, Gordon, & Welch, on- and off-center cortical inputs. We conclude that the 1988), in multiunit recordings (Kremkow et al., 2014; ability to discriminate between orientations is Xing, Yeh, & Shapley, 2010), and in single cortical intimately connected with the properties of subcortical neurons (Komban et al., 2014; Kremkow, Jin, Wang, & channels. Alonso, 2016; Lee, Huang, & Fitzpatrick, 2016; Liu & Yao, 2014; Samonds, Potetz, & Lee, 2012; Veit, Bhattacharyya, Kretz, & Rainer, 2014; Wang et al., 2015; C. I. Yeh et al., 2009). Each of these studies has
Citation: Luo-Li, G., Alais, D., & Freeman, A. W. (2016). Orientation discrimination requires coactivation of on- and off- dominated visual channels. Journal of Vision, 16(15):18, 1–13, doi:10.1167/16.15.18.
doi: 10.1167/16.15.18 Received June 1, 2016; published December 22, 2016 ISSN 1534-7362
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License. Downloaded From: https://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/935912/ on 11/29/2018
60
Journal of Vision (2016) 16(15):18, 1–13 Luo-Li, Alais, & Freeman 2
found that responses to darks are stronger, faster, subjects were tested with a cathode-ray-tube (CRT) subserved by greater numbers of neurons, or more monitor. Concerns have been raised about the fitness precisely retinotopically mapped, than responses to of such a monitor for testing asymmetries in responses lights. The asymmetry is present in species from tree to lights and darks (Gawne & Woods, 2003); we shrews to humans, suggesting that it may play a useful therefore used a liquid-crystal display (LCD) monitor visual role. The measured response difference between for the remaining subjects. The two equipment sets are lights and darks is, however, often small compared with described in turn. the magnitude of the responses themselves. We wondered whether this small difference is due in part to the use of suprathreshold stimuli, leading to non- CRT monitor linearities such as response saturation. We therefore Stimuli were presented on a CRT monitor (Philips measured the light/dark asymmetry in human subjects 105S, Philips, Amsterdam, the Netherlands) driven by presented with low-contrast stimuli. an ATI Radeon HD 5770 video card. The card was The presence of this asymmetry in cortical responses controlled, and responses collected, with the Psycho- suggests that it may influence an important visual physics Toolbox software (Brainard, 1997; Pelli, 1997) capability: orientation discrimination. This ability is extended by a low-level kernel driver (Kleiner, Brainard, thought to rely on orientation-selective neurons (Wilson & Pelli, 2007). The monitor had a spatial resolution of & Wilkinson, 2004). The cortical origin of orientation 77 pixels/8 and a frame rate of 75 Hz. Stimuli were selectivity is still controversial (Vidyasagar & Eysel, monochromatic (x 0.340, y 0.329) with a mean 2015), but the strongest line of evidence indicates that it luminance of 42 cd/m¼ 2. Luminance¼ was modulated with depends on the convergence of on- and off-center 10 bits/gun, yielding a contrast resolution of 1/(0.53 210) subcortical inputs onto single cortical neurons (Jin, 0.0020. Larger contrasts were linearized using a Wang, Swadlow, & Alonso, 2011b; Reid & Alonso, ¼lookup table. Subjects used a chin rest which stabilized 1995; Tanaka, 1983). Given that on- and off-dominated eye-to-monitor distance at 1.14 m, and they signaled neurons convey responses to lights and darks, respec- their responses by using two buttons on an RTbox (Li, tively, it is to be expected that the asymmetry between Liang, Kleiner, & Lu, 2010). Experiments were con- channels will affect orientation discrimination. We ducted in a darkened room so that the only light visible tested this idea by breaking gratings into light and dark to the subjects was from the monitor. bars, which were presented asynchronously. We therefore had two aims. The first was to measure response asymmetries between lights and darks in the LCD monitor absence of nonlinearities such as response saturation, The second set of equipment differed from the first in masking, and afterimages. The second aim was to see the following respects. Stimuli were presented on a what part this asymmetry plays in orientation dis- VIEWPixx monitor (VPixx Technologies, Saint-Bruno, crimination. We have previously presented a summary Canada). Monitor resolution was 73 pixels/ at 120 Hz, of parts of this work (Freeman, Luo-Li, & Alais, 2015). 8 mean luminance was 48 cd/m2, and chromaticity was (x 0.294, y 0.312). Eye-to-monitor distance was 0.92 ¼m, and subjects¼ signaled responses with a ResponsePixx Methods button box (VPixx).
Subjects Comparison between setups Sixteen human subjects (11 women, five men) took Figure 1 shows consecutive frames from each of our part in these experiments; they were aged 18 to 49. monitors. The LCD monitor offers at least two Subjects had normal vision in that visual acuity was at advantages over the CRT: It is faster (120 Hz compared least 6/6 in each eye and stereo threshold was 1 min or with 75 Hz), and the decay time of a video frame is the better. One subject (GL-L) is an author of this article; same as the onset time. We therefore quantitatively all other subjects were unaware of the aims of the study compared the psychophysical data obtained from the and were paid for their time. All procedures conformed two monitors. An analysis of variance on the data in with the Declaration of Helsinki. Figure 5C showed no effect from the monitor (factors: subject, odd powers of contrast up to 5, monitor)—a 0.05, r2 0.95, F(1, 141) 0.0071, p 0.79—nor did¼ Equipment such an¼ analysis on the data¼ in Figure¼ 8C (factors: subject, contrast and its square, contrast polarity, Two sets of equipment were used, distinguished monitor)—a 0.05, r2 0.92, F(1, 141) 0.35, p 0.55. mainly by the type of visual stimulator. Eleven Accordingly,¼ the data¼ are combined without¼ distin-¼
Downloaded From: https://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/935912/ on 11/29/2018
61 Chapter 5. Responses to light and dark stationary stimuli
Journal of Vision (2016) 16(15):18, 1–13 Luo-Li, Alais, & Freeman 3
Figure 1. Video-frame timing on the stimulus monitors. Two stimulus monitors were used, one a cathode-ray tube and the other a liquid-crystal display. The graphs show the voltage recorded by a photosensor placed at the middle of the screen in the absence of any stimulus. Two video frames are shown for the cathode-ray-tube monitor and three for the liquid-crystal display. The liquid-crystal display’s backlight successively lit eight horizontal strips down the screen, producing the eight discontinuities visible within each frame. Figure 3. Stimuli in Experiment 2. (A) Gabors were broken into a sum of components lighter and darker than the background. (B) Light and dark bars were presented asynchronously. The duration of a video frame was fixed at 13 ms because only the cathode-ray-tube monitor was used for this experiment. (C) On a specific trial, both light and dark bars were tilted 28 from vertical. The subjects’ task was to indicate whether the tilt was clockwise or counterclockwise from vertical.
guishing between monitors. The fitness of our meth- odology for measuring light/dark asymmetries is taken up in the Discussion.
Experiment 1: Detection of lights and darks
We performed two experiments. The first compared the detection of contrast increments with detection of decrements.
Stimuli Figure 2A shows the spatial form of the stimulus, a raised Gabor: 0.5 3 contrast 3 (1 cosine) 3 Gaussian. More precisely, the stimulus is þ x2 y2 s x y 0 5 3 c 1 cos 2 fy exp 0 5 3 Figure 2. Stimuli in Experiment 1. (A) On each trial, a Gabor ; : p : þ2 ; ð Þ¼ þ ð Þ À rstim patch was presented on either the left or right side of the viewing area, with either positive or negative contrast. The where x and y are the distances across the visual field subjects’ task was to indicate the side on which the patch (in degrees) in the horizontal and vertical directions, appeared. (B) The time course during a single trial. The test respectively. Contrast is defined as stimulus was presented between 1 and 2 s from the start. c lcenter lbackground =lbackground; Incorrect responses were signaled with auditory feedback. (C) If ¼ð À Þ the subject did not respond within 1 s of the test onset, an where lcenter is luminance at the center of the stimulus auditory prompt was presented and a further second allowed and lbackground is background luminance. The spatial for the response. Reaction times in these prompted trials were frequency of the cosine function was f 3c/8 and the ¼ discarded. standard deviation of the Gaussian function was rstim
Downloaded From: https://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/935912/ on 11/29/2018
62
Journal of Vision (2016) 16(15):18, 1–13 Luo-Li, Alais, & Freeman 4
Figure 5. Psychometric functions, Experiment 1. (A) The proportion of correct responses as a function of stimulus contrast for a single subject. (B) The psychometric function in (A) is made more monotonic by plotting the probability that the subject chooses the lighter side during stimulation: This has the effect of inverting the points at negative contrast. (C) Psychometric functions for all subjects. (D) The circles show the Figure 4. Model for results of Experiment 1. (A) The model psychometric function obtained by pooling data across subjects, assumes that four populations of cortical neurons are involved: and the line shows the best fit of the signal-detection model on- and off-dominated neurons on each side of the viewing described in the text. area. (B) Probability density of membrane potential, relative to the action-potential threshold, for each population. Unstimu- density spanning the range 1 to 2 s. The stimulus had lated neurons (green) have a mean potential which is less than sudden onset and faded linearly over 0.2 s. Subjects the action-potential threshold. A stimulus with positive contrast shifts the on-dominated neurons (red) rightward and the off- responded in the 1 s following the start of the stimulus dominated neurons (blue) leftward from the resting position. or, if they failed to respond (usually because the (C) Impulse rate is obtained by thresholding the membrane stimulus was too faint), a medium-pitch auditory potential. prompt was delivered and they were given another
0.38 in both the horizontal and vertical directions. ¼On each trial, the stimulus was presented with equal probability in two locations separated by 0.88. Contrast on each trial was sampled from a Gaussian probability density with zero mean and a standard deviation of 0.05. Thus, as shown in Figure 2A, each trial contained a contrast increment or decrement: The subject’s task was to indicate on which side the stimulus was delivered. The area containing the stimulus was 2.58 wide and high, bordered by a black Figure 6. Contrast sensitivity, Experiment 1. (A) Contrast line 0.258 wide. sensitivity was calculated from the psychometric functions in Figure 5C by taking the difference between neighboring Procedure contrast bins. Each line represents a single subject. (B) Mean contrast sensitivity across subjects, with 95% confidence The time course of a single experimental trial is intervals. The error bar at a contrast of zero includes zero shown in Figure 2B. It started with an interval whose contrast sensitivity, indicating that subjects are insensitive to a duration was sampled from a uniform probability range of contrasts about zero.
Downloaded From: https://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/935912/ on 11/29/2018
63 Chapter 5. Responses to light and dark stationary stimuli
Journal of Vision (2016) 16(15):18, 1–13 Luo-Li, Alais, & Freeman 5
Figure 7. Psychometric functions versus contrast magnitude, Figure 8. Chronometric functions, Experiment 1. (A) Reaction Experiment 1. (A–B) The psychometric function for (A) a single time was measured along with the proportion of correct subject and (B) all subjects. The function for contrast increments responses. This graph shows reaction time for a single subject, is shown in red, and for decrements in blue. (C) Circles show with responses to contrast increments in red and to decrements pooled data, and lines give the predictions of the signal- in blue. (B–C) Mean reaction time for (B) a single subject and (C) detection model. Subjects are more sensitive to decrements all subjects. (D) The pooled data, given by the circles, show that than to increments: The thresholds for the increment and responses to contrast decrements are faster than those to decrement curves, measured as the contrast at which the increments. The lines show the best-fitting model. (interpolated) proportion correct is 0.75, are 0.043 and 0.037, respectively. (D) Contrast threshold for contrast increments Model divided by that for decrements. Each circle represents one subject. We modeled the results of Experiment 1 as follows. Simple receptive fields in primary visual cortex typically second in which to respond (Figure 2C). In both have multiple subfields of two types: On-subfields unprompted and prompted cases a low-pitch signal was respond to positive contrast and off-subfields to sounded if their choice was incorrect. The next trial negative contrast (Hubel & Wiesel, 1962). The two then started. Each run of trials lasted 60 s, and subjects polarities of subfield typically differ in strength, so that rested between runs if they wished. a neuron can be either on- or off-dominant (Komban et al., 2014). We assume four neuronal populations, with on- and off-dominant neurons at each of the two Experiment 2: Orientation discrimination stimulus locations. Let p be the membrane potential relative to the action-potential threshold. As illustrated The second experiment tested the ability to discrim- in Figure 4B, membrane potential in the absence of a inate between gratings of differing orientation. This stimulus is assumed to be experiment differed from the previous one as follows. p pr n; As shown in Figure 3A, the Gaussian envelope of the ¼ þ Gabor stimuli had a standard deviation of 0.458 and the where the resting potential pr is negative, indicating Gabors were split into light bars (0.5 3 contrast 3 (1 that the neuron is hyperpolarized relative to threshold. þ sine) 3 Gaussian) and dark bars ( 0.5 3 contrast 3 (1 The noise component n represents a Gaussian proba- À À sine) 3 Gaussian). The two polarities were presented bility density with zero mean and standard deviation r. asynchronously, for one video frame each, as shown in Given that our stimuli are weak, we assume that Figure 3B. On each trial, light and dark bars were membrane potential depends linearly on contrast c: aligned with each other but tilted from vertical by 28 p pr sc n; (Figure 3C). The subjects’ task was to indicate whether ¼ þ þ the tilt was clockwise or counterclockwise. where s is contrast sensitivity. On- and off-dominant
Downloaded From: https://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/935912/ on 11/29/2018
64
Journal of Vision (2016) 16(15):18, 1–13 Luo-Li, Alais, & Freeman 6
neurons have positive and negative sensitivities al., 2014; Krauskopf, 1980; Kremkow et al., 2014, 2016; sn and sf, respectively. As shown in Figure 4C, the Lee et al., 2016; Liu & Yao, 2014; Lu & Sperling, 2012; action-potential rate a, is proportional to the part of Samonds et al., 2012; Veit et al., 2014; Wang et al., membrane potential that exceeds the ,threshold: 2015; Xing et al., 2010; C. I. Yeh et al., 2009; Zemon et al., 1988). We aimed to describe this asymmetry by a kh pp; ¼ ð Þ minimizing response nonlinearities and thereby deter- where mining how the asymmetry affects orientation dis- crimination. The stimuli in our first experiment, shown 1 p 0 h p in Figure 2A, consisted of Gabors with a raised carrier: ð Þ¼ 0 p, 0 A raised cosine function of distance was multiplied by a and k is the (constant) conversion factor from potential Gaussian profile so that each presentation either to impulse rate. Reaction time t is obtained by incremented or decremented light, but not both. The integrating the action-potential rate to a criterion Gabor appeared on either the left or right of the number of action potentials m: midline, and the subjects’ task was to indicate on which side it appeared. m Contrast increments had fast onset and faded offset, t tmin ; ¼ þ a as shown in Figure 2B, in order to favor on-dominant neurons. Contrast decrements had the same timing, but where tmin is the minimum reaction time (and therefore includes fixed sensory and motor delays). with the opposite polarity, to preferentially stimulate The subject chooses the side of the viewing area off-dominated neurons. We recorded the proportion of containing the population that first reaches the correct responses as well as reaction time. Subjects criterion, and reaction time is set equal to the time at frequently failed to see test stimuli with the lowest which the population reaches this criterion. When all contrasts and were therefore prompted for a response, four populations have zero impulse rate (typically as shown in Figure 2C. These prompted responses because of a very low contrast), the subject chooses contributed to the measurement of proportion correct either side with equal probability, and reaction time is but were not included with the reaction-time data. assumed to be greater than 1 s. The model was solved by Monte Carlo simulation. For each contrast a total of 5,000 trials were run, each using a noise sample Psychometric functions independent of other trials. The model was fitted to the data by minimizing least squared error. Denoting the The results for one subject are shown in Figure 5A. gap between resting potential and action-potential Contrast on each trial was sampled from a Gaussian threshold as one unit of potential difference, the probability density centered on 0, and the psychometric resulting parameters were function was computed by compiling responses into bins that contained, as nearly as possible, equal sn 23:7 potential-unit=contrast-unit ¼ numbers of samples. As expected, the proportion of correct responses increases with contrast magnitude. It sf 26:8 potential-unit=contrast-unit ¼À is convenient to make this U-shaped function more monotonic, as shown in Figure 5B. The vertical axis r 0:372 potential-unit here plots the probability that the subject chooses the ¼ lighter side of the bordered area. Figure 5C and D show m=k 0:0292 potential-unit:s data from all subjects and responses pooled across ¼ subjects, respectively. The line in Figure 5D is the tmin 0:364 s: prediction of a signal-detection model that will be ¼ described later. The psychometric functions in Figure 5 have two notable features. First, there is a plateau centered on Results zero contrast. This feature is explored in more detail in Figure 6A, which shows the slope of the psychometric function (obtained by differencing neighboring bins in Experiment 1: Detection of lights and darks the data of Figure 5C). Figure 6B shows the mean over subjects, with 95% confidence intervals. The error bar It is clear that psychophysical and neurophysiolog- at zero contrast includes zero contrast sensitivity, ical responses to light decrements are not simple mirror indicating that the sensitivity is not significantly images of the responses to light increments (Bowen et different from zero. Thus, there is a range of contrasts al., 1989; Dannemiller & Stephens, 2001; Komban et in which the subject has little or no information about
Downloaded From: https://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/935912/ on 11/29/2018
65 Chapter 5. Responses to light and dark stationary stimuli
Journal of Vision (2016) 16(15):18, 1–13 Luo-Li, Alais, & Freeman 7
