In psychology, there is a lot of bad research out there by all estimates. The poor quality of this research can be attributed to concerns about ideology-driven research agendas, research bias, demand characteristics, lack of any real theory guiding the research itself, p-hacking, file-drawer effects, failures to replicate, small sample sizes, and reliance on undergraduate samples, among others. Arguably, there is more bad (or at least inaccurate) research than good research floating around as, in principle, there are many more ways of being wrong about the human mind than there are of being right about it (even given our familiarity with it); a problem made worse by the fact that being (or appearing) wrong or reporting null findings does not tend to garner one social status in the world of academia. If many of the incentives reside in finding particular kinds of results – and those kinds are not necessarily accurate – the predictable result is a lot of misleading papers. Determining what parts of the existing psychological literature are an accurate description of human psychology can be something of a burden, however, owing to the obscure nature of some of these issues: it’s not always readily apparent that a paper found a fluke result or that certain shady research practices have been employed. Thankfully, it doesn’t take a lot of effort to see why some particular pieces of psychological research are silly; criticizing that stuff can be as relaxing as a day off at the beach.

Kind of like this, but indoors and with fewer women

The last time I remember coming across some of the research that can easily be recognized as silly was when one brave set of researchers asked if leaning to the left made the Eiffel tower look smaller. The theory behind that initial bit of research is called, I think, number line theory, though I’m not positive on that. Regardless of the name, the gist of the idea seems to be that people - and chickens, apparently - associate smaller numbers with a relative leftwardly direction and larger numbers with a rightwardly one. For humans, such a mental representation might make sense in light of our using certain systems of writing; for nonhumans, this finding would seem to make zero sense. To understand why this finding makes no sense, try and place it within a functional framework by asking (a) why might humans and chickens (and perhaps other animals as well) represent smaller quantities with their left, and (b) why might leaning to the left be expected to bias one’s estimate of size? Personally, I’m coming up with a blank on the answer to those questions, especially because biasing one’s estimate of size on the basis of how one is leaning is unlikely to yield more accurate estimates. A decrease in accuracy seems like that could only carry costs in this case; not benefits. So, at best, we’re left calling those findings a development byproduct for humans and likely a fluke for the chickens. In all likelihood, the human finding is probably a fluke as well.

Thankfully, for the sake of entertainment, silly research is not to be deterred. One of the more recent tests of this number line hypothesis (Anelli et al, 2014) makes an even bolder prediction than the Eiffel tower paper: people will actually get better at performing certain mathematical operations when they’re traveling to the left or the right: specifically, going right will make you better at addition and left better at subtraction. Why? Because smaller numbers are associated with the left? How does that make one better at subtraction? I don’t know and the paper doesn’t really go into that part. On the face of it, this seems like a great example of what I have nicknamed “dire straits thinking”. Named after the band’s song, “money for nothing” this type of thinking leads people to hypothesizing that others can get better (or worse) at tasks without any associated costs. The problem with this kind of thinking is that if people did possess the cognitive capacities to be better at certain tasks, one might wonder why people ever perform worse than they could. This would lead me to pose questions like, “why do I have to be traveling right to be better at addition; why not just be better all the time?” Some kind of trade-offs need to referenced to explain that apparent detriment/bonus to performance, but none ever are in dire straits thinking.

In any case, let’s look at the details of the experiment, which was quite simple. Anelli et al, (2014) had a total of 48 participants walk with an experimenter (one at a time; not all 48 at once). The pair would walk together for 20 seconds in a straight line, at which point the experimenter would call out a three-digit number, tell the participants to add or subtract from it by 3 aloud for 22 seconds, give them a direction to turn (right or left), and tell them to begin. At that point, the participant would turn and start doing the math. Each participant completed four trials: two congruent (right/addition or left/subtraction) and two incongruent (right/subtraction or left/addition). The researchers hoped to uncover a congruency effect, such that more correct calculations would be performed in the congruent, relative to incongruent, trials.

Now put the data into to the “I’m right” program and it’s ready to publish

Indeed, just such an effect was found: when participants were moving in a congruent direction as their mathematical operations, they performed more correct calculations on average (M = 10.1), relative to when they were traveling in an incongruent direction (M = 9.6). However, when this effect was broken down by direction, it turns out that the effect only exists when participants were doing addition (M = 11.1 when going right, 10.2 when going left); there was no difference for subtraction (M = 9.0 and 9.1, respectively). Why was there no effect for subtraction? Well, the authors postulate a number of possibilities – one of which being that perhaps participants needed to be walking backwards – though none of them include the possibility of the addition finding being a statistical fluke. It’s strange how infrequently this possibility is ever mentioned in published work, especially in the face of inconsistent findings.

