Progress in Neurobiology 68 (2003) 409–437 Fundamental mechanisms of visual motion detection: models, cells and functions C.W.G. Clifford a,∗, M.R. Ibbotson b,1 a Colour, Form and Motion Laboratory, Visual Perception Unit, School of Psychology, The University of Sydney, Sydney 2006, NSW, Australia b Centre for Visual Sciences, Research School of Biological Sciences, Australian National University, Canberra 2601, ACT, Australia Received 8 May 2002; accepted 12 November 2002 Abstract Taking a comparative approach, data from a range of visual species are discussed in the context of ideas about mechanisms of motion detection. The cellular basis of motion detection in the vertebrate retina, sub-cortical structures and visual cortex is reviewed alongside that of the insect optic lobes. Special care is taken to relate concepts from theoretical models to the neural circuitry in biological systems. Motion detection involves spatiotemporal pre-filters, temporal delay filters and non-linear interactions. A number of different types of non-linear mechanism such as facilitation, inhibition and division have been proposed to underlie direction selectivity. The resulting direction-selective mechanisms can be combined to produce speed-tuned motion detectors. Motion detection is a dynamic process with adaptation as a fundamental property. The behavior of adaptive mechanisms in motion detection is discussed, focusing on the informational basis of motion adaptation, its phenomenology in human vision, and its cellular basis. The question of whether motion adaptation serves a function or is simply the result of neural fatigue is critically addressed. Crown Copyright © 2003 Published by Elsevier Science Ltd. All rights reserved. Contents 1. Introduction ............................................................................... 410 2. General motion detector mechanisms ........................................................ 410 2.1. Fundamentals of motion detection ...................................................... 410 2.2. Pre-filtering ........................................................................... 410 2.2.1. On- and Off-channels ........................................................... 410 2.2.2. Temporal characteristics of pre-filters ............................................. 412 2.3. Temporal delay filtering................................................................ 413 2.4. Non-linear interactions................................................................. 415 2.4.1. Facilitation ..................................................................... 416 2.4.2. Inhibition ...................................................................... 417 2.4.3. Speed-tuned motion detectors .................................................... 418 3. Evidence for the cellular mechanisms of motion detection..................................... 419 3.1. Retinal motion detectors in vertebrates .................................................. 419 3.2. Sub-cortical motion processing ......................................................... 421 3.3. Cortical motion processing ............................................................. 422 3.4. Motion detectors in insect optic lobes ................................................... 424 Abbreviations: AOS, accessory optic system; APB, 2-amino-4-phosphonobutyric acid; DAE, direction aftereffect; DS, direction selective; DTN, dorsal terminal nucleus; fMRI, functional magnetic resonance imaging; GABA, ␥-aminobutyric acid; ISI, inter-stimulus interval; LGN, lateral geniculate nucleus; LTN, lateral terminal nucleus; MAE, motion aftereffect; MST, medial superior temporal area; MT, middle temporal area (V5); MTN, medial terminal nucleus; NOT, nucleus of the optic tract; PMLS, posteromedial lateral supersylvian area; RGC, retinal ganglion cell; STOLF, space–time oriented linear filter; TFRF, temporal filter response function; V1, primary visual cortex (area 17); V5, middle temporal area (MT); WIM, weighted intersection model ∗ Corresponding author. Tel.: +61-2-9351-6810; fax: +61-2-9351-2603. E-mail addresses: [email protected] (C.W.G. Clifford), [email protected] (M.R. Ibbotson). 1 Tel.: +61-2-6125-4118; fax: +61-2-6125-3808. 0301-0082/03/$ – see front matter Crown Copyright © 2003 Published by Elsevier Science Ltd. All rights reserved. doi:10.1016/S0301-0082(02)00154-5 410 C.W.G. Clifford, M.R. Ibbotson / Progress in Neurobiology 68 (2003) 409–437 4. Adaptive mechanisms in motion detection ................................................... 426 4.1. Perceptual consequences of motion adaptation ........................................... 426 4.2. Function or fatigue? ................................................................... 