Response Selection Multitasking Bottleneck: Can it be bypassed? Part 2

In the yesterday's post, I summarized the conclusions of a seminal study that helps explain the extent to which people can perceive multiple stimuli, select multiple responses, and perform multiple actions concurrently--i.e. to multitask. In the cases of perceiving and performing, it appears people can multitask. However, in the case of selecting responses, it appears people cannot multitask. Response selection is said to be a bottleneck in the multitasking thoroughfare.

Assuming that model is correct, one alternative to explore is whether it's possible to bypass response selection. A commonly pursued route to bypass response selection is automatization through training and practice. According to the literature review of the October issue of the journal Memory and Cognition article titled "Bypassing the central bottleneck after single-task practice in the psychological refractory period paradigm: Evidence for task automatization and greedy resource recruitment," several studies have been performed to determine the feasibility of training people thoroughly enough that they can bypass it (Maquestiaux et al.). As often seems to be the case in soft sciences like psychology, the evidence is not clear, but the general trend hints it is possible under certain conditions.

In the 1960's, at least three studies were conducted to discover whether subjects could be trained well enough that they would generate responses automatically without the need to select one. The results of all three studies were negative--automatization was not achieved by any of the subjects. Maquestiaux et al. believe the one confounding factor that needed to be controlled in those studies was their reliance on subjects producing manual responses, rather than a combination of verbal and manual. In 1999, a study by researchers Van Selst, Ruthruff, and Johnston showed that allowing for verbal responses enabled subjects to nearly eliminate the response delay, but it was still there--except for one subject. That subject participated in another study in 2003, and it was found that the reason the delay was not observed was because the perception time for the second task happened so quickly that it was able to be processed before the first task, thus there was no delay observed for the performance of the second task. When the order of the tasks was reversed, the delay materialized. So the bottleneck was still there. In the last 8 years, additional studies have controlled for the confounding factors of the previous studies, and some have shown the bottleneck to have been bypassed. Maquestiaux et al. point out, though, it may be that the response delays are just small enough to escape detection by the equipment being used to take the measurements (among other experimental design flaws).

In 2006, researchers named Ruthruff, Van Selst, Johnston, and Remington conducted a study in which 4 out of 18 subjects appeared to bypass the response selection bottleneck--even though all of the aforementioned design weaknesses were controlled for. However, those 4 subjects were only successful in one of two sets of trials. One explanation for the subjects' failure in one experiment but success in another is that the one of the tasks exhibited a "greedy" nature when it was presented first. Ruthruff et al. hypothesized that even though the greedy task could be performed automatically, it had the property of monopolizing all processing resources if it was perceived first, but it would revert to automatic processing if it was perceived second.

With all of that background out of the way, I will be ready to discuss Maquestiaux et al.'s study in my next post. In Maquestiaux et al.'s study, they attempt to determine the correctness of Ruthruff et al.'s hypothesis that certain tasks are greedy, thereby also determining whether or not tasks can be automatized (and bypass the response selection bottleneck) as long as the greedy task is presented second.

I know you're as excited as me to find out the conclusion of this nailbiter, so stay tuned!

Response Selection Multitasking Bottleneck: Can it be bypassed? Part 1

For the next few posts, I will be discussing a study by researchers François Maquestiaux, Maude Laguë-Beauvais, Eric Ruthruff, and Louis Bherer titled "Bypassing the central bottleneck after single-task practice in the psychological refractory period paradigm: Evidence for task automatization and greedy resource recruitment" published in October of 2008 in the journal Memory and Cognition. This study represents an "aha moment" for me, as I believe I am beginning to understand the extent to which people are capable of performing multiple tasks simultaneously. The article forced this epiphany on me during its literature review section.

While reviewing the history of dual-task research, Maquestiaux et al. mentioned a seminal study titled "The 'psychological refractory period' and the timing of high-speed performance: A review and a theory" that was published in a 1952 issue of the British Journal of Psychology. The author of that study, Alan Welford, observed that the closer to simultaneously that two stimuli are presented, the greater the amount of time it takes for subjects to respond to the stimulus that was presented second. Welford referred to the delayed response to the second stimulus as a "psychological refractory period." Further, he observed that subjects were capable of perceiving multiple stimuli simultaneously and performing multiple responses simultaneously, but something happened between the perception of the stimuli and the performance of the responses that always caused a delay in the second response. From these observations, he inferred that a processing bottleneck occurred in the middle of his subjects' task activities. This bottleneck, he explained, resulted from the brain's need to select a response to each stimulus, and that selection of responses had to occur consecutively, not concurrently.

