ATTENTIONAL CONTROL

Relationships Matter

Becker, S. I., Folk, C. L., & Remington, R. W. (2013). Attentional capture does not depend on feature similarity, but on target–nontarget relations. Psychological Science, 24(5), 634–647.

We can direct attention to something like a red target very efficiently, so we took it for granted that, computationally, this task would be implemented in the most straightforward way: prioritizing the feature “red” and deprioritizing everything else. However, Becker, Folk, and Remington (2013) argue for a more sophisticated strategy. The mechanism of attentional selection may find the target based on its relationship with the context. We may not direct our attention to some absolute “red” but rather to the “reddest” items in the display. An observer, set of “reddest” might deploy attention to an orange item if it is redder than anything else in the display.

The difference between these two implementations of selection (by feature vs. by relation) can be revealed by measuring contingent capture by another set of task-irrelevant stimuli. In Becker et al (2013), the observers attempted to find a color-defined target among a total of 4 items. Each item was preceded by a four-dot cue surrounding the item. These 4 cues were also colored, with the color of one of them different from the rest. If the selection of the target was facilitated by the odd-color cue compared to when that cue surrounded another item, then we know the odd-color cue has attracted attention.

Becker et al (2013) used four colors: red, orange, gold, and yellow, which were evenly distributed along a line in color space from red to yellow. In Experiment 1, when the search task was “orange among gold”, both a “red among orange” cue and a “gold among yellow” cue attracted attention because they shared the same relation (redder) with the search task. Sensitivity to absolute feature values would not predict this. In a search for orange among gold the “gold” in “gold among yellow” condition should be inhibited while the “orange” in the “red among orange” condition should have been facilitated. On the other hand even though the target was orange among gold, neither an orange singleton cue among red cue nor a yellow among gold cue attracted attention because the unique cue was yellower while the search target was redder.

Experiment 2 found that, more strikingly, attentional selection also relies on relationships even when the task is selection of an intermediate feature value. For example, when the search task was “orange among red + yellow”, a “gold among orange + yellow” cue attracted attention because the cue and search task shared the same relation (the middle one). On the other hand, an “orange among gold + yellow” cue did not attract attention because the relationship was inconsistent between the cues (the reddest) and in the search task (the middle one).

To summarize, the experiments demonstrated that attention system often makes selections based on the relationship between the target and the context rather than on the absolute feature values. This finding has important implications on the studies of visual attention because it showed that “feature-based selection” really refers to the relationships between features. This might lead to an entirely different way of formulating the algorithms of visual selection.—L.Q.H.

SPATIAL VISION

Steppin’ out

Westrick, Z. M., Henry, C. A., & Landy, M. S. (2013) Inconsistent channel bandwidth estimates suggest winner-take-all nonlinearity in second-order vision. Vision Research, 81(58), 58–68.

Neurons identified by Hubel and Weisel (1959;) have different preferences for dark and light stripes. There are neurons that like thin stripes, neurons that like thick stripes, and neurons that like everything in between. The strength of those preferences can be quantified in terms of “bandwidth.” Low bandwidth implies strong preferences; high bandwidth implies weak preferences. It’s relatively easy to infer the bandwidth from neurons that respond linearly with the difference (i.e. the contrast) between dark and light. Starting with the neuron’s preferred thickness, adjust the stripes’ size until the neuron’s response has been halved. “Full bandwidth at half height” is twice the difference between log stripe widths before and after adjustment. (For “octave” bandwidths, use log base 2.)

Visual consequences of these neural preferences are legion and there are several psychophysical techniques for estimating bandwidth. Of course, those estimates cannot be ascribed to any particular neuron, so psychophysicists instead use the term “channel” to describe the putative manifestation of neurons having similar preferences. Virtually all of those techniques produce bandwidth estimates between 1 and 2 octaves.