the stimulus. Neurons in primary visual cortex are hyperpolarized at rest (Tan, Chen, Scholl, Seidemann, & Priebe, 2014) and therefore require a contrast magnitude greater than some fixed minimum to generate action potentials. It seems likely that the plateau in our psychometric functions indicates the range of contrasts in which membrane potential has not reached the action-potential threshold. The other notable feature in both Figures 5 and 6 is the asymmetry between the left and right sides of the Figure 9. Reaction-time differences between contrast incre- curves. This is shown more clearly in Figure 7, where ments and decrements, Experiment 1. (A–B) Reaction time for the proportion of correct responses is shown as a contrast decrements was subtracted from that for increments; function of contrast magnitude for a single subject the difference is shown for (A) all subjects and (B) the mean (Figure 7A), for all subjects (Figure 7B), and pooling across subjects. The 95% confidence intervals in (B) show that data across all subjects (Figure 7C). Red and blue reaction times for decrements are significantly less than for curves derive from contrast increments and decrements, increments across most of the contrast range. respectively. In general, the blue curves lie above the red ones, indicating that our subjects were more sensitive to contrast decrements than to increments of variance on the data in Figure 8C showed that the the same magnitude. An analysis of variance on the difference is highly significant (factors: subject, contrast magnitude and its square, contrast polarity), a 0.05, data in Figure 7B showed that the difference in 2 10 ¼ r 0.92, F(1, 141) 46, p 2.9 3 10À . This result proportion correct was significant (factors: subject, ¼ ¼ ¼ powers of contrast magnitude from 1 to 3, contrast matches previous psychophysics (Komban et al., 2011, polarity), a 0.05, r2 0.94, F(1, 140) 8.7, p 0.0038. 2014) and the finding that off-dominated neurons in cat Figure 7D was¼ obtained¼ by finding the¼ contrast¼ at primary visual cortex respond faster than do on- which the proportion of correct responses was 0.75 and dominated neurons (Komban et al., 2014). The lines in dividing this threshold for contrast increments by that Figure 8D, which give the predictions of the model to for decrements: The ratio averages 1.15 over all be described later, are based on single-neuron proper- subjects. ties such as these. The reaction-time difference between contrast in- crements and decrements is shown for all subjects in Chronometric functions Figure 9A and for the mean across subjects in Figure 9B. There are two points of particular interest in this Along with the proportion of correct responses, we last plot. First, the biggest difference between lights and measured the reaction times of our subjects. Figure 8A darks is obtained with a contrast magnitude of 0.05, shows reaction times for one subject, where each circle and the difference declines with higher contrasts. gives the time taken to respond to a single presentation Second, the maximum reaction-time difference we of the test stimulus. For contrasts close to zero, subjects obtain is about 45 ms. This is a substantially larger were often not aware that the test had been delivered difference that that found in single-neuron studies and therefore had to be prompted to respond. These (Komban et al., 2014), possibly because we used points give the reaction time to an auditory prompt 1 s smaller contrasts. after the visual stimulus, and reaction times greater than 1 s have therefore been discarded. For the remaining points, reaction time decreases with contrast Model magnitude, as expected. We were particularly interested in comparing reac- Our psychophysical demonstrations of response tion times to lights and darks. Figure 8B through D asymmetries between lights and darks have clear neural make this comparison by plotting the data in various correlates. We find, for example, that the contrast ways against contrast magnitude. They show plots in sensitivity for darks is greater than that for lights. This which the data are binned by contrast magnitude and matches the finding that the areas of cat (Jin et al., each point provides mean reaction time. Figure 8B 2008) and monkey (Xing et al., 2010; C. I. Yeh et al., gives the results for a single subject, Figure 8C gives 2009) visual cortex dominated by off-responses are results for all subjects with one line for each subject and larger than on-dominated areas. We therefore wished contrast polarity, and Figure 8D shows pooled data. to see whether a signal-detection model based on The last plot shows that reaction times for contrast known physiology could reproduce our psychometric decrements are less than for increments; an analysis of and chronometric functions.
Downloaded From: https://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/935912/ on 11/29/2018
66
Journal of Vision (2016) 16(15):18, 1–13 Luo-Li, Alais, & Freeman 8
The model assumes two neuronal populations—on- and off-dominated cortical cells—on each side of the viewing area. Membrane potential in each of these four populations is assumed to have a Gaussian probability density. Unstimulated, the mean of this density lies below the action-potential threshold, as shown in Figure 4B. Stimulated, the mean changes by the product of contrast and contrast sensitivity, where the sensitivity is positive for on- and negative for off- dominated neurons. Figure 4B shows the resulting shifts in the density. Impulse rate is obtained by half- wave rectifying the membrane potential, as shown in Figure 4C. On any trial, a sample is taken from the thresholded density in each of the populations and integrated over time toward a criterion number of impulses. The subject chooses the side of the viewing area containing the population that first reaches the criterion, and reaction time is set equal to the time at which the population reaches this criterion. The model was fitted to the observations by optimizing five parameters: contrast sensitivity for on- and off-dominated neurons, standard deviation of the Figure 10. Experiment 2. (A) The subjects’ task in this probability density, criterion level, and minimum experiment was to discriminate between gratings with differing reaction time. The results are shown by the lines in tilts. On each trial, the light and dark components of a grating Figures 5D, 7C, and 8D. The fits are good in that the were presented asynchronously, as shown by the legend at fraction of variance accounted for by the model is right. Psychometric functions calculated for the pooled data are 99.4%, 98.7%, and 98.2%, respectively. It seems, shown. (B) The effect of onset asynchrony was determined by therefore, that a relatively simple model can account interpolating the psychometric functions to find the contrast for both the proportion of correct responses and the threshold at a proportion correct of 0.75. Contrast sensitivity is reaction time, provided that contrasts are small. the reciprocal of this threshold. Error bars give 95% confidence intervals determined by bootstrap resampling of the pooled data. Performance is best when light bars precede dark bars by Experiment 2: Orientation discrimination 13 ms. (C) Chronometric functions for the pooled data. (D) Circles show contrast sensitivity at a reaction time of 0.55 s, and Human subjects are highly sensitive to contour error bars give 95% confidence intervals. Performance is again orientation, in that they can reliably judge the best for asynchronies close to zero. alignment of two contours to within about 0.58 (Westheimer, Shimamura, & McKee, 1976). This was to indicate the direction of tilt. The pooled results capability is thought to depend on orientation-selective for 10 subjects are shown in Figure 10A. Seven neurons in primary visual cortex (Wilson & Wilkinson, psychometric functions are shown, with one for each 2004). It is known that on- and off-center geniculate onset asynchrony between light and dark bars. relay cells can converge onto the same cortical neuron Orientation discrimination is best when the light and (Alonso, Usrey, & Reid, 1996; Tanaka, 1983), and dark bars are presented at about the same time. there is increasing evidence that this convergence We quantified this effect by finding the (interpolated) provides the foundation for orientation selectivity (Jin contrast at a proportion correct of 0.75, then taking the et al., 2011b; Reid & Alonso, 1995). Given the timing reciprocal of this contrast to obtain contrast sensitivity. differences between on- and off-pathways that we have The result is shown in Figure 10B, along with 95% already described, we wondered whether orientation confidence intervals. Two aspects of this graph are discrimination might also depend on the relative notable. First, discrimination is substantially poorer timings of contrast increments and decrements. when the asynchrony between light and dark bars is To test this idea, we decomposed a grating into light large in magnitude than when it is close to zero. and dark bars as shown in Figure 3A and presented Second, there is a clear asymmetry between the left and these two components asynchronously at the same right sides of the graph: Discrimination is better when location (Figure 3B). On each trial, both sets of bars light bars lead dark bars. In particular, the two points were tilted 28 clockwise or counterclockwise from with an asynchrony magnitude of 13 ms do not fall vertical, as shown in Figure 3C, and the subjects’ task within each other’s error bars, indicating that this
Downloaded From: https://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/935912/ on 11/29/2018
67 Chapter 5. Responses to light and dark stationary stimuli
Journal of Vision (2016) 16(15):18, 1–13 Luo-Li, Alais, & Freeman 9
difference is significant. Given that on-pathways have a addressed these concerns by including a relatively fast longer response latency than do off-pathways, this (120 Hz) stimulator. More conclusively, we find that result suggests that orientation discrimination is the peak latency difference between dark and light optimal when on- and off-signals reach cortex at about responses averages 45 ms, which covers more than the same time. three frames in the case of the CRT monitor and more Reaction time provides further evidence for this than five frames for the LCD. This rules out the finding via the chronometric functions in Figure 10C. possibility that our findings are artifacts of intraframe The tuning curve in Figure 10D was calculated by effects or of interactions between consecutive frames. finding the contrast at which reaction time equaled 0.55 Another methodological issue is the definition of s. This measure of discriminability again falls with contrast. The use of Michelson contrast, increasing magnitude of asynchrony. The tuning curve c lmaximum lminimum = lmaximum lminimum , is prob- is asymmetric, but in this case the asymmetry narrowly lematic¼ð whenÀ comparingÞ ð luminanceþ incrementsÞ with fails to reach significance. decrements: The denominator for luminance patterns lying above background is greater than for patterns in which luminance is less than the background. This could result in light/dark asymmetries due to the Discussion contrast definition rather than to visual processing. We have been careful, therefore, to use Weber contrast, Our results demonstrate a robust difference between c lcenter lbackground =lbackground. In this case, stimulus responses to lights and darks, and the influence of this luminance¼ð À is proportionalÞ to contrast, so that light/ difference on a fundamental visual capability: orienta- dark asymmetries cannot be ascribed to the contrast tion discrimination. The key new findings are as definition. follows. There is a range of contrasts centered on zero for Psychometric function which the stimulus is invisible Previous studies have shown that responses to light The psychometric function in Figure 5D is uncon- decrements are stronger than responses to incre- ventional in that it displays responses to both positive ments. Other studies have found that responses to and negative contrasts. The upper right quadrant, for darks are faster than to lights. We have shown in a which contrast is positive, is more familiar in that it is single experiment that responses to light decrements sigmoidal in shape (Nachmias & Sansbury, 1974; Naka have an advantage over increments in both sensitivity & Rushton, 1966). What accounts for the low contrast and speed sensitivity at near-zero contrast? Foley and Legge These data are reproduced by a signal-detection (1981) provided evidence for a noisy signal followed by model in which off-dominated neurons have higher a threshold but did not suggest a neural locus for the contrast sensitivity than their on-dominated neigh- threshold. Dean (1981) showed that neurons in cat bors but both classes of neuron are hyperpolarized at primary visual cortex do not respond to contrasts rest. below a fixed magnitude, and it has recently been found Optimal orientation discrimination requires that on- (Tan et al., 2014) that unstimulated neurons in the and off-inputs to visual cortex be activated at about primary visual cortex of awake monkeys are hyperpo- the same time. larized relative to the action-potential threshold. It We review our methodology and then discuss each of seems likely, therefore, that the contrast-response these results in turn. function has a low gradient near the origin because of resting hyperpolarization in primary cortex. Methodology Light/dark asymmetries Several concerns have been raised about the use of video stimulators in vision experiments (Zele & Asymmetries between responses to lights and darks, Vingrys, 2005), particularly in studies of response which thread through our results, are now well asymmetries to lights and darks (Gawne & Woods, established in the literature. Zemon et al. (1988) 2003). These concerns include the time taken to paint, recorded visually evoked cortical potentials in human and the asymmetry of the luminance profile during, a subjects presented with positive- and negative-contrast single video frame (Zele & Vingrys, 2005) and the checkerboards, and found that contrast gain was interval between the offset of one frame and the onset greater in the latter case. Bowen et al. (1989) used a of the next (Gawne & Woods, 2003). We have partially stimulus with a sawtooth time course to show that
Downloaded From: https://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/935912/ on 11/29/2018
68
Journal of Vision (2016) 16(15):18, 1–13 Luo-Li, Alais, & Freeman 10
psychophysical detection was better when the rapid receive convergent input from spatially separated change in luminance was downward than when it was populations of on- and off-center geniculate neurons upward. More recently, three studies have found that (Jin et al., 2011b; Reid & Alonso, 1995; Tanaka, 1983). off-dominant neurons are more prevalent than on- Given that on-pathways to cortex are slower than off- dominant in primary visual cortex of cat (Jin et al., pathways (Jin et al., 2011a), light components of a 2008) and monkey (Xing et al., 2010; C. I. Yeh et al., stimulus apparently need to be delivered before dark 2009). Off-center inputs to cortex have a lower response components in order for the two response components latency than do their on-center counterparts (Jin et al., to arrive at the cortex at about the same time. Only 2011a), a difference that is reflected in the findings that then are orientation-selective neurons optimally stim- off-dominant cortical neurons respond faster that do ulated. This finding raises an obvious puzzle: Natural on-dominant (Komban et al., 2014) and that psycho- stimulation, in which light and dark components of a physical reaction times are less for darks than for lights stimulus are delivered simultaneously, will result in (Komban et al., 2011, 2014). A number of other studies suboptimal stimulation. Is the asymmetry of light and (Dannemiller & Stephens, 2001; Krauskopf, 1980; dark responses a bug or a feature in the visual system? Kremkow et al., 2014, 2016; Lee et al., 2016; Liu & One answer to this question may be that natural Yao, 2014; Lu & Sperling, 2012; Samonds et al., 2012; viewing typically involves durations substantially lon- Veit et al., 2014; Wang et al., 2015) also show that ger than the asynchronies used in our experiment. responses to negative contrasts have an advantage over Another answer may lie in motion sensitivity. The on- those to positive contrasts. Two studies (Chichilnisky & and off-center geniculate-neuron populations that Kalmar, 2002; Smith, Whitney, & Fitzpatrick, 2015) converge on a cortical neuron are spatially separated present contrary findings, but they derive from and have differing timing, providing the essential stimulation in the peripheral visual field, suggesting substrate for motion-direction selectivity: A stimulus that distance from the center of vision may be an moving in one direction will almost certainly evoke a important factor. cortical response that differs from that evoked by Where do light/dark asymmetries originate in the movement in the opposite direction. Indeed, a model visual system? Komban et al. (2014) have suggested the assuming that cortical inputs are separated in both photoreceptor-bipolar synapse as the source. The space and time reproduces a number of published synapses between photoreceptors and on- and off- observations on motion-direction selectivity (Hesam center bipolar calls are metabotropic and ionotropic, Shariati & Freeman, 2012). Neurophysiological con- respectively, and the former type of synapse is known firmation of this idea is not yet available. In the to be slower (Snellman, Kaur, Shen, & Nawy, 2008). meantime, the possibility remains that light/dark Direct evidence for or against the proposal that light/ asymmetries play a major role in the foundation of dark asymmetries originate at this synapse is lacking. motion-direction selectivity. Another possible source is light adaptation. Increased (decreased) luminance decreases (increases) visual sensitivity (Shapley & Enroth-Cugell, 1984), and much Stimulus appearance of the sensitivity change occurs within 100 ms of a luminance change (Hayhoe, Benimoff, & Hood, 1987; We have shown that responses to lights are weaker T. Yeh, Lee, & Kremers, 1996). The relationship than responses to darks. How might this influence the between sensitivity and luminance is a power law with appearance of a pattern containing both lights and an exponent between 0.5 and 1 (Shapley & Enroth- darks? Peli (1997) presented subjects with two Gabors Cugell, 1984). Figure 7CÀ indicatesÀ a sensitivity increase of differing phase and adjusted the physical contrast of of 15% when luminance drops by 8% (at a contrast the patterns until they had the same apparent contrast. magnitude of around 0.04). This implies a power law The match was best when the Michelson contrasts of with an exponent more negative than 1, meaning that the patterns were equal. Michelson contrast, using our if adaptation produces asymmetries betweenÀ light and terminology, is dark responses it cannot be solely responsible. cMichelson lcenter lbackground = lcenter lbackground : ¼ð À Þ ð þ Þ The denominator here is less for darks than for lights, Orientation discrimination suggesting an advantage for responses to darks relative to lights. We therefore replotted the data in Figures 7C We have also shown that orientation discrimination and 8D in terms of Michelson contrast instead of depends on the timing of light and dark components of Weber contrast. The results, illustrated in Figure 11, the stimuli being discriminated. This makes sense given show that the advantage of darks over lights is still the known physiology of the geniculocortical synapse. significant—left side (factors: subject, powers of con- Orientation-selective neurons in primary visual cortex trast magnitude from 1 to 3, contrast polarity), a ¼
Downloaded From: https://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/935912/ on 11/29/2018
69 Chapter 5. Responses to light and dark stationary stimuli
Journal of Vision (2016) 16(15):18, 1–13 Luo-Li, Alais, & Freeman 11 Acknowledgments
The authors declare no competing financial interests. Funding was provided by the Australian Research Council (DP130102336). We thank Paul Martin for his valuable comments on an earlier version of this article.