Another way of looking at this is to say the senses are capable of receiving (and temporarily storing) multiple inputs simultaneously, and the body is capable of performing (and delaying) multiple outputs simultaneously, but the brain only seems to be capable of making one decision at a time. That, in a nutshell, is my "aha moment." I think it's quite an elegant explanation for the observed phenomena. Moreover, that's how my senses, brain, and body "feel" like they work.

As the title for this post series indicates, what I'm interested in now is whether that bottleneck can be bypassed, and if so what the limits are. In my next post, I will discuss the progress that researchers have made toward answering that question by sharing the findings of a few more studies mentioned in the research review of the Memory and Cognition article that led up to the 2008 study; in part 3 of this series, I will share its results.

I'm anxious to get started.

Response Selection, Multitasking's Arch Enemy

During the last few weeks, I've shared several studies that suggest dual task performance--of both concurrent and switching varieties--can be improved through training, technology-based training specifically. I even wrote yesterday that I have accepted that conclusion as a fact. No sooner did I think that thought, than I stumbled upon a study describing a scenario in which training appears not to improve dual task performance. You might think I'd feel dismayed, but I'm actually pleased. You see, most of the studies I've shared so far have examined very low-level, unpractical examples of multitasking, and I haven't felt like they are sufficiently representative of complex, real-world, thought-demanding situations--"multitasking lite," so to speak. The study I will share today is very practical and sufficiently represents a complex, real-world, thought-demanding situation: Talking on a cell phone while driving.

The study, "Effects of Simulator Practice and Real-World Experience on Cell Phone-Related Driver Distraction," was conducted by Professor of Psychology David Strayer and his research assistant, Joel Cooper, at the University of Utah and published in the journal Human Factors in December of 2008. The study suggests that training does not improve one's ability to concurrently talk on the phone and drive. One of the strongest pieces of evidence in favor of their conclusion was that subjects who claimed to have extensive prior experience driving while talking on the phone performed equally as poorly as subjects who reported having little or no experience doing so. Another strong piece of evidence was that, after four days of training, the subjects did not demonstrate improved competency with talking on the phone while driving. Strayer and Cooper also observed that even when subjects practiced a driving scenario repeatedly (so all dangers were predictable), the only aspect of their driving that improved was crash avoidance, while all other factors--following distance, speed, and braking--remained unimproved; at the same time, when the subjects completed the scenario without talking on the phone, all factors did improve.

Strayer and Cooper conceded that the results of other experiments contradict theirs. They explained that the other experiments failed to mirror the unpredictable nature of driving in the real world. "(T)he findings (of other experiments) may not apply to less predictable aspects of driving, such as strategic vehicle control and response to sudden-onset events" (Strayer & Cooper, 2008, p 894). The reason those unpredictable events are important is because they require drivers to make rapid, thoughtful decisions. Decisions seem to be resistant to simultaneous processing, as Strayer and Cooper pointed out when they wrote, "Although driving and cell phone conversation are thought to be resource compatible (Horrey & Wickens, 2003), simultaneous task performance may be limited by the irascible central processing bottleneck thought to exist in response selection" (p 894). They accept the hypothesis that automatization of tasks through practice enables people to perform multiple tasks concurrently, but they believe making the decisions inherent in conversing and driving might not be automatizable.

One thing I learned from this study that was particularly enlightening is that "response selection" seems to limit the types of tasks capable of being performed concurrently. Now I wonder what other classes of tasks exist that are also resistant to concurrent performance. I think I am beginning to figure out how researchers could clinically tell the difference between dual task performance of the concurrent and switching varieties (but I'll not reveal my thoughts yet). Lastly, I hypothesize that rapid response selection is positively correlated with task switching efficiency. I wonder if there is any research related to my hypothesis.

Strategy Video Games + Senior Citizens = Increased Task Switching Competence

According to a study the April 2008 issue of the journal Psychology and Aging, senior citizens can improve their ability to switch tasks rapidly by playing "real time strategy video games." In this study, subjects who were approximately 70 years old, in good health, with good vision, and had little/no experience with electronic games were selected to learn and practice Rise of Nations, a real time strategy video game. After 23.5 hours of training and practice, the researchers discovered that the subjects' ability to switch tasks rapidly had improved in two ways: Their speed increased by 33% while their accuracy increased by 9%. The increase in speed was statistically significant with a p-value of 0.01 (Basak et al., p 772).

