Skip to main content

Following a path of breadcrumbs: How fish recognize landmarks during navigation

Periodic Reporting for period 1 - FISHNAV (Following a path of breadcrumbs: How fish recognize landmarks during navigation)

Reporting period: 2015-08-31 to 2017-08-30

The Action “Following a path of breadcrumbs: How fish recognise landmarks during navigation” uses the natural navigational behaviour of fish to explore how they process complex visual information given their particular brain anatomy. All animals that rely on vision face similar problems when detecting, recognising, and responding to visual information, yet how different species have adapted to solve these problems can vary considerably. One common challenge is in the recognition of three-dimensional (3D) objects. The appearance of a single object can vary dramatically based on its relative orientation to the viewer, including shading, lighting, shape and colour. As has been demonstrated by computer vision programs, achieving a high degree of accuracy is a computationally difficult task. Yet object recognition underpins a range of important animal behaviours, including predator and prey identification, sexual selection and navigation, making recognition mistakes costly. While mammals, and increasingly computers, can devote considerable processing power to the task, animals with different brain anatomy may have to rely on simpler recognition methods.

Recognition of 3D object can be completely flexible, whereby an object can be recognised regardless of the viewing position of the observer (view-invariant), partly flexible in which recognition is somewhat limited by viewing position (view-dependent), or completely inflexible and recognition can only occur when the observed view matches that of a learned template (view-dependent template matching). When presented with new objects, humans have a flexible, but view-dependent system and there are limits to how much the appearance of an object can change before recognition breaks down. Some insects on the other hand, appear to use a template matching system. Using template matching, a large number of snapshots of an image would need to be stored in order for the observer to recognise the object under different conditions. This would require a significant memory capacity. Yet insects have demonstrated that they can reliably recognise objects for important tasks such as navigation. To achieve this, insects appear to reduce the number of snapshots required when navigating using a behavioural adaptation called ‘active vision,’ in which they follow previously learned paths between landmarks and reduce the number of views that they actually encounter.

The objectives of this Marie Skłodowska Curie Action (MSCA) has been to determine a) whether fish have a flexible recognition system, and b) if they can use behavioural adaptations such as active vision to reduce the processing burden of the task. Fish were used as they lack a cortex, the area of the brain associated with complex mammalian behaviour, yet they have also demonstrated other behaviours typically associated with advanced processing abilities (e.g. social learning, numeracy). This project used behavioural experiments with a species of coral reef fish, Rhinecanthus aculeatus, to test the two primary research questions. The long-term impact of this fundamental research is to improve our understanding of the brain; particularly how brain architecture influences behavioural capabilities. This project may also inform computer vision systems as it suggests there are methods of recognition that don’t require extensive processing capacity.
Research questions in this project were addressed via two work packages (WP). WP1 tested to what extent fish could discriminate complex rotated objects. It was found that they could not only discriminate complex images (human faces) but they could continue to recognise them when rotated, despite never previously seeing different orientations. However, fish became less accurate and slower to make decisions the more it was rotated. This led to a study examining how fish respond when recognition breaks down, and it was found that they can use alternate sources of information when their main source becomes difficult to distinguish. This work resulted in two research publications, one conference publication, and one invited talk. At least one more manuscript is in preparation.

WP2 explored the movement trajectories of fish in relation to local objects. Although the use of active vision was not explicitly tested, it was found that fish do alter their movement trajectories based on the visibility of targets and that they are far less efficient at locating food when visibility is poor. This work has resulted in one submitted manuscript and is expected to produce further conference presentations and manuscripts, including one detailing new analysis methods.

A further four publications were produced during this project through collaborations, and two review articles, as part of the career development plan of the Fellow. The Fellow undertook training in programming as part of this project, which has led to the development of a follow-on project examining the movement behaviour of fish in their natural environment and the development of two state-of-the-art fish tracking and terrain mapping systems. Throughout this project, the Fellow has contributed to the transfer of knowledge through outreach with the general public, teaching, and advising of students.
This MSCA has produced new information about object recognition in non-cortical brains and indicate that similar recognition behaviour can be produced without a specialised visual cortex. These results are expect to be of relevance to the fields of neuroscience, animal behaviour, vision science, and engineering. Although external interruptions limited the ability to explicitly test active vision, novel methods of tracking and analysis of animal movements underwater were developed. These methodological advancements are particularly important as they can be used to analyse a wide range of aquatic species and be applied to different biological questions.

This project has also resulted in two review papers which describe specialisations in the visual system of a species of fish with a unique hunting strategy, and what is currently known about abstract concept learning in fish. Both of these reviews not only synthesise information from a large number of other studies, but detail research directions for future work. An additional four papers were published on related topics as part of international collaborations. These studies add to our knowledge of fish visual behaviour and visual signal processing.

The impact of this project was increased through sharing results within the academic community as well as the general public and students. During this MSCA, the Fellow captured the interest of the general public with work showing that fish could learn to differentiate pictures of human faces. The Fellow participated in over 30 interviews with national and international media sources and results were discussed in over 100 outlets in print, TV, and radio. This news reached a range of audiences from all over the world including the U.K. Canada, Australia, U.S. Ireland, India, Germany, Russia, and many more. The impacts of this coverage are likely to be long lasting as it can foster interest and empathy in science and nature in the general public. The Fellow also communicated more directly with students through public lectures, personal video chats, and the University of Oxfords Pathways Program. A website has been created to provide non-specialist explanations of research.
Study species, Rhinecanthus aculeatus