Note October 2018. This post was written before the LTEM model was developed. It therefore refers to the Four-Level Model’s Levels instead of LTEM’s eight tiers. The same principle still applies and I am leaving the article as written originally. To learn more about LTEM, click here or copy and paste: https://www.worklearning.com/ltem/. To learn why it’s now called the Kirkpatrick-Katzell model, click here.

The Kirkpatrick-Katzell four-level model of evaluation includes Level 1 learner reactions, Level 2 learning, Level 3 behavior, and 4 Level results. Because of the model’s ubiquity and popularity, many learning professionals and organizations are influenced or compelled by the model to measure the two higher levels—Behavior and Results—even when it doesn’t make sense to do so and even if poor methods are used to do the measurement. This pressure has led many of us astray. It has also enabled vendors to lie to us.

Let me get right to the point. When we ask learners whether a learning intervention will improve their job performance, we are getting their Level 1 reactions. We are NOT getting Level 3 data. More specifically, we are not getting information we can trust to tell us whether a person’s on-the-job behavior has improved due to the learning intervention.

Similarly, when we ask learners about the organizational results that might come from a training or elearning program, we are getting learners’ Level 1 reactions. We are NOT getting Level 4 data. More specifically, we are not getting information we can trust to tell us whether organizational results improved due to the learning intervention.

One key question is, “Are we getting information we can trust?” Another is, “Are we sure the learning intervention caused the outcome we’re targeting—or whether, at least, it was significant in helping to create the targeted outcomes?”

Whenever we gather learner answers, we have to remember that people’s subjective opinions are not always accurate. First there are general problems with human subjectivity; including people’s tendencies toward wanting to be nice, to see themselves and their organizations in a positive light, to believing they themselves are more productive, intelligent, and capable than they actually are. In addition, learners don’t always know how different learning methods affect learning outcomes, so asking them to assess learning designs has to be done with great care to avoid bias.

The Foolishness of Measuring Level 3 and 4 with Learner-Input Alone

There are also specific difficulties in having learners rate Level 3 and 4 results.

  • Having learners assess Level 3 is fraught with peril because of all the biases that are entailed. Learners may want to look good to others or to themselves. They may suffer from the Dunning-Kruger effect and rate their performance at a higher level than what is deserved.
  • Assessing Level 4 organizational results is particularly problematic. First, it is very difficult to track all the things that influence organizational performance. Asking learners for Level 4 results is a dubious enterprise because most employees cannot observe or may not fully understand the many influences that impact organizational outcomes.

Many questions we ask learners in measuring Level 3 and 4 are biased in and of themselves. These four questions are highly biasing, and yet sadly they were taken directly from two of our industry’s best-known learning-evaluation vendors:

  • “Estimate the degree to which you improved your performance related to this course?” (Rated on a scale of percentages to 100)
  • “The training has improved my job performance.” (Rated on a numeric scale)
  • “I will be able to apply on the job what I learned during this session.” (rated with a Likert-like scale)
  • “I anticipate that I will eventually see positive results as a result of my efforts.” (rated with a Likert-like scale)

At least two of our top evaluation vendors make the case explicitly that smile sheets can gather Level 3 and 4 data. This is one of the great lies in the learning industry. A smile sheet garners Level 1 results! It does not capture data at any other levels.

What about delayed smile sheets—questions delivered to learners weeks or months after a learning experience? Can these get Level 2, 3, and 4 data? No! Asking learners for their perspectives, regardless of when their answers are collected, still gives us only Level 1 outcomes! Yes, learners answers can provide hints, but the data can only be a proxy for outcomes beyond Level 1.

On top of that, the problems cited above regarding learner perspectives on their job performance and on organizational results still apply even when questions are asked well after a learning event. Remember, the key to measurement is always whether we can trust the data we are collecting! To reiterate, asking learners for their perspectives on behavior and results suffers from the following:

  • Learners’ biases skew the data
  • Learners’ blind spots make their answers suspect
  • Biased questioning spoils the data
  • The complexity in determining the network of causal influences makes assessments of learning impact difficult or impossible

In situations where learner perspectives are so in doubt, asking learners questions may generate some reasonable hypotheses, but then these hypotheses must be tested with other means.

