Working memory (WM) capacity reflects an important constraint on complex cognitive processing. Experts can overcome this constraint in their area of expertise, demonstrating enhanced WM capacity for domain-specific information. This increased WM capacity might either arise automatically from practice with structured information or reflect strategic organization of domain knowledge. The current study used a new sequential visuospatial (SeVi) WM task to test whether information-specific gains in WM would occur when the structured information was covertly embedded in practice on 20% of trials. After 2 hours of training, participants exhibited overall improvement in the task and dramatic gains in WM capacity when information to be held in mind was drawn from the practiced structure. The increase in capacity for practiced information that was observed here suggests that the increased domain-specific WM capacity seen in experts likely results from their increased experience with information in their areas of expertise.

Gigler, K.L. & Reber, P.J.

Problem/Major Purpose: Recent research demonstrating improvements in working memory (WM) capacity has challenged the idea that WM capacity is an immutable cognitive trait. Indeed, relatively modest training protocols have been shown to lead to significant improvement. Because WM is a core cognitive process, increasing capacity has the potential to enhance performance on a wide range of cognitive functions. A critical question is the degree to which gains exhibited on the WM training task transfer from the specifically-trained skill to other tasks, including other cognitive functions. The current work examined the ideas of training and transfer through an experiment which utilized a newly-designed working memory task.

Procedure: Participants were 9 individuals between the ages of 21 and 36 (7 female). Participants received 10 hour-long sessions of training (2000 trials total) on a novel visuospatial working memory task. Each trial consists of two phases: the presentation phase, during which participants see a sequence of moving visual cues and must hold that sequence in WM, and the response phase, during which participants attempt to replicate the sequence. The training is adaptive, adjusting the length of presented sequences based on performance in order to keep training near each individual’s WM span. Transfer to other cognitive functions was assessed through the use of a battery of cognitive tasks completed by participants both before and after training. Improvement on all tasks was examined through the use of paired t-tests.

Results: Participants demonstrated significant and continued improvement on the WM task across training. Reliable improvement was also demonstrated on several other WM tasks, as well as on tests of processing speed, attention and long-term memory.
Implications/Conclusions: These results indicate that the WM task used in this work is a viable candidate for WM training, and that further, such training may improve not only WM-specific performance, but also other cognitive functions and abilities. That WM training can improve performance on cognitive functions such as processing speed and long-term memory suggests its use could be beneficial in work with older adults, children with attentional deficits, and many other clinical populations.

Gigler, K.L. & Reber, P.J.

Cognitive training to slow or reverse age-related cognitive decline is based on the premise that core cognitive functions can be strengthened by challenging, repetitive practice. A good candidate cognitive process for improvement is working memory (WM), which supports effective problem solving, complex language comprehension and encoding into long-term memory; increasing WM capacity may therefore produce gains in these related cognitive functions. Across two experiments, we show the promise of a new Sequential Visuospatial (SeVi) WM task for cognitive training. Each trial of the SeVi task consists of a presentation phase, during which participants see a sequence of moving visual cues and hold a series of response locations in WM, and a response phase in which the series is reproduced. The training is adaptive, increasing the length of presented sequences as performance improves in order to keep training near each individual’s WM span. In Experiment 1, young adult participants completed 10 hour-long training sessions (2000 trials) and not only improved on the SeVi task, but showed transfer via improved measures of long-term memory and processing speed on the CogState assessment battery. In Experiment 2, older adult participants (healthy and MCI) completed 6 45minute-long training sessions (approximately 900 trials). Participants improved significantly on the SeVi task, and reported enjoying the WM training. Transfer to other measures of cognitive function is expected. It is hoped that gains made in training lead not only to transfer to other cognitive functions, but also to improvements in real-world memory performance and quality of life.

Link if embed doesn’t work: http://www.youtube.com/watch?v=LyU5v0ZYMjI

The point is not to make fun of this kind of skill acquisition but to raise the question of why we do this.  In particular, the first people who started racing through stacking cups had to have been doing it because it provided some intrinsic reward for them for motivation.  I think skill learning does have some intrinsic reward (and maybe more for some people than others), possibly related to the involvement of cortico-striatal loops that use or are at least near reward processing (and the common dependence on dopamine).

Emi M Nomura and Paul J Reber, in press Brain Sciences

Abstract: Considerable evidence has argued in favor of multiple neural systems supporting human category learning, one based on conscious rule inference and one based on implicit information integration. However, there have been few attempts to study potential system interactions during category learning. The PINNACLE (Parallel Interactive Neural Networks Active in Category Learning) model incorporates multiple categorization systems that compete to provide categorization judgments about visual stimuli. Incorporating competing systems requires inclusion of cognitive mechanisms associated with resolving this competition and creates a potential credit assignment problem in handling feedback. The hypothesized mechanisms make predictions about internal mental states that are not always reflected in choice behavior, but may be reflected in neural activity. Two prior functional magnetic resonance imaging (fMRI) studies of category learning were re-analyzed using PINNACLE to identify neural correlates of internal cognitive states on each trial. These analyses identified additional brain regions supporting the two types of category learning, regions particularly active when the systems are hypothesized to be in maximal competition, and found evidence of covert learning activity in the “off system” (the category learning system not currently driving behavior). These results suggest that PINNACLE provides a plausible framework for how competing multiple category learning systems are organized in the brain and shows how computational modeling approaches and fMRI can be used synergistically to gain access to cognitive processes that support complex decision-making machinery.

Prof Dario Maestripieri from U. Chicago is on promoting his new book: Games Primates Play.  I’m not entirely sure why they invited me — the book is about comparative psychology, evolutionary biology and the “game” reference in the title is more game theory/economics.  Fortunately, although those aren’t actually my research area, I can probably stay engaged in the conversation because I tend to read in those areas.

