ORA (Round 5): The active observer

Lead Research Organisation: CARDIFF UNIVERSITY
Department Name: Sch of Psychology

Abstract

Every time we move, the image of the world at the back of the eye changes. Despite this, our perception is of an unchanging world. How does the brain translate a continually changing image into a percept of a stable, stationary, rigid world? Does the brain use a map of the external environment (an "allocentric map") and the position of the observer within it, built up over time, to underpin the perception of stability? Does the brain continually update a map of where scene objects are relative to the observer (an "egocentric map"; e.g. there is an object straight ahead of me, if I walk forward I should expect it to get closer to me)? Does the brain not create a map but just divide up the image motion into that which is likely due to movement of the observer (and which can consequently be ignored) and that which is likely due to objects moving within the scene (which become a focus of attention)? The hypothesis that underpins this research project is that it is not a single one of these mechanisms that underpins perceptual stability, but that all of them, their contribution dependent on the task being performed by the observer. In some cases the task will require a fast estimate to support an ongoing action which might favour one mechanism, on another task, where timing is not so critical, a slower, but more accurate, mechanism might be more appropriate. This collaborative project, which combines complementary expertise in Psychology, Movement Sciences, and Computing from Germany, The Netherlands and the United Kingdom, and importantly, researchers that start from different theoretical perspectives, will test this hypothesis. We will study a diverse series of tasks that present a range of challenges to the moving observer. We will make use of various innovative experimental paradigms that exploit recent technological advances such as virtual reality combined with simultaneous motion tracking. Understanding where and how different mechanisms of perceptual stability play a role advances not only our scientific understanding, but also has the potential to inform industry as well as medicine about the circumstances in which disorientation or nausea in real or virtual environments can be minimised.

Planned Impact

Industrial impact
Perceptual stability is a key problem in the development of artificial virtual environments, as is illustrated in this recent account of the problems that people have moving through virtual environments (https://blog.google/products/google-vr/daydream-labs-locomotion-vr/). Members of the team have established contacts with industrial groups (e.g. Oculus, Sony) working on these problems and have a long track record of working in or with industry (Hewlett-Packard, Nissan). A better understanding of the different possible visual representations and when we use them is als likely to make systems visually more intuitive, and therefore easier to use (as has been demonstrated for haptics: Mugge et al., 2016).
A 4 day "impact" residential workshop will be held in the UK in the second year of the project. This is modelled on a similar workshop run last year with Microsoft Hololens and HP labs participants.

Clinical impact
An understanding of the underlying mechanisms of perceptual stability has important implications for several clinical conditions of visual vertigo and nystagmus. For example, we will share our findings with Prof. Jelte Bos, an expert on motion sickness who advises on motion sickness in many contexts, including space travel, seasickness, entertainment (3D movies) and self-driving cars. The UK PI is also collaborating with clinical and academic colleagues on understanding visual vertigo, a clinical condition in which visual motion is perceived as overwhelming in some situations (http://psych.cf.ac.uk/engagementimpact/visualvertigo/ , http://www.bbc.co.uk/news/uk-wales-south-east-wales-38715719). Any findings with clinical relevance will be presented at a suitable UK based meeting.

Popular interest
The topics covered in this proposal will likely raise public interest. Perception and action in real and virtual environments attracts attention of a broad audience and can be comprehensible even for laypeople. Likewise these topics are well received in popular science magazines, national newspapers and television programmes. The 3 PIs have contributed to such public outreach in the past, e.g. Gehirn und Geist (German PI), BBC Wales Today and BBC Points West (UK PI). Moreover, yearly science fairs organized by the universities and local museums (e.g. Cardiff University's Brain Games, http://sites.cardiff.ac.uk/cubric/public-outreach/brain-games which attracted almost 4000 people last year) provide an excellent opportunity to reach a broad audience. We will make use of such opportunities to communicate our scientific findings to the non-academic community.

Training Highly Skilled Researchers
We will train post-doctoral fellows who will have substantial input into the organization and design of the project. These postdoctoral fellows should gain the skills necessary to run their own laboratories and translate the techniques and findings into applied settings.
 