The Ethics of the Practice

It is unfair to call Level 1 data Level 3 data or Level 4 data.

In truth, it is not only unfair, it is deceptive, disingenuous, and harmful to our learning efforts.

How Widespread is this Misconception?

If two of are top vendors are spreading this misconception, we can be pretty sure that our friend-and-neighbor foot soldiers are marching to the beat.

Last week, I posted a Twitter poll asking the following question:

If you ask your learners how the training will impact their job performance, what #Kirkpatrick level is it?

Twitter polls only allow four choices, so I gave people the choice of choosing Level 1 — Reaction, Level 2 –Learning, Level 3 — Behavior, or Level 4 — Results.

Over 250 people responded (253). Here are the results:

  • Level 1 — Reaction (garnered 31% of the votes)
  • Level 2 — Learning (garnered 15% of the votes)
  • Level 3 — Behavior (garnered 38% of the votes)
  • Level 4 — Results (garnered 16% of the votes)

Level 1 is the correct answer! Level 3 is the most common misconception!

And note, given that Twitter is built on a social-media follower-model—and many people who follow me have read my book on Performance-Focused Smile Sheets, where I specifically debunk this misconception—I’m sure this result is NOT representative of the workplace learning field in general. I’m certain that in the field, more people believe that the question represents a Level 3 measure.

Yes, it is true what they say! People like you who read my work are more informed and less subject to the vagaries of vendor promotions. Also better looking, more bold, and more likely to be humble humanitarians!

My tweet offered one randomly-chosen winner a copy of my award-winning book. And the winner is:

Sheri Kendall-DuPont, known on Twitter as:

Thanks to everyone who participated in the poll…

This is a guest post by Robert O. Brinkerhoff (www.BrinkerhoffEvaluationInstitute.com).

Rob is a renowned expert on learning evaluation and performance improvement. His books, Telling Training’s Story and Courageous Training, are classics.

______________________________

70-20-10: The Good, the Bad, and the Ugly

The 70-20-10 framework may not have much if any research basis, but it is still a good reminder to all of us in in the L&D and performance improvement professions that the work-space is a powerful teacher and poses many opportunities for practice, feedback, and improvement.

But we must also recognize that a lot of the learning that is taking place on the job may not be for the good. I have held jobs in agencies, corporations and the military where I learned many things that were counter to what the organization wanted me to learn: how to fudge records, how to take unfair advantage of reimbursement policies, how to extend coffee breaks well beyond their prescribed limits, how to stretch sick leave, and so forth.

These were relatively benign instances. Consider this: Where did VW engineers learn how to falsify engine emission results? Where did Well Fargo staff learn how to create and sell fake accounts to their unwitting customers?

Besides these egregiously ugly examples, we have to also recognize that in the case of L&D programming that is intended to support new strategic and other change initiatives, the last thing the organization needs is more people learning how to do their jobs in the old way. AT&T, for example, worked very hard to drive new beliefs and actions to enable the business to shift from landline technologies to wireless; on-the-job learning dragged them backwards, and creates problems still today. As AllState Insurance tries to shift sales focus away from casualty policies to financial planning services, the old guard teaches the opposite actions, as they continue to harvest the financial benefits of policy renewals. Any organization that has to make wholesale and fundamental shifts to execute new strategies will have to cope with the negative effects of years of on-the-job learning.

When strategy is new, there are few if any on-the-job pockets of expertise and role models. Training new employees for existing jobs is a different story. Here, obviously, the on-job space is an entirely appropriate learning resource.

In short, we have to recognize that not all on-the-job learning is learning that we want. Yet on the job learning remains an inexorable force that we in L&D must learn how to understand, leverage, guide and manage.

Two and a half years ago, in writing a blog post on learning styles, I did a Google search using the words “learning styles.” I found that the top 17 search items were all advocating for learning styles, even though there was clear evidence that learning-styles approaches DO NOT WORK.

Today, I replicated that search and found the following in the top 17 search items:

  • 13 advocated/supported the learning-styles idea.
  • 4 debunked it.