Because I like distant analogies, I’m strongly tempted to claim in response: Implicit learning is the dark matter of memory

To ground the analogy, the idea is that what we normally think of memory (the visible, regular matter) is our conscious experience of memory retrieval.  Implicit learning and memory is this broad set of invisible memory processes that shape our perceptions, actions and thoughts that aren’t directly observable by our subjective experience.

To stretch the analogy, dark matter is thought to constitute 83% of the universe (via Wikipedia).  Is it even possible to quantify the percentage of “memory” that is implicit so that we could compare numbers?

One approach would be based on neurobiology.  Under our “plasticity principle” model of implicit learning, virtually all synapses in the brain have some inherent ability to be modified by experience. Those changes that depend materially on the medial temporal lobe are thought to contribute to explicit, conscious memory while all the other changes are reflected in implicit memory phenomena.  Just based on brain size, we might make the claim that implicit memory should reflect the majority of these plasticity events, perhaps even around 80% of all memory events in the brain.  However, we’d really need some estimate of the rate of change events (or even magnitude) and they might be much more common and significant in the MTL, potentially reducing the percentage by a lot.

Another approach would be based on some quantification of the behavioral consequence of memory and try to carve up changes apportioned based on the relative roles of explicit and implicit memory processes.  After brief consideration of this problem, I’m going to recommend the biological approach.

I’m tempted to invite argument for and against the dark matter hypothesis for students of memory.  Probably the concept is a bit too bizarrely abstract for good debate, though.

http://newsfeedresearcher.com/data/articles_m9/game-study-cognitive.html

The source lab looks to be doing some interesting stuff: http://www.gainsthroughgaming.org/index.html

The published report isn’t quite officially out yet (I think this link goes to a corrected proof): http://www.sciencedirect.com/science/article/pii/S0747563212000143

Some things to note — participants were 60-77 years old, played 14 hours of game play and only the lowest scoring participants exhibited reliable gains.  (Bonus statistical question — why would this raise some concerns about the reliability of the result?).  Looks like you can get a lot of media attention even for not quite totally awesome data in this area.  It’s probably deserved since it’s probably a real effect.  I suspect you’d want a bit more time in-game to see something more robust and I don’t see a real measure of “processing speed” in their assessments.

One additional thing to note — this is actually a hard study to pull off.  Getting ~40 adults in this age range to play WoW for a decent chunk of hours reflects a significant amount of work just in logistics and recruiting.  I think these gaming studies look easy from the outside, but when you try to run a training study of course you realize it’s not.

Projects
Depletion: This data is very peculiar. It seems like we do have something happening with the depletion manipulation, but I think there might be too much variance in pre-experimental depletion (i.e. depletion prior to manipulation) to make sense of things.
Single-Quad: Ramping up data collection.
Implicit Explicit: I’ve been ramping up editing and should be shooting this back to you shortly. Since we’re going to submit to cognition next, we can be somewhat liberal with word count. Obviously, concise is good, but at least we can expand on ideas.
Wiitar: Sweet Mary, it’s done. Now it just needs to be analyzed…
mTurk Delay: Draft mode.
mTurk Fragments: What’s happening with this one?
A/B/C: Dave is going to rescore his Sound vs. No-Sound and let us know if he finds anything interesting. This is sitting on the backburner while I hammer out the I/E manuscript – that is definitely at the forefront right now.

Projects
Depletion: Collecting Data
Single-Quad: Collecting Data
Implicit Explicit: I’ve been working on editing this. Few Notes: The in-text results only report means for trained sequence and foil sequences separately, but not the subtraction score. However, the figures only have subtraction scores. I think I’ll edit this to be more consistent (i.e. add in-text subtraction numbers). Also, the correlation between recognition score and SISL score in Experiment 1 dropped from .3 to .22 and is no longer significant. The difference between the high-rec and low-rec post-hoc groups has also dropped so the significance is sitting at p = .09. Upon closer inspection, the correlations between recognition and SISL score in Experiment 1 are hugely driven by the explicit pre-training group (.39) as opposed to the implicit group (.03). I’m thinking we’ll want to reframe this in the manuscript (or possibly just get rid of it?). The motivation for Experiment 2 is still valid even without the post-hoc results (more robust explicit training).
Wiitar: The manual rescoring is going well. I keep finding new categories that my old code didn’t catch. For instance, someone would always press the next button without lifting off the previous button (so, essentially, their button press would look like D, DF) and then lift off the previous button (D) before strumming. Completely legitimate and allowed, but a very different left-hand response profile. However, I think this will essentially collapse into categories as if the person had released the previous key like every other normal person who was in the experiment. Interesting none-the-less (at least to me).
mTurk Delay: Still being drafted.
mTurk Fragments: Totally cool data. Very interested to see the next round. However, I was thinking about the “precognitive” trigrams in the 3-fragment experiment. Essentially, the first two cues are predictive of the third. From a response stand-point, that’s not argued. However, if you take spatial layout into account, that on-screen triplet is unique and viewable prior to responding to any of the cues. It’s possible that the appearance of that learned spatial layout (in this instance, a small triplet of circles) is cuing the knowledge in a way that is allowing for the expression of knowledge across all three items in the 3-item fragment.
A/B/C: I’ve got ABC data AND Sound vs. No Sound data from Dave. If I get tired of editing manuscripts and hand-scoring data, I might write a quick script to analyze these. ABC is clearly on the back-burner; but the sound data might be interesting (especially if it’s only a small time investment on my part).