Description This ESRC project was part of a larger ORA project with collaborators in Amsterdam (Netherlands) and Giessen (Germany) who were funded through their national agencies, NWO and DFG respectively. The aim of the project was to examine how the human brain deals with an ever changing image on the back of the eye, how it is that despite the continually changing retinal image we perceive stale and rigid scene. This is a problem that has intrigued scientists and philosophers since the earliest times. Across the three sites we explored a number of potential solutions, the use of "egocentric representations" (that means knowledge of where things are relative to your body, e.g. the ball is to the right of my foot), "allocentric representations" (that means knowledge of where things are relative to the surrounding scene or other things, e.g. the ball is on the floor, to the right of the door), "compensation" for action (or predicting the visual consequences of action), and perception of action relevant information on the fly.

Three lines of work were planned at Cardiff, the examination of "Action Gist" (investigating how quickly information for action could be acquired), "Predicting the visual consequences of observer translation" (investigating mechanisms for updating visual representations of objects in the scene to compensate for movement, e.g. a forward step, of the observer), "Use of a stable allocentric representation in walking" (investigating whether we use a representation of our location relative to the scene as we walk towards target objects).

The first has proven to be a fruitful topic. We have quantified how quickly information for guiding a grasp and for guiding a step is picked up by observers. Across a series of experiments we have probed deeply into this question and discovered that humans can pick up this information remarkably quickly (<30ms). This opens the possibility that action does not require the acquisition and updating of an internal representation of the scene and opportunities for interaction with objects within it, but that the necessary information can be picked up very rapidly on the fly.

For the second we planned to explore visual perception as participants physically moved forward. The constraints of the pandemic forced a change in approach for the core work in Cardiff. Instead of having observers physically move in laboratory experiments we created experimental displays that simulated self-movement and that could be run by isolated participants, in particular the research team who would be prepared to generate very large amounts of data on their own. We found evidence that the visual system is "tuned for motion". That the first, lowest level, processes that detect objects are based on what are termed "directionally selective units", what that means is that the low level mechanisms that integrate signals over time on the retina are are effectively motion detectors. This means the visual system is built in such a way that it requires no additional mechanism to deal with the fact the eye is never still, it is always, for one reason or another, moving. A second strand to this work was conducted in Giessen as a collaboration between the sites. We explored how well observers could predict the visual consequences of taking a step forward (the most basic and common action humans perform). We found that they are very poor. This raises important questions about the timescales over which the brain may predict the visual consequences of action.

The third did not prove to be practical to pursue due to the constraints of the pandemic, it required testing, in an enclosed space, of many participants, each one requiring close interaction as sensors were fitted. Therefore a decision was made to instead explore perception by the active observer from a different perspective, through use of neuroimaging data. By necessity, brain scanners require participants to remain still and so can be argued to tell us little about vision of the active observer. However an interesting approach is to ask observers to watch films or "movies" while laid in a scanner. Movies are a great stimulus, they provide naturalistic visual input with rich scenes, moving people, movements of the viewpoint (camera) through space, etc. When the movies are Hollywood action movies they are also very effective stimuli for holding observers' attention and controlling their gaze. We made use of some data collected 7-8 years earlier by an EPSRC funded PhD student. The traditional view of the visual perception is that it involves two separate systems or "pathways", a "dorsal" pathway that is primarily concerned with the visual guidance of action, and a "ventral" pathway that is primarily concerned with visual perception (recognising objects etc). At least four groups have independently proposed that there is a third pathway but they disagree as to the location and purpose of the pathway. Using a technique called "Independent Component Analysis" applied to data collected while participants watched an action movie, which puts the brain into an active state, we have been able to isolate a third functional pathway that is not present when the brain is in a passive state.