Commercial relationships: none. Corresponding author: Alan W. Freeman. Figure 11. Michelson contrast. The data in Figures 7C and 8D are Email: [email protected]. replotted in terms of Michelson (rather than Weber) contrast. Address: School of Medical Sciences, The University of The advantage of responses to darks over responses to lights is Sydney, Sydney, Australia. reduced. This result helps to explain the apparent contrast of patterns containing both light and dark (Peli, 1997).
0.05, r2 0.94, F(1, 140) 9.2, p 0.0029; right side References (factors:¼ subject, contrast¼ magnitude¼ and its square, 2 contrast polarity), a 0.05, r 0.93, F(1, 141) 269, p Alonso, J.-M., Usrey, W. M., & Reid, R. C. (1996). 34 ¼ ¼ ¼ 1.8 3 10À —but effect sizes are reduced by this ¼ Precisely correlated firing in cells of the lateral procedure (left side: threshold ratio 1.10; right side: geniculate nucleus. Nature, 383, 815–819. maximum reaction-time difference ¼36 ms). This Blakemore, C., & Campbell, F. W. (1969). On the means that our subjects respond similarly¼ to differing existence of neurones in the human visual system patterns if contrast is defined so as to reinforce the selectively sensitive to the orientation and size of contribution of lights. Conversely, the result in Figure retinal images. The Journal of Physiology, 203, 237– 11 suggests that Peli’s observations on stimulus 260. appearance can be at least partly explained by the response asymmetry between lights and darks. Bowen, R. W., Pokorny, J., & Smith, V. C. (1989). Sawtooth contrast sensitivity: Decrements have the edge. Vision Research, 29 , 1501–1509. Low-contrast stimulation Brady, N., & Field, D. J. (2000). Local contrast in natural images: Normalisation and coding effi- We have used very small contrast magnitudes in our ciency. Perception, 29 , 1041–1055. measurements, mostly less than 0.1. The use of low Brainard, D. H. (1997). The Psychophysics Toolbox. contrast has at least two advantages. First, it minimizes Spatial Vision, 10, 433–436. nonlinearities such as response saturation, masking, Campbell, F. W., & Kulikowski, J. J. (1966). Orienta- and afterimages. Using low contrast therefore leaves us tional selectivity of the human visual system. The with three essential nonlinearities: the ‘‘dead zone’’ in Journal of Physiology, 187 , 437–445. sensitivity around zero contrast, and the higher gain and lower latencies seen with contrast decrements. The Chichilnisky, E. J., & Kalmar, R. S. (2002). Functional finding that these nonlinearities cannot be removed by asymmetries in ON and OFF ganglion cells of using weak stimuli suggests that they play a critical role primate retina. Journal of Neuroscience, 22, 2737– in visual function. A second advantage of using low 2747. contrasts is that they correspond with the most Dannemiller, J. L., & Stephens, B. R. (2001). Asym- prevalent contrasts in the natural environment (Brady metries in contrast polarity processing in young & Field, 2000). Frequency histograms of contrast in human infants. Journal of Vision, 1(2):5, 112–125, natural scenes show a steady decline away from zero doi:10.1167/1.2.5. [PubMed] [Article] contrast. Our visual environments contain large but Dean, A. F. (1981). The relationship between response informative areas of low contrast such as surfaces, amplitude and contrast for cat striate cortical textures, and shadowed regions. Our low-contrast neurones. The Journal of Physiology, 318, 413–427. results therefore throw light on the information Foley, J. M., & Legge, G. E. (1981). Contrast detection available from these ubiquitous features of the visual and near-threshold discrimination in human vision. scene. Vision Research, 21, 1041–1053. Keywords: visual cortex, light/dark asymmetry, ori- Freeman, A. W., Luo-Li, G., & Alais, D. (2015, May). entation discrimination, psychophysics, model The psychophysicist’s microscope: Weak stimuli
Downloaded From: https://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/935912/ on 11/29/2018
70
Journal of Vision (2016) 16(15):18, 1–13 Luo-Li, Alais, & Freeman 12
reveal neuron-like response properties. Paper pre- tial resolution for darks than lights. Proceedings of sented at the Annual Meeting of the Vision Sciences the National Academy of Sciences, USA, 111, 3170– Society, Tampa, FL. 3175. Gawne, T. J., & Woods, J. M. (2003). Video-rate and Kremkow, J., Jin, J., Wang, Y., & Alonso, J. M. (2016). continuous visual stimuli do not produce equiva- Principles underlying sensory map topography in lent response timings in visual cortical neurons. primary visual cortex. Nature, 533, 52–57. Visual Neuroscience, 20, 495–500. Lee, K. S., Huang, X., & Fitzpatrick, D. (2016). Hayhoe, M. M., Benimoff, N. I., & Hood, D. C. (1987). Topology of ON and OFF inputs in visual cortex The time-course of multiplicative and subtractive enables an invariant columnar architecture. Nature, adaptation process. Vision Research, 27, 1981– 533, 90–94. 1996. Li, X., Liang, Z., Kleiner, M., & Lu, Z. L. (2010). Hesam Shariati, N., & Freeman, A. W. (2012). A multi- RTbox: A device for highly accurate response time stage model for fundamental functional properties measurements. Behavior Research Methods, 42, in primary visual cortex. PLoS One, 7, e34466. 212–225. Hubel, D. H., & Wiesel, T. N. (1962). Receptive fields, Liu, K., & Yao, H. (2014). Contrast-dependent OFF- binocular interaction and functional architecture in dominance in cat primary visual cortex facilitates the cat’s visual cortex. The Journal of Physiology, discrimination of stimuli with natural contrast 160, 106–154. statistics. European Journal of Neuroscience, 39, Hubel, D. H., & Wiesel, T. N. (1968). Receptive fields 2060–2070. and functional architecture of monkey striate Lu, Z. L., & Sperling, G. (2012). Black-white asym- cortex. The Journal of Physiology, 195, 215–243. metry in visual perception. Journal of Vision, Jin, J., Wang, Y., Lashgari, R., Swadlow, H. A., & 12(10):8, 1–21, doi:10.1167/12.10.8. [PubMed] Alonso, J. M. (2011a). Faster thalamocortical [Article] processing for dark than light visual targets. Nachmias, J., & Sansbury, R. V. (1974). Grating Journal of Neuroscience, 31, 17471–17479. contrast: Discrimination may be better than detec- Jin, J., Wang, Y., Swadlow, H. A., & Alonso, J. M. tion. Vision Research, 14, 1039–1042. (2011b). Population receptive fields of ON and Naka, K. I., & Rushton, W. A. H. (1966). S-potentials OFF thalamic inputs to an orientation column in from colour units in the retina of fish (Cyprinidae). visual cortex. Nature Neuroscience, 14, 232–238. The Journal of Physiology, 185, 536–555. Jin, J. Z., Weng, C., Yeh, C.-I., Gordon, J. A., Peli, E. (1997). In search of a contrast metric: Matching Ruthazer, E. S., Stryker, M. P., . . . Alonso, J. M. the perceived contrast of Gabor patches at different (2008). On and off domains of geniculate afferents phases and bandwidths. Vision Research, 37, 3217– in cat primary visual cortex. Nature Neuroscience, 3224. 11, 88–94. Pelli, D. G. (1997). The VideoToolbox software for Kleiner, M., Brainard, D., & Pelli, D. (2007, August). visual psychophysics: Transforming numbers into What’s new in Psychtoolbox-3 Paper presented at movies. Spatial Vision, 10, 437–442. the European Conference on Visual Perception, Phillips, G. C., & Wilson, H. R. (1984). Orientation Arezzo, Italy. bandwidths of spatial mechanisms measured by Komban, S. J., Alonso, J. M., & Zaidi, Q. (2011). masking. Journal of the Optical Society of America Darks are processed faster than lights. Journal of A, 1, 226–232. Neuroscience, 31, 8654–8658. Reid, R. C., & Alonso, J. M. (1995). Specificity of Komban, S. J., Kremkow, J., Jin, J., Wang, Y., monosynaptic connections from thalamus to visual Lashgari, R., Li, X., . . . Author, J. (2014). cortex. Nature, 378, 281–284. Neuronal and perceptual differences in the tempo- Samonds, J. M., Potetz, B. R., & Lee, T. S. (2012). ral processing of darks and lights. Neuron, 82, 224– Relative luminance and binocular disparity prefer- 234. ences are correlated in macaque primary visual Krauskopf, J. (1980). Discrimination and detection of cortex, matching natural scene statistics. Proceed- changes in luminance. Vision Research, 20, 671– ings of the National Academy of Sciences, USA, 109, 677. 6313–6318. Kremkow, J., Jin, J., Komban, S. J., Wang, Y., Shapley, R., & Enroth-Cugell, C. (1984). Visual Lashgari, R., Li, X., . . . Alonso, J. M. (2014). adaptation and retinal gain controls. In N. Osborne Neuronal nonlinearity explains greater visual spa- & G. Chader (Eds.), Progress in retinal research
Downloaded From: https://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/935912/ on 11/29/2018
71 Chapter 5. Responses to light and dark stationary stimuli
Journal of Vision (2016) 16(15):18, 1–13 Luo-Li, Alais, & Freeman 13
(Vol. 3, pp. 263–346). Oxford, UK: Pergamon tion of spatial phase in visual cortex. Nature Press. Neuroscience, 18, 97–103. Smith, G. B., Whitney, D. E., & Fitzpatrick, D. (2015). Westheimer, G., Shimamura, K., & McKee, S. P. Modular representation of luminance polarity in (1976). Interference with line-orientation sensitivi- the superficial layers of primary visual cortex. ty. Journal of the Optical Society of America, 66, Neuron, 88, 805–818. 332–338. Snellman, J., Kaur, T., Shen, Y., & Nawy, S. (2008). Wilson, H. R., & Wilkinson, F. (2004). Spatial channels Regulation of ON bipolar cell activity. Progress in in vision and spatial pooling. In L. M. Chalupa & J. Retinal and Eye Research, 27, 450–463. S. Werner (Eds.), The visual neurosciences (Vol. 2, Tan, A. Y. Y., Chen, Y., Scholl, B., Seidemann, E., & pp. 1060–1068). Cambridge, MA: MIT Press. Priebe, N. J. (2014). Sensory stimulation shifts Xing, D., Yeh, C. I., & Shapley, R. M. (2010). visual cortex from synchronous to asynchronous Generation of black-dominant responses in V1 states. Nature, 509, 226–229. cortex. Journal of Neuroscience, 30, 13504–13512. Tanaka, K. (1983). Cross-correlation analysis of Yeh, C. I., Xing, D., & Shapley, R. M. (2009). ‘‘Black’’ geniculostriate neuronal relationships in cats. Journal of Neurophysiology, 49, 1303–1318. responses dominate macaque primary visual cortex V1. Journal of Neuroscience, 29, 11753–11760. Veit, J., Bhattacharyya, A., Kretz, R., & Rainer, G. (2014). On the relation between receptive field Yeh, T., Lee, B. B., & Kremers, J. (1996). The time structure and stimulus selectivity in the tree shrew course of adaptation in macaque retinal ganglion primary visual cortex. Cerebral Cortex, 24, 2761– cells. Vision Research, 36, 913–931. 2771. Zele, A. J., & Vingrys, A. J. (2005). Cathode-ray-tube Vidyasagar, T. R., & Eysel, U. T. (2015). Origins of monitor artefacts in neurophysiology. Journal of feature selectivities and maps in the mammalian Neuroscience Methods, 141, 1–7. primary visual cortex. Trends in Neurosciences, 38, Zemon, V., Gordon, J., & Welch, J. (1988). Asymme- 475–485. tries in ON and OFF visual pathways of humans Wang, Y., Jin, J., Kremkow, J., Lashgari, R., Komban, revealed using contrast-evoked cortical potentials. S. J., & Alonso, J. M. (2015). Columnar organiza- Visual Neuroscience, 1, 145–150.
Downloaded From: https://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/935912/ on 11/29/2018
72
Chapter 6. Responses to light and dark moving stimuli
The previous chapter studied light/dark asymmetries for stationary stimuli. There is no guarantee that the same asymmetries apply to moving stimuli. Given the ubiquity of moving stimuli in our visual environments, I wished to extend my study to such stimuli. Chapter 6, therefore, is devoted to measuring responses to light and dark stimuli in motion.
Contributions of authors
This chapter reproduces a paper I published as part of my PhD. My contributions to the paper are stated on its fifth page. In full, the statement is as follows.
Gloria Luo-Li, Qasim Zaidi, and Alan Freeman designed the psychophysical
experiments. Gloria Luo-Li performed those experiments. Reece Mazade and
Jose-Manuel Alonso performed the electrophysiological experiments. All
authors analysed and discussed the results. Gloria Luo-Li, Reece Mazade, Jose-
Manuel Alonso, and Alan Freeman wrote the manuscript.
73 Chapter 6. Responses to light and dark moving stimuli
ARTICLE DOI: 10.1038/s42003-018-0066-y OPEN Motion changes response balance between ON and OFF visual pathways
Gloria Luo-Li1, Reece Mazade2, Qasim Zaidi2, Jose-Manuel Alonso2 & Alan W. Freeman 1 1234567890():,; Humans are faster at detecting dark than light stationary stimuli, a temporal difference that originates early in the visual pathway. Here we show that this difference reverses when stimuli move, making detection faster for moving lights than darks. Human subjects judged the direction of moving edges and bars, and made faster and more accurate responses for light than for dark stimuli. This light/dark asymmetry is greatest at low speeds and dis- appears at high speeds. In parallel experiments, we recorded responses in the cat visual cortex for moving bars and again find that responses are faster for light bars than for dark bars moving at low speeds. We show that differences in the luminance-response function between ON and OFF pathways can reproduce these findings, and may explain why ON pathways are used for slow-motion image stabilization in many species.