Now one obvious criticism of this research is that the participants were never traveling right or left; they were walking straight ahead in all cases. Right or left, unlike East or West, depends on perspective. When I am facing my computer, I feel I am facing ahead; when I turn around to walk to the bathroom, I don’t feel like I’m walking behind me. The current research would thus rely on the effects of a momentary turn affecting participant’s math abilities for about half a minute. Accordingly, participants shouldn’t even have needed to be walking; asking them to turn and stand in place should be expected to have precisely the same effect. If the researchers wanted to measure walking to the right or left, they should have had participants moving to the side by sliding, rather than turning and walking forward.

Other obvious criticisms of the research could include the small sample size, the small effect size, the inconsistency of the effect (works for addition but not subtraction and is inconsistent with other research they cite which was itself inconsistent – people being better at addition when going up in an elevator but not walking up stairs, if I understand correctly), or the complete lack of anything resembling a real theory guiding the research. But let’s say for a moment that my impression of these results as silly is incorrect; let’s assume that these results accurately describe the workings of human mind in some respect. What are the implications of that finding? What, in other words, happens to be at stake here? Why would this research be published, relative to the other submissions received by Frontiers in Psychology? Even if it’s a true effect – which already seems unlikely, given the aforementioned issues – it doesn’t seem particularly noteworthy. Should people be turning to the right and left while taking their GREs? Do people need to be doing jumping jacks to improve their multiplication skills so as to make their body look more like the multiplication symbol? If so, how could you manage to do them while you’re supposed to be sitting down quietly while taking your GREs without getting kicked out of the testing site? Perhaps someone more informed on the topic could lend a suggestion, because I’m having trouble seeing the importance of it.

Maybe the insignificance of the results is supposed to make the reader feel more important

Without wanting to make a mountain out of a mole hill, this paper was authored by five researchers and presumably made it passed an editor and several reviewers before it saw publication. At a minimum, that’s probably about 8 to 10 people. That seems like a remarkable feat, given how strange the paper happens to look on its face. I’m not just mindlessly poking fun at the paper, though: I’m bringing attention to it because it seems to highlight a variety of problems in the world of psychological research. There are, of course, many suggestions as to how these problems might be ferreted out, though many of them that I have seen focus more on statistical solutions or combating researcher degrees of freedom. While such measures might reduce the quantity of bad research (like pre-registering studies), they will be unlikely to increase the absolute quality of good work (since one can pre-register silly ideas like this), which I think is an equally valuable goal. For my money, the requirement of some theoretical functional grounding for research would likely be the strongest candidate for improving work in psychology. I imagine many people would find it harder to propose such an idea in the first place if they needed to include some kind of functional considerations as to why turning right makes you better at addition. Even if such a feat was accomplished, it seems those considerations would make the rationale for the paper even easier to pick apart by reviewers and readers.

Instead of asking for silly research to be conducted on larger, more diverse samples, it seems better to ask that silly research not be conducted at all.

References: Anelli, F., Lugli, L., Baroni G., Borghi, A., & Nicoletti, R. (2014). Walking boosts your performance in making additions and subtractions. Frontiers in Psychology, 5, doi: 10.3389/fpsyg.2014.01459

Over the course of my time at various public schools and universities I have encountered a great many teachers. Some of my teachers were quite good. I would credit my interest in evolutionary psychology to one particularly excellent teacher – Gordon Gallup. Not only was the material itself unlike anything I had previously been presented with in other psychology courses, but the way Gordon taught his classes was unparalleled. Each day he would show up and, without the aid of any PowerPoints or any apparent notes, just lecture. On occasion we would get some graphs or charts drawn on the board, but that was about it. What struck me about this teaching style is what it communicated about the speaker: this is someone who knows what he’s talking about. His command of the material was so impressive I actually sat through his course again for no credit in the follow years to transcribe them (and the similarity from year-to-year was remarkable, given that lack of notes). It was just a pleasure listening to him do what we did best.

A feat I was recently recognized for

That I say Gordon was outstanding is to say he was exceptional, relative to his peers (even if many of those peers, mistakenly, believe they are exceptional as well). The converse to that praise, then, is that I have encountered many more professors who were either not particularly good at what they did or downright awful at it (subjectively speaking, of course). I’ve had some professors who act, more or less, as an audio guide to the textbook that, when questioned, didn’t seem to really understand the material they were teaching; I’ve had another tell his class “now, we know this isn’t true, but maybe it’s useful” as he reviewed Maslow’s hierarchy of needs for what must have been the tenth time in my psychology education – a statement which promptly turned off my attention for the day. The number of examples I could provide likely outnumber my fingers and toes, so there’s no need to detail each one. In fact, just about everyone who has attended school has had experiences like this. Are these subjective evaluations of teachers that we have all made accurate representations of their teaching ability, though?

According to some research by Braga et al (2011), that answer is “yes”, but in a rather perverse sense: teacher evaluations tend to be negatively predictive of actual teaching effectiveness. In other words, at the end of a semester when a teacher receives evaluations from their students, the better these evaluations, the less effective the teacher tends to be. As someone who received fairly high evaluations from my own students, this should either be cause for some reflection as to my methods (since I am interested in my students learning; not just their being satisfied with my course) or a hunt for why the research in question must be wrong to make me feel better about my good reviews. In the interests of prioritizing my self-esteem, let’s start by considering the research and seeing if any holes can be poked in it.

“Don’t worry; I’m sure those good reviews will still reflect well on you”

Braga et al (2011) analyzed data from a private Italian university offering programs in economics, business, and law in 1998/9. The students in these programs had to take a fixed course of classes with fixed sets of materials and the same examinations. Additionally, students were randomly assigned to professors, making this one of the most controlled academic settings for this kind of research I could imagine. At the end of the terms, students provided evaluations of their instructors, allowing their ratings of instructors to be correlated – at the classroom level, as the evaluations were anonymous – with their performance in being effective teachers.

Teaching effectiveness was measured by examining how students did in subsequent courses, (controlling for a variety of non-teacher factors, like class size) the assumption being that students with better professors in the first course would do better in future courses, owing to their more proficient grasping of the material. These non-teacher factors accounted for about 57% of the variance in future course grades, leaving plenty of room for teacher effects. The effect of teachers was appreciable, with an increase of one standard deviation in effectiveness led to gain of about 0.17 standard deviations of grade in future classes (about a 2.3% bump up). Given the standardized materials and the gulf which could exist between the best and worst teachers, it seems there’s plenty of room for teacher effectiveness to matter. Certainly no students want to end up at a disadvantage because of a poor teacher; I know I wouldn’t.

When it came to the main research question, the results showed that teachers who were the least effective in providing future success for their students tended to receive the highest evaluations. This effect was sizable as well: for each standard deviation increase in teaching effectiveness, student evaluation ratings dropped by about 40% of a standard deviation. Perhaps unsurprisingly, grades were correlated with teaching evaluations as well: the better grades the students received, the better the evaluations they tended to give the professors. Interestingly, this effect did not exist in classes comprised of 25% or more of the top students (as measured by their cognitive entrance exams); the evaluations of those classes were simply not predictive of effectiveness.

That last section is the part of the paper that most everyone will cite: the negative relationship between teacher evaluations and future performance. What fewer people seem to do when referencing that finding is consider why this relationship exists and then use that answer to inform their teaching styles (as I get the sense this information will quite often be cited to excuse otherwise lackluster evaluations, rather than to change anything). The authors of the paper posit two main possibilities for explaining this effect: (1) that some teachers make class time more entertaining at the expense of learning, and/or (2) that some teachers might “teach for the test”, even if they do so at the expense of “true learning”. While neither possibility is directly tested in the paper, the latter possibility strikes me as most plausible: students in the “teaching for the test” classes might simply focus on the particular chunks of information relevant for them at the moment, rather than engaging it as a whole and understanding the subject more broadly.

In other words, vague expectations encourage cramming with a greater scope

With that research in mind, I would like to present a section of my philosophy when it came to teaching and assessment. A question of interest that I have given much thought to is what, precisely, are grades aimed at achieving? For many professors – indeed, I’d say the bulk of them – grades serve the ends of assessment. The grades are used to tell people – students and others – how well the students did at understanding the material come test time. My answer to this question is a bit different, however: as an instructor, I had no particular interest in the assessment of students per se; my interest was in their learning. I only wanted to assess my students as a means of pushing them to the end of learning. As a word of caution, my method of assessment demands substantially more effort from those doing the assessing, be it a teacher or assistant, than is typical. It’s an investment of time many might be unwilling to make.

My assessments were all short-essay style questions, asking students to apply theories they have learned about to novel questions we did not cover directly in class; there were no multiple choice questions. According to the speculations of Braga et al (2011), this would put me firmly in the “real teaching” camp, instead of the “teaching to the test” one. There are a few reasons for my decision: first, multiple choice questions don’t allow you to see what the students were thinking when answering the question. Just because someone gets an answer correct on a multiple choice exam, it doesn’t mean they got the correct answer for the right reasons. For my method to be effective, however, it does mean someone needs to read the exams in depth instead of just feeding them through a scantron machine, and that reading takes time. Second, essay exams force students to confront what they do and do not know. Having spent many years as a writer (and even more as a student), I’ve found that many ideas that seem crystal clear in my head do not always translate readily to text. The feeling of understanding can exist in lack of actual understanding. If students find they cannot explain an idea as readily as felt they understood it, that feeling might be effectively challenged, yielding a new round of engagement with the material.

After seeing where the students were going wrong, the essay format allowed me to make notes on their work and hand it back to them for revisions; something you can’t do very well with multiple choice questions either. Once the students had my comments on their work, they were free to revise it and hand it back into me. The grade they got on their revisions would be their new grade: no averaging of the two or anything of the sort. The process would then begin again, with revisions being made on revisions, until the students were happy with their grade or stopped trying. In order for assessment to serve the end of learning, assessment needs to be ongoing if you expect learning to be. If assessment is not ongoing, students have little need to fix their mistakes; they’ll simply look at their grade and then toss their test in the trash as many of them do. After all, why would they bother putting in the effort to figure out where they went wrong and how to go right if doing so successfully would have no impact whatsoever on the one thing they get from the class that people will see?

Make no mistake: they’re here for a grade. Educations are much cheaper than college.

I should also add that my students were allowed to use any resource they wanted for the exams, be that their notes, the textbook, outside sources, or even other students. I wanted them to engage with the material and think about it while they worked, and I didn’t expect them to have it all memorized already. In many ways, this format mirrors the way academics function in the world outside the classroom: when writing our papers, we are allowed to access our notes and references whenever we want; we are allowed to collaborate with others; we are allowed – and in many cases, required – to make revisions to our work. If academics were forced to do their job without access to these resources, I suspect the quality of it would drop precipitously. If these things all improve the quality of our work and help us learn and retain material, asking students to discard all of them come test time seems like a poor idea. It does require test questions to have some thought put into their construction, though, and that means another investment of time.

Some might worry that my method makes things too easy on the students. All that access to different materials means they could just get an easy “A”, and that’s why my evaluations were good. Perhaps that’s true, but just as my interest is not on assessment, my interest is also not on making a course “easy” or “challenging”; it’s on learning, and tests should be as easy or hard as that requires. As I recall, the class average for each test started at about a 75; by the end of the revisions, the average for each test had risen to about a 90. You can decide from those numbers whether or not that means my exams were too easy.

Now I don’t have the outcome measures that Braga et al (2011) did for my own teaching success. Perhaps my methods were a rousing failure when it came to getting students to learn, despite the high evaluations they earned me (in the Braga et al sample, the average teacher rating was 7 out of 10 with a standard deviation of 0.9; my average rating would be around a 9 on that scale, placing my evaluations about two standard deviations above the mean); perhaps this entire post reflects a defensiveness on my part when it comes to, ironically, having to justify my positive evaluations, just as I suspect people who cite this paper might use the results to justify relatively poor evaluations. In regards to the current results, I think both myself and others have room to be concerned: just because I received good evaluations, it does not mean my teaching method was effective; however, just because you received poor evaluations, it does not mean your teaching method is effective either. Just as students can get the right answer for the wrong reason, they can also give a teacher a good or bad evaluation for the right or wrong reasons. Good reviews should not make teachers complacent, just as poor reviews should not be brushed aside. The important point is that we both think about how to improve on our effectiveness as teachers.

References: Braga, M., Paccagnella, M., & Pellizzari, M. (2011). Evaluating students’ evaluations of professors. Economics of Education Review, 41, 71-88.  

