User:Eugene M. Izhikevich/Proposed/Change blindness
Change blindness is the failure to detect changes that are easily seen once noticed. Such failures can happen even when the change is large, constantly repeated, and the observer knows that it occurs (e.g.,#F1). When detected, the change immediately jumps into a viewer’s awareness with complete clarity. This indicates that change blindness cannot be due to physical factors such as poor visibility, but to mechanisms internal to the observer. It also indicates that even though we have a strong impression that we see everything that happens in front of us, this impression is false—there are limits to what we consciously perceive at any given time.
Change blindness can be induced in many different ways (see Rensink, 2002). One is shown in #F1 a. In this one-shot paradigm, an image is briefly presented, followed by a brief blank or mask, and then followed by a second display, possibly containing a changed version of the first. Performance is measured by the accuracy of change detection. #F1 b shows a related approach, the flicker paradigm, where the two displays continually alternate until the observer reports the presence or absence of the change. The measure here is the time taken to detect the change. In both cases, performance (measured either by accuracy or by response time) is usually far worse than might be expected.
Change blindness can be explained by the proposal that focused attention is needed to see change (Rensink et al, 1997). If a change is not attended as it occurs, it will pass by unnoticed and so be effectively invisible. This explanation has two important implications. First, it implies that change blindness may be a useful way of studying visual attention (and memory), with limits on our ability to see change reflecting limits on the mechanisms involved. Second, it suggests that our conscious experience of a scene does not involve a representation in which all events are faithfully represented; instead, representations of objects and events may be created on a just-in-time basis by the careful management of attention. And the careful study of change blindness can provide insights as to how this is done.
The fact that we can be blind to large changes in front of our eyes is counterintuitive not only subjectively, but also on more objective grounds. The ability to see change is extremely useful. In our ancestral environment, for example, it was important to keep track of all possible predators and prey—any sudden change in their behavior was potentially relevant to our survival. This ability is equally useful in the modern world. For example, when going to a destination outside our home, we need to keep accurate track of nearby pedestrians or automobiles. Given the importance of perceiving change, it might be expected that we would perceive it well enough that few events would escape our notice.
Indeed, our perception of change appears so effortless and effective that historically there was little interest in understanding how it was done. Occasional failures to see a change could occur, of course, but these were usually considered rare glitches in a system that otherwise worked well. It was only with the arrival of film technology in the early 20th century that this situation began to change. For example, it was found that changes made during a film cut often went unnoticed, even when clearly visible (see Dmytryk, 1984). Such continuity errors were serious enough that specialists were employed to find and eliminate them. However, this phenomenon did not appear to interest academic researchers of the time.
Instead, the scientific study of change blindness began in the 1950s, with work that explored the nature of visual memory. The one-shot paradigm (#F1 a) was developed at that time, using position changes in dot arrays and other simple stimuli. This line of work uncovered a limit to the ability to detect change; later developments (e.g., Phillips, 1974) eventually led to the proposal of a limited-capacity visual short-term memory (vSTM). A second line of study investigated position changes made during an eye movement (e.g., Bridgeman et al, 1979). Here too, observers were found to be surprisingly poor at detecting these, and this was also found to have its origins in the limited capacity of vSTM (Irwin, 1991).
Two related extensions of this work became popular in the mid 1990s, based on technological advances made possible by the personal computer. The first used realistic images of real-world scenes (e.g., Grimes, 1996). Often, the changes were repeated rather than occurring just once, allowing the use of time to measure performance (cf. #F1 b). In addition, blindness was often induced via new kinds of manipulation, such as making the change during a:
- sudden shift of the image (Blackmore et al, 1995)
- sudden appearance of splats elsewhere in the image (Rensink et al, 2000)
- brief eyeblink (O’Regan et al, 2000)
- brief occlusion (Simons & Levin, 1998)
- movie cut (Levin & Simons, 1997)
- sudden reversal of image contrast polarity (Turatto et al, 2003)
- gradual fade-out (Simons et al, 2000).
The second approach used a refined one-shot paradigm, with simple stimuli examined over various temporal gaps (e.g., Luck & Vogel, 1997; Pashler, 1988). It focused primarily on the nature of the mechanisms involved in the perception of change itself (particular visual attention and vSTM) rather than how these mechanisms were controlled or coordinated. However, the results obtained via these two approaches are consistent in areas where they overlap. As such, they are better seen as complementary rather than competing paradigms.
A phenomenon closely related to change blindness is inattentional blindness. This is the failure to experience an object or event that is easily seen once noticed. It typically occurs when the observer is engaged in an attentionally-demanding task elsewhere, and does not expect the object or event. In many ways, both forms of blindness are similar. However, inattentional blindness is concerned with first-order information—the presence of quantities—while change blindness involves second-order information—the transitions between quantities. The two therefore refer to somewhat different aspects of the visual world, with somewhat different mechanisms likely involved in each (Rensink, 2000).
In addition, several different aspects of change perception itself can be distinguished: detection (determining that a change of some sort has been made), identification (determining what kind of change it is), and localization (determining where it is). There is also the related ability of determining which item changed. Although failures of each aspect can be found, change blindness primarily refers to the failure of detection, since this is the aspect that has been most studied.
A final important distinction is between change and difference. Change refers to the transformation of a single structure; difference to the comparison of two or more separate structures. Spotting the difference between two side-by-side stimuli is unlikely to engage the same mechanisms as detecting a change in successively-presented stimuli (Rensink, 2002). Consequently, when discussing change blindness, it is important to make sure that—psychologically at least—the same structure is perceived to transform.
Implications for visual perception
Change blindness has turned out to be a powerful tool for investigating the mechanisms by which we see. Particular details concerning the conclusions from such studies are still the subject of debate, but the general outlines are becoming clear.
Visual attention and memory
All results to date are consistent with the proposal that focused attention is needed to see (i.e., consciously experience) change. Experiments on carefully controlled stimuli suggest that no more than 4-5 items can be seen to change at a time (e.g., Pashler, 1988), a limit similar to that for visual attention. The exact way that attention is involved is somewhat unclear. The perception of change involves—at least in principle—a sequence of several steps: enter information into a memory store, consolidate it into a form usable by subsequent processes, hold it across a temporal interval, compare it to the current stimulus at that location, clear the memory store, and then shift to the next item. A limit on any of these would limit the entire process. This makes it difficult to determine which is the relevant operation in a given situation, and thus, which mechanism the measured characteristics would apply to.
Although some proposed explanations for change blindness involve a limit on the comparison process (Scott-Brown et al, 2000), most involve a limit on the amount (and type) of information held. These are often expressed in terms of limits on visual attention, although they may also be described as limits on vSTM (sometimes referred to as working memory). The difference may be largely one of terminology (see Rensink, 2002), depending on whether emphasis is placed on the selective nature of the process (attention) or its ability to hold information over time (memory). The result is often taken to be the formation of representations coherent over space and time; such representations are similar to those posited as the basis of vSTM.
In any event, interesting results have been found concerning the nature of these representations. There is evidence that the basic elements may be localized structures formed rapidly in the absence of attention (proto-objects), with properties based on the binding of features in the item—for example, its color correctly linked with its orientation (Luck & Vogel, 1997). If so, the function of (focused) attention would not be to bind such properties, but to create a representation with extended spatial and temporal coherence. (See binding problem.) A related—and currently controversial—proposal is whether the features of each item held in memory are necessarily bound, or whether each feature is separately maintained in memory, and bound at a later stage (Wheeler & Treisman, 2002).
If attention is needed to see change, and if only a few items can be attended at any time, our impression of seeing everything that happens in front of us cannot be correct. To account for this impression—and the fact that we do react to most events that are important to us—it has been proposed that scene perception is mediated by a virtual representation (see Rensink, 2002). Here, coherent representations (which support the perception of change) are created on a just-in-time basis, formed whenever they are needed for a task and dissolved afterwards when attention is withdrawn. Meanwhile, a sparse schematic map of the scene—a brief description of perhaps a dozen or so items—is formed independently, in the absence of attention. This map can guide attention to particular items on the basis of high-level control (e.g., knowledge about likely objects) and low-level control (e.g., motion signals drawing attention to unexpected events). If the resulting co-ordination is successful, all parts of a scene will appear to have a coherent representation, even though the amount of information in coherent form at any one time is limited. It is only when these guidance mechanisms are disrupted (by the appearance of an sudden flash, say) that the sparse nature of this representation becomes apparent.
Since change blindness deals only with dynamic aspects of scene perception, it cannot say anything definite about the static aspects that may (or may not) be represented (see Simons & Rensink, 2005). In principle, a dense description of static properties could be built up that matches our subjective impression. But such a description is unnecessary: a virtual representation could handle most if not all aspects of importance. Moreover, other studies of visual perception (such as those on eye movements) have provided more direct evidence that relatively little information is accumulated.
No results to date indicate accumulation of information much beyond the relatively sparse amounts used for guidance and the contents of focused attention and/or vSTM. It has been found that information about a particular item or scene can be stored in a long-term memory of some kind. Although the ultimate capacity of this is not known, current estimates are not high—on the order of 20 bits (Konkle et al, 2010). It may be that many scenes can be stored this way, resulting in a considerable amount of stored information. But the information density for any individual scene would still be quite low-—perhaps just enough for a schematic map.
Given the dependence of visual experience on the allocation of attention, and the dependence of allocation on what is believed, it follows that each individual can literally see a given scene differently. Change blindness can provide a useful way to investigate how particular individuals—and cultures—perceive the world around them.