1 Sydney Medical School, The University of Sydney, Lidcombe, NSW 1825, Australia. 2 Graduate Center for Vision Research, College of Optometry, State University of New York, 33 West 42nd Street, New York, NY 10036, USA. These authors contributed equally: Gloria Luo-Li, Reece Mazade. Correspondence and requests for materials should be addressed to J.-M.A. (email: [email protected]) or to A.W.F. (email: [email protected])
COMMUNICATIONS BIOLOGY | (2018) 1:60 | DOI: 10.1038/s42003-018-0066-y | www.nature.com/commsbio 1
74
ARTICLE COMMUNICATIONS BIOLOGY | DOI: 10.1038/s42003-018-0066-y
he visual system is organized in parallel ON and OFF As expected from the strong OFF dominance of the visual cor- pathways that process light and dark targets in visual tex13,14, cortical responses were stronger for dark than light sti- Tscenes. ON and OFF pathways are present in many species, muli (Fig. 2a). However, when the stimulus speed was reduced, from flies1 to humans2. These pathways originate at the first the responses to lights became faster than the stronger responses synapse of the visual pathway in mammals and the second to darks (Fig. 2a, 5 deg s−1; see Supplementary Fig. 2a,b for synapse in flies. The kinetics of this synapse are slower for ON additional comparisons of response strength and latency between than OFF pathways and, consistently, humans are also slower and lights and darks). The response latency differences between less accurate at detecting lights than darks3–5. However, while the slowly-moving darks and lights were an order of magnitude larger OFF pathway has better temporal resolution than the ON path- than any possible latency artifacts due to the display monitor way6,7, slow-motion image stabilization is strongly dominated by (Supplementary Fig. 3). the ON pathway8 through an ON directional selective cell9,10 that is among the best preserved retinal ganglion cells in mammals11. Slow-motion image stabilization is needed in almost all verte- Modelling. The different speed sensitivity to lights and darks could result from differences in the luminance-response functions brates to keep a steady image on the retina during self-movement, 15–17 which is important to maximize visual acuity12. It’s unclear why between ON and OFF visual pathways . To test this possible the ON pathway dominates slow-motion image stabilization. Our mechanism, we passed a Gaussian function through different ON results provide a possible explanation by demonstrating that the and OFF luminance-response functions. Due to higher contrast ON pathway is faster and more sensitive at processing slow sensitivity and response saturation for ON than OFF, this motion compared to the OFF pathway. simulation made the Gaussian response broader and reach threshold faster for light than dark stimuli (Fig. 2d, 5 deg s−1), a difference that decreased with increasing stimulus speed (Fig. 2d, Results 60 deg s−1). At the highest speeds, the differences in response Psychophysics. To quantify the speed at which humans perceive amplitude become more pronounced than the differences in moving objects in psychophysical experiments, adult human response width and responses become faster to darks. This simple subjects viewed edges or bars moving through a Gaussian window simulation reproduced the differences in response latency (Fig. 1a). On each trial the stimulus was lighter or darker than the between lights and darks measured in the cat visual cortex background, with equal probability, and moved left or right, again (Fig. 2e). with equal probability. The subject was asked to indicate the perceived direction of motion (Fig. 1a, b). Surprisingly, all six subjects were more accurate (Fig. 1c, p < = 0.006, ANOVA test) Discussion and faster (Fig. 1d, p < 0.001, ANOVA test) at detecting light than In summary, our results demonstrate a functional advantage of dark motion. This result was consistent across subjects and could the ON pathway in processing slow motion in humans and be reproduced with both moving edges and bars (Fig. 1c, d, carnivores. The ON pathway is also more effective at processing 18 Supplementary Fig. 1a-d). The finding that humans are not slow motion in flies , and zebrafish cannot correctly stabilize 19 always faster and more accurate at detecting dark than light sti- slow motion when the ON pathway is inactivated . Therefore, muli is in contrast to what was previously thought3,4. the different response to speed between lights and darks that we The difference in sensitivity between lights and darks was most demonstrate in humans and carnivores could be general across pronounced at low speeds and was reduced as the speed increased animals. The ON pathway has higher luminance sensitivity than (Fig. 1e, f, ANOVA tests, p < 0.001). The difference was reversed the OFF pathway and its slower kinetics allows a more effective at high speeds (Supplementary Fig. 1e-f). On average, human temporal integration of low-contrast edges during slow motion. subjects more accurately detected the slow motion of light than The general properties of the ON pathway in animal vision (high they did dark stimuli (Fig. 1g, 1 deg s−1 edge: 80 vs. 61%, 1 deg s sensitivity, slow kinetics) may therefore also explain why it is the −1 bar: 92 vs. 79%, p < 0.001, paired t-tests). Detection was also chosen pathway for image motion stabilization. faster for lights than darks (Fig. 1h, 1 deg s−1 edge: 312 vs. 371 ms, p = 0.017, 1 deg s−1 bar: 161 vs. 248 ms, p < 0.001, paired t- fl Methods tests). Conversely, when the stimuli were stationary and ashed Human psychophysics. A total of three female and three male human subjects (the fastest possible entrance of the stimulus in the receptive took part in the experiments. Their ages ranged from 21 to 28 and they all had field), the detection was more accurate and faster for darks normal, or corrected to normal, vision in that acuity was 6/6 or better in each eye compared to lights (Fig. 1g, 73 vs. 67%, p < 0.001; Fig. 1h, 432 vs. and stereo-threshold was 1 min or less. Subjects were unaware of the aims or results of the experiments. Stimuli were presented on a cathode-ray-tube monitor driven 446 ms, p = 0.007, paired t-tests). Importantly, these differences by an ATI Radeon HD 5770 video card. The card was controlled, and responses were not due to a trade-off between accuracy and reaction time collected, with the Psychophysics Toolbox software extended by a low-level kernel since the motion direction of lights was reported with both higher driver20. Luminance was modulated with 10 bits per gun. Contrast is defined by accuracy (Fig. 1c, e) and faster reaction time (Fig. 1d, f). From (Lstim−Lbg)/Lbg where Lstim and Lbg are the stimulus luminance (maximum for lights, minimum for darks) and background luminance, respectively. Measure- these experiments, we conclude that humans see slow motion ments of small luminance increments and decrements around the background level better for lights than darks. (obtained using a PR-650 photometer, Photo Research, Inc.) showed that measured contrast differed from the contrast setting by an average of 0.0018 (Supplementary Fig. 4a). Larger contrasts were linearized using a look-up table: fitting a straight line Electrophysiology. To investigate the neuronal mechanisms to the relationship between measured and set contrast yielded an adjusted r2 of underlying the low-speed advantage for lights, we recorded 0.999, indicating a near-perfect fit (Supplementary Fig. 4b). −1 responses to light and dark bars swept across the receptive fields Most experiments used a monitor with a spatial resolution of 77 pixels deg ,a 75 Hz video frame rate, a background luminance of 40 cd m−2, and a chromaticity of neurons in cat primary visual cortex. As with human reaction of x = 0.34, y = 0.33. Supplementary Fig. 4c, d shows that light and dark moving times, cortical responses were faster for light than dark bars at low bars had identical timing on this monitor, ruling out the possibility that reaction speeds but the difference was reduced when the bar speed time differences were due to the monitor. For the experiments that required a increased. This trend could be demonstrated in recordings from stimulus speed of 30 deg s−1, the monitor characteristics were 46 pixels deg−1, 120 Hz, 32 cd m−2, and x = 0.31, y = 0.35. Subjects viewed the monitor from a distance single cortical sites (Fig. 2a), the average cortical response of of 1.14 m, used a chinrest to stabilize the head, and responded to stimuli by individual animals (Fig. 2b), and the average cortical response pressing one of two buttons on an RTBox21. Experiments were conducted in a across animals (Fig. 2c, p < 0.0001, bootstrap confidence interval). quiet room in which the only substantial source of light was the monitor.
2 COMMUNICATIONS BIOLOGY | (2018) 1:60 | DOI: 10.1038/s42003-018-0066-y | www.nature.com/commsbio
75 Chapter 6. Responses to light and dark moving stimuli
COMMUNICATIONS BIOLOGY | DOI: 10.1038/s42003-018-0066-y ARTICLE
a b 1–2 s or Edge Light Test
< 1 s Response or Bar Dark Feedback
c Moving edge (3 deg/s) Moving bar (3 deg/s) d Moving edge (3 deg/s) Moving bar (3 deg/s) 1 1 0.5 Light < Dark 0.5 Light < Dark P = 2.3 × 10–6 P = 9.2 × 10–9
0.75 0.75 0.35 0.35
6 subjects 6 subjects Reaction time (s) Reaction time (s)
Proportion correct Light > Dark Proportion correct Light > Dark 0.5 P = 0.006 0.5 P = 0.004 0.2 6 subjects0.2 6 subjects
0 0.025 0.05 0 0.025 0.05 00.030.06 00.030.06 Contrast magnitude Contrast magnitude Contrast magnitude Contrast magnitude
efMoving edge Moving bar Moving edge Moving bar 90 150 50 70
45 75 25 35
Light > Dark Light > Dark Light > Dark Light > Dark –6 –6 –5 –7 Contrast sensitivity (PC) Contrast sensitivity (RT) Contrast sensitivity (PC) P = 6.3 × 10 P = 2.0 × 10 P = 1.3 × 10 P = 1.6 × 10 0 0 0 Contrast sensitivity (RT) 0 1 3 10 13 1013 1013 10 Speed (deg/s) Speed (deg/s) Speed (deg/s) Speed (deg/s)
ghDarks Darks 1.0 Lights *** 0.5 Lights * ** *** *** *
*** 0.75 0.25 Reaction time (s) time Reaction Proportion correct Proportion
0.5 0 1 deg/s 1 deg/s 30 deg/s Stationary 1 deg/s 1 deg/s 30 deg/s Stationary edge bar bar flash edge bar bar flash
Fig. 1 Humans see low speeds better with lights than darks. a, b Subjects detected the motion direction of dark/light drifting stimuli (edges/bars) presented at different contrasts. c Psychometric functions for accuracy obtained with light (red) or dark (blue) edges or bars. An analysis of variance showed that accuracy was better for light than for dark stimuli: F(1, 49) = 8.2 and 9.3 for edges and bars, respectively, where the factors were subject, contrast polarity, and powers of contrast magnitude from 1 to 4. P-values are shown on the graphs. d Same as (c) but for reaction time. Reactions were faster for light stimuli than for dark: F(1, 51) = 28 and 47 for edges and bars, respectively, with factors subject, contrast polarity, contrast magnitude and its square. e Contrast sensitivity was calculated by pooling psychometric functions across subjects (Supplementary Fig. 1a, c), finding the contrast at which proportion correct (PC) was 0.75, and taking the reciprocal of this contrast. Three stimulus speeds are shown. f Similarly, contrast sensitivity for reaction times (RT) was obtained from the contrast at which the pooled times (Supplementary Figure 1b, d) were halfway between their highest and lowest values. Contrast sensitivity for light stimuli is significantly greater than that for dark: F(1, 26) = 32, 37, 29, and 50 for the four graphs left to right, with factors subject, contrast polarity, speed and its square, and contrast polarity × speed. g Pair comparisons in the detection of light and dark stimuli presented at different speeds. Data for the stationary flash come from a reanalysis of results in Luo-Li et al.4. h Same as (g) for reaction time. Stars in (g) and (h) show significant values calculated with paired t-tests: *p < 0.05, **p < 0.01, ***p < 0.001
Experimental procedures were approved by the University of Sydney Human responded, whichever was sooner. The moving bar, shown in the lower part of Research Ethics Committee and informed consent was obtained from all subjects. Fig. 1a, was windowed with the same Gaussian and was terminated in the same way All stimuli were presented in a bordered area with inner diameter 2.5 deg × 2.5 as the edge. Stimulus speed was 1, 3, 10, or 30 deg s−1. Bar width was 0.1 deg except deg. Borders were black and 0.25 deg wide. The upper part of Fig. 1a shows three for the experiments using 30 deg s−1, in which case the width was 0.2 deg. On each frames from the moving edge stimulus. The image was multiplied by a Gaussian trial the motion was leftward or rightward with equal probability. The subject’s task profile centered in the bordered area; the standard deviation was 0.3 deg. The was to indicate motion direction. stimulus was present until it was 0.25 deg from the border, or the subject
COMMUNICATIONS BIOLOGY | (2018) 1:60 | DOI: 10.1038/s42003-018-0066-y | www.nature.com/commsbio 3
76
ARTICLE COMMUNICATIONS BIOLOGY | DOI: 10.1038/s42003-018-0066-y
a 5 deg/s 15 deg/s 30 deg/s 60 deg/s 100 200 200 Darks 150 Lights
100 50 100 100
50 Response (spk/s) Response (spk/s) Response (spk/s) Response (spk/s)
0 0 0 0 –2.12 0 2.12 –0.710 0.71 –0.350 0.35 –0.180 0.18
Time (s) Time (s) Time (s) Time (s) e b Individual c All d Model animals animals 5 deg/s 60 deg/s 0.1 0.1 0.1 Data P < 0.0001 161213 Model Lum. 170507 170711 Time Time 0.05 0.05 0.05 170808 Resp. Lum. latency difference (s) latency difference (s) latency difference (s) L L 0 L 0 0 - - - Darks D D D Lights
510 30 60 510 30 60 Resp. 56010 30 Stimulus speed (deg/s) Stimulus speed (deg/s) Time Time Stimulus speed (deg/s)
Fig. 2 Different luminance-response functions between ON and OFF visual pathways explain the speed switch between lights and darks. a Cortical responses to light (red) and dark (blue) bars moving at different speeds. Notice the different time scales. The dotted line indicates the response level (half- amplitude for light bar) at which latency was measured. b Response latency differences between darks (D) and lights (L) measured in different animals. c Average difference in response latency. Error bars show 95% confidence intervals. d Latency differences explained with a model that uses different luminance-response functions for lights and darks (middle) to transform the stimulus (top) into peristimulus time histograms (dotted line is static threshold). e The model reproduces the data illustrated in (c)
Contrast on each trial was sampled from a Gaussian probability density with zero Electrophysiological recordings and data acquisition. Two 32-channel linear mean. Each stimulus therefore had equal probability of being lighter or darker than multielectrode arrays (0.1 mm inter-electrode distance, Neuronexus) were intro- the background. The time course of a single trial is shown in Fig. 1b. Reaction time duced horizontally and one 32-channel linear multielectrode array was introduced was measured relative to the midpoint time, that is, the time at which the edge or bar vertically in primary visual cortex to measure cortical multiunit activity; horizontal reached the center of the bordered area. The interval between the start of a trial and its and vertical arrays were in opposite hemispheres. The horizontal multielectrode midpoint time was sampled from a uniform probability density spanning the range 1 arrays were introduced into the cortex with <5° angle and centered in layer 4. The to 2 s. Subjects responded up to 1 s following the midpoint time or, if they failed to spike recordings were filtered between 250 Hz and 8 kHz, sampled at 40 kHz and respond (usually because the stimulus was too faint), a medium-pitch auditory collected by a computer running Omniplex (Plexon), as previously described. prompt was delivered and they were given another second in which to respond. Prompted responses contributed to the calculation of proportion correct but not to reaction time. This explains why the horizontal axis in Fig. 1d differs from that in 1c: Visual stimuli and data analysis. Custom MATLAB code (Mathworks) with low-contrast reaction times are missing and the bins are therefore centered on higher Psychtoolbox extensions was used to present visual stimuli on a 24-inch LCD gamma- −2 contrasts. In both unprompted and prompted cases, a low-pitch signal was sounded if corrected monitor (BenQ XL2420-B, 120 Hz, mean luminance: 120 cd m ). The the subject’s choice was incorrect. The next trial then started. Each run of trials lasted gamma correction was performed by measuring the input voltage and output lumi- 60 s and subjects rested between runs if they wished. nance of the monitor (power function with a gamma exponent). A function in Response differences between lights and darks were tested using analyses of Psychtoolbox 3 was generated to correct for the gamma nonlinearity, thereby line- variance. Factors were subject, contrast polarity, and powers of contrast magnitude arizing the relationship between contrast setting and displayed contrast. from 1 to 4 (Fig. 1c and Supplementary Fig. 1a, c and e); subject, contrast polarity, The preferred direction/orientation of each cortical site was determined using −2 −2 contrast magnitude, and its square (Fig. 1d and Supplementary Fig. 1b, d and f); and moving light (240 cd m ) and dark (0.27 cd m ) bars (2.1 deg width) on a mid- −2 subject, contrast polarity, speed and its square, and contrast polarity × speed (Fig. 1e, f). gray background (120 cd m ). The bars began at one side of the monitor and moved across the entire screen. The orientation of the moving bar was one of 16 possible values. For each dataset, the moving bars were presented at 6 speeds: 5, 10, 15, 17.3, 30, 60 deg s−1. Responses at 15 and 17.3 deg s−1 were similar and Animal physiology. Surgery and preparation. All surgical and recording pro- therefore combined; they are referred to as 16.15 deg s−1 (average of 15 and 17.3). cedures were performed in accordance with the US Department of Agriculture Each dark and light oriented bar was presented four times, and peristimulus time guidelines and were approved by the Institutional Animal Care and Use Com- histograms were compiled into 35 bins covering a stimulus sweep. To find the mittee (IACUC) at the State University of New York, State College of Optometry. direction selectivity index for a recording site, we combined responses to light and Complete details of the surgical procedures have been described previously13,22. dark for each motion direction, lightly smoothed impulse rate (Gaussian weighting Briefly, adult male cats (Felis catus, n = 4) were tranquilized with an intramuscular with a standard deviation of 1 bin), and determined the maximum response, R, −1 injection of acepromazine (0.2 mg kg ) and anesthetized with an intramuscular over all directions. The index was calculated as (Rpref – Ropp)/Rpref where Rpref is the −1 injection of ketamine (10 mg kg ). Two intravenous catheters were inserted into response for the direction in which response was maximal, and Ropp is the response each hind limb to administer continuous infusions of propofol (5–6 mg kg−1 h−1), in the opposite direction. Only cortical recording sites that showed clear direction sufentanil (10–20 ng kg−1 h−1), vecuronium bromide (0.2 mg kg−1 h−1), and saline selectivity were selected for analysis (direction selectivity index > 0.25). (1–3 ml h−1). The animal was intubated, ventilated, and pupils dilated with 1% We wished to average time courses over recording sites in order to look for atropine sulfate with the nictitating membranes retracted with 2% neosynephrine. light/dark asymmetries. Each site, however, receives its input from a slightly The eyes were fitted with contact lenses with a 3 mm pupil to focus on the monitor different location in the visual field and responses therefore peak at differing times. placed 57 cm in front of the animal. Throughout the surgery and recordings, the To compensate for these time differences, we fitted the model shown in animal vital signs including temperature, electrocardiogram (EKG), expired CO2, Supplementary Fig. 2c and d (a Gaussian profile added to a linear function of time). electroencephalogram (EEG), pulse oximetry, and blood pressure were monitored The time course fitted with this model was then shifted laterally by subtracting the and carefully maintained within normal physiological limits to ensure adequate model’s peak time, and the shifted time course for the preferred direction was anesthesia and ventilation. averaged over recording sites. Response latency was obtained by interpolating on
