Training of Executive Control Functions: Negative Transfer and Far Transfer

The Cognitive Neuroscience Society 2010 Annual Meeting was held last week in Montreal, Québec. Unfortunately, many European registrants were unable to attend because of the Eyjafjallajokull volcano. The meeting website has links to the PDFs for 67 of these "Volcano Posters".

One of those unable to attend was Dr. Jonas Persson of Stockholm University. He was scheduled to speak in the final symposium of the conference, which was on control of executive control, or who controls the "controller" in the brain (without resorting to a homunculus or an infinite regress of Mini-Me's). His co-author, Dr. Patricia Reuter-Lorenz of the University of Michigan, gave an interesting presentation about their work on the ups and downs of cognitive training: gains that transfer to other tasks across sessions and fatigue that transfers to other tasks within a session. The abstract is reprinted below.

---------------
Symposium Session 5
Tuesday, April 20, 1:00 - 3:00 pm, Westmount et al Ballroom

What Controls Executive Control? The Influence of 'Control Context'

Chair: Amishi Jha, University of Pennsylvania

. . .

Talk 3: Training and Depletion of Executive Functions: The Case of Interference Control

Jonas Persson¹, Patricia Reuter-Lorenz²; ¹Stockholm University, ²University of Michigan

Brain imaging reveals overlapping sites of prefrontal activation for different cognitive tasks suggesting they may share core executive processes. We tested this hypothesis by measuring behavioral interactions between memory tasks presumed to require interference control - a putative executive process that mediates selection from competing representations. Behavioral data show that different training regimens produce either negative or positive transfer from working memory to semantic and episodic memory task performance. We show that eight days of training on high interference versions of three different working memory tasks increased the efficiency of interference control on the training tasks and on untrained tasks in new memory domains. In contrast we have also demonstrated negative transfer and process-specific “fatigue” effects indicating that control efficiency in a second task is diminished by high control demands in a prior task immediately preceding it in time. This suggests that interference control is a finite resource that can be temporarily depleted. Functional magnetic resonance imaging (fMRI) was used to elucidate the mechanisms associated with decreasing efficiency or resource depletion of the interference control process. Along with reduced performance, fMRI indicates negative transfer is associated with reduced process-specific activation, and increased homologous activation that may be compensatory. In sum, this suggests that interference control is an executive function that is both resource limited and plastic making it possible for training to alter its efficiency.
---------------

These findings became extremely relevant (albeit ignored) in light of the ultra-high impact paper by Adrian Owen et al. (2010) published in BBC/Nature on April 20, informing us there was "No gain from brain training" in a massive group of 11,430 volunteers. The research volunteers participated in online training exercises that tapped (1) reasoning, planning, problem solving; (2) short-term memory, attention, visuospatial processing, mathematics; or (3) ability to answer obscure questions using the internet.



Numerous other outlets have already summarized these results, so I won't attempt to do so here. The main issue was whether the gains obtained through simple practice effects transferred to tasks that weren't in the training set. The short answer is no.

Brain training doesn't boost brain power, work suggests (BBC)

Brain-training games don't work (Guardian)

Brain training games don't work (BPS Research Digest)

Brain-training games get a D at brain-training tests (Not Exactly Rocket Science)

Brain Test Britain - Results in-depth (BBC - Lab UK)

I am not a proponent of brain training games that make unsound claims lacking credible scientific evidence to back them up. But in light of the spectacularly negative findings of Owen et al., the question arises of whether their training regimen (10 min 3 times a week for 6 weeks) was adequate to produce significant effects. Predictably, those with a stake in the matter said no, the training was neither sufficient in duration nor applicable to other commercially available products.

BBC “Brain Training” Experiment: the Good, the Bad, the Ugly

Posit Science Disputes Results of the BBC Brain Training Study

A final issue concerns the population under study, a presumably "normal" group of 18-60 year olds without cognitive impairments or cognitive decline. Would the same training exercises help an older population at risk for dementia? On the other hand, seniors might be especially vulnerable to persuasive false claims from unscrupulous brain training vendors. Thus it's important to have scientifically valid and accessible research on whether cognitive training might help an aging population (Lustig et al., 2009). Coincidentally, the NIH is currently trying to reach a consensus on:

NIH State-of-the-Science Conference
Preventing Alzheimer’s Disease and Cognitive Decline

April 26–28, 2010
Bethesda, Maryland

Live webcast, also information on archived video and publicly available consensus statement.

Let us now return to the studies of Persson and Reuter Lorenz (2008). This work examined whether the training of a specific cognitive process (overcoming interference from competing representations in memory) in one task will improve the ability to overcome interference in a different task. For example, a set of four letters would be presented in a working memory task, followed by a delay and then a probe letter, which requires a yes/no response (was this letter a part of the set you just studied?). The interference arises when participants must reject a probe letter that was a member of the memory set on the preceding trial, but is not a member of the set on the current trial ("Recent Negative", illustrated below).