At the Dutch and German sites much progress was made on understanding the role of "egocentric" representations in the guidance of action and the storage of information about object locations. Again the pandemic impacted on plans to collaborate in person but three pieces of joint work, between the Cardiff and Amsterdam and Cardiff and Giessen teams were conducted, the first was the study mentioned about on predicting the visual consequences of action., the other two were on the use of egocentric reference frames in interceptive actions.
Exploitation Route All three lines of work have the potential to be taken forward and used. The "action gist" findings point to a different way to think about the visual guidance of action, that it need not rely on internal models of the environment that it can be based on information that is processed very rapidly on the fly. The two strands of the work on "Predicting the visual consequences of action" raise questions about the timescale over which predictive processing can occur - over the order of 10s of milliseconds through directionally selective units, but not over the order of seconds, or the duration it takes to take a single step. The last strand of work on visual pathways raises supports the idea that vision involves more than two pathways and points to a way to study natural, or ecological vision in a brain scanner.

point to the idea that low-level compensation for movement of the eye is built in from the lowest level of visual processing and it suggest that "direction selectivity" (motion processing) is core to perception. The second strand, the finding that we cannot predict the visual consequences of a step forward suggests fundamental limits on predictive processing has its limits, although it may occur over the the order of 10s or 100s of milliseconds, there is no evidence that it occurs over the interval of a second or so, even during what must be the most fundamental human action, taking a step. This may lead others to consider the limits of predictive processing.
Sectors Aerospace

Defence and Marine

Education

 
Title Spatial and Temporal Visual Integration 
Description Spatial and temporal integration of retinal signals. Data from a psychophysical detection task. 
Type Of Material Database/Collection of data 
Year Produced 2024 
Provided To Others? Yes  
Impact Publications in prep. 
 
Description High Phi 
Organisation University of Paris - Descartes
Country France 
Sector Academic/University 
PI Contribution Study design, stimulus generation, data collection, data analysis.
Collaborator Contribution Study design, stimulus generation, data collection, data analysis.
Impact Ongoing data collection.
Start Year 2020
 
Description Integration of signals over time and space 
Organisation Cardiff University
Department School of Optometry and Vision Sciences
Country United Kingdom 
Sector Academic/University 
PI Contribution Design and set up of experiments. Data collection Data processing Interpretation of data
Collaborator Contribution Design of experiments Interpretation of data
Impact Data deposited at reshare.
Start Year 2021
 
Description Motion silencing 
Organisation York University Toronto
Country Canada 
Sector Academic/University 
PI Contribution In a collaborative project I designed and ran an experiment looking at motion silencing in the presence of full-field optic flow. In collaboration I analysed the data and presented it at ECVP.
Collaborator Contribution My collaborators at York contributed to the design, execution, analysis and write-up stages of the project.
Impact Presentation at ECVP. [insert REF]
Start Year 2016
 
Description Neural processing of naturalistic visual input 
Organisation University of Auckland
Country New Zealand 
Sector Academic/University 
PI Contribution Joint effort to explore the neural architecture of the cortical visual system during eye movements.
Collaborator Contribution Joint effort to explore the neural architecture of the cortical visual system during eye movements.
Impact Tangtartharakul, G., Morgan, C. A., Rushton, S. K., & Schwarzkopf, D. S. (2023). Retinotopic connectivity maps of human visual cortex with unconstrained eye movements , Human Brain Mapping, 44, 5221-5237
Start Year 2021
 
Description Neural processing of naturalistic visual input 
Organisation University of Auckland
Country New Zealand 
Sector Academic/University 
PI Contribution Joint effort to explore the neural architecture of the cortical visual system during eye movements.
Collaborator Contribution Joint effort to explore the neural architecture of the cortical visual system during eye movements.
Impact Tangtartharakul, G., Morgan, C. A., Rushton, S. K., & Schwarzkopf, D. S. (2023). Retinotopic connectivity maps of human visual cortex with unconstrained eye movements , Human Brain Mapping, 44, 5221-5237
Start Year 2021
 
Description Visual Vertigo 
Organisation Cardiff and Vale University Health Board
Country United Kingdom 
Sector Public 
PI Contribution Along with internal collaborators (Sumner) and clinical collaborators (Derry and Rajenderkumar) at the University Hospital Wales, Cardiff I have begun to look at Visual Vertigo. This is a problem with motion perception (sometimes called "supermarket syndrome" or "visual dependency") that can arise following damage to the vestibular system. In collaboration we have collected preliminary data on the relationship between symptoms of visual vertigo and visual processing in the student population. Next we will examine clinical patients.
Collaborator Contribution The partners have collaborated with us on data collection.
Impact Presentation at ECVP (Powell et al) in 2016.
Start Year 2015