Why students fail to transfer what they learn

by Stephanie Chasteen on November 17, 2008

We had a visit from Stanford education researcher Dan Schwartz last week, and what he told us about how people learn just rocked my world. I always enjoyed his work (and it was a real pleasure to tell him how much he’s influenced my thinking about education), and have blogged before about his A Time for Telling paper. Still, spending many hours with him over two days was a transformative experience for me. Let me try to tell you why.

All images in this post are courtesy of Dan Schwartz. His research website is here.

This is a problem of transfer (or so we say).

For example, students can learn how to do permutation problems using, say, cars or marbles as examples. But when you ask the marble-trained students to answer a question using cars, they struggle. Similarly, we see our own students do well on homework and chapter-level tests, but not on the final. They know the formulas and ideas, but don’t know how to apply them. This has been called a problem of transfer from one domain to another.

But, he argued, how is it that we could fail to transfer what we know? After all, we learn something at home and we bring it to work, or we learn something at home, and we bring it to school. We seem to transfer all the time. And, as my colleague Noah Finkelstein has argued, in order to believe in transfer, we have to believe that there is some thing to be transferred. What is transferring? Some static little packet of knowledge? There’s no tangible chunk of knowledge that we can bring from place A to place B. Knowledge is about skills and process and understanding, it’s not a static thing.

Efficiency over Adaptation

So, Schwartz argued, it’s not really a problem of transfer. The problem is that we’ve trained people to do things quickly – efficiently – not to adapt to new situations. We train people to recall lists of words quickly, or take timed tests, or tell us what causes the seasons when asked on the street. So we’ve trained efficiency over adaptation. While efficiency is important for routine tasks, experts readily adapt their knowledge to a new situation.

Preparation for Future Learning

Schwartz did a fascinating study to see what helps students learn to adapt to new situations.

  1. One set of students read a chapter and then hear a lecture about it
  2. Another set of students analyze and graph data, deciding what they think is important to graph
  3. A third set played around with graphing the data and then heard a lecture about it.bar-chart

So, how did they each do on assessment? On a traditional factual test, group 1 (reading and lecture) and group 3 (graphing then lecture) do equally well. Group 2 (graphing only) did very poorly — without some expert guidance they didn’t really learn much from just playing around. Those data are to the right.

OK, so does that mean that it’s equally good to have studentsbarchart2 read and hear a lecture as to play around with the data before hearing the lecture?

Nope… he then gave them all a test that required them to use their understanding in a new situation, and those data are to the right here. Those who first played around with the data and then heard the expert lecture did much better on that test. They were approaching adaptive expertise more quickly than the others! The differences in performance on this test, above, are stunning — these students (who, he argues, were prepared to be able to learn during the test by the instruction they were given) did more than twice as well on this test.

So the message here is that there is a time for telling (ie., lecture) — just not too soon!

This is particularly appropriate to remember in math and science. Math, for example, is usually presented as an efficient way to solve problems. What if, instead, students found that math helped them understand science and manage complex problems? For example, he took a class of 9th grade students and taught them statistics by asking them to find a way to rate the reliability of pitching machines. Below are two examples of student approaches to this problem.

untitled2untitled1

This forced them to create ways to deal with variability in data before being given a formula for computing variability (eg., standard deviations). He found that, even a year later, these students did better than college students in understanding formulas for variability, and were much better able to understand variability in data and its importance. Below is the task that he gave these students — those who had struggled with variability before hearing the lecture were able to recognize that the IQ scores of the green people had more spread, even though the IQ of the blue people was higher, on average.

untitled3

He argued that this emphasis on efficiency is very American. We train people to become expert at routine tasks, but what we need to emphasize instead is innovative experiences. Let go of what you’re told, and try something new. For one, when students innovate a solution first, then they have a context for what they’re learning. When given the solution first, they don’t have a context for it. Telling people the answer works if they have a lot of prior knowledge (and that’s why talks at conferences, for example, can still be decent ways to get a lot of information across). But when you’re learning something new, don’t tell too soon!

You can find out more about Dan Schwartz’s work here.