The speed of task switching was measured like this: Subjects viewed a single digit on top of a solid-colored background of blue or pink. "If the background was blue, participants used one hand to report as quickly as possible whether (the number) was high or low. If the background was pink, participants used their other hand to report as quickly as possible whether the number was odd or even" (Basak et al., p 768). Naturally, response times were faster when two consecutive digits were on top of the same colored background, since they didn't have to switch tasks. However, when the background color changed between two consecutive digits, subjects responded slower because they had to queue up the other set of instructions--i.e. they switched tasks. The difference between the faster and slower responses was calculated and called the task switching "cost." As subjects gained more experience with the video game, the difference in their response times decreased (because their slower responses got faster).

Having read this study as well as the others already discussed in this blog, I have begun to accept as fact that technology can be used to improve task switching competency. Now, among many other things, I wonder how many hours of training it would take before the benefits plateaued, how long the benefits last after training stops, and how task switching affects the quality of learning. I've also begun to wonder to what degree task switching competency will be important for success in the "real world" and how that importance could be quantified. One thing's certain: I've got a lot more research to do.

Multitasking: From Unlikely to Unavoidable?

Today is March 27th. This is not an April Fool's joke.

According to Science Daily, in the upcoming April 14th issue of Current Biology, researchers at the University of Texas-Houston Medical School will publish a study titled "Attention Alters Visual Plasticity during Exposure-Based Learning." The study's lead author, Valentin Dragoi, says that "(I)gnoring the stimuli presented over days of exposure (is) more effective than actually attending them...(T)his finding can be explained by the fact that, typically, attention filters out unwanted stimuli so they are not consciously processed. However, in the absence of attention, stimuli are able to escape the attentional mechanisms" that would have otherwise filtered them out "and induce robust learning after multiple exposures" (Science Daily, 2009, para 8).

You must be wondering, "Did I understand that right? Trying to devote my undivided attention to something causes me to miss lots of details, and not paying attention enables my mind to perceive and learn more? Huh?" That was my reaction too.

On the basis of this study, it seems that focused attention functions like a microscope: It allows people to see a subset of details within a very narrow field of view. In the periphery beyond the microscope's lens, of course, there is much more that could be observed, but it is beyond one's conscious attention. All that data might not completely escape notice, though. The study suggests that one's sensory system and mind are capable of concurrently perceiving an untold portion of the multitude of stimuli beyond the object on which one has focused his attention.

Sounds too good to be true, doesn't it? When you look at the details of the study, you realize that it takes quite a bit of extrapolation to get from the experimental conditions to anything resembling "real" learning situations. Here's a rough summary of the procedures:

Six subjects fixated on a certain position of a computer screen at which circles filled with parallel lines flashed rapidly. From time to time the circle's orientation or color changed. Subjects were instructed to discriminate these changes by pressing a button. Concurrently, on the periphery, another set of similar circles flashed, but subjects were forced to ignore them (in order to be able to identify changes to the circle on which they were instructed to fixate). After many sessions of being habituated to watching the circles and becoming familiar with the angle at which the lines slanted, then the subjects were tested to see if they could identify circles in which the lines were slanted at variety of different angles compared to the orientation on which they had been trained, and that's where the interesting thing happened. The subjects were able to discriminate a much wider range of changes in the orientation of the circle on the periphery. Without trying, they had learned about it's orientation, and their knowledge of it was more generalizable than their knowledge of the circle on which they had fixated. There is more to the experiment, but that gives you a general idea of how it worked.

Knowing the details of the experiment, it doesn't seem quite as dramatic as Dragoi's characterization made it sound. It's not dramatic that my senses monitor my surroundings beyond the focus of my attention. It's also not dramatic that repeated exposure to certain stimuli, even unattended, would cause my sensory system to send messages to my brain, yielding increasingly stronger neural connections. I suppose in the strictest sense, these phenomena "count" as multitasking and learning, but I am interested in more demanding forms of multitasking and learning. I am struggling to think of any practical ways I might apply the findings of this study. Here's a creazy idea: Instead of explicit teaching, I could broadcast educational subliminal messages in my classroom while we all watch television and play video games! (Just kidding, of course.)