The same psychophysical techniques used to estimate the bandwidth of channels preferring light and dark stripes can also be used to estimate the bandwidths of channels for which there are no clear neural correlates. Westrick, Henry, & Landy used two of these techniques to estimate the bandwidth of channels preferring orientation-defined stripes like these:

\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\

/////////////////////////////////////////

\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\

/////////////////////////////////////////

In this case, however, the two techniques suggested very different bandwidths. Identification at threshold, in which observers are given the dual task of a) selecting which of two stimuli contained orientation-defined stripes and b) reporting whether those stripes were thick or thin, produced bandwidth estimates in the range of 1–1.5 octaves. The critical-band-masking technique, on the other hand, suggested bandwidths in excess of 4 octaves.

To resolve this apparent discrepancy, Westrick et al inferred that the so-called “second-order” channels sensitive to orientation-defined stripes must not respond linearly with the difference between the stripes’ orientation contents. Instead, second-order channels might respond linearly with locally thresholded orientation contrast. Local thresholding (with a “step” function) makes the input to second-order channels effectively binary. Westrick et al demonstrated that this thresholding could be responsible for both sets of their results: identification at threshold, and critical band masking.

In summary, this paper forces us to reject the popular “filter-rectify-filter” model for extracting orientation-defined signals. At least one additional non-linearity is required, and step function now seems to be the leading candidate.—J.A.S.

Additional References

Hubel, D. & Wiesel, T. (1959). Receptive fields of single neurons in the cat’s striate cortex. Journal of Physiology 148, 574–591.

PHOTORECEPTION

Even ‘simple’ pupil reflexes aren’t simple

Lucas, R. J., Peirson, S. N., Berson, D. M., Brown, T. M., Cooper, H. M., Czeisler, C. A., et al. (2014). Measuring and using light in the melanopsin age. Trends Neurosci, 37(1), 1–9.

Photoreception is a complicated business and it has been getting more complicated in recent years. We used to understand that there were two types of photoreceptors: the rods and the cones. There was one type of rod photopigment and three types of cone photopigments, allowing the cones to subserve color vision. Rods were connected to rod bipolar cells and cones to cone bipolar cells and these, in turn gave rise to an ever increasing variety of retinal ganglion cells. This was quite complicated enough as many undergraduates in our sensation and perception classes will tell you.

There were some mysteries. We have circadian clocks that cause many physiological functions to oscillate over the course of an approximately 24-hour cycle. When we change time zones, the clock’s phase can be reset, allowing us to recover from jet lag. Light is the predominate signal for this reset. However, some individuals, without photoreceptors and otherwise blind, showed an ability to reset that clock. Mice were bred, genetically modified to lack photoreceptors. They showed similar abilities, making it possible to rule out non-photic cues. How could this be? The basic mystery was solved in the early part of the present century, when it was discovered that there is a population of “intrinsically photosensitive Retinal Ganglion Cells” (ipRGCs). As the name suggests, these ipRGCs are stimulated by light directly. They contain melanopsin, a photopigment, different from the opsins in rods or cones. Their output modulates a variety of non-visual responses to light from the pupillary light reflex to heart rate to core body temperature.

Now suppose that you want to know how the pupil responds to “light”. Most of us probably assumed that an increment of N units of light produces a pupillary contraction of X%. In fact, it is more complex. The ipRGCs that send the relevant signals get input from five photopigments (the rods, 3 cones types, and the melanopsin in the ipRGC, itself). Cones are fast but relatively high threshold. Rods are slower but more sensitive. Rods and cones drive the initial response. Once the light stays constant, the state of the pupil becomes more and more driven by the still slower and relatively insensitive melanopsin response. Since the peak of the melanopsin absorption spectrum is about 480 nm, that means that those shorter wavelengths come to have a stronger influence on the pupil over time (on a scale of minutes).

The fine details of the pupil response may not have vast behavioral consequences. Circadian effects, on the other hand, have an impact on everything from emotions to industrial accidents and they are under the influence of similarly complex signals from ipRGCs. So, how should you light your factory for your nightshift workers? Do you want to deliberately stimulate the ipRGCs in order to shift the workers’ circadian phase? Do you want to spare the ipRGCs so as not to shift them? And what mix of wavelengths of different intensities will do what you want? We don’t know the answers yet and, as Lucas et al. warn us, finding those answers will require rethinking how we measure that light.—J.M.W.

AGING

The elixir of slowed cognitive aging?

