Change blindness
From Wikipedia, the free encyclopedia

Change blindness is a phenomenon in visual perception where apparently large changes within a visual scene are undetected by the viewer. Typically for change blindness to occur, the change in the scene has to coincide with some visual disruption such as an eye movement or a brief obscuration of the observed scene or image.

When change blindness was first explored systematically by George McConkie and his colleagues in the late 1970s, the phenomenon was largely limited to the study of changes introduced to words and text during eye movements. A student of McConkie's, John Grimes, was the first to extend this phenomenon to the domain of scene perception (in a conference presentation in 1992, not published until a book chapter in 1996). Grimes showed that people miss large changes to scenes when the changes are introduced during an eye movement. For example, many people failed to notice when two people in a scene exchanged heads! In these saccade-contingent change blindness studies, changes to the scene were synchronized with measured movements of the observer's eyes, so that the changes occurred only when the eyes were moving. Under these conditions, changes are often hard to detect. A number of studies since then have explored saccade-contingent change blindness (e.g., Henderson & Hollingworth, 1999; McConkie & Currie, 1996).

Later experiments showed that change blindness was not specifically related to eye movements -- other forms of visual disruption could also induce change blindness. Rensink et al, popularized the "flicker" technique in which two images alternate repeatedly with a brief (80ms) blank screen after each image (giving the display a flickering appearance). With the blank screen in place, surprisingly large changes could be made to the scene without the observer noticing. Rensink et al (1997) also introduced the term "change blindness."

Other studies showed that change blindness occurs when the change is introduced during a cut or pan in a motion picture, even when the change is to the central actor in a scene (Levin & Simons, 1997). People also regularly fail to notice editing errors in commercial movies, despite the intense scrutiny of movies during the production process.

Change blindness can be particularly dramatic when changes occur unexpectedly, with many observers even failing to notice when a person they were talking to was surreptitiously replaced by a different actor (Simons & Levin, 1998). Change blindness has now been shown to occur with a wide variety of visual disruptions (e.g., blinks, transient noise flashed on a display, etc).

Change blindness is related to other induced failures of awareness, such as inattentional blindness. A crucial difference is that successful change detection in the presence of a visual disruption requires a comparison of one image to another one held in memory. Consequently, change blindness can occur due to a failure to store the information in the first place or to a failure to compare the relevant information from the current scene to the representation (hence models of visual short term memory may be important for understanding the phenomenon). In contrast, inattentional blindness reflects the failure to detect an unexpected stimulus that is fully visible in a single display -- it does not require a comparison to memory.

In the real world, change blindness may be responsible for traffic accidents. If drivers fail to notice significant changes around them (e.g., the presence of a pedestrian in their path).

Choice blindness

In psychology, choice blindness is a phenomenon in which subjects fail to detect conspicuous mismatches between their intended (and expected) choice and the actual outcome.

Writing in Science, psychologist Petter Johansson and coworkers describe choice blindness in an ingenious experiment.

The subject is presented with two cards, on which different (female) faces appear. The subject is asked to choose which one he finds more attractive. In the non-manipulated (NM) version, the subject is handed the card that he chose and asked to say why he chose that one. In the manipulated (M) version, the experimenter uses sleight of hand techniques to switch the cards without the subject's knowledge and give the subject the other card.
The workers found that most subjects failed to notice the switch, and furthermore justified their decision using post-hoc confabulated evidence. For example, in a M trial, a subject might say "I preferred this one because I prefer blondes" when he had in fact chosen (and pointed to) the dark-haired woman, but was handed a blonde.

They point out that his experiment allows one to investigate the relationship between choice and introspection.

Johansson concludes that he has found that some normal participants unequivocally produce confabulatory reports when asked to describe the reasons behind their choices and suggests that choice blindness affords some insight into the mechanisms behind truthful report.

Inattentional blindness, closely related to the subject of change blindness, is observed phenomena of the inability to perceive features in a visual scene if they are not being attended to. That is to say that humans have a limited capacity for attention which thus limits the amount of information processed at any particular time. Any otherwise salient feature within the visual field will not be observed if not processed by attention.