That’s progress, but clearly Google is not up to the task of providing valid information on learning styles.

Scientific Research that clearly Debunks the Learning-Styles Notion:

  • Kirschner, P. A. (2017) Stop propagating the learning styles myth. Computers & Education, 106, 166-171.
  • Willingham, D. T., Hughes, E. M., & Dobolyi, D. G. (2015). The scientific status of learning styles theories. Teaching of Psychology, 42(3), 266-271.
  • Pashler, H., McDaniel, M., Rohrer, D., & Bjork, R. (2008). Learning styles: Concepts and evidence. Psychological Science in the Public Interest, 9(3), 105-119.
  • Rohrer, D., & Pashler, H. (2012). Learning styles: Where’s the evidence? Medical Education, 46(7), 634-635.

Follow the Money

  • Still no one has come forward to prove the benefits of learning styles, even though it’s been over 10 years since $1,000 was offered, and 3 years since $5,000 was offered.

The Debunker Club (http://www.debunker.club/), one of my hobbies, is hosting a debate about the potency/viability of the 70-20-10 model. For more information, go directly to the Debunker Club’s event page.

A new meta-analysis on debunking was released last week, and I was hoping to get clear guidelines on how to debunk misinformation. Unfortunately, the science still seems somewhat equivocal about how to debunk. Either that, or there’s just no magic bullet.

Let’s break this down. We all know misinformation exists. People lie, people get confused and share bad information, people don’t vet their sources, incorrect information is easily spread, et cetera. Debunking is the act of providing information or inducing interactions intended to correct misinformation.

Misinformation is a huge problem in the world today, especially in our political systems. Democracy is difficult if political debate and citizen conversations are infused with bad information. Misinformation is also a huge problem for citizens themselves and for organizations. People who hear false health-related information can make themselves sick. Organizations who have employees who make decisions based on bad information, can hurt the bottom line.

In the workplace learning field, there’s a ton of misinformation that has incredibly damaging effects. People believe in the witchcraft of learning styles, neuroscience snake oil, traditional smile sheets, and all kinds of bogus information.

It would be nice if misinformation could be easily thwarted, but too often it lingers. For example, the idea that people remember 10% of what they read, 20% of what they hear, 30% of what they see, etc., has been around since 1913 if not before, but it still gets passed around every year on bastardized versions of Dale’s Cone.

A meta-analysis is a scientific study that compiles many other scientific studies using advanced statistical procedures to enable overall conclusions to be drawn. The study I reviewed (the one that was made available online last week) is:

Chan, M. S., Jones, C. R., Jamieson, K. H., & Albarracin, D. (2017). Debunking: A meta-analysis of the psychological efficacy of messages countering misinformation. Psychological Science, 28(11), 1531–1546. Available here (if you have journal access: http://journals.sagepub.com/doi/10.1177/0956797617714579).

This study compiled scientific studies that:

  1. First presented people with misinformation (except a control group that got no misinformation).
  2. Then presented them with a debunking procedure.
  3. Then looked at what effect the debunking procedure had on people’s beliefs.

There are three types of effects examined in the study:

  1. Misinformation effect = Difference between the group that just got misinformation and a control group that didn’t get misinformation. This determined how much the misinformation hurt.
  2. Debunking effect = Difference between the group that just got misinformation and a group that got misinformation and later debunking. This determined how much debunking could lesson the effects of the misinformation.
  3. Misinformation-Persistence effect = Difference between the group that got misinformation-and-debunking and the control group that didn’t get misinformation. This determined how much debunking could fully reverse the effects of the misinformation.

They looked at three sets of factors.

First, the study examined what happens when people encounter misinformation. They found that the more people thought of explanations for the false information, the more they would believe this misinformation later, even in the face of debunking. From a practical standpoint then, if people are receiving misinformation, we should hope they don’t think too deeply about it. Of course, this is largely out of our control as learning practitioners, because people come to us after they’ve gotten misinformation. On the other hand, it may provide hints for us as we use knowledge management or social media. The research findings suggest that we might need to intervene immediately when bad information is encountered to prevent people from elaborating on the misinformation.