Rebok, G.W., Ball, K., Guey, L.T., Jones, R.N., Kim, H.K., King, J.W.,…Willis, S.L. (2014). Ten-year effects of the Advanced Cognitive Training for Independent and Vital Elderly cognitive training trial on cognition and everyday functioning in older adults. Journal of the American Geriatrics Society, 62(1), 16–24.

Everybody talks about cognitive aging, but nobody does anything about it. Or at least nothing effective. We all know that certain mental faculties decline with age. While some might seek remedies in the pharmaceutical realm (http://www.salon.com/2013/12/29/sciences_obsession_the_search_for_a_smart_pill/), psychologists, understandably, have tried to devise cognitive training regimens to stave off Alzheimer’s, mild cognitive impairment, and normal cognitive aging processes. However, the results have been disappointing. While training regimens can produce immediate benefits on the trained tasks, long-term benefits are elusive, and the gains do not generalize (Melby-Lervåg & Hulme, 2013; Papp, Walsh, & Snyder, 2009); playing Call of Duty: Ghosts is unlikely to keep you sharp in your old age (Boot, Blakely, & Simons, 2011).

However, a new study may have overcome the limitations of previous work in this area. Rebok and colleagues (2014) report the results from a ten-year intervention study in healthy, independently-living elderly participants. These participants were recruited in 1998 and 1999 for the ACTIVE (Advanced Cognitive Training for Independent and Vital Elderly) study. 2832 participants, aged 65 and older at the time, living in six US metropolitan areas, were randomized to four different treatment groups: Memory Training; Reasoning Training; Speed Training; and Control (no training, no contact). Those in the training groups received ten hour-long sessions over a 5–6 week period; a randomly selected subset received booster training at 1 and 3 years. Participants were tested immediately, and again at 1, 2, 3, 5, and 10 years after the initial training session. In addition to the large sample size and long follow-up period, what makes the study especially useful is that, in addition to the usual neuropsychologically-based outcome measures, they included a set of measures intended to determine whether the training regimens generalized to participant’s daily life, including self-report and performance measures.

The three training interventions produced improvements in their specific domain at immediate test, and advantages over the control group persisted to the five year retest. At ten years, reasoning and speed training were still effective, but the memory training group’s advantage over control participants had dissipated. However, all three training groups led to improvements in self-reported daily life activities, with 60–70% as good or better than they had been at the baseline testing. Interestingly, the performance-based tests showed no advantage for the training groups at ten years. Thus, the promise of the study hinges on whether self-report measures or “objective” performance tests should be considered a more valid measure of cognitive function in everyday life.—T.H.

Additional References

Boot, W. R., Blakely, D. P., & Simons, D. J. (2011). Do action video games improve perception and cognition? Frontiers in Cognition, 2, 226. doi:10.3389/fpsyg.2011.00226

Melby-Lervåg, M., & Hulme, C. (2013). Is working memory training effective? A meta-analytic review. Developmental Psychology, 49(2), 270–291. doi:10.1037/a0028228

Papp, K. V., Walsh, S. J., & Snyder, P. J. (2009). Immediate and delayed effects of cognitive interventions in healthy elderly: A review of current literature and future directions. Alzheimer’s & Dementia, 5(1), 50–60. doi:10.1016/j.jalz.2008.10.008

EVENT PERCEPTION

Parsing the stream of experience

Paine, L.E., & Gilden, D.L., (2013). A class of temporal boundaries derived by quantifying the sense of separation. Journal of Experimental Psychology: Human Perception and Performance 39(6), 1581–1597.

To make sense of the flow of experience, we need to chop it into singular events, each with a beginning, a middle, and an end. This parsing of experience is, itself, hierarchical. Although songs, stories, and gestures are singular wholes in themselves, often they too need to be analyzed into parts, and those parts have their own sorts of wholeness. But what really ARE the rules that govern how people parse the stream of sensory stimulation? And how can we get an experimental handle on this? Newtson (1973) asked people to view videos and press a button when they observed boundaries between events, and this sort of approach has been used often since those initial studies. However, a theory of the processes that control the structuring of events has proven elusive. Newtson (1976) hypothesized that observers segment the stream of action when striking changes occur in the features of the ongoing sensory flow. Zacks et al’s (2007) “event segmentation theory” (EST) proposes that event segmentation is controlled by the maintenance of event models in memory. A given model is retained as long as its predictions match unfolding experience. When experience diverges too far from the predictions of the currently active model, an event boundary occurs. At this point, the slate of event memory is wiped clean.