Experiments demonstrating inattentional blindness

The most well known study demonstrating inattentional blindness was conducted by Daniel Simons of the University of Illinois at Urbana-Champaign and Christopher Chabris of Harvard University. In their study, subjects are asked to watch a short video [1] in which two groups of people (wearing black and white t-shirts) pass a basketball back among themselves. The subjects are told to either count the number of passes made by one of the teams, or to keep count of bounce passes vs. aerial passes. In different versions of the video a woman walks through the scene carrying an umbrella, or wearing a full gorilla suit. In one version the woman in the gorilla suit even stops in the middle, faces the camera, and pounds her chest before walking out of the scene. After watching the video the subjects are asked if they saw anything out of the ordinary take place. In most groups 50% of the subjects did not report seeing the gorilla. Simons interprets this by stating that we are mistaken with regard to how important events will automatically draw our attention away from current tasks or goals. This result indicates that the relationship between what is in our visual field and perception is based much more significantly on attention than was previously thought.

Another research finding was done by Steve Most, Chabis and Scholl. They had objects moving up and down on a computer screen. Participants were instructed to attend to the black objects and ignore the white, or vice versa. After several trials, a red cross unexpectedly appeared and traveled across the display, remaining on the computer screen for five seconds. The results of the experiment showed that even though the cross was distinctive from the black and white objects both in color and shape, about a third of participants nonetheless missed it. They had found that people may be attentionally tuned to certain perceptual dimensions, such as brightness or shape.

Thus the common saying "out of sight, out of mind", meaning that people usually don't think about what they do not see, could be reversed into: "out of mind, out of sight".

Visual short term memory
From Wikipedia, the free encyclopedia

In the study of vision, visual short-term memory (VSTM) is one of three broad memory systems including iconic memory and long-term memory. VSTM is a type of short-term memory, but one limited to information within the visual domain.

The Visuospatial Sketchpad is a VSTM subcomponent within the theoretical model of working memory proposed by Alan Baddeley. However, the term VSTM refers in a theory-neutral manner to the non-permanent storage of visual information over an extended period of time.

Whereas iconic memories are fragile, decay rapidly, and are unable to be actively maintained, visual short-term memories are robust to subsequent stimuli and last over many seconds.

Overview

The introduction of stimuli which were hard to verbalize, and unlikely to be held in long-term memory, revolutionized the study of visual short-term memory (VSTM) in the early 1970s (Cermak, 1971; Phillips, 1974; Phillips & Baddeley, 1971). The basic experimental technique used required observers to indicate whether two matrices (Phillips, 1974; Phillips &Baddeley, 1971), or figures (Cermak, 1971), separated by a short temporal interval, were the same. The finding that observers were able to report that a change had occurred, at levels significantly above chance, indicated that they were able to encode some aspect of the first stimulus in a purely visual store, at least for the period until the presentation of the second stimulus. However, as the stimuli used were complex, and the nature of the change relatively uncontrolled, these experiments left open various questions, such as: (1) whether only a subset of the perceptual dimensions comprising a visual stimulus are stored (e.g., spatial frequency, luminance, or contrast); (2) whether some perceptual dimensions are maintained in VSTM with greater fidelity than others; and (3) the nature by which these dimensions are encoded (i.e., are perceptual dimensions encoded within separate, parallel channels, or are all perceptual dimensions stored as a single bound entity within VSTM.

Psychophysical approaches to VSTM

In a typical psychophysical VSTM experiment, observers' ability to discriminate between sequentially presented test and reference patterns are measured using a two-interval forced-choice (2-IFC) paradigm. For example, in a study involving spatial frequency, observers might be required to make a judgment as to whether the first or second pattern presented was of higher (or lower) spatial frequency. Typically the test and reference patterns are separated by ISIs in the range of 0 s to 30 s. The properties of stimulus pairs can be controlled by using a psychophysical staircase procedure, or via the method of constant stimuli (for details of these techniques, see Regan, 2000). When a staircase procedure is used, the properties of the stimulus pairs are altered until a criterion threshold level of performance is achieved (e.g., 75% correct).

Fidelity of memory representations

A series of studies over the last decade and a half (for good reviews, see Magnussen, 2000; Magnussen & Greenlee, 1999) have demonstrated that VSTM stores various perceptual dimensions (e.g., spatial frequency, orientation, hue) with a remarkable degree of fidelity and stability (Magnussen & Greenlee, 1992; Magnussen, Greenlee, Asplund, & Dyrnes, 1991; Magnussen, Idas, & Myhre, 1998; Regan, 1985). It has been shown, for instance, that with a reference frequency of 10 c/deg, spatial frequency thresholds tested with ISIs of up to several seconds (measured as Weber fractions, Df/f) differ by only three-to-six percent from those recorded when gratings are presented simultaneously (Regan, 1985). With a period difference of 360 arcseconds, a threshold of 0.04 Df/f implies that observers are able to distinguish spatial frequency differences of 14.4 arcsec (Magnussen & Greenlee, 1999). As this is approximately half the average cone spacing on the fovea, it implies that observers are able to store spatial frequency information within the hyperacuity range for upwards of 60 s (Bennett & Cortese, 1996).