Second, the meta-analysis examined whether debunking messages that included procedures to induce people to make counter-arguments to the misinformation would outperform debunking messages that did not include such procedures (or that included less potent counter-argument-inducing procedures). They found consistent benefits to these counter-argument inducing procedures. These procedures helped reduce misinformation. This suggests strongly that debunking should induce counter-arguments to the misinformation. And though specific mechanisms for doing this may be difficult to design, it is probably not enough to present the counter-arguments ourselves without getting our learners to fully process the counter-arguments themselves to some sufficient level of mathemagenic (learning-producing) processing.

Third, the meta-analysis looked at whether debunking messages that included explanatory information for why the misinformation was wrong would outperform debunking messages that included just contradictory claims (for example, statements to the effect that the misinformation was wrong). They found mixed results here. Providing debunking messages with explanatory information was more effective in debunking misinformation (getting people to move from being misinformed to being less misinformed), but these more explanatory messages were actually less effective in fully ridding people of the misinformation. This was a conflicting finding and so it’s not clear whether greater explanations make a difference, or how they might be designed to make a difference. One wild conjecture. Perhaps where the explanations can induce relevant counter-arguments to the misinformation, they will be effective.

Overall, I came away disappointed that we haven’t been able to learn more about how to debunk. This is NOT these researchers’ fault. The data is the data. Rather, the research community as a whole has to double down on debunking and persuasion and figure out what works.

People certainly change their minds on heartfelt issues. Just think about the acceptance of gays and lesbians over the last twenty years. Dramatic changes! Many people are much more open and embracing. Well, how the hell did this happen? Some people died out, but many other people’s minds were changed.

My point is that misinformation cannot possibly be a permanent condition and it behooves the world to focus resources on fixing this problem — because it’s freakin’ huge!

————

Note that a review of this research in the New York Times painted this in a more optimistic light.

————

Some additional thoughts (added one day after original post).

To do a thorough job of analyzing any research paradigm, we should, of course, go beyond meta-analyses to the original studies being meta-analyzed. Most of us don’t have time for that, so we often take the short-cut of just reading the meta-analysis or just reading research reviews, etc. This is generally okay, but there is a caveat that we might be missing something important.

One thing that struck me in reading the meta-analysis is that the authors commented on the typical experimental paradigm used in the research. It appeared that the actual experiment might have lasted 30 minutes or less, maybe 60 minutes at most. This includes reading (learning) the misinformation, getting a ten-minute distractor task, and answering a few questions (some treatment manipulations, that is, types of debunking methods; plus the assessment of their final state of belief through answers to questions). To ensure I wasn’t misinterpreting the authors’ message that the experiments were short, I looked at several of the studies compiled in the meta-analysis. The research I looked at used very short experimental sessions. Here is one of the treatments the experimental participants received (it includes both misinformation and a corrective, so it is one of the longer treatments):

Health Care Reform and Death Panels: Setting the Record Straight

By JONATHAN G. PRATT
Published: November 15, 2009

WASHINGTON, DC – With health care reform in full swing, politicians and citizen groups are taking a close look at the provisions in the Affordable Health Care for America Act (H.R. 3962) and the accompanying Medicare Physician Payment Reform Act (H.R. 3961).

Discussion has focused on whether Congress intends to establish “death panels” to determine whether or not seniors can get access to end-of-life medical care. Some have speculated that these panels will force the elderly and ailing into accepting minimal end-of-life care to reduce health care costs. Concerns have been raised that hospitals will be forced to withhold treatments simply because they are costly, even if they extend the life of the patient. Now talking heads and politicians are getting into the act.

Betsy McCaughey, the former Lieutenant Governor of New York State has warned that the bills contain provisions that would make it mandatory that “people in Medicare have a required counseling session that will tell them how to end their life sooner.”

Iowa Senator Chuck Grassley, the ranking Republican member of the Senate Finance Committee, chimed into the debate as well at a town-hall meeting, telling a questioner, “You have every right to fear…[You] should not have a government-run plan to decide when to pull the plug on Grandma.”

However, a close examination of the bill by non-partisan organizations reveals that the controversial proposals are not death panels at all. They are nothing more than a provision that allows Medicare to pay for voluntary counseling.

