Memory begins with perception. We can’t remember what we don’t perceive, and our memory of things is influenced by how we perceive them.
Our ability to process visual scenes has been the subject of considerable research. How do we process so many objects? Some animals do it by severely limiting what they perceive, but humans can perceive a vast array of features. We need some other way of filtering the information. Moreover, it’s greatly to our advantage that we can process the environment extremely quickly. So that’s two questions: how do we process so much, and so fast?
Brain region behind the scene-facilitation effect identified
A critical factor, research suggests, is our preferential processing of interacting objects — we pick out interacting objects more quickly than unrelated objects. A new study has now identified the region of the brain responsible for this ‘scene-facilitation effect’. To distinguish between the two leading contenders, the lateral occipital cortex and the intraparietal sulcus, transcranial magnetic stimulation was used to temporarily shut down each region in turn, while volunteers viewed brief flashes of object pairs (half of which were interacting with each other) and decided whether these glimpsed objects matched the presented label. Half of the object pairs were shown as interacting.
The scene-facilitation effect was eliminated when the lateral occipital cortex was out of action, while the non-performance of the intraparietal sulcus made no difference.
The little we need to identify a scene
The scene-facilitation effect is an example of how we filter and condense the information in our visual field, but we also work in the opposite direction — we extrapolate.
When ten volunteers had their brains scanned while they viewed color photographs and line drawings of six categories of scenes (beaches, city streets, forests, highways, mountains and offices), brain activity was nearly identical, regardless of whether participants were looking at a color photo or a simple line drawing. That is, researchers could tell, with a fair amount of success, what category of scene the participant was looking at, just by looking at the pattern of brain activity in the ventral visual cortex — regardless of whether the picture was a color photo or a line drawing. When they made mistakes, the mistakes were similar for the photos and the drawings.
In other words, most of what the brain is responding to in the photo is also evident in the line drawing.
In order to determine what those features were, the researchers progressively removed some of the lines in the line drawings. Even when up to 75% of the pixels in a line drawing were removed, participants could still identify what the scene was 60% of the time — as long as the important lines were left in, that is, those showing the broad contours of the scene. If only the short lines, representing details like leaves or windows, were left, participants became dramatically less accurate.
The findings cast doubt on some models of human visual perception which argue that people need specific information that is found in photographs to classify a scene.
Consistent with previous research, activity in the parahippocampal place area and the retrosplenial cortex was of greatest importance.
The brain performs visual search near optimally
Visual search involves picking out a target in a sea of other objects, and it’s one of the most important visual tasks we do. It’s also (not surprisingly, considering its evolutionary importance) something we are very very good at. In fact, a new study reveals that we’re pretty near optimal.
Of course we make mistakes, and have failures. But these happen not because of our incompetence, but because of the complexity of the task.
In the study, participants were shown sets of lines that might or might not contain a line oriented in a particular way. Each screen was shown for only a fraction of a second, and the contrast of each line was randomly varied, making the target easier or more difficult to detect. The variation in contrast was designed as a model for an important variable in visual search — that of the reliability of the sensory information. Optimally, an observer would take into consideration the varying reliability of the items, giving the information different weights as a result of that perceived reliability. That weighted information would then be combined according to a specific integration rule. That had been calculated as the optimal process, and the performance of the participants matched that expectation.
The computer model that simulated this performance, and that matched the human performance, used groups of (simulated) neurons that responded differently to different line orientations.
In other words, it appears that we are able, very quickly, to integrate information coming from multiple locations, while taking into account the reliability of the different pieces of information, and we do this through the integration of information coming from different groups of neurons, each group of which is responding to different bits of information.
Another recent study into visual search has found that, when people are preparing themselves to look for very familiar object categories (people or cars) in natural scenes, activity in their visual cortex was very similar to that shown when they were actually looking at the objects in the scenes. Moreover, the precise activity in the object-selective cortex (OSC) predicted performance in detecting the target, while preparatory activity in the early visual cortex (V1) was actually negatively related to search performance. It seems that these two regions of the visual cortex are linked to different search strategies, with the OSC involved in relatively abstract search preparation and V1 to more specific imagery-like preparation. Activity in the medial prefrontal cortex also reflected later target detection performance, suggesting that this may be the source of top-down processing.
The findings demonstrate the role of preparatory and top-down processes in guiding visual search (and remind us that these processes can bias us against seeing what we’re looking for, just as easily as they help us).
'Rewarding' objects can't be ignored
Another aspect of visual search is that some objects just leap out at us and capture our attention. Loud noises and fast movement are the most obvious of the attributes that snag our gaze. These are potential threats, and so it’s no wonder we’ve evolved to pay attention to such things. We’re also drawn to potential rewards. Prospective mates; food; liquids.
What about rewards that are only temporarily rewarding? Do we move on easily, able to ignore previously rewarding items as soon as they lose their relevance?
In a recent study, people spent an hour searching for red or green circles in an array of many differently colored circles. The red and green circles were always followed by a monetary reward (10 cents for one color, and 1 cent for the other). Afterwards, participants were asked to search for particular shapes, and color was no longer relevant or rewarded. However, when, occasionally, one of the shapes was red or green, reaction times slowed, demonstrating that these were distracting (even though the participants had been told to ignore this if it happened).
This distraction persisted for weeks after the original learning session. Interestingly, people who scored highly on a questionnaire measuring impulsivity were more likely to be distracted by these no-longer-relevant items.
The findings indicate that stimuli that have been previously associated with reward continue to capture attention regardless of their relevance to the task in hand, There are implications here that may help in the development of more effective treatments for drug addiction, obesity and ADHD.
People make an image memorable
What makes an image memorable? It’s always been assumed that visual memory is too subjective to allow a general answer to this question. But an internet study has found remarkable consistency among hundreds of people who viewed images from a collection of about 10,000 images, some of which were repeated, and decided whether or not they had seen the image before. The responses generated a memorability rating for each image. Once this had been collated, the researchers made "memorability maps" of each image by asking people to label all the objects in the images. These maps were then used to determine which objects make an image memorable.
In general, images with people in them were the most memorable, followed by images of human-scale space — such as the produce aisle of a grocery store — and close-ups of objects. Least memorable were natural landscapes, although those could be memorable if they featured an unexpected element, such as shrubbery trimmed into an unusual shape.
Computer modeling then allowed various features for each image (such as color, or the distribution of edges) to be correlated with the image's memorability. The end result was an algorithm that can predict memorability of images the computational model hasn't "seen" before.
The researchers are now doing a follow-up study to test longer-term memorability, as well as working on adding more detailed descriptions of image content.