One approach is based on the priority given to particular items in an image. If attention is allocated first to those items believed to be interesting, the speed at which a change is detected will indicate the extent to which the associated item is considered important. Thus, for instance, observers addicted to a substance such as alcohol or cannabis are faster and more likely to detect changes to items associated with that substance than are observers who are not (Jones et al, 2003).
Individuals also differ in the way they encode items. Experts at American football are better able to spot meaningful changes to football scenes than novices, indicating that the experts have learned to see aspects not perceived by others (Werner & Thies, 2000). Other studies show that Americans are less able to detect changes in a surrounding context than are Japanese, indicating a difference in the encoding of foreground and background elements (Masuda & Nisbett, 2006).
Finally, individuals may also differ in the mechanisms at their disposal. For example, some can have a gut feeling of something changing before they experience a picture of this change (Rensink, 2004). Although consensus has not yet been reached as to the basis of this phenomenon, the fact that observers can show differences in the kinds of experiences they have may provide an interesting new approach to investigating perception.
- Blackmore SJ, Brelstaff G, Nelson K, & Troscianko T. (1995). Is the richness of our visual world an illusion? Transsaccadic memory for complex scenes. Perception, 24:1075-1081.
- Bridgeman B, Hendry D, & Stark L. (1975). Failure to detect displacement of the visual world during saccadic eye movements. Vision Research, 15:719-722.
- Dmytryk E. (1984). On Film Editing. Boston, MA: Focal Press.
- Grimes J. (1996). On the failure to detect changes in scenes across saccades. In Perception (Vancouver Studies in Cognitive Science, vol. 5), ed. K Akins, pp. 89-109. New York : Oxford University Press.
- Irwin DE. (1991). Information integration across saccadic eye movements. Cognitive Psychology, 23:420-456.
- Jones BT, Jones BC, Smith H, & Copley N. (2003). A flicker paradigm for inducing change blindness reveals alcohol and cannabis information processing biases in social users. Addiction, 98:235-244.
- Konkle T, Brady TF, Alvarez GA, & Oliva A. (2010). Scene memory is more detailed than you think: The role of scene categories in visual long-term memory. Psychological Science, 21(11):1551-1556.
- Levin DT, & Simons DJ. (1997). Failure to detect changes to attended objects in motion pictures. Psychonomic Bulletin & Review, 4:501-506.
- Luck SJ, & Vogel EK. (1997). The capacity of visual working memory for features and conjunctions. Nature, 390:279-280.
- Masuda T, & Nisbett RE. (2006). Culture and Change Blindness. Cognitive Science, 30:381-399.
- O'Regan JK, Deubel H, Clark JJ, & Rensink RA. (2000). Picture changes during blinks: Looking without seeing and seeing without looking. Visual Cognition, 7:191-211.
- Pashler HE. (1988). Familiarity and visual change detection. Perception & Psychophysics, 44:369-378.
- Phillips WA. (1974). On the distinction between sensory storage and short-term visual memory. Perception & Psychophysics, 16:283-290.
- Rensink RA (2000). When good observers go bad: Change blindness, inattentional blindness, and visual experience. Psyche, 6. [link]
- Rensink RA. (2002). Change detection. Annual Review of Psychology, 53:245-277.
- Rensink RA. (2004). Visual sensing without seeing. Psychological Science, 15: 27-32.
- Rensink RA, O'Regan JK, & Clark JJ. (1997). To see or not to see: The need for attention to perceive changes in scenes. Psychological Science, 8:368-373.
- Rensink RA, O'Regan JK, & Clark JJ. (2000). On the failure to detect changes in scenes across brief interruptions. Visual Cognition, 7:127-145.
- Scott-Brown KC, Baker MR, & Orbach HS. (2000). Comparison blindness. Visual Cognition, 7:253-267.
- Simons DJ, Franconeri SL, & Reimer RL. (2000). Change blindness in the absence of visual disruption. Perception, 29:1143-1154.
- Simons DJ, & Levin DT. (1998). Failure to detect changes to people during a real-world interaction. Psychonomic Bulletin & Review, 5:644-649.
- Simons DJ, & Rensink RA. (2005). Change blindness: Past, present, and future. Trends in Cognitive Sciences, 9:16-20.
- Turatto M, Bettalla S, Umilta C, & Bridgeman B. (2003). Perceptual conditions necessary to induce change blindness. Visual Cognition, 10:233-255.
- Werner S, & Thies B. (2000). Is "change blindness" attenuated by domain-specific expertise? An expert-novices comparison of change detection in football images. Visual Cognition, 7:163-173.
- Wheeler ME, & Treisman A. (2002). Binding in short-term visual memory. Journal of Experimental Psychology: General, 131:48-64.
- Simons DJ. (2000). (ed.), Change Blindness and Visual Memory. Hove, East Sussex, UK: Psychology Press.