A series of psychophysical studies have found that many perceptual dimensions (i.e., spatial frequency, hue, orientation, speed) are stored with little or no loss in VSTM. As already mentioned, spatial frequency can be stored for upwards of 60 s with no increase in thresholds (Bennett & Cortese, 1996; Magnussen & Greenlee, 1997; Magnussen et al., 1991; Magnussen, Greenlee, & Thomas, 1996; Regan, 1985). Other studies have shown that colour (Nilsson & Nelson, 1981), speed (Magnussen & Greenlee, 1992), and orientation (Magnussen et al., 1998), are also stored in VSTM for upwards of 10 s with no significant decay.

The one notable exception to this rule is contrast. Several studies have shown that thresholds for contrast discrimination grow rapidly as ISIs increase, with thresholds doubling as ISIs are raised from 0 s to 10 s (Lee & Harris, 1996; Magnussen et al., 1991; Magnussen et al., 1996). This appears to be due to a loss of information about contrast as ISI increases, which correspondingly makes it increasingly unlikely that a change will be reported as ISIs increase. The decay in contrast information is likely to underlie the apparent decay in information for VSTM experiments using matrix patterns (e.g., Phillips, 1974; Phillips & Baddeley, 1971).

Structure of memory representations

With the exception of contrast, basic perceptual attributes are similar in terms of both the accuracy and stability with which they are stored in VSTM. However, it is unclear whether the information from each perceptual stream is encoded separately within parallel channels, or whether information for different perceptual dimensions is represented within VSTM as a single, bound set of features. Two different lines of evidence – one derived from the experimental paradigm known as memory masking, the other associated with the differential effects observed for decisions made either within or between perceptual dimensions – suggest that VSTM stores information within multiple parallel perceptual channels.

Memory masking

Memory masking refers to an experimental technique in which the addition of a "masking" grating, placed between the reference and test stimuli in a psychophysical VSTM experiment (Bennett & Cortese, 1996; Magnussen & Greenlee, 1992; Magnussen et al., 1991), leads to an increase in psychophysical thresholds. It is important to note, however, that the use of the term "masking" here is somewhat misleading, as the temporal placement of the additional grating is such that it acts neither as a pattern mask nor as an energy mask for the test or reference stimuli (Breitmeyer, 1984).

If the masking grating matches the test or reference grating in the perceptual dimension being discriminated, no increase in threshold is observed relative to a no-mask control condition. However, the more the masking stimulus differs from the reference grating on the dimension being discriminated, the more thresholds increase, until thresholds are approximately double those recorded in the absence of a mask (Bennett & Cortese, 1996; Magnussen & Greenlee, 1992; Magnussen et al., 1991).

The short presentation times of the masking stimuli (e.g., 200 ms), coupled with the relatively long time periods between the mask and both the test and reference stimuli (e.g., Magnussen et al., 1991) argue against the possibility that spatial adaptation is an explanation for the increase in thresholds caused by the presence of the mask (e.g., Blakemore & Campbell, 1969).

Another feature of the memory-masking paradigm is that the effects of the mask are specific to the type of discrimination being made. For instance, when performing a spatial frequency judgment, the orientation of the masking grating has no effect on threshold levels. Likewise, the spatial frequency of the masking grating does not alter thresholds obtained when orientation is being discriminated (Magnussen et al., 1991). This specificity of the masking effect on thresholds is evidence against its being mediated either through distracting the observer, or by adding an additional non-specific burden to memory. Since orientation and spatial frequency are conjointly coded early in the visual system (DeValois & DeValois, 1990), this result supports the view that the neurophysiological site affected by memory masking occurs post-V1, at a locus where orientation and spatial frequency information are coded into independent perceptual channels. This argument is supported by the finding that masking by spatial frequency follows perceptual, rather than retinal coordinates (Bennett & Cortese, 1996), as size constancy is also thought to occur at a point post-V1 in the visual processing hierarchy (Magnussen, 2000), perhaps in V4 (see, for instance, Schiller, 1995).