Figure 1 (D'Esposito et al., 1999). Target items were presented for 950 msec followed by a 7,050-msec delay period. This interval was followed by a probe letter for 1,500 msec that either was recently presented (Recent trial) or was not recently presented (Nonrecent trial). Note that, in the Recent Negative trial, the probe letter “P” is not in the target set of that trial but that it is in the target set of the two previous trials.

In the current experiment (Persson & Reuter-Lorenz, 2008), 48 young control participants were randomly assigned to one of three groups:

1. Interference Training (working memory requirements and interference)
2. Control 1 (working memory requirements only)
3. Control 2 (minimal memory requirements and no interference)

Training consisted of eight 40 min sessions over a two week period. Benchmark tests were administered on the days before and after the training program to evaluate interference resolution performance on non-trained tasks of working memory (item recognition with concrete nouns), semantic memory (verb generation), and episodic memory (paired associates). The experimental design is shown below.


Fig. 1 (Persson & Reuter-Lorenz, 2008). Schematic depiction of the basic experimental design. The illustration shows the tasks performed by each subject group during each phase of the experiment: pretraining (Session 1), training (Sessions 2–9), and posttraining (Session 10). All groups performed the same three transfer tasks in the pretraining and posttraining sessions. During training, the interference-training group performed tasks involving working memory (WM) and interference; Control Group 1 performed noninterference versions of the same tasks, and Control Group 2 performed similar tasks that required little WM and involved no interference.

Results indicated that Interference Training reduced reaction time (RT) measures of interference in each of the three transfer tasks (left bars), but control training did not (right bars).

Fig. 3. (Persson & Reuter-Lorenz, 2008). Mean interference-resolution scores (response time in the high-interference condition minus response time in the low-interference condition) on the transfer tasks as a function of group (interference-training group vs. control groups) and time (pretraining vs. posttraining). Results are shown separately for the (a) verb-generation task (semantic memory), (b) item-recognition task (working memory), and (c) paired-associates task (episodic memory). Error bars show standard errors of the means.

A previous study demonstrated the opposite sort of process-specific interactions, in the form of within-session fatigue effects (i.e., negative transfer) across a subset of these tasks (Persson et al., 2007):

The idea that tasks with overlapping neural representations may involve similar executive components was ... critical to our approach. Of particular interest were tasks requiring resolution of interference among competing representations. Within a single experimental session intensive training reduced the ability to resolve interference on a transfer task if the training task placed high demands on interference resolution. Negative transfer was absent when interference resolution was minimally required by the task, or when the training and transfer tasks did not rely on overlapping neural representations.

Based on these results, the authors concluded that targeted training -- which identifies and trains a specific cognitive process that is implemented by a known network in the brain -- can produce improvements (across sessions) that transfer to tasks other than those in the training set. Conversely, fatigue effects (within session) can result in negative transfer to related tasks. In the case of interference resolution for items held in memory (or "proactive interference"), the neuroimaging literature suggests that a particular area in the inferior frontal gyrus of the left frontal lobe is a key region for overcoming interference among competing memory representations in various types of tests (Jonides & Nee, 2006).

What does all this mean for the brain training biz and for the BBC/Nature study? Adequate training time and a brain-based approach to target highly specific cognitive processes and tasks will yield the best results.


ADDENDUM (April 10, 2011): The 2008 Psychological Science paper by Persson and Reuter-Lorenz has been retracted:

This article has been retracted by both of the authors. During further extension of this work, it was discovered that a mistake had been made in the programming of the working memory training tasks used in this study. This error resulted in the presentation of repeated trial sequences, which were then practiced repeatedly over the 8-day training period. This practice likely resulted in learning of the sequences, rather than training of interference control as we had originally inferred. The first author takes responsibility for this error, and both authors regret the publication of invalid results. In a separate replication (N = 16) of the working memory training condition that used randomized, novel sequences for each of the 8 training days, we were not able to reproduce the main findings of the original report...

via Retraction Watch

References

Jonides J, Nee DE. (2006). Brain mechanisms of proactive interference in working memory. Neuroscience 139:181-93.

Lustig C, Shah P, Seidler R, Reuter-Lorenz PA (2009). Aging, training, and the brain: a review and future directions. Neuropsychol Rev 19:504–522.

Owen AM, Hampshire A, Grahn JA, Stenton R, Dajani S, Burns AS, Howard RJ, Ballard CG. (2010). Putting brain training to the test. Nature Apr 20. [Epub ahead of print]. PDF

Persson, J., & Reuter-Lorenz, P. (2008). Gaining Control: Training Executive Function and Far Transfer of the Ability to Resolve Interference. Psychological Science 19: 881-888. DOI: 10.1111/j.1467-9280.2008.02172.x

PERSSON, J., WELSH, K., JONIDES, J., & REUTER-LORENZ, P. (2007). Cognitive fatigue of executive processes: Interaction between interference resolution tasks. Neuropsychologia 45: 1571-1579. DOI: 10.1016/j.neuropsychologia.2006.12.007