The Truth about Digital Natives and Multitasking

In 2001 Marc Prensky, founder of Games2Train and self-proclaimed "visionary and futurist," coined the meme "Digital Natives." Among other qualities, Prensky wrote that Digital Natives learn better if teachers present content "faster, less step-by-step, more in parallel, and with more random access" (p 4). Part of his rationale for making such an assertion is based on the idea of neuroplasticity. He thinks--because Digital Natives have spent countless hours surfing the Web while carrying on various e-conversations, listening to music, and maybe even studying--that plasticity has rendered their brains uniquely capable of learning in a multitasking environment. Is he right?

TIME magazine published an article titled "The Multitasking Generation" in March of 2006 that sheds some light on the question. In addition to presenting anthropological vingettes of Digital Natives and their relationship with technology, the article included interviews with several scientists who study multitasking. They seem to disagree with Prensky's assertion. For example, chief of cognitive neuroscience at the National Institute of Neurological Disorders and Strokes Jordan Grafman says "Kids that are instant messaging while doing homework, playing games online and watching TV, I predict, aren't going to do well in the long run" (para 17). Another expert, David E. Meyer, director of the Brain, Cognition and Action Laboratory at the University of Michigan says "The toll in terms of slowdown (due to task switching and multitasking) is extremely large--amazingly so" (para 22). The article also mentions that Meyer often performs multitasking experiments with Digital Natives as his subjects, and he observes that they are perform poorly trying to multitask.

The article does not claim that people are incapable of multitasking. It reminds readers that people multitask frequently, when driving for example. In fact, it even points out that sometime in late adolescence and early adulthood, multitasking competence peaks. However, similar to previous posts on this page, the article explains that multitasking ability--even for Digital Natives--is limited. Performing multiple automated tasks like walking, talking, and keeping an eye out for danger is no problem. But when people try to perform multiple tasks that require serious thought--like learning--then bottlenecks form and information is lost. So, should teachers follow Prensky's advice? Probably not.

Computer Training Involving Multitasking and Working Memory and Their Effect on Fluid Intelligence

About one year ago, an article titled "Improving Fluid Intelligence with Training on Working Memory," was published by the Proceedings of the National Academy of Sciences. The article described an experiment in which researchers used a computer program to improve subjects' working memory. The reason for the experiment was to test the researchers' hypothesis that a relationship between working memory capacity and Fluid Intelligence exists.

The article explains that Fluid Intelligence, notated in research literature as Gf, involves one's ability to identify patterns, solve puzzles, learn new things, and to process information rapidly (p 1). It is contrasted with Crystallized Intelligence (Gc) which is characterized by the total quantity of facts one has learned in his lifetime. If one wishes to increase his Crystallized Intelligence, he would need simply to put forth the effort necessary to learn new concepts, skills, and relationships. Experts have suggested that Fluid Intelligence, on the other hand, cannot be increased intentionally (p 1). If the results of this study are correct, however, those experts might have to revise their position. The results of the study showed a positive correlation between the quantity of time subjects invested training their working memories with the computer program and their performance on a test of Fluid Intelligence.

The purpose of this blog is to explore issues related to multitasking, task switching, and technology--not intelligence. The article referred to above is relevant to this blog. The computer program used to train the subjects required them to multitask. Here is roughly how it worked: The computer screen that subjects viewed displayed a small square that changed to 1 of 6 different positions every 3 seconds. Simultaneously, the subjects' headphones played 1 of 6 spoken letters. Users were instructed to observe the position of the small square and to associate that position with the spoken letter. If the same letter and position recurred, subjects were to press a button indicating they noticed a repeated combination. Each time the subjects identified a match correctly, the computer increased the level of difficulty by placing more distractors between matches. Clearly subjects were required to perform and attend to multiple cognitive processes concurrently for this experiment.

The study also showed that each training session produced improved performances by the subjects. The researchers believed that their subjects' working memories were improving. It also seems reasonable, however--given the nature of the training task--to say that the subjects' capacity for multitasking improved at each training session. As yesterday's post implied, working memory capacity seems to be the arena in which multitasking is performed.

With each post, it seems more evidence suggests multitasking is possible and that it can be improved with computer-based training, as can task switching. What are the limits? What about training that is not so contrived, like video games? There's only one way to find out...back to the virtual books!