Dual discrimination costs

It is well established that observers are able to make independent decisions about multiple stimulus dimensions (e.g., spatial frequency, contrast, orientation) with little or no cost (e.g., Chua, 1990; Greenlee & Thomas, 1993; Vincent & Regan, 1995). These studies support the view that spatial frequency, orientation, and contrast are encoded within independent, parallel channels. Since individual neurons in striate cortex conjointly code spatial frequency and orientation (DeValois & DeValois, 1990), these channels are likely to exist at a point later in the visual processing hierarchy than V1.

The memory masking literature supports the view that different perceptual properties are encoded independently within parallel channels. This view is further supported by evidence from experiments examining the costs of making dual decisions for attributes that are encoded either within the same or between different perceptual channels (Greenlee & Thomas, 1993; Magnussen & Greenlee, 1997; Magnussen et al., 1996; Thomas, Magnussen, & Greenlee, 2000). The Greenlee-Thomas model assumes that different perceptual dimensions (e.g., spatial frequency, contrast, orientation, movement) are encoded within independent channels (Greenlee & Thomas, 1993). According to this model, making dual judgments about different perceptual dimensions will lead to only a moderate increase in threshold, associated with the increased uncertainty of making two independent judgments (i.e., decision-noise). However, if the two judgments made are not independent — as might be expected if observers were required to make two decisions which draw on the same limited resource — thresholds are predicted to increase to a greater extent than can be explained on the basis of decision-noise alone.

Magnussen and Greenlee (1997) performed a series of VSTM experiments in which the relative costs associated with making dual discriminations within and between stimulus dimensions were compared. Their results can be summarized as follows: (1) when making discriminations regarding both contrast and spatial frequency, observers' thresholds rise by an amount predicted by the additional uncertainty in making two independent decisions; (2) when judgments are made within the same perceptual dimension, there is a much greater increase in associated thresholds than predicted on the basis of the increased uncertainty associated with making multiple independent decisions, suggesting that these judgments are not made independently.

Summary of results from psychophysical experiments

Psychophysical experiments in VSTM suggest that most perceptual dimensions (e.g., spatial frequency, orientation, colour, speed) are stored with remarkable fidelity over relatively long periods of time (Magnussen, 2000). The one exception to this rule is contrast, which has been shown to decay rapidly in VSTM (Lee & Harris, 1996). Converging evidence, drawn both from experiments using memory masking (Magnussen & Greenlee, 1992), and from a comparison of single and dual-discrimination costs (Magnussen & Greenlee, 1997), suggests that information is encoded in VSTM in the form of multiple independent channels, each channel representing a different perceptual dimension. Further evidence suggests that this information is encoded at a level in the visual hierarchy later than V1 (e.g., Bennett & Cortese, 1996).

Set-size effects in VSTM

In a typical VSTM experiment, observers are presented with two arrays, composed of a number of stimuli. The two arrays are separated by a short temporal interval, and the task of observers is to decide if the first and second arrays are composed of identical stimuli, or whether one item differs across the two displays (e.g., Luck & Vogel, 1997). Increasing the number of stimuli present within the two arrays leads to a monotonic decrease in the sensitivity of observers to differences in stimuli across the two arrays (Luck & Vogel, 1997; Pashler, 1988). This capacity limit has been linked to the posterior parietal cortex, the activity of which increases with the number of stimuli in the arrays, but only up to the capacity limit of about four stimuli (Todd & Marois, 2004). There are a number of frameworks that attempt to explain the effect of increasing set-size on performance in VSTM. These can be broadly grouped under three categories: (1) psychophysical frameworks (e.g., Magnussen & Greenlee, 1997); (2) sample size models (e.g., Palmer, 1990); and (3) urn models (e.g., Pashler, 1988).

Problems with psychophysical explanations

Psychophysical experiments suggest that information is encoded in VSTM across multiple parallel channels, each channel associated with a particular perceptual attribute (Magnussen, 2000). Within this framework, a decrease in an observer's ability to detect a change with increasing set-size can be attributed to two different processes: (1) if decisions are made across different channels, decreases in performance are typically small, and consistent with decreases expected when making multiple independent decisions (Greenlee & Thomas, 1993; Vincent & Regan, 1995); (2) if multiple decisions are made within the same channel, the decrease in performance is much greater than expected on the basis of increased decision-noise alone, and is attributed to interference caused by multiple decisions within the same perceptual channel (Magnussen & Greenlee, 1997).