The American Medical Association and the National Hospice and Palliative Care Organization support the provision. For years, federal laws and policies have encouraged Americans to think ahead about end-of-life decisions.

The bills allow Medicare to pay doctors to provide information about living wills, pain medication, and hospice care. John Rother, executive vice president of AARP, the seniors’ lobby, repeatedly has declared the “death panel” rumors false.

The new provision is similar to a proposal in the last Congress to cover an end-of-life planning consultation. That bill was co-sponsored by three Republicans, including John Isakson, a Republican Senator from Georgia.

Speaking about the end of life provisions, Senator Isakson has said, “It’s voluntary. Every state in America has an end of life directive or durable power of attorney provision… someone said Sarah Palin’s web site had talked about the House bill having death panels on it where people would be euthanized. How someone could take an end of life directive or a living will as that is nuts.”

That’s it. That’s the experimental treatment.

Are we truly to believe that such short exposures are representative of real-world debunking? Surely not! In the real world, people who get misinformation often hold that misinformation over months or years while occasionally thinking about the misinformation again or encountering additional supportive misinformation or non-supportive information that may modify their initial beliefs in the misinformation. This all happens and then we try our debunking treatments.

Finally, it should be emphasized that the meta-analysis also only compiled eight research articles, many using the same (or similar) experimental paradigm. This is further inducement to skepticism. We should be very skeptical of these findings and my plea above for more study of debunking — especially in more ecologically-valid situations — is reinforced!

As professionals in the learning field, memory is central to our work. If we don’t help our learners preserve their memories (of what they learned), we have not really done our job. I’m oversimplifying here — sometimes we want to guide our learners toward external memory aids instead of memory. But mostly, we aim to support learning and memory.

Glacier View

You might have learned that people who take photographs will remember less than those who did not take photographs. Several research studies showed this (see for example, Henkel, 2014).

The internet buzzed with this information a few years ago:

  • The Telegraph — http://www.telegraph.co.uk/news/science/science-news/10507146/Taking-photographs-ruins-the-memory-research-finds.html
  • NPR — http://www.npr.org/2014/05/22/314592247/overexposed-camera-phones-could-be-washing-out-our-memories
  • Slate — http://www.slate.com/blogs/the_slatest/2013/12/09/a_new_study_finds_taking_photos_hurts_memory_of_the_thing_you_were_trying.html
  • CNN — http://www.cnn.com/2013/12/10/health/memory-photos-psychology/index.html
  • Fox News — http://www.foxnews.com/health/2013/12/11/taking-pictures-may-impair-memories-study-shows.html

Well, that was then. This is now.

Research Wisdom

There are CRITICAL LESSONS to be learned here — about using science intelligently… with wisdom.

Science is a self-correcting system that, with the arc of time, bends toward the truth. So, at any point in time, when we ask science for its conclusions, it tells us what it knows, while it apologizes for not knowing everything. Scientists can be wrong. Science can take wrong turns on the long road toward better understanding.

Does this mean we should reject scientific conclusions because they can’t guarantee omniscience; they can’t guarantee truth? I’ve written about this in more depth elsewhere, but I’ll say it here briefly — recommendations from science are better than our own intuitions; especially in regards to learning, given all the ways we humans are blind to how learning works.

Memory With Photography

Earlier studies showed that people who photographed images were less able to remember them than people who simply examined the images. Researchers surmised that people who off-loaded their memories to an external memory aid — to the photographs — freed up memory for other things.

We can look back at this now and see that this was a time of innocence; that science had kept some confidences hidden. New research by Barasch, Diehl, Silverman, and Zauberman (2017), found that people “who could freely take photographs during an experience recognized more of what they saw” and that those “with a camera had better recognition of aspects of the scene that they photographed than of aspects they did not photograph.

Of course, this is just one set of studies… we must be patient with science. More research will be done, and you and will benefit in knowing more than we know now and with more confidence… but this will take time.

What is the difference between the earlier studies and this latest set of studies? As argued by Barasch, Diehl, Silverman, and Zauberman (2017), the older studies did not give people the choice of which objects to photograph. In the words of the researchers, people did not have volitional control of their photographing experience. They didn’t go through the normal process we might go through in our real-world situations, where we must decide what to photograph and determine how to photograph the objects we target (i.e., the angles, borders, focus, etc.).

In a series of four experiments, the new research showed that attention was at the center of the memory effect. Indeed, people taking photographs “recognized more of what they saw and less of what they heard, compared with those who could not take any photographs (I added the bold underlines).

Interestingly, some of the same researchers, just the year before had found that taking photographs actually improved people’s enjoyment of their experiences (Diehl, Zauberman, & Barasch, 2016).

Practical Considerations for Learning Professionals

You might be asking yourself, “How should I handle the research-based recommendations I encounter?” Here is my advice:

  1. Be skeptical, but not too skeptical.
  2. Determine whether the research comes from a trusted source. Best sources are top-tier refereed scientific journals — especially where many studies find the same results. Worst sources are survey-based compilations of opinions. Beware of recommendations based on one scientific article. Beware of vendor-sponsored research. Beware of research that is not refereed; that is, not vetted by other researchers.
  3. Find yourself a trusted research translator. These people — and I count myself among them — have spent enough substantial time exploring the practical aspects of the research that they are liable to have wisdom about what the research means — and what its boundary conditions might be.
  4. Pay your research translators — so they can continue doing their work.
  5. Be good and prosper. Use the research in your learning programs and test it. Do good evaluation so you can get valid feedback to make your learning initiatives maximally effective.

Inscribed in My High School Yearbook in 1976

Time it was, and what a time it was, it was
A time of innocence, A time of confidences
Long ago, it must be, I have a photograph
Preserve your memories; They’re all that’s left you

Written by Paul Simon

The Photograph Above

Taken in Glacier National Park, Montana, USA; July 1, 2017
And incidentally, the glaciers are shrinking permanently.

Research Cited

Barasch, A., Diehl, K., Silverman, J., & Zauberman, G. (2017). Photographic Memory: The Effects of Volitional Photo Taking on Memory for Visual and Auditory Aspects of an Experience. Psychological Science, early online publication.

Diehl, K., Zauberman, G., & Barasch, A. (2016). How taking photos increases enjoyment of experiences. Journal of Personality and Social Psychology, 111, 119–140.

Henkel, L. A. (2014). Point-and-shoot memories: The influence of taking photos on memory for a museum tour. Psychological Science, 25, 396–402.

I must be in a bad mood — or maybe I’ve been unlucky in clinking on links — but this graphic is horrifying. Indeed, it’s so obviously flawed that I’m not even going to point out it’s most glaring problem. You decide!

One more editorial comment before the big reveal:  Why, why, why is the gloriously noble and important field of learning besieged by such crap!!!!

 

 

.

.

.

.

.

.

.

.

.

.

Why is the goal of a learning-focused game, “fun?”

Updated July 11, 2016. An earlier version was more apocalyptic.

==============================

THIS IS HUGE. A large number of studies from the last 15 years of neuroscience research (via fMRI) could be INVALID!

A recent study in the journal PNAS looked at the three most commonly used software packages used with fMRI machines. Where they expected to find a normal familywise error rate of 5%, they found error rates up to 70%.

Here’s what the authors’ wrote:

“Using mass empirical analyses with task-free fMRI data, we have found that the parametric statistical methods used for group fMRI analysis with the packages SPM, FSL, and AFNI can produce FWE-corrected cluster P values that are erroneous, being spuriously low and inflating statistical significance. This calls into question the validity of countless published fMRI studies based on parametric clusterwise inference. It is important to stress that we have focused on inferences corrected for multiple comparisons in each group analysis, yet some 40% of a sample of 241 recent fMRI papers did not report correcting for multiple comparisons (26), meaning that many group results in the fMRI literature suffer even worse false-positive rates than found here (37).”

In a follow-up blog post, the authors estimated that up to 3,500 scientific studies may be affected, which is down from their initial published estimate of 40,000. The discrepancy results because only studies at the edge of statistical reliability are likely to have results that might be affected. For an easy-to-read review of their walk-back, Wired has a nice piece.

The authors also point out that there is more to worry about than those 3,500 studies. An additional 13,000 studies don’t use any statistical correction at all (so they’re not affected by the software glitch reported in the scientific paper). However, these 13,000 studies use an approach that “has familywise error rates well in excess of 50%.” (cited from the blog post)

Here’s what the authors say in their walk-back:

“So, are we saying 3,500 papers are “wrong”? It depends. Our results suggest CDT P=0.01 results have inflated P-values, but each study must be examined… if the effects are really strong, it likely doesn’t matter if the P-values are biased, and the scientific inference will remain unchanged. But if the effects are really weak, then the results might indeed be consistent with noise. And, what about those 13,000 papers with no correction, especially common in the earlier literature? No, they shouldn’t be discarded out of hand either, but a particularly jaded eye is needed for those works, especially when comparing them to new references with improved methodological standards.”

 

Some Perspective

Let’s take a deep breadth here. Science works slowly and we need to see what other experts have to say in the coming months.

The authors reported that there were about 40,000 published studies in the last 15 years that might be affected. Of this amount, only some of 3,500 + 13,000 = 16,500 are affected. That’s 41% of published articles with a potential to have invalid results.

But, of course, in the learning field, we don’t care about all these studies as most of them have very little to do with learning or memory. Indeed, a search of the whole history of PsycINFO (a social-science database) finds a total of 22,347 articles mentioning fMRI at all. Searching for articles that have a learning or memory aspect culls this number down to 7,056. This is a very rough accounting, but it does put the overall findings in some perspective.

As the authors warn, it’s not appropriate to dismiss the validity of all the research articles, even if they’re in one of the suspect groups of studies. Instead, when looking at the potentially-invalidate articles, each one has to be examined individually to know whether it has problems.

Despite these comforting caveats, the findings by the scientists have implications for many neuroscience research studies over the past 15 years (when the bulk of neuroscience research has been done).

On the other hand, there truly haven’t been many neuroscience findings that have much practical relevance to the learning field as of yet. See my review for a critique of overblown claims about neuroscience and learning. Indeed, as I’ve argued elsewhere, neuroscience’s potential to aid learning professionals probably rests in the future. So, being optimistic, maybe these statistical glitches will end up being a good thing. First, perhaps they’ll propel greater scrutiny to research methodologies, improving future neuroscience research. Second, perhaps they’ll put the brakes on the myth-creating industrial complex around neuroscience until we have better data to report and utilize.

Still, a dark cloud of low credibility may settle over the whole neuroscience field itself, hampering researchers from getting funding, and making future research results difficult for practitioners to embrace. Time will tell.

 

Popular Press Articles Citing the Original Article (Published Before the Walk-Backs).

Here are some articles from the scientific press pointing out the potential danger:

  • http://arstechnica.com/science/2016/07/algorithms-used-to-study-brain-activity-may-be-exaggerating-results/
  • http://cacm.acm.org/news/204439-a-bug-in-fmri-software-could-invalidate-15-years-of-brain-research/fulltext
  • http://www.wired.co.uk/article/fmri-bug-brain-scans-results
  • http://www.zmescience.com/medicine/brain-imageflaw-57869/

==============================

Notes:

From Wikipedia (July 11, 2016): “In statistics, family-wise error rate (FWER) is the probability of making one or more false discoveries, or type I errors, among all the hypotheses when performing multiple hypotheses tests.”

Neuroscience and Learning

The Debunker Club, formed to fight myths and misconceptions in the learning field, is currently seeking public comment on the possibility that so-called neuroscience-based recommendations for learning and education are premature, untenable, or invalid.

 

Click here to comment or review the public comments made so far…

 

Click here to join The Debunker Club…

 

A year and a half ago, three esteemed researchers (and me) published a series of articles debunking the meme that people remember 10% of what they read, 20% of what they hear, and 30% of what they read, etc…

Here was my review of those research articles.

Unfortunately, until now, the articles themselves were not available online. To subscribe to the originating journal, Educational Technology, click here.

Now, we the authors are able to share a copy with you.

 

Click here to get a copy of the four articles…