Inspired by EST, Paine and Gilden (2013) introduce an ingenious new method for analyzing the effectiveness with which event boundaries are triggered by different features in the stream of experience. The method makes use of an empirical effect that is interesting in its own right. Consider the following rudimentary task: Suppose that on each trial a disk (either red or green) is presented in the middle of the screen and your task is to press “1” as quickly as possible if the disk is red or “2” if it’s green. If the red and the green are equally salient with respect to the background, then mean RTs should be roughly equal on red- vs green-disk trials. Now let’s complicate the situation slightly by introducing an irrelevant feature to the disk: suppose that on some trials the disk is presented above fixation, and on other trials the disk is presented below fixation. Even though the location of the stimulus is irrelevant to the task, it does in fact influence response times. Specifically, it turns out that RTs are (1) fast on trials in which the color and the location of the disk are either both the same as they were or else both different than they were on the previous trial and (2) slow on trials in which one of the two features (color or location) is the same as it was on the previous trial and the other feature is different. So what we have is a powerful interaction across successive trials between the to-be-attended feature of color and the to-be-ignored feature of location in determining response time. This effect is not specific to color and location; it holds generally for stimuli that vary in two perceptual dimensions, one to be used in producing responses and the other to be ignored.

The key discovery of Paine and Gilden (2013) is that this interaction disappears if an event boundary intervenes between the two successive trials. They demonstrate this first by imposing a rhythmic structure on the trial sequence so that the interval between successive trials alternates between short and long. The interaction shows up powerfully for RTs on trials after the short interval but not for RTs on trials after the long interval. The authors hypothesize that this is because the trials closely grouped in time in this sequence are chunked together into single events; however, event boundaries are injected between trials separated by “rests” in the sequence. Through a careful sequence of control experiments, the authors succeed in firmly establishing the validity of this account.

What does this mean? It means we can measure the effectiveness of various sorts of features in the sensory flux at triggering event boundaries. Suppose, for example, that we suspect that a flash of light at a random location in the peripheral visual field produces an event boundary. We can investigate this question by running an experiment of the sort described above with the disks. Between some pairs of trials we present a flash of light at a random location in the periphery. If the interaction across trials between the to-be-attended and to-be-ignored stimulus features is disrupted by the flashes, we can conclude that the flashes do in fact insert event boundaries.—C.C.

Additional References

Newtson, D., & Engquist, G. (1976). The perceptual organization of ongoing behavior. Journal of Experimental Social Psychology, 12(5), 436–450.

Zacks, J. M., Speer, N. K., Swallow, K. M., Braver, T. S., & Reynolds, J. R. (2007). Event perception: A mind/brain perspective. Psychological Bulletin, 133, 273–293.

NUMEROSITY PERCEPTION

The Neural Organization of Numerical Magnitude

Harvey, B.M., Klein, B.P, Petridou, N., Dumoulin, S.O. (2013). Topographic representation of numerosity in the human parietal cortex. Science, 341, 1123–1126.

There can be little doubt that the development of mathematics into a formal system of thought is one of the crowning achievements of the human mind. And yet, there is increasing evidence to suggest that our ability to understand precise mathematical operations may be founded on a much more intuitive and approximate sense of numerical magnitude that is shared by infants and animals alike. Moreover, there is increasing evidence that this intuitive sense of numerical magnitude functions in a manner similar to other basic feature dimensions such as the perception of color, luminance, and orientation. For instance, just as observers can selectively adapt to bars of different orientations, they can also selectively adapt to arrays of dots expressing different magnitudes. And, just as observers are less sensitive to a vertical bar after staring at a vertical bar for 60 seconds, observers are more likely to underestimate the magnitude of a medium-sized array of dots after staring at a large-size array of dots for 60 seconds. Conversely, they are more likely to overestimate the magnitude of a medium-sized array of dots after staring at small-sized array of dots for 60 seconds. In addition, as with other feature dimensions, there is also evidence that the psychological scaling of numerical magnitude is seriously non-linear. Specifically, the psychological distance between small magnitudes such as one and two is much greater than the psychological distance between larger magnitudes such as seven and nine.