However, the Greenlee-Thomas model (Greenlee & Thomas, 1993) suffers from two failings as a model for the effects of set-size in VSTM. First, it has only been empirically tested with displays composed of one or two elements. It has been shown repeatedly in various experimental paradigms that set-size effects differ for displays composed of a relatively small number of elements (i.e., approximately = 4 items), and those associated with larger displays (i.e., approximately > 4 items). The Greenlee-Thomas (1993) model offers no explanation for why this might be so. Second, while Magnussen, Greenlee, and Thomas (1997) are able to use this model to predict that greater interference will be found when dual decisions are made within the same perceptual dimension, rather than across different perceptual dimensions, this prediction lacks quantitative rigor, and is unable to accurately anticipate the size of the threshold increase, or give a detailed explanation of its underlying causes.

In addition to the Greenlee-Thomas model (Greenlee & Thomas, 1993), there are two other prominent approaches for describing set-size effects in VSTM. These two approaches are can be referred to as sample size models (Palmer, 1990), and urn models (e.g., Pashler, 1988). They differ from the Greenlee-Thomas (1993) model by: (1) ascribing the root cause of set-size effects to a stage prior to decision making; and (2) making no theoretical distinction between decisions made in the same, or across different, perceptual dimensions.

Models of capacity limits in VSTM

If observers are asked to report on the quality (e.g., color) of an item stored in memory, while performance might be perfect when only a few items are encoded (the number of items that can be perfectly encoded varies depending on the attribute being encoded, but is usually less than five), after which performace invariably declines in a monotonic fashion as more items are added. Different theoretical models have been put forward to explain this decline in performance.

Slot models

A prominent class of model proposes that observers are limited by the total number of items which can be encoded, either because the capacity of VSTM itself is limited (e.g., Cowan, 2001; Luck & Vogel, 1997; Pashler, 1988), or because of a bottleneck in the number of items which can be attended to prior to encoding. This type of model has obvious similarities to urn models used in probability theory (see, for example, Mendenhall, 1967). In essence, an urn model assumes that VSTM is restricted in storage capacity to only a few items, k (often estimated to lie in the range of three-to-five). The probability that a suprathreshold change will be detected is simply the probability that the change element is encoded in VSTM (i.e., k/N). Although urn models are used commonly to describe performance limitations in VSTM (e.g., Luck & Vogel, 1997; Pashler, 1988; Sperling, 1960), it is only recently that the actual structure of items stored has been considered. Luck and colleagues have reported a series of experiments designed specifically to elucidate the structure of information held in VSTM (Luck & Vogel, 1997). This work provides evidence that items stored in VSTM are coherent objects, and not the more elementary features of which those objects are composed.

Noise models

A much more controversial framework has more recently been put forward by Wilken and Ma (2004) who suggest that apparent capacity limitations in VSTM are caused by a monotonic decline in the quality of the internal representations stored (i.e., monotonic increase in noise) as a function of set size. In this conception capacity limitations in memory are not caused by a limit on the number of things that can be encoded, but by a decline in the quality of the representation of each thing as more things are added to memory.

In their 2004 experiments, they varied color, spatial frequency, and orientation of objects stored in VSTM using a signal detection theory (SDT) approach. The participants were asked to report difference between the visual stimuli presented to them in consecutive order. The invesigators found that different stimuli were encoded independently and in parallel, and that the major factor limiting discrimination performance was neuronal noise.

Sample size models

Sample size models (Palmer, 1990) propose that the monotonic decrease in performance with increasing set-size in VSTM experiments is a direct outcome of a limit in the amount of information observers can extract from a visual display.

In the sample size model, each perceptual attribute of a stimulus is associated with an internal, unidimensional percept, formed by the collection of a finite number of discrete samples. It is assumed that the total number of samples that can be collected across the entire visual scene is fixed. Assuming that equal attention is paid to each stimulus, it follows that the total number of samples taken from each element in an array will be inversely proportional to the number of stimuli present, N. Central limit theorem implies that the mean of the samples taken, and therefore the mean of the internal percept, will have a variance inversely proportional to N. Signal detection theory defines sensitivity (i.e., d') as being inversely proportional to the standard deviation of the underlying representation to be discriminated (Macmillan & Creelman, 1991). Therefore according to the sample size model, in a VSTM experiment an observer's sensitivity to a stimulus change, d', will be inversely proportional to square-root of N.

Unfortunately, few studies have directly tested this prediction of the sample size model. Some evidence has been provided by Palmer (1990), who performed a VSTM experiment using arrays composed of lines of varying length, and set-sizes of one, two or four. The task of observers was to determine whether there had been a change in the length of one of the lines. It was found that observers' thresholds increased proportional to square-root of N, in accordance with the predictions of the sample size model.