426 4.3. Informational basis of motion adaptation ................................................ 427 4.4. Dynamics of motion adaptation......................................................... 428 4.5. Directionality of motion adaptation ..................................................... 429 4.6. Distinguishing motion adaptation from contrast adaptation ................................ 430 5. Concluding remarks........................................................................ 431 References ................................................................................... 431 1. Introduction 2. General motion detector mechanisms While numerous visual animals lack color or binocular 2.1. Fundamentals of motion detection vision, the ability to see motion is ubiquitous and, next to the detection of light and dark, may be the oldest and Exner (1894) was the first person to discuss the require- most basic of visual capabilities (Nakayama, 1985). Conse- ments necessary for generating a motion signal from neural quently, visual motion processing is of fundamental interest circuitry. He presented a drawing of a neural network that to systems neuroscience and has been the subject of in- can be regarded as the first attempt at a motion detector tense research. The present paper reviews recent advances model (Fig. 1A). However, it was another German scientist, in our understanding of motion detection in biological Reichardt (1961), who promoted the first computation- systems in the context of the large body of work that has ally based model of motion detection (Hassenstein and gone before. In particular, we emphasize the contribution Reichardt, 1956), a model that has subsequently been given made by comparative studies of motion detection to the his name (Fig. 1B). Although motion detector models vary broader understanding of the topic. The modern theoretical in their detailed structure, the Reichardt detector is useful framework for motion detection was developed from behav- in setting out the basic framework necessary for motion ioral experiments on the Chlorophanus beetle (Hassenstein detection (e.g. Borst and Egelhaaf, 1989). and Reichardt, 1956; Reichardt, 1961). The relevance of Detecting the direction of motion requires that the im- this early work to subsequent studies of motion detection, age be sampled at more than one position or spatial phase, including primate cortical physiology and human psy- that these samples be processed asymmetrically in time, and chophysics, indicates the importance of a broad biological that they be combined in a non-linear fashion (Poggio and approach. A key feature of motion processing to emerge at Reichardt, 1973; Borst and Egelhaaf, 1989). This is a se- both the cellular and systems levels is its dynamic nature rial process that involves computation at multiple synaptic and adaptive plasticity. Consequently, motion adaptation levels. These stages will be covered in three sub-sections: will be a major focus of this review. Perhaps the next great pre-filtering, delay filtering and non-linear interactions. challenge in understanding motion detection is to reconcile the wide range of theoretical approaches with the cellu- 2.2. Pre-filtering lar basis. Given the important advances that have already been made in this direction, we review progress on both of While Reichardt’s (1961) original model included spatial these levels. and temporal pre-filters (Fig. 1B), many subsequent mod- The moving world is projected onto the retina in the form els of motion detection have neglected the importance of of a spatiotemporal pattern of light intensity. From this dy- pre-filtering (although see van Santen and Sperling, 1984, namic signal, recovery of the direction of image motion is 1985; Ibbotson and Clifford, 2001a,b). Pre-filters are impor- the first stage of extracting behaviorally relevant informa- tant because their properties affect the tuning characteristics tion. Section 2 of this review will cover motion processing of the motion detectors they feed. What do we know about from initial non-directional filtering strategies up to the the pre-filters of biological motion detectors? point where a directional signal is produced. Section 3 then deals with the evidence for cellular mechanisms that might 2.2.1. On- and Off-channels perform these tasks in a range of brain areas and species. It is well established that vertebrate photoreceptors are Specifically, we look at motion processing in the vertebrate hyperpolarized by light and their outputs are fed into bipo- and insect visual systems. Section 4 deals with the im- lar cells. Sign conserving synapses feed Off-bipolar cells portant role performed by adaptive mechanisms in motion while sign inverting synapses feed On-bipolar cells (Werblin processing. and Dowling,