Another characteristic feature of perceptual (and motor) systems is that they are topographically organized in the brain, often in a manner that mirrors the receptor surface. For instance, visual information is said to be organized in a retinotopic fashion in area V1 of the brain (and elsewhere) because V1 neurons with adjacent receptive fields reflect adjacent locations on the retina. Although our intuitive sense of number does not derive from a dedicated sense organ, B.M Harvey and colleagues (Harvey et al., 2013) recently used fMRI scanning technology to investigate the extent to which the perception of numerical magnitude in human observers was systematically organized in the brain. Following other researchers, Harvey et al. focused their analyses on posterior regions in the parietal cortex. Observers passively viewed dot arrays ranging in magnitude from one to seven. After controlling for a variety of potentially important visual features such as dot size, surface area, luminance, density, and the amount of contour, Harvey et al. found evidence to suggest that numerosity was topographically organized in the posterior parietal cortex. In addition, consistent with the psychological scaling differences observed between large and small magnitudes, Harvey et al. also found evidence to suggest that the amount of cortical area devoted to smaller magnitudes was magnified relative to the amount of cortical area devoted to larger magnitudes. Together, these findings are important because they provide strong evidence that the foundations of higher-level cognition share certain organizational features in the brain with more basic perceptual and motor systems.—B.G.

SENSATION OF AGENCY

Did I do that?

Timm, J., Schönwiesner, M., SanMiguel, I., & Schröger, E. (2014). Sensation of agency and perception of temporal order. Consciousness and cognition, 23, 42–52.

Did you ever have the experience of performing an action, such as pressing a button on the remote control, immediately before some unrelated event, like the ringing of a doorbell? If so, you might have felt “responsible” for the unrelated event, even though you knew perfectly well that your action was not the cause. The feeling that a specific event is a consequence of one’s own action is known as the “sensation of agency”. It is probably intuitively clear that the sensation of agency is tightly coupled to the temporal order of the action and event. Specifically, it seems reasonable to expect that your action has to precede the event by a short interval of time in order to generate the sensation of agency. However, such coupling of sensation of agency and temporal order judgment might not be so simple. There are many different demonstrations of illusory perception of temporal order, and it is not obvious that our sensation of agency is similarly susceptible to illusions.

To test whether sensation of agency is tightly coupled to temporal order judgment, Timm and colleagues employed a temporal order illusion in which the participants performed a button-press action and were presented with a stimulus (visual or auditory). There were two conditions. On the majority of the trials of the real-time condition, the stimulus immediately followed the action. On the majority of the trials of the delay-time condition the stimulus was presented 100 ms after the action. On the critical minority of the trials in both conditions, the stimulus appeared at an unpredictable, variable time before or after the action. The participants were asked either to judge the temporal order of the action and stimulus (i.e., which was first the action or the stimulus), or to indicate their sensation of agency (i.e., whether or not their action generated the stimulus). These conditions produce a temporal order illusion. In both real-time and delay-time conditions, the stimulus had to appear before the actual button press in order to be perceived as appearing simultaneously with the action. The size of this effect, the time interval by which the stimulus had to precede the action, was shorter in the delay-time condition, probably because the participants were adapted to a longer inter-stimulus-interval between action and stimulus. The important finding in the Timm et al paper is that there is a similar pattern of results for the sensation of agency. Thus, if a stimulus was perceived as following the action, it also generated a sensation of agency. This held true even if, in reality, the physical stimulus actually preceded the action. The fact that temporal order judgment and sensation of agency varied in a very similar fashion as a function of the time between the action and stimulus suggests that these two aspects of perception may be mediated by the same mechanisms, and provides support to the claim that the sensation of agency is highly related to perception of causality.—Y.Y.