{ 9 comments… read them below or add one }

Lorn November 18, 2008 at 8:03 am

My arguments against The Industrialist Who Hires Blue People:

0. your logic is deeply flawed, Sir, because:
1. the difference (98 to 101) is too small for any conclusion especailly since
2. the sample population is too small and
3. sorting people by color is unwarranted and
4. not correlated with the (small) data and
5. offensive and furthermore
6. IQ and “smarter” are not synonyms and
7. you clearly need to revise what you expect from your employees or the smart ones will stay away from you.

Great post. I couldn’t resist solving the task. Did I get it right?

sciencegeekgirl November 18, 2008 at 9:42 am

Hi Lorn,

I’m so glad you liked the post! And actually your responses are a perfect example of the usefulness of the Preparation for Future Learning activities. Your arguments are very similar to what college students argue when given this task, without having first struggled with issues of variance in another task. They say that IQ tests aren’t valid, that there are other things important than hiring in IQ, etc.

However, if you look at the SPREAD in the scores for the blue versus the green people, you’ll see that while, on average, the blue people have a higher IQ, the green people actually have more people with an IQ above 104. In fact if you took the average of the top 20% of the green people, it would likely be higher than the blue people. Another thing to look at is the variability around the median. If you find the middle person (that’s #22), you’ll see that in the blue people, all the people around #22 have IQ’s of 101. The green people around #22 have a much higher spread of IQ.

Most people who haven’t been prepared to think about variability give more meta-arguments about whether or not it’s useful to compare people on IQ, as you did, and don’t look at the data. I’m not saying your arguments aren’t good ones, but there are other arguments based on the data that can be made. Your comment about sample size is a good one based on the data, but there are other data-based arguments to be made.

Thanks for your post!

Stephanie

Dianne November 20, 2008 at 4:39 pm

The post about Dan Schwarz really hit home with me. I see this problem with transference daily. My science students have trouble applying what they learned in math to science formulas and vice versa. They also have trouble moving past the examples in the lectures and reading. My school district a few years ago began pushing inquiry science which sounds very similar to what Dan Scwharz is saying. In my own teaching, I find that when I use a laboratory activity for their notes, the students seem to catch on to the material faster.

I plan on looking into Dan Schwarz’s research more closely. I think it could have a profound impact on my teaching.

sciencegeekgirl November 24, 2008 at 11:36 am

I’m so glad you find this stuff as compelling as I do, Dianne. If you have any questions, please feel free to post them. In particular, I think his focus on contrasting cases to highlight the deep structure that experts take for granted is key, and a bit tough to grasp.

Bill Ward November 24, 2008 at 1:34 pm

Hire the blue people. You can always fire the dumb ones, or assign them to tasks that don’t require intelligence. But you’ll end up with a few really smart folks who can make the company more successful.

Bill Ward November 24, 2008 at 1:36 pm

Oh wait, I didn’t read carefully enough. Nevermind. Green FTW!

Frank Noschese December 13, 2008 at 1:42 pm

This research, and Dianne’s observations about activities as notes, is quite similar to Modeling Physics:

http://modeling.asu.edu/

sciencegeekgirl December 14, 2008 at 9:38 am

Yes, there are definitely some similarities, in that both approaches require students to wrestle with an idea and come up with their own model of something before the “expert” instructor comes and guides them in any way.

There are some differences, however. Modeling is explicitly a way to train students in scientific thinking and approach to problems. The “models” they are coming up with are quite a bit broad — eg., a mathematical and physical formulation of an understanding of an Atwood Machine — supported by data and applied to different situations. Schwartz’s Invention Activities are a bit more targeted to key concepts, such as “density” or or “variance” or “vectors.” It’s a way to get students to think about a situation and (like modeling) recognize the key elements that go into creating a definition, before the expert comes and gives the common way to approach that term. His activities tend to be very visual.

So, in a way, invention activities are helping students wrestle with the key terms that would go into creating a model in Modeling Physics. They’re smaller, more fundamental chunks than are dealt with in Modeling, and less tied to phenomena.

Here’s a nice summary from the website that Frank included:
http://modeling.asu.edu/modeling/mod_cycle.html

Watch Year One Online Free June 20, 2009 at 7:48 pm

This is very up-to-date info. I’ll share it on Twitter.
p.s. Year One is already on the Internet and you can watch it for free.

Leave a Comment

{ 2 trackbacks }

Previous post:

Next post: