Theories + evidence o Structural-Descrition (Marr & Nishihara, Biederman) o Image Based
Structural-Descriptive models (SD) o entail building a description of the object from retinal image and o matching the description to a representation stored in memory> Evidence for and against both theories specifically and for SD models in general as they share some characteristics. Also: image based theories- much evidence against SD models supports image-based models
Marr & Nishihara (1978) o Proposed series of reps/descriptions. Primal sketch 2D Info about light intensity changes, edges, contours, blobs of visual surfaces Viewpoint dependent 2D Sketch: info re texture, motion, etc. info re depth and texture Viewpoint dependent 3D Sketch: Viewpoint independent Takes into account the relationships between the elements of the object Concavities v important for segmenting parts Canonical coordinate frame: primitive units for describing object = cylinders with a major axis o Hierarchically organised o Motivated by computer graphics. o Challenges: If OR is viewpoint independent, then performance on OR tasks should be independent of viewer position or object orientation. However, when rotated in depth, 3D objects change their 2D retinal projection. Authors offered no empirical support for theory Purely computational and not empirical model. o Support: According to the theory, establishing the location of the central axis is key in OR. Study by Lawson and Humphreys (1996): participants had to identify objects from line drawings. Rotation didnt have any effect unless major axis was foreshortened, had effect on performance Warrington & Taylor (1978) Brain damage to right posterior cerebral hemisphere results in inability to recognise objects from unusual views o Possible explanation: inability to transform 2D to 3D representation. Possibly due to key feature being hidden: o Humphreys & Riddoch (1984): gave images to patients with damage to right hemisphere and found that performance was more affected by foreshortening of central axis, not due to blocking of key feature. Overall, appears to support the view that locating the central axis is key in OR.
Biederman (1987): Recognition by Components Theory (RBC) o Objects can be broken down into simpler objects 36 Geons- blocks, cylinders, arcs, spheres, etc. o Relationship between objects very important and provides scope for many combinations of these geons o Challenges: No empirical support for geons being the building blocks of object recognition Is 36 geons enough? There are only 44 phonemes in the English language, but they can be combined in any way, same as geons. Theory focuses more on edge than surface info ie it ignores colour Joseph and Proffitt (1996) found that colour helps OR o Support: Biederman, Ju & Clapper (1985) Complex objects can still be detected even when some components/geons are missing. Presented line drawing with up to of the components or geons missing, accuracy was 90% in recognising objects. Biederman (1987) Discussed study where ps were given degraded line drawings OR was much poorer when parts of contour info relating to concavities were missing than when other contour info was. SD Theories in general: o Focus on coding of edges, grouping or encoding into higher-order features, matching to stored structural knowledge and access to semantic knowledge o Challenges Some objects do not have easily identifiable constituent parts e.g. clouds Sanocki et al (1998): edge information often insufficient to allow OR- undermines strong focus on importance of edge extraction in low-level OR. Participants were worse at identifying line drawings than colour photographs, especially when they were presented in context. SD models/theories do not account for within description descriminations e.g. we know a chair is a chair, but how do we discriminate between two chairs? Hardly any emphasis on context in OR Palmer (1975) presented pictures in neutral context, appropriate and inappropriate contexts. Found that performance was best in app. and worst in inapp. contexts. Discrepancy in viewpoint dependency Support for both viewpoint dependence and independence Tarr et al. (1998) conducted OR experiments using the same 3D objects under various orientations and conditions. Found that performance was close to viewpoint invariant when task was easy e.g. a between category distinction, a well-known object, feedback given after each trial and viewpoint dependent or variant when task was hard e.g. no feedback given. Image-Based Models o Cost of new orientations- Greebles (Gauthier & Tarr, 2002) Similar pattern to mental rotation tasks o Pure existence of canonical views Palmer et al. (1981)- scale from 1-7 on how canonical or typical the view of an object was Naming latency longer for views that were less canonical o Patient studies Humphreys and Riddoch (1984) some patients can only recognise objects from particular views o Context effects
Question 5 Pain exists within the brain Touch subtopic: focus on nocioceptors, spino-thalamic pathway- describe and illustrate. o All info received as touch until it reaches the brain, where it can be detected as pain once it has been processed. Define, illustrate, evaluate Bottom up: pathway, top down: expectation, prior learning, affective factors. Studies looked at in class. o Basic outline of Pain Topic: o Pain is an unpleasant sensory and emotional experience associated with actual or potential tissue damage International Association for the Study of Pain. o 3 types of pain: Nocioceptive: activation of skin nocioceptors (heat, chemical, pressure, cold) Inflammatory: damage to tissues, inflammation of joint or tumour cells Neuropathic: damage to nervous system o Direct Pathway Model of Pain Stimulation of nocioceptors -> spinal cord -> brain. Simplistic Does not take into account the affective or cognitive side of pain perception. Mental state, attention, phantom limb pain (no stimulation of receptors) o Gate Control Model of Pain (Melzack & Wall, 1965) Basic Outline: Pain signals enter the brain via the spinal cord by the spinothalamic pathway Small-diameter or S fibres carry information from nocioceptors and excite T-cells, leading to pain increase Large-diameter or L-fibres carry general non-pain haptic information Can inhibit T-cell activity Element of central control: from cortex, carrying info re expectation, attention, distraction, etc. Closes gate, inhibits T-cell activity, decreases pain. o Implications of Gate Control Model Pain determined by balance between input from nocioceptors and normal non-pain haptic input. o Effects of Cognitive Factors on Pain Perception: Expectation: Post surgery patients told what to expect re pain were sent home nearly 3 days earlier than those not told (Finniss & Benedetti, 2005) Perhaps some adaptive mechanism, similar to placebo effect (Weisenberg, 2007) Attention: Use of virtual reality games in burns patients during bandage changing significantly reduced reported pain (Hoffman et al., 2002) Content of Emotional Distraction: Participants shown positive images able to keep hand in cold water for longer than ps shown negative or neutral images (deWeid & Verbaten, 2001). o Pain and the Brain: Pain matrix- areas stimulated by pain Example of the distributed representation on pain in the brain Includes many subcortical structures e.g. amygdala, thalamus, as well as the somatosensory cortex, insula, anterior cingulate cortex (ACC), prefrontal cortex. Pain is multimodal -> served by two different areas of the brain Chemicals in the brain: opiods linked with pain perception They act on receptors in the brain that are activated by pain-reducing endorphins Possible link with placebos: may induce endorphin release Pain threshold may possibly be linked to levels of opiod release in the brain (Zubieta et al., 2003) o Pain in Social Situations: Social loss may activate similar areas as stimulated by physical pain (pain matrix) Eisenberger et al. (2003): simulated group exclusion and found reported distress correlated with activation of the ACC Similar findings by Singer et al. (2004): people, especially women, watching their partners receive shocks showed activation of their own ACC o Especially in women who reported high empathy. o Pain Perception Tutorial: Main points: Pain is not perceived as more than touch until it reaches the brain Stimulation of skin receptors is not necessary for pain perception Cognitive and emotional factors can play a part Describe and illustrate spinothalamic pathway Signals from nocioceptors in the skin sent to the spinal cord by axons with their cell bodies (somas) in the dorsal root ganglion. Axons enter the SC via the dorsal horn and ascend 1- 2 levels forming the Tract of Lissauer Axons synapse in the grey matter of the dorsal horn ipsilaterally Secondary neurons may ascend further ipsilaterally or decussate the other side of the SC via the anterior white commissure The axons then enter the spinothalamic fasciculus and synapse in the ventral post lateral nucleus of the thalamus The neurons then project to the post central gyrus of the cortex in the somatosensory cortex Others extend to the reticular formation of the medulla and to areas included in the pain matrix- the thalamus, amygdala, hippocampus, prefrontal cortex and ACC o These areas are related the brains central control of pain perception and from here the descending pathways can block the T cell activity and reduce perceived pain in line with the Gate Control Theory.