4 COMMUNICATIONS BIOLOGY | (2018) 1:60 | DOI: 10.1038/s42003-018-0066-y | www.nature.com/commsbio
77 Chapter 6. Responses to light and dark moving stimuli
COMMUNICATIONS BIOLOGY | DOI: 10.1038/s42003-018-0066-y ARTICLE
the mean time course at the midpoint of the response with lower amplitude. The 7. Jin, J., Wang, Y., Lashgari, R., Swadlow, H. A. & Alonso, J.-M. Faster latency difference was calculated as the latency for the response to a dark bar less thalamocortical processing for dark than light visual targets. J. Neurosci. 31, fi than for a light bar. The con dence intervals in Fig. 2c were calculated with a 17471–17479 (2011). bootstrap procedure using 200 subsamples. 8. Simpson, J. I. The accessory optic system. Annu. Rev. Neurosci. 7,13–41 (1984). 9. Barlow, H. B., Hill, R. M. & Levick, W. R. Retinal ganglion cells responding Latency model. A simple model was constructed to replicate the light-dark latency selectively to direction and speed of image motion in the rabbit. J. Physiol. 173, differences measured in our recordings. A moving bar was first modeled as a 377–407 (1964). Gaussian function to simulate the changes in luminance over time at the receptive 10. Sun, L. O. et al. Functional assembly of accessory optic system circuitry critical field center, with narrower Gaussians simulating faster movement. Then, the for compensatory eye movements. Neuron 86, 971–984 (2015). Gaussian stimuli were passed through different ON and OFF luminance-response 11. Berson, D. M. in The senses: a comprehensive reference Vol. Vision 1 functions simulated as Naka-Rushton equations. (A. I. Basbaum ed) 491–519 (Elsevier, 2008). The parameters that varied in the model were the saturation of the luminance- 12. Schweigart, G., Mergner, T., Evdokimidis, I., Morand, S. & Becker, W. Gaze response function and the spike threshold. The saturation of the luminance-response stabilization by optokinetic reflex (OKR) and vestibulo-ocular reflex (VOR) function was only allowed to vary within a narrow range that matched the during active head rotation in man. Vision Res. 37, 1643–1652 (1997). physiological measurements obtained in cat visual cortex for light and dark stimuli 13. Jin, J. Z. et al. On and off domains of geniculate afferents in cat primary visual 15,23 presented on gray backgrounds . The luminance-response saturation had to be cortex. Nat. Neurosci. 11, 88–94 (2008). always larger for lights than darks (as observed experimentally) and had to be 14. Yeh, C. I., Xing, D. & Shapley, R. M. “Black” responses dominate macaque constrained within a narrow physiological range (1 to 2.5 for the exponent and 0.3 to primary visual cortex V1. J. Neurosci. 29, 11753–11760 (2009). 0.5 for the L 50 of the luminance-response function). The spike threshold was not 15. Kremkow, J. et al. Neuronal nonlinearity explains greater visual spatial resolution needed to simulate the reduction (or elimination) of light-dark latency differences as for darks than lights. Proc. Natl Acad. Sci. USA 111,3170–3175 (2014). velocity increased. However, using a spike threshold allowed a more precise match of 16. Chichilnisky, E. J. & Kalmar, R. S. Functional asymmetries in ON and OFF the data and it was needed to simulate the sharp drop in latency differences at 10 deg s ganglion cells of primate retina. J. Neurosci. 22, 2737–2747 (2002). −1. The rest of the parameters used in the model were fixed including the stimulus 17. Zaghloul, K. A., Boahen, K. & Demb, J. B. Different circuits for ON and OFF time course (Gaussian width matching stimulus velocity used in experiments) and retinal ganglion cells cause different contrast sensitivities. J. Neurosci. 23, R (20% lower for light than dark stimuli as measured experimentally). max 2645–2654 (2003). These parameters generated luminance-response functions with higher contrast 18. Leonhardt, A. et al. Asymmetry of Drosophila ON and OFF motion detectors sensitivity and saturation for ON than OFF pathways, as found in our experimental enhances real-world velocity estimation. Nat. Neurosci. 19, 706–715 (2016). measurements from cats, monkeys, and humans15,23. In turn, the different luminance- 19. Emran, F. et al. OFF ganglion cells cannot drive the optokinetic reflex in response functions made Gaussian responses wider and weaker for light than dark fi stimuli. The differences in light-dark latency were calculated as the difference between zebra sh. Proc. Natl Acad. Sci. USA 104, 19126–19131 (2007). the response time for lights and darks at half the maximum response to lights, as in the 20. Kleiner, M., Brainard, D. & Pelli, D. What’s new in Psychtoolbox3? European experimental measures. Below, we describe in more detail the equations for the Conference on Visual Perception, Arezzo, Italy (2007). different parts of the model and the parameters used in the simulation. 21. Li, X., Liang, Z., Kleiner, M. & Lu, Z. L. RTbox: a device for highly accurate The inputs of the model were Gaussian functions (stim) whose width matched the response time measurements. Behav. Res. Methods 42, 212–225 (2010). duration of the stimulus at different speeds. The shape of the Naka-Rushton function 22. Kremkow, J., Jin, J., Wang, Y. & Alonso, J. M. Principles underlying sensory was determined by the exponent (n) and the luminance that generated 50% of the map topography in primary visual cortex. Nature 533, 52–57 (2016). maximum response (L 50). The response magnitude was scaled by Rmax (Eq. 1) and the 23. Pons, C. et al. Neuronal mechanisms underlying differences in spatial spike threshold (th) was simulated as a power function. resolution between darks and lights in human vision. J. Vision 17, 5 (2017). stimn th 1 R Rmax n n ¼ L 50 stim ð Þ Acknowledgements þ This study was supported by NIH grants EY05253 (JMA) and EY027157 (RM).
The Rmax, L 50, and n were lower for ON than OFF pathways (Rmax: 80, L 50: 0.4, n: 1.4 for ON; Rmax: 100, L 50: 0.5, n: 2.3 for OFF) to make the saturation of the Author contributions luminance-response function more pronounced for ON than OFF pathways, G.L., Q.Z. and A.F. designed the psychophysical experiments. G.L. performed those consistently with previous measurements in cats, monkeys, and humans15,23. The spike threshold values for each speed were arbitrarily chosen with the only experiments. R.M. and J.A. performed the electrophysiological experiments. All authors requirement being that they had to increase with stimulus velocity (5 deg s−1: 1.0, analyzed and discussed the results. G.L., R.M., J.A. and A.F. wrote the manuscript. 10 deg s−1: 2.0, 16.15 deg s−1: 2.5, 30 deg s−1: 3.0, 60 deg s−1: 3.5). Both the reduction in the dark-light difference-latency with velocity and the convergence of Additional information the light-dark difference at zero could be simulated without the spike threshold. Supplementary information accompanies this paper at https://doi.org/10.1038/s42003- 018-0066-y. Code availability. Computer code used to analyze data in this study is available from https://github.com/AlanFreeman/anaMotion. Competing interests: The authors declare no competing interests.
Reprints and permission information is available online at http://npg.nature.com/ Data availability. The data that support the findings of this study are available from the corresponding author upon reasonable request. reprintsandpermissions/ Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in Received: 26 October 2017 Accepted: 3 May 2018 published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons References Attribution 4.0 International License, which permits use, sharing, 1. Joesch, M., Schnell, B., Raghu, S. V., Reiff, D. F. & Borst, A. ON and OFF adaptation, distribution and reproduction in any medium or format, as long as you give pathways in Drosophila motion vision. Nature 468, 300–304 (2010). appropriate credit to the original author(s) and the source, provide a link to the Creative 2. Dacey, D. M. & Petersen, M. R. Dendritic field size and morphology of midget Commons license, and indicate if changes were made. The images or other third party and parasol ganglion-cells of the human retina. Proc. Natl Acad. Sci. USA 89, material in this article are included in the article’s Creative Commons license, unless 9666–9670 (1992). indicated otherwise in a credit line to the material. If material is not included in the 3. Komban, S. J., Alonso, J. M. & Zaidi, Q. Darks are processed faster than lights. ’ J. Neurosci. 31, 8654–8658 (2011). article s Creative Commons license and your intended use is not permitted by statutory 4. Luo-Li, G., Alais, D. & Freeman, A. W. Orientation discrimination requires regulation or exceeds the permitted use, you will need to obtain permission directly from coactivation of on- and off-dominated visual channels. J. Vision 16, 18 (2016). the copyright holder. To view a copy of this license, visit http://creativecommons.org/ 5. Nam, J. H. & Chubb, C. Texture luminance judgments are approximately licenses/by/4.0/. veridical. Vision Res. 40, 1695–1709 (2000). 6. Komban, S. J. et al. Neuronal and perceptual differences in the temporal © The Author(s) 2018 processing of darks and lights. Neuron 82, 224–234 (2014).
COMMUNICATIONS BIOLOGY | (2018) 1:60 | DOI: 10.1038/s42003-018-0066-y | www.nature.com/commsbio 5
78
79 Chapter 6. Responses to light and dark moving stimuli
80
3
81 Chapter 6. Responses to light and dark moving stimuli
4
82
Chapter 7. Mechanisms underlying motion direction
selectivity
Abstract
Aims. The previous chapter investigated responses in cat primary visual cortex to moving light and dark bars (Luo-Li, Mazade, Zaidi, Alonso, & Freeman, 2018). It concluded that responses to light bars moving slowly in the preferred direction precede responses to dark bars. The model proposed to account for this observation was based on the differing contrast-response functions of on- and off-dominated neurons. Here I propose an alternative model. Simple cell receptive fields in primary visual cortex typically comprise spatially offset on- and off-subfields. I propose that the preferred direction of motion for a simple cell is from the on- to the off-subfield. A bar moving in the preferred direction would then meet the on- before the off-subfield and would evoke an earlier response if the bar were light than if it were dark. The aim was to test this hypothesis. Methods. The data used to test the hypothesis were multi- unit responses from cat primary visual cortex. These data were provided by Jose-
Manuel Alonso. Sparse noise data were used to plot receptive fields and the preferred directions were determined from responses to moving bars. Results. Preferred direction of motion was predicted by applying the hypothesis to the measured receptive fields. The predicted preferred direction was positively correlated with measured preferred direction for one dataset but uncorrelated for a second dataset.
Conclusion. The data provide only limited support for the hypothesis.
83 Chapter 7. Mechanisms underlying motion direction selectivity
Introduction
Motion perception is a vital component of the mammalian sensory repertoire. We use it to navigate through the environment, sense objects moving around us, and to communicate. All of these capabilities ultimately depend on motion-sensitive neurons in primary visual cortex. These neurons owe their sensitivity to direction selectivity: the response is stronger for one direction of stimulus motion than for all other directions. Sixty years after the discovery of direction-selective neurons (Hubel &
Wiesel, 1959) the mechanisms underlying direction selectivity are still obscure.
I aimed to test a relatively simple hypothesis for the preferred motion direction. The previous chapter showed that responses to light moving bars precede those to dark bars. I therefore propose that the preferred direction of a simple cell is from the on- subfield to the off-subfield. The data used to test the hypothesis come from the laboratory of Jose-Manuel Alonso; I thank him and his colleagues for providing me with access to these data. I first reanalyse the neuronal data in the previous chapter to find the timing of responses not only for the preferred direction of motion but also for the opposite direction. I then compute responses to a sparse noise stimulus, in which a square of light or dark is presented at a variety of locations to map out the receptive field. These responses are analysed to find the on- and off-subfield for each recorded channel, their spatial offset and therefore the direction from on- to off-subfield. For the same recording channels, I then examine responses to a bar moved in a variety of directions and find the preferred direction, and check whether there is correlation across channels between on-to-off direction and preferred direction.
84 Introduction
Methods
This section provides a brief summary of the methods used for the animal physiology experiments, the details of which can be found in Chapter 6.
Animal preparation
Anaesthetised adult male cats (n = 4) were used to investigate responses to brief static and moving stimuli. The animal was intubated, ventilated, and the pupils were dilated. Contact lenses with artificial pupils of 3 mm were used to focus the eyes on a monitor screen placed 57 cm away from the eyes. The skull and dura overlying cortical area 17 were removed. The animal’s vital signs were monitored and carefully maintained to ensure the optimal anaesthesia during the surgical preparation and recordings (Jin et al., 2008; Kremkow et al., 2016).
Electrophysiology recordings and data collection
Two 32-channel linear multielectrode arrays (100 !m inter-electrode distance,
Neuronexus), as shown in Figure 7.1A, were inserted into the cortex parallel with the surface and centred in layer 4. Spike recordings were filtered, sampled and collected by a computer using the Omniplex (Plexon) program.
85 Chapter 7. Mechanisms underlying motion direction selectivity
Figure 7.1. Data acquisition and visual stimuli. A. Two 32-channel linear arrays (Neuronexus) were inserted into the cortex as nearly as possible parallel with the surface and centred in layer 4. Recording channels were spaced 100 μm apart, and the distance between the two shanks was 300 μm. B. Sparse noise was used to map receptive fields. Squares of light or dark were presented briefly at a variety of locations. C. Moving bars were used to determine the preferred direction of motion. Bars were light
(240 cd/m2) or dark (0.27 cd/m2) and were presented on a grey background (120 cd/m2). The bars appeared on one side of the monitor and moved across the entire screen. Orientation and speed of the moving bars were varied.
86 Methods
Visual stimuli
The visual stimuli were generated using custom Matlab (Mathworks, Inc.) code in conjunction with the Psychtoolbox extension, and presented on a 24-inch LCD monitor. Figure 7.1B and C illustrate the stimuli used in the experiments. The purpose and data analysis of each stimulus type are described in the following sections.
Sparse noise mapping
On- and off- subfield responses were mapped using sparse noise stimuli. Brief square light and dark stimuli (Figure 7.1B) were delivered at a variety of locations across the visual field. Recording sites differ in their contrast sensitivity so contrast was set at maximum to make sure that low-sensitivity sites were adequately stimulated.
Analysis of sparse noise responses
The sparse noise square at each visual field location (�, �) was delivered many times.
Each action potential that occurred between 10 ms before and 250 ms after the square’s onset was assigned an interval, t, from stimulus onset. These intervals were compiled into a peri-stimulus time histogram �(�, �, �) , as illustrated in Figure 7.3. The hypothesis testing in this chapter requires robust representation of a recording channel’s receptive field. This requires, in turn, that the peri-stimulus time histogram be collapsed over time to obtain a purely spatial image. To do this, I used the chi- square statistic. For a specific interval, � , after square onset, the response is a two- dimensional array, �(� , �, �) of spike counts. This array is largely formless at the onset of the square (because of conduction delays between eye and cortex) and many milliseconds after onset (because any response eventually dies away). At intervening times, a subset of locations within the array differs significantly from other locations.
87 Chapter 7. Mechanisms underlying motion direction selectivity
A chi-square test of independence reveals these times, with an elevated chi-square, as illustrated in Figure 7.4..
I therefore calculated the chi-square, � (�), for each time, �, after onset and also the
value, � , at the 5% significance level. I then weighted each spatial array by the difference between its chi-square statistic and the significance level. This weighted array was then averaged over all times to obtain the receptive field:
∑ �(�)�(�, �, �) �(�, �) = ∑ �(�) where � is the non-negative weight
�(�) = max (0, � (�) − � )
Channels differ in the strength of their responses. Some channels, for example, are not driven by the stimulated eye. In measuring the relative location of on- and off- subfields, some channels will therefore be more informative than others. Plotting the on-to-off direction across channels will result in a noisy plot if this variability between channels is not taken into account. I down-weighted unresponsive channels as follows. � was calculated as a function of time for each channel, as described above.
The reliability of a channel’s data can be quantified by the maximum across time:
� , = max (� (�)) where � is the channel number. The weight assigned to a channel is then:
� = max (0, � , − � )
I then smoothed the weighted receptive field maps across neighbouring channels with a Gaussian weighting function:
88 Methods
� � (�, �) � , (�, �) = ∗ �(�, �) mean (� ) where
�(��) = � . ( ⁄ ) ,
�� is the difference between channel numbers, � is the standard deviation of the
Gaussian function and ∗ represents convolution.
Moving bar responses
The preferred direction/orientation of each cortical site was determined using moving light (240 cd/m2) and dark (0.27 cd/m2) bars (2.1 deg width) (Figure 7.1C) on a mid- grey background (120 cd/m2) as described in the previous chapter (Luo-Li et al., 2018).
Moving bars began at one side of the monitor and moved across the entire screen, with speeds of 5, 10, 15, 17.3, 30 and 60 deg/s.
Results
This chapter tests the hypothesis that the preferred motion direction for a simple cell is from its on-subfield to the off-subfield. I start by showing the data that led to this hypothesis.
Response latency
Figure 7.2 shows the response to a bar moving at 5 deg/s in the preferred direction.
This time course was obtained by averaging data from all channels in a dataset, using the methods described in Luo-Li et al. (2018). Four graphs are shown, the preferred direction (Figure 7.2A) and antipreferred direction (Figure 7.2B) for the left and right eye. This result is consistent with the conclusions in Luo-Li et al. (2018), in that
89 Chapter 7. Mechanisms underlying motion direction selectivity
responses to light moving bars lead responses to dark. The result is also suggestive: if the preferred stimulus reaches the on-subfield before the off-subfield, the response to a light bar will precede that to a dark bar. Conversely, for motion in the antipreferred direction, a light bar will reach the on-subfield after the off-subfield and its response will therefore lag the response to a dark bar. This prediction is borne out by the data in Figure 7.2B. I therefore adopt the hypothesis that the preferred motion direction for a simple cell is from its on-subfield to its off-subfield. This hypothesis is tested using receptive field data in what follows.
Figure 7.2. A. Response time course for motion in the preferred direction. A bar was moved in the preferred direction at 5 deg/s and responses were averaged across channels (Luo-Li et al., 2018). For stimulation through either eye, responses to a light bar precede those to a dark bar. B. Response time
90 Results course for motion in the antipreferred direction. The analysis here is the same as in the preceding panel except that the bar moved in the opposite direction. In this case, responses to dark bars precede those to light bars. Experiment 170711, dataset 1-2, channel 1-64.
Sparse noise
Response time course
The sparse noise stimulus presents a square of light or dark at a variety of spatial locations in the visual field. As described in the Methods section, a peristimulus time histogram of the response is compiled across presentations at a specific location, and the histograms are then compiled into spatial maps such as those shown in Figure 7.3.
The colour bar shows the coding of impulse rate by colour. The top left map, for example, shows the response at the onset of a dark sparse noise element at each spatial location. The response was obtained from channel 12 in dataset 1 of experiment
170711, as noted at the bottom of the figure. The figure shows maps for dark and light elements at three times. The top and bottom graphs show, as expected, that the map is largely formless at the onset of the element and 100 ms after the onset, where the response has died away. The middle row shows a time near the peak of the response, and the area of heightened impulse rate in this case shows the receptive field of the recorded channel. The response to a light element is weaker than the dark response, as expected of dark dominance in visual cortex.
91 Chapter 7. Mechanisms underlying motion direction selectivity
Figure 7.3. Sparse noise response. Each point in a map shows the impulse rate evoked at a specific visual field location at a specific time after sparse noise element onset. Impulse rate is coded by colour, as shown on the colour bar. Reponses to dark elements are shown on the left, responses to lights on the right. The three times after stimulus onset are 0, 37, and 100 ms. As expected, there is no significant response at stimulus onset and the response has died away at 100 ms. In between there is a response at one location, the receptive field for this channel. Light responses are weaker than dark, as expected of dark dominance in primary visual cortex. Experiment 170711, dataset 1, channel 12.
92 Sparse noise
My aim was to compare the spatial location of on-subfields with that of off-subfields, and I therefore wished to collapse the response across times into a single map. I used the chi-square statistic to do this. The chi-square test of independence checks whether the two variables in a table are statistically independent. In my case, chi-square takes a low value if the map is formless and a higher value if the response in one spatial dimension depends on the response in the other dimension, that is, if there is a
“hotspot”. Chi-square is shown as a function of time after element onset in Figure 7.4, and the chi-square 5% significance level is shown by the dashed line in part B of the figure. Responses are significant for only a brief interval around 37 ms after element onset, which is why this time was chosen for display in the previous figure. I therefore collapsed the response over time by weighting each map with the difference between its chi-square and the significance level and averaging the weighted maps across time.
93 Chapter 7. Mechanisms underlying motion direction selectivity
Figure 7.4. Response significance as a function of time. The horizontal axis is the time after onset of a sparse noise element. The vertical axis shows the chi-square statistic for a receptive field map such as those in Figure 7.3. Formally, the chi-square independence test checks whether the response in the vertical direction is independent of that in the horizontal direction. More colloquially, it tests whether there is a significant “hotspot”. A. Data for two channels. B. Data taken from A, but with an expanded time axis. Points above the significance level show times at which there is a receptive field hotspot.
Channel 12, for example shows significant values for both lights and darks at around 37 ms, which is the reason for showing that time in Figure 7.3. Experiment 170711, dataset 1, channel 12, 13.
94 Sparse noise
Receptive field map
The result is shown in Figure 7.5A, with the dark subfield shown at left and the light subfield at right. It is difficult to compare the spatial locations of the two subfields in this representation, so I calculated contours for each subfield and combined the contours into a single plot as shown in part B of the figure. The colour bars to the right of each plot show the impulse rates represented by each contour. The offset between subfields is now clearly visible; an expanded version of the plot is shown on the right.
For some channels, however, chi-square values were barely above significance at any times after stimulus onset. This was usually due to ocular dominance: the recorded neurons were poorly driven by the stimulated eye. The strength of a channel’s response can be conveniently quantified by the maximum chi-square across its temporal response. Further, the subfield offset can be smoothed across neighbouring channels by using that maximum. This is illustrated in Figure 7.6: unsmoothed data are shown on the left and smoothed data on the right. Smoothing was performed by weighting each channel’s map by its maximum chi-square and then averaging, with a
Gaussian profile, across the weighted maps. For the three channels shown, subfield offset is more consistent across channels for the smoothed case.
95 Chapter 7. Mechanisms underlying motion direction selectivity
Figure 7.5. Averaging over time. A. Sparse noise response maps were weighted by their chi-square statistics (less the significance level) and summed. The result, shown for channel 12, is independent of time. B. Contour plots were calculated from A so that light and dark responses could be shown together.
The plot on the right is an expanded version of that on the left. The spatial offset between on- and off- subfields is clear. Experiment 170711, dataset 1, channel 12.
96 Sparse noise
Figure 7.6. Smoothing across channels. Some channels have weak responses because, for example, they are not driven by the stimulated eye. Each channel was therefore weighted by its maximum value of chi-square and weighted maps were then averaged across channels with a Gaussian function. The standard deviation of the Gaussian was one channel. Left side: Receptive field maps, without smoothing, for channels 11 to 13. Right side: Smoothing results in greater consistency of the map from one channel to the next. Experiment 170711, dataset 1, channel 11-13.
97 Chapter 7. Mechanisms underlying motion direction selectivity
On-to-off direction
To quantify the on-to-off direction I fitted a model to each subfield. The model consisted of a Gaussian mass for which the spreads in the horizontal and vertical dimensions were uncorrelated: the long axis of the mass was therefore parallel with the horizontal or vertical axis of the map. The empirical and fitted data are shown for one channel in Figure 7.7A. On-to-off direction was then calculated as the direction of the line connecting the centres of the two subfield masses; the arrow on the right-side plot shows this direction. On-to-off direction is shown as a function of channel in
Figure 7.7B. Unsmoothed and smoothed data are shown on the left and right, respectively. Smoothing produces a much clearer trend from one channel to the next.
The left side of Figure 7.8A shows on-to-off direction for dataset 1, for which stimulation was through the left eye. The right side shows data for dataset 2, which represents the same channels stimulated through the right eye. There are gaps on the left side graph that are filled in on the right-side graph, and vice versa. This is as expected of ocular dominance columns in visual cortex. Figure 7.8B combines the left and right eye data onto one graph, with the two eyes represented by green and red respectively. Each graph shows data from one electrode shank. There is a trend in direction with channel location apart from a discontinuity in the latter where the dominant eye switches.
98 Sparse noise
Figure 7.7. Calculating on-to-off direction. A. A Gaussian volume was fitted to each subfield; the
Gaussian was uncorrelated and therefore had its long axis parallel to one of the graphical axes.
Empirical data for channel 12 are shown on the left, the fitted model on the right. The on-to-off direction was calculated from the vector running between the centres of the two Gaussian volumes, as illustrated.
B. On-to-off directions are shown for channels 1 to 32 with (right) and without (left) cross-channel smoothing. The darkness of a circle codes the significance: chi-square increases with darkness.
Smoothing produces a clearer trend across channels. Experiment 170711, dataset 1, channel 1-32.
99 Chapter 7. Mechanisms underlying motion direction selectivity
Figure 7.8. On-to-off directions for both eyes. A. Directions are shown for the left and right eyes separately. Channels 1 to 14 are driven by the left eye and 12 to 24 by the right. B. Left and right eye data are shown with green and red symbols, respectively. Saturation increases with chi-square. Both shanks are included as indicated by the labels above the graphs. The right graph shows a discontinuity in direction in the transition from one eye to the other. Experiment 170711, dataset 1-2, channel 1-32.
100 Sparse noise
Moving bar
Variation with direction and speed
The second stimulus in this study was a bar moved in directions evenly spaced across the 360° range. Figure 7.9 shows responses in one channel for six directions. The passage of the bar over the channel’s receptive field evokes an increase in impulse rate for both light and dark bars, shown in red and blue respectively. The increase is biggest for one direction which is therefore the preferred one. The bar also varied in speed, as shown in Figure 7.10, where the preferred direction was used. Responses tend to increase with speed. Figure 7.11 provides tuning curves for motion direction across all speeds and three channels. The larger peak in each tuning curve indicates the preferred direction and the other peak the antipreferred direction. Some channels, such as the first, show only a weak preference for direction.
Preferred direction
The preferred direction appeared to change little with speed, and data were therefore pooled across speeds. Figure 7.12A shows an average across speeds for two channels and Figure 7.12B indicates how preferred direction was calculated. A model was fitted to the unaveraged data; the model comprised the sum of two von Mises functions separated by 180°. I set the preferred direction equal to that of the larger peak. The direction selectivity index was calculated as the difference of the amplitudes of the two peaks divided by their sum. Figure 7.13 shows preferred direction as a function of channel number, using the same format as Figure 7.8. Unlike on-to-off direction, preferred direction shows many discontinuities across channel location. One reason for this is that all channels are shown, including those with weak direction selectivity.
101 Chapter 7. Mechanisms underlying motion direction selectivity
Figure 7.9. Responses to moving bars. Bars were moved in 16 directions evenly distributed over the full
360° range. Shown here are the responses of a single channel to the directions indicated by the insets.
Impulse rate increases as the bar passes over the channel’s receptive field. Responses to light and dark bars are shown in red and blue, respectively. One direction is preferred, in that it evokes the biggest responses. Experiment 170711, dataset 1, channel 7.
102 Moving bar
Figure 7.10. Responses to varying speeds. Bars were moved in the preferred direction at six speeds, one for each of graphs shown here. Responses become briefer, and amplitudes tend to increase, as speed increases. Experiment 170711, dataset 1, channel 7.
103 Chapter 7. Mechanisms underlying motion direction selectivity
Figure 7.11. Tuning to motion direction. The horizontal axis shows the direction in which the bar moved and the vertical axis the maximum response amplitude for that direction. Each graph shows these tuning curves for all six bar speeds, with dark bars in the left column and lights on the right. The top, middle and bottom rows provide responses for channels 7, 8 and 9 respectively. Two peaks appear in each graph, one for the preferred direction and the other for the antipreferred. The preferred direction is downward and to the right for all channels, but channels differ in their direction selectivity, as can be seen from the relative height of the two peaks. Experiment 170711, dataset 1, channel 7-9.
104 Moving bar
Figure 7.12. Calculation of preferred direction. A. The previous figure showed that preferred direction differed little between speeds. Tuning curves were therefore averaged across speeds, as shown here for channels 7 and 8. B. A model was fitted to the unaveraged data; the model and data are shown by line and circles, respectively. The model comprised the sum of two von Mises functions separated by 180°, and preferred direction was set equal to the direction of the higher peak. Experiment 170711, dataset 1, channel 7-8.
105 Chapter 7. Mechanisms underlying motion direction selectivity
Figure 7.13. Preferred direction across channels. A. The preferred motion direction is shown for each channel. The left and right-side graphs give preferred directions when the left and right eyes, respectively, are stimulated. The darkness of a circle increases with direction selectivity index. B. These graphs combine data for the two eyes, with left- and right-eye data shown in blue and red respectively.
Data from both electrode shanks are represented. There are numerous 180° discontinuities between neighbouring channels. Experiment 170711, dataset 1-2, channel 1-32.
106 Moving bar
Comparing preferred and on-to-off direction
Figure 7.14 shows preferred direction and on-to-off direction on the vertical and horizontal axes, respectively. Each circle represents one channel, but channels with a direction selectivity index less than 0.2 have been excluded from the plot. The two datasets are shown, one on each graph. The hypothesis being tested, that preferred direction is equal to on-to-off direction, requires that points fall along the line of identity. The alternate hypothesis, that preferred direction equals off-to-on direction, requires that points fall on diagonal lines 180° above and below the line of identity.
Accordingly, the plots have been banded with green and red areas representing, respectively, the predictions of the tested and alternate hypotheses. Given that the two areas each take up half of the total area, the tendency of the points to fall in the green area can be tested under the assumption of the binomial probability density in which the probability of success is 0.5: p-values are shown above each graph. Only the left graph shows a significant result, rejecting the alternate hypothesis at the 5% level. The other graph fails to reach significance. This, therefore, is limited evidence in favour of the tested hypothesis.
107 Chapter 7. Mechanisms underlying motion direction selectivity
Figure 7.14. Comparison between on-to-off and preferred direction. The horizontal axes show on-to-off directions obtained from sparse noise stimulation. On the vertical axes is preferred direction calculated from stimulation with a moving bar. Each circle represents one channel, and channels are only included if the direction selectivity index for the moving bar is greater than 0.2. A perfect correlation between the two direction measures requires that points fall on the identity line. An imperfect correlation would yield points in the green area. Points demonstrating a correlation between preferred direction and off- to-on direction would fall in the red area. The p-value above each graph shows the result of a binomial test that preferred direction correlates with on-to-off direction. There is support for this hypothesis in the left graph, representing shank 1. In the right graph, representing shank 2, the evidence provides no support for a correlation between preferred direction and either on-to-off or off-to-on direction.
Experiment 170711, dataset 1-2, channel 1-32.
Discussion
Hubel and Wiesel (1962) surmised that optimally oriented moving stimuli were very effective in evoking simple cell responses “probably because of the synergistic effects of leaving an inhibitory area and simultaneously entering an excitatory area”.
Goodwin et al. (1975) ruled out this explanation on the grounds that direction selectivity can be shown within a receptive field subfield. In the intervening decades,
108 Discussion there has been no clear demonstration of a link between receptive field and preferred direction. I have re-examined this issue, given the availability of simultaneous recordings of both receptive field and preferred direction in multiple contiguous cortical sites. Unfortunately, the evidence for my hypothesis, that the preferred direction is from the on- to the off-subfield, is weak. We must therefore look at alternatives.
A mechanism based on inhibition
Soon after the discovery of direction-selective neurons it was proposed that inhibition plays a major role in this selectivity. Goodwin et al. (1975) provided evidence that direction selectivity is due to inhibition spreading from the stimulus edge in the nonpreferred direction of motion. Emerson and Gerstein (1977), who also studied cat primary visual cortex, used bar stimuli placed at a variety of positions across the receptive field. Successive placement of neighbouring bars in the nonpreferred direction resulted in inhibition of the response to the second bar. The authors therefore inferred a short-lasting forward inhibition in the nonpreferred direction.
More recently, intracellular recording in cat primary visual cortex has shown clearer evidence that direction selectivity involves inhibition. Priebe and Ferster (2005) measured excitatory and inhibitory conductance in simple cells responding to moving gratings. They found that the two types of conductance were tuned in the same way but differed in their timing, for motion in the preferred and nonpreferred directions.
Summation of excitatory and inhibitory synaptic inputs was linear. They concluded that the nonlinearity in direction selectivity lay in the conversion from membrane potential to action potential rate.
109 Chapter 7. Mechanisms underlying motion direction selectivity
Priebe and Ferster did not find a source for the asymmetry in synaptic potentials evoked by the two motion directions. Another cat study, Cardin, Palmer, and
Contreras (2007) provides a clue as to the source. Cardin et al. showed that fast-spiking neurons, which are assumed to be inhibitory, have receptive fields similar to simple cells. Further, they are selective for stimulus orientation with a tuning only slightly worse than that of excitatory neurons. If inhibitory neurons inhabit a map of preferred orientations, we can imagine a simple cell to be surrounded by inhibitory neurons, each with its own preferred orientation. This means that a grating that stimulates the simple cell will stimulate only some of the surrounding inhibitory neurons as shown in Figure 7.15. In other words, inhibition will be spatially asymmetric and could provide the source of the asymmetry that Priebe and Ferster demonstrated. Whether this model of motion direction selectivity is realistic remains to be tested in the laboratory.
Figure 7.15. Motion direction selectivity model. Inhibitory neurons are orientation selective (Cardin et al., 2007). A grating capable of stimulating the simple cell will stimulate inhibitory neurons with similar orientation preference but won’t stimulate inhibitory neurons with differing preference. This will potentially produce spatially asymmetric inhibition and therefore motion direction selectivity.
110
Chapter 8. Discussion
Two parallel pathways – on-centre and off-centre – are very prominent in the subcortical visual system. It is only at their destination, primary visual cortex, that the two pathways converge. The fact that the two pathways are kept separate and largely independent through three processing stages (bipolar cells, ganglion cells, thalamic relay cells) before they interact at the cortex illustrates the very different properties before and after convergence. Subcortical visual signals are processed at the level of the retina and are relatively linear (Kaplan, Purpura, & Shapley, 1987). Multiple sources of evidence indicate that the convergence results in orientation selectivity (Jin,
Wang, Swadlow, et al., 2011; Reid & Alonso, 1995; Sedigh-Sarvestani et al., 2017) and the consequent emergence of feature-based visual signals. Further, the hyperpolarisation of geniculorecipient cells in the cortex (Anderson, Carandini, &
Ferster, 2000; Scholl, Tan, Corey, & Priebe, 2013) introduces an essential nonlinearity which sharpens the selectivity of cortical cells (Jagadeesh et al., 1997).
Despite the convergence of on- and off-signals at the cortex, the evidence of sign-based signals does not disappear in the cortex. It has been known for several decades that psychophysical responses are stronger to darker stimuli than to lighter stimuli (Bowen et al., 1989). This result was extended to neuronal recordings that showed that cortical responses to darks are stronger, faster, and occupy more cortical surface than light responses (Jin et al., 2008; Komban et al., 2014; Yeh et al., 2009). My aim in this thesis has been to explore the response asymmetry evoked by dark and light stimuli, and to explore the consequences of this asymmetry for visual perception. My findings include the following.
111 Chapter 8. Discussion
Chapter 5
• Subjects were presented briefly with a low contrast grating and indicated
whether it appeared to the left or right of centre. Accuracy was at chance levels
for a range of contrasts centred on zero. I attribute this plateau to the resting
hyperpolarisation found in simple cells (Anderson et al., 2000).
• Accuracy was better and responses were faster when gratings were dark relative
to the background than when they were lighter. This result extends previous
psychophysics by measuring both accuracy and reaction time in a single
experiment.
• These results were reproduced by a model in which off-dominated neurons have
higher contrast sensitivity than on-dominated.
• I decomposed gratings into light and dark components, presented them
asynchronously, and asked the subjects to judge grating orientation. Orientation
discrimination was best when light bars were presented before dark, suggesting
that performance is optimised when on- and off-inputs to cortex coincide in
time.
Chapter 6
• Instead of stationary stimuli I here used moving bars, which were either lighter
or darker than the background. Against expectations, human subjects
responded sooner and more accurately to light bars, provided that low speeds
were used.
• I analysed cortical responses from anesthetised cats presented with similar
stimuli. These recordings were supplied by a colleague’s laboratory. The results
112 Results
were similar in that responses to slowly moving light bars preceded those to
dark bars.
• These results were explicable in terms of a contrast-response function that
differs between on- and off-dominated cortical cells.
Chapter 7
• There is an alternative hypothesis for the results of Chapter 6: direction-selective
simple cells prefer motion from the on-subfield to the off-subfield. I tested this
hypothesis on measurements made in our colleague’s laboratory.
• Cells were stimulated with stationary squares of light or dark which allowed the
reconstruction of the receptive field. They were also stimulated with moving
light and dark bars, which provided preferred direction.
• Measured and predicted preferred direction were compared. A significant
positive correlation was found for one dataset, but there was no correlation in
the other dataset. There is therefore only limited support for the hypothesis.
Biological advantage of parallel subcortical pathways
This thesis is largely devoted to comparing on-centre and off-centre subcortical pathways, and their central connections. Why has the visual system evolved with parallel subcortical pathways rather than a single pathway that can signal both increments and decrements of light? After all, parallel pathways will approximately double the number of subcortical axons, requiring valuable neural real estate and energy supplies. I suggest that the answer lies in orientation selectivity. This property requires that responses from adjacent spatial locations be differenced, to form a bipartite receptive field, and that the selective cell be hyperpolarised at rest, to produce
113 Chapter 8. Discussion the iceberg effect. Both of these operations will reduce signal amplitude. If they occurred in the retina or thalamus, they would result in low signal-to-noise inputs to cortex. Instead, subcortical pathways in carnivores and primates have relatively high firing rates, resulting in largely linear signal processing (Kaplan et al., 1987), delaying the signal-reduction operations until cortex.
Biological advantage of light/dark asymmetries
Published data and the results in this thesis indicate that light/dark response asymmetries are ubiquitous. Do they serve a useful purpose? One possibility is motion direction selectivity. This selectivity requires sampling the visual image at two locations and a processing delay at one location relative to the other (Adelson &
Bergen, 1984). The convergence of on- and off-centre pathways onto simple cells would seem to be ideal for this purpose: the inputs are spatially offset and differ in their timing (Komban et al., 2014). Indeed a computational model (Hesam Shariati &
Freeman, 2012) showed that these properties did yield direction selectivity. The results in Chapter 7 argue against this possibility, however, because the preferred direction only weakly correlates with receptive field structure.
Another possible use for light/dark asymmetry is image stabilisation, which keeps images steady on the retina during self-motion and thereby improves visual acuity
(Schweigart, Mergner, Evdokimidis, Morand, & Becker, 1997). Stabilisation is subserved by a directionally selective on-centre ganglion cell that is found in several species (Barlow et al., 1964; Sun et al., 2015). The results in Chapter 7 reveal an on- dominated pathway that could serve to stabilise slow motion in the cat and human visual systems.
114 Results
One other possible biological advantage of light/dark asymmetry is its organising power in visual cortex. It has been shown that off-dominated cortical neurons have receptive fields that change steadily in retinotopy across the cortex surface, whereas on-dominated receptive fields vary in position around their off-neighbours (Kremkow et al., 2016; Lee et al., 2016). In other words, off-dominated neurons anchor retinotopy.
It could be, therefore, that having a dominant sign of response provides stable retinotopy.
Source of light/dark response asymmetries
Responses to stationary dark stimuli dominate responses to lights. Despite the substantial evidence supporting this observation, the source of the dominance is not clear. It has been suggested (Luo-Li et al., 2018) that this asymmetry originates in the synapse – between photoreceptors and bipolar cells – at which the on- and off-centre pathways start. The off-centre synapse is ionotropic whereas the on-centre synapse is metabotropic, making the former faster. Chichilnisky and Kalmar (2002), however, found that on-centre ganglion cells in the monkey retina were faster than their off- centre neighbours, throwing this hypothesis into doubt. Another possibility lies in ganglion cell numbers. Wässle, Boycott, and Illing (1981) found that off-centre beta ganglion cells in the cat retina outnumbered on-centre cells, but did not provide statistics for the significance of this imbalance. The source of the asymmetry is therefore still unknown.
115 Chapter 8. Discussion
Methodology
I have used two techniques in my psychophysical experiments that have proven critical in obtaining useful results. The first is the use of very low contrasts. Figure 5 in
Chapter 5, for example, shows contrast magnitudes less than 0.1, and a response plateau for magnitudes less than about 0.02. These low contrasts avoid complications such as response saturation, masking and afterimages that can make responses difficult to interpret. Conversely, low contrasts reveal essential nonlinearities such as the response plateau. It has also been shown that the most prevalent contrasts in the natural environment are low (Brady & Field, 2000), similar to the range I have used.
This gives extra relevance to results obtained at low contrast. A disadvantage with low-contrast experiments, however, is that calibration is critical. I have therefore been careful to measure the contrast-luminance function in detail, as shown in Figure 3.4.
A second technique essential to my experiments is simultaneous recording of accuracy and reaction time. Obtaining the same result with the two response variable strengthen the interpretation. Chapter 5, for example, shows that darks increase accuracy and decrease reaction time relative to lights, and that both response changes are consistent with higher contrast sensitivity in off- than in on-centre neurons. Further, I would not have been able to match psychophysics with neuronal data in Chapter 6 if I had only measured accuracy in the psychophysical experiments.
Contrast sensitivity
There is a very interesting difference between the psychometric curves I measured with stationary stimuli (Figure 7 in Chapter 5) and with moving stimuli (Figure 1 in
Chapter 6). Contrast sensitivity is higher for dark than light stimuli in the former case, but light stimuli have the advantage in the latter case. Why is it so? There is a possible
116 Results answer in the mechanism for motion direction selectivity described in the Discussion section of Chapter 7. According to that mechanism, cortical responses are modulated by the inhibitory neurons surrounding, and connecting to, each excitatory neuron. But the most prevalent neurons in primary visual cortex are off-dominated, meaning that a moving dark stimulus should evoke more inhibition than a light stimulus. This chain of reasoning suggests, then, that contrast sensitivity will be lower for moving dark stimuli. Whether this suggestion is true remains to be tested empirically.
117
References
Adelson, E. H., & Bergen, J. R. (1984). Spatiotemporal energy models for the perception
of motion. Journal of the Optical Society of America A-Optics & Image Science, 2(2),
284-299.
Anderson, J. S., Carandini, M., & Ferster, D. (2000). Orientation tuning of input
conductance, excitation, and inhibition in cat primary visual cortex. Journal of
Neurophysiology, 84(2), 909-926.
Barlow, H. B., Hill, R. M., & Levick, W. R. (1964). Retinal ganglion cells responding
selectively to direction and speed of image motion in the rabbit. Journal of
Physiology, 173, 377-407.
Barlow, H. B., & Levick, W. R. (1965). The mechanism of directionally selective units
in rabbit's retina. Journal of Physiology, 178, 477-504.
Behnia, R., Clark, D. A., Carter, A. G., Clandinin, T. R., & Desplan, C. (2014). Processing
properties of ON and OFF pathways for Drosophila motion detection. Nature,
512(7515), 427–430.
Borg-Graham, L. J., & Grzywacz, N. (1992). A model of the direction selectivity circuit
in retina: transformations by neurons singly and in concert. In Single Neuron
Computation. Edited by McKenna T, Davis J, Zornetzer SF. San Diego: Academic
Press, 347-375.
Bowen, R. W., Pokorny, J., & Smith, V. C. (1989). Sawtooth contrast sensitivity:
decrements have the edge. Vision Research, 29, 1501-1509.
Brady, N., & Field, D. J. (2000). Local contrast in natural images: normalisation and
coding efficiency. Perception, 29, 1041-1055.
118
Brainard, D. H. (1997). The Psychophysics Toolbox. Spatial Vision, 10(4), 433-436.
Briggman, K. L., Helmstaedter, M., & Denk, W. (2011). Wiring specificity in the
direction-selectivity circuit of the retina. Nature, 471, 183–188.
Cardin, J. A., Palmer, L. A., & Contreras, D. (2007). Stimulus feature selectivity in
excitatory and inhibitory neurons in primary visual cortex. Journal of
Neuroscience, 27(39), 10333–10344.
Casanova, C., Nordmann, J. P., Ohzawa, I., & Freeman, R. D. (1992). Direction
selectivity of cells in the cat's striate cortex: differences between bar and grating
stimuli. Vision Neuroscience, 9, 505-513.
Chichilnisky, E. J., & Kalmar, R. S. (2002). Functional asymmetries in on and off
ganglion cells of primate retina. The Journal of Neuroscience, 22(7), 2737–2747.
Cleland, B. G., & Levick, W. R. (1974). Properties of rarely encountered types of
ganglion cells in the cat's retina and an overall classification. Journal of
Physiology, 240, 457-492.
Dacey, D. M., & Petersen, M. R. (1992). Dendritic field size and morphology of midget
and parasol ganglion cells of the human retina. Proceeding of the National
Academy of Sciences, 89, 9666–9670.
Dannemiller, J. L., & Stephens, B. R. (2001). Asymmetries in contrast polarity
processing in young human infants. Journal of Vision, 1, 112-125.
Dean, A. F. (1981). The relationship between response amplitude and contrast for cat
striate cortical neurones. Journal of Physiology, 318, 413-427.
119
DeAngelis, G. C., Ohzawa, I., & Freeman, R. D. (1993a). Spatiotemporal organization
of simple-cell receptive fields in the cat's striate cortex. I. General characteristics
and postnatal development. Journal of Neurophysiology, 69(4), 1091-1117.
DeAngelis, G. C., Ohzawa, I., & Freeman, R. D. (1993b). Spatiotemporal organization
of simple-cell receptive fields in the cat's striate cortex. II. Linearity of temporal
and spatial summation. Journal of Neurophysiology, 69(4), 1118-1135.
DeAngelis, G. C., Ohzawa, I., & Freeman, R. D. (1995). Receptive-field dynamics in the
central visual pathways. Trends in Neurosciences, 18(10), 451-458.
Emerson, R. C., & Gerstein, G. L. (1977). Simple striate neurons in the cat. II.
Mechanisms underlying directional asymmetry and directional selectivity.
Euler, T., Detwiler, P. B., & Denk, W. (2002). Directionally selective calcium signals in
dendrites of starburst amacrine cells. Nature, 418, 845–852.
Felten, J., Hall, H., Jaumot, J., Tauler, R., de Juan, A., & Gorzsas, A. (2015). Vibrational
spectroscopic image analysis of biological material using multivariate curve
resolution-alternating least squares (MCR-ALS).(PROTOCOL)(Report). Nature
Protocols, 10(2), 217(224).
Foley, J. M., & Legge, G. E. (1981). Contrast detection and near-threshold
discrimination in human vision. Vision Research, 21(7), 1041-1053.
Fukuda, Y., & Stone, J. (1974). Retinal distribution and central projection of Y-, X-, and
W-cells of the cat's retina. Journal of Neurophysiology, 37, 749-772.
Gizzi, M. S., Katz, E., Schumer, R. A., & Movshon, J. A. (1990). Selectivity for
orientation and direction of motion of single neurons in cat striate and
extrastriate visual cortex. Journal of Neurophysiology, 63(6), 1529-1543.
120
Goodwin, A. W., Henry, G. H., & Bishop, P. O. (1975). Direction selectivity of simple
striate cells: Properties and mechanism. Journal of Neurophysiology, 38(6), 1500-
1523.
Hesam Shariati, N., & Freeman, A. W. (2012). A multi-stage model for fundamental
functional properties in primary visual cortex. PLoS One, 7(4):e34466(Epub 2012
Apr 9).
Hubel, D. H., & Wiesel, T. N. (1959). Receptive fields of single neurones in the cat's
striate cortex. Journal of Physiology, 148, 574-591.
Hubel, D. H., & Wiesel, T. N. (1961). Intergrative action in the cat's lateral geniculate
body. Journal of Physiology, 155, 385-398.
Hubel, D. H., & Wiesel, T. N. (1962). Receptive fields, binocular interaction and
functional architecture in the cat's visual cortex. Journal of Physiology, 160, 106-
154.
Hubel, D. H., & Wiesel, T. N. (1965). Binocular interactions in striate cortex of kittens
reared with artificial squint. Journal of Neurophysiology, 28, 1041-1059.
Hubel, D. H., & Wiesel, T. N. (1968). Receptive fields and functional architecture of
monkey striate cortex. Journal of Physiology, 195(1), 215-243.
Humphrey, A. L., & Saul, A. B. (2002). The emergence of direction selectivity in cat
primary visual cortex In B. R. Payne & A. Peters (Eds.), The cat primary visual
cortex. (New York/Academic Press) (pp. 343–386).
Jagadeesh, B., Wheat, H. S., Kontsevich, L. L., Tyler, C. W., & Ferster, D. (1997).
Direction selectivity of synaptic potentials in simple cells of the cat visual cortex.
Journal of Neurophysiology, 78(5), 2772-2789.
121
Jin, J., Wang, Y., Lashgari, R., Swadlow, H., & Alonso, J.-M. (2011). Faster
thalamocortical processing for dark than light visual targets. Journal of
Neuroscience, 31, 17471-17479.
Jin, J., Wang, Y., Swadlow, H., & Alonso, J. (2011). Population receptive fields of ON
and OFF thalamic inputs to an orientation column in visual cortex. Nature
Neuroscience, 14(2), 232-238.
Jin, J. Z., Weng, C., Yeh, C. I., Gordon, J. A., Ruthazer, E. S., Stryker, M. P., et al. (2008).
On and off domains of geniculate afferents in cat primary visual cortex. Nature
Neuroscience, 11(1), 88-94.
Joesch, M., Schnell, B., Raghu, S. V., Reiff, D. F., & Borst, A. (2010). ON and OFF
pathways in Drosophila motion vision. Nature, 468, 300–304.
Jones, J. P., & Palmer, L. A. (1987). The two-dimensional spatial structure of simple
receptive fields in cat striate cortex. Journal of Neurophysiology, 58, 1187-1211.
Kandel, E., Schwartz, J., & Jessell, T. E. (Eds.). (2000). Principles of neural science (4th ed).
New York ; London : McGraw-Hill, Health Professions Division.
Kaplan, E., Purpura, K., & Shapley, R. M. (1987). Contrast affects the transmission of
visual information through the mammalian lateral geniculate nucleus. Journal
of Physiology, 391, 267-288.
Kleiner, M., Brainard, D., & Pelli, D. G. (2007). What's new in Psychtoolbo—3. Paper
presented at the the European Conference.
Komban, S., Kremkow, J., Jin, J., Wang, Y., Lashgari, R., Li, X., et al. (2014). Neuronal
and perceptual differences in the temporal processing of darks and lights.
Neuron, 82(1), 224-234.
122
Komban, S. J., Alonso, J. M., & Zaidi, Q. (2011). Darks are processed faster than lights.
Journal of Neuroscience, 31(23), 8654-8658.
Krauskopf, J. (1980). Discrimination and detection of changes in luminance. Vision
Research, 20, 671-677.
Kremkow, J., Jin, J., Komban, S. J., Wang, Y., Lashgari, R., Li, X., et al. (2014). Neuronal
nonlinearity explains greater visual spatial resolution for darks than lights. The
National Academy of Sciences, 111, 3170-3175.
Kremkow, J., Jin, Y., Wang, Y., & Alonso, J. M. (2016). Principles underlying sensory
map topography in primary visual cortex. Nature, 533:, 52-57.
Lee, K. S., Huang, X., & Fitzpatrick, D. (2016). Topology of ON and OFF inputs in
visual cortex enables an invariant columnar architecture. Nature, 533(7601), 90-
94.
Li, X., Liang, Z., Kleiner, M., & Lu, Z. L. (2010). RTbox: A device for highly accurate
response time measurements. Behavior Research Methods, 42, 212-225.
Liu, K., & Yao, H. (2014). Contrast-dependent OFF-dominance in cat primary visual
cortex facilitates discrimination of stimuli with natural contrast statistics.
European Journal of Neuroscience, 39, 2060-2070.
Livingstone, M. (1998). Mechanisms of diretion selectivity in macaque V1. Neuron, 20,
509-526.
Luo-Li, G., Alais, D., & Freeman, A. W. (2016). Orientation discrimination requires
coactivation of on- and off-dominated visual channels. Journal of Vision, 16, 18
(2016).
123
Luo-Li, G., Mazade, R., Zaidi, Q., Alonso, J.-M., & Freeman, A. W. (2018). Motion
changes response balance between ON and OFF visual pathways.
Communications Biology, 1:60.
Martin, P. R., White, A. J., Goodchild, A. K., Wilder, H. D., & Sefton, A. E. (1997).
Evidence that blue-on cells are part of the third geniculocortical pathway in
primates. European Journal of Neuroscience, 9(7), 1536-1541.
Martinez, L. M., Wang, Q., Reid, R. C., Pillai, C., Alonso, J. M., Sommer, F. T., et al.
(2005). Receptive field structure varies with layer in the primary visual cortex.
Nature Neuroscience, 8(3), 372-379.
Mastronarde, D. N. (1987). Two classes of single-input X-cells in cat lateral geniculate
nucleus. II. Retinal inputs and the generation of receptive-field properties.
Journal of Neurophysiology, 57(2), 381-413.
Mauss, A. S., Vlasits, A., Borst, A., & Feller, M. (2017). Visual circuits for direction
selectivity. Annual Review of Neuroscience, 40, 211-230.
McLean, J., & Palmer, L. A. (1989). Contribution of linear spatiotemporal receptive
field structure to velocity selectivity of simple cells in area 17 of cat. Vision
Research, 29, 675–679.
Nachmias, J., & Sansbury, R. V. (1974). Grating contrast: discrimination may be better
than detection. Vision Research, 14(10), 1039-1042.
Peterson, M. R., Li, B., & Freeman, R. D. (2004). The derivation of direction selectivity
in the striate cortex. Journal of Neuroscience, 24(14), 3583-3591.
Priebe, N. J., & Ferster, D. (2005). Direction Selectivity of Excitation and Inhibition in
Simple Cells of the Cat Primary Visual Cortex. Neuron, 45, 133–145.
124
Priebe, N. J., Lampl, I., & Ferster, D. (2010). Mechanisms of direction selectivity in cat
primary visual cortex as revealed by visual adaptation. Journal of
Neurophysiology, 104(5), 2615-2623.
Purves, D., Augustine, G. J., Fitzpatrick, D., Hall, W. C., LaMantia, A. S., Mooney, R.
D., et al. (2017). Neuroscience 6th Edition: Sinauer Associates is an imprint of
Oxford University Press.
Reid, R. C., & Alonso, J. M. (1995). Specificity of monosynaptic connections from
thalamus to visual cortex. Nature, 378(6554), 281-284.
Reid, R. C., Soodak, R. E., & Shapley, R. M. (1987). Linear mechanisms of directional
selectivity in simple cells of cat striate cortex. Proceedings of the National Academy
of Science of the United States of America, 84, 8740-8744.
Scholl, B., Tan, A. Y., Corey, J., & Priebe, N. J. (2013). Emergence of orientation
selectivity in the Mammalian visual pathway. Journal of Neuroscience, 33(26),
10616-10624.
Schweigart, G., Mergner, T., Evdokimidis, I., Morand, S., & Becker, W. (1997). Gaze
stabilization by optokinetic reflex (OKR) and vestibulo-ocular reflex (VOR)
during active head rotation in man. Vision Research, 37, 1643–1652.
Sedigh-Sarvestani, M., Vigeland, L., Fernandez-Lamo, I., Taylor, M. M., Palmer, L. A.,
& Contreras, D. (2017). Intracellular, in vivo, dynamics of thalamocortical
synapses in visual cortex. The Journal of Neuroscience, 37(21), 5250 –5262.
Single, S., & Borst, A. (1998). Dendritic integration and its role in computing image
velocity. Science (New York, N. Y.), 281, 1848-1850.
Solomon, S. G., & Lennie, P. (2007). The machinery of colour vision. Natural Reviews
Neuroscience, 8(4), 276-286.
125
Sun, L. O., Brady, C. M., Cahill, H., Al-Khindi, T., Sakuta, H., Dhande, O. S., et al.
(2015). Functional assembly of accessory optic system circuitry critical for
compensatory eye movements. Neuron, 86(4), 971-984.
Tan, A. Y., Chen, Y., Scholl, B., Seidemann, E., & Priebe, N. J. (2014). Sensory
stimulation shifts visual cortex from synchronous to asynchronous states.
Nature, 509, 226-229.
Tanaka, K. (1983). Cross-correlation analysis of geniculostriate neuronal relationships
in cats. Journal of Neurophysiology, 49(6), 1303-1318.
Vaney, D. I., & Taylor, R. W. (2002). Direction selectivity in the retina. Current Opinion
in Neurobiology, 12 (4), 405–410.
Vlasits, A. L., Morrie, R. D., Tran-Van-Minh, A., Bleckert, A., Gainer, C. F., DiGregorio,
D. A., et al. (2016). A role for synaptic input distribution in a dendritic
computation of motion direction in the retina. Neuron, 89, 1317–1330.
Wässle, H., Boycott, B. B., & Illing, R.-B. (1981). Morphology and mosaic of on- and off-
beta cells in the cat retina and some functional considerations. Proceedings of the
Royal Society of London B, 212, 177-195.
Watson, A. B., & Ahumada, A. J. (1985). Model of human visual-motion sensing.
Journal of the Optical Society of America A-Optics & Image Science, 2(2), 322-341.
Xing, D., Yeh, C. I., & Shapley, R. M. (2010). Generation of black-dominant rsponses in
V1 cortex. Journal of Neuroscience, 30, 13504-13512.
Yamane, S., Maske, R., & Bishop, P. O. (1985). Direction selectivity of simple cells in
cat striate cortex to moving light bars. Experimental Brain Research, 57, 523-536.
126
Yeh, C. I., Xing, D., & Shapley, R. M. (2009). “Black” responses dominate macaque
primary visual cortex V1. Journal of Neuroscience, 29, 11753–11760.
Yoshida, K., Watanabe, D., Ishikane, H., Tachibana, M., Pastan, I., & Nakanishi, S.
(2001). A key role of starburst amacrine cells in originating retinal directional
selectivity and optokinetic eye movement. Neuron, 30, 771-780.
Zemon, V., Gordon, J., & Welch, J. (1988). Asymmetries in ON and OFF visual
pathways of humans revealed using contrast-evoked cortical potentials. Visual
Neuroscience, 1, 145-150.
127
Publications
Journal paper
Luo-Li, G., Alais, D., & Freeman, A. W. (2016). Orientation discrimination requires coactivation of on- and off-dominated visual channels. Journal of Vision, 16, 18 (2016).
Luo-Li, G., Mazade, R., Zaidi, Q., Alonso, J.-M., & Freeman, A. W. (2018). Motion changes response balance between ON and OFF visual pathways. Communications
Biology, 1:60. doi: 10.1038/s42003-018-0066-y
Abstracts and conference proceedings
Luo-Li, G., Alais, D., & Freeman, A. W. (2013a). Measuring perception without introspection. Paper presented at the Neuroscience Conference, San Diego California.
Luo-Li, G., Alais, D., & Freeman, A. W. (2013b). Objective measurement of binocular rivalry suppression. Paper presented at the UWS sensory neuroscience symposium,
UWS Campbelltown.
Luo-Li, G., Alais, D., & Freeman, A. W. (2014a). Behavioural responses shaped by single- neuron properties. Paper presented at the 5th Bosch Young Investigator Neuroscience
Symposium 2014, The University of Sydney.
Luo-Li, G., Alais, D., & Freeman, A. W. (2014). Psychophysical responses to weak visual stimuli reveal neuron-like properties. Paper presented at the 5th UWS Sensory
Neuroscience Symposium, The University of Western Sydney. Australia.
Freeman, A. W., Luo-Li, G., & Alias, D. (2014). Behavioural indicators of perceptual state.
Paper presented at the Australasian Neuroscience Society, Adelaide Australia.
128
Luo-Li, G., Alais, D., & Freeman, A. W. (2015). Human responses to weak visual stimuli reveal neuron-like properties. Paper presented at the Australasian Experimental
Psychology Conference 2015, School of Psychology, University of Sydney.
Freeman, A. W., Luo-Li, G., & Alais, D. P. (2015). Orientation discrimination depends on co-activation of on- and off dominated visual channels. Paper presented at the Monash
Symposium 2015, Melbourne.
Freeman, A. W., Luo-Li, G., & Alais, D. (2015). The psychophysicist’s microscope: weak stimuli reveal neuron-like properties. Paper presented at the Vision Sciences Society 2015,
Florida, USA.
Freeman, A. W., Luo-Li, G., & Alais, D. (2015). The psychophysicist's microelectrode: weak visual stimuli reveal neuron-like response properties. Paper presented at the Society for
Neuroscience Conference 2015, Chicago, IL.
Luo-Li, Freeman, & Alais. (2016). Light / dark response asymmetries for static and moving visual stimuli. Paper presented at the Bosch Institute Young Investigators Symposium
2016, The University of Sydney.
Freeman, A. W., Luo-Li, G., & Alais, D. P. (2016). Orientation discrimination depends on co-activation of on- and off-centre visual channels. Paper presented at the Vision Sciences
Society 2015, Florida, USA.
Freeman, Luo-Li, & Alais. (2016). Orientation discrimination depends on co-activation of on- and off-dominated visual channels. Paper presented at the 46th Society for
Neuroscience Conference 2016, San Diego, California.
Luo-Li, G. Q., Mazade, R., Alonso, J.-M., & Freeman, A. W. (2017). The light and dark of visual motion sensitivity. Paper presented at the 37th Annual Scientific Meeting of the
Australasian Neuroscience Society, The International Convention Centre Sydney.
129
Freeman, A. W., & Luo-Li, G. (2017). Light/dark asymmetries in motion perception. Paper presented at the 47th Society for Neuroscience Conference 2017, Washington, DC,
USA.
Freeman, A. W., Luo-Li, G. Q., Mazade, R., & Alonso, J.-M. (2018). Motion switches the leading visual response from off to on. Paper presented at the 45th Annual Conference of
The Australasian Society for Experimental Psychology, Hobart, Tasmania.
130