Let’s make a few assumptions about teaching ability. The first of these is that the ability to be an effective teacher (a broad trait, to be sure, compromised of many different sub-traits) – as measured by your ability to, roughly, put knowledge into people’s head in such a manner as so they can recall it later – is not an ability that is evenly distributed throughout the human population. Put simply, some people will make better teachers than others, all else being equal.The second assumption is that teaching ability is approximately normally distributed: a few people are outstanding teachers, a few people are horrible, and most are a little above or below average. This may or may not be true, but let’s just assume that it is to make things easy for us. Given these two assumptions, we might wonder how many of those truly outstanding-tail-end teachers end up being instructors at the college level. The answer to that question depends, of course; on what basis are teachers being hired?

Glasses AND a sweater vest? Seems legitimate enough for me.

Now, having never served on any hiring committees myself, I can offer little data or direct insight on that matter. Thankfully, I can offer anecdotes. From what I’ve been told, many colleges seem to look at two things when considering how to make their initial cut of the dozens or hundreds of resumes they receive for the single job they are offering: publications in academic journals (more publications in “better” journals is a good thing) and grant funding (the more money you have, the better you look, for obvious reasons). Of course, those two factors aren’t everything when it comes to who gets hired, but they at least get your foot in the door for consideration or an interview. The importance of those two factors doesn’t end post-hiring either, as far as I’ve been told, later becoming relevant for such minor issues like “promotions” and “tenure”. Again, this is all gossip, so take it with a grain of salt.

However, to the extent that this resembles the truth of the matter, it would seem to game the incentive system away from investing time and effort into becoming a “good” teacher, as such investments in teaching (as well as the teaching itself) would be more of a “distraction” from other, more-important matters. How does this bear on our initial question? Well, if college professors are being hired primarily on their ability to do things other than teach, we ought to expect that the proportion of professors being drawn from the upper-tail of that distribution in teaching ability might end up being lower than we would prefer (that is, unless teaching ability correlates pretty well with one’s ability to do research and get grants, which is certainly an empirical matter). I’m sure many of you can relate to that issue, having both had teachers who inspired you to pursue an entirely new path in life, as well as teachers who inspired you to get an extra hour of sleep instead of showing up to their class.The difference between a good teacher (and you’ll know them when you see them, just like porn) and a mediocre or poor one can be massive.

So why ask this questions about teaching ability? It has to do with a recent meta-analysis by Freeman et al (2014) examining what the empirical research has to say about the improvements in education outcomes that active learning classes have over traditional lecture teaching in STEM fields. For those of you not in the know, “active learning” is a rather broad, umbrella term for a variety of classroom setups and teaching styles that go beyond strictly lecturing. As the authors put it, the term, “...included approaches as diverse as occasional group problem-solving, worksheets or tutorials completed during class, use of personal response systems with or without peer instruction, and studio or workshop course designs“. Freeman et al (2014) wanted to see which instruction style had better outcomes for both (1) standardized tests and (2) failure/withdrawal rates from the classes.

“Don’t lecture him, dear; just let the active learning happen”

The results found that, despite this exceedingly-broad definition for active learning, the method seemed to have a marked increase in learning outcomes, relative to lecture classes. With respect to the standardized test scores, the average effect size was 0.47, meaning that, on the whole, students in active learning classes tended to score about half a standard deviation higher than students in lecture based classes. In simpler terms, this means that students in the active learning classes should be expected to earn about a B on that standardized test, relative to the lecture student’s B-. While that might seem neat, if not terribly dramatic, the effect of the failure rate was substantially more noteworthy: specifically, students in lecture-only classes were 1.5 times more likely to fail than a student in an active learning class (roughly 22% failure rate in active learning classes, relative to lecture’s 34%). These effects were larger in small classes, relative to large ones, but held regardless of class size or subject matter. Active learning seemed to be better.

The question of why active learning seems to have these benefits is certainly an interesting one, especially given the diversity of methods that fall under the term. As the authors note, “active learning” could refer both to a class that spent 10% of its time on “clicker” questions (real-time multiple choice questions) or a class that was entirely lecture-free. One potential explanation is that active learning per se doesn’t actually have too much of a benefit; instead, the results might be due to the “good” professors being more likely to volunteer for research on the topic of teaching or likely to adopt the method. This explanation, while it might have some truth to it, seems to be contradicted by the fact that the data reported by Freeman et al (2014) suggests that the active learning effect isn’t diminished even when it’s the same professor doing the teaching in both kinds of courses.

We might also consider that there’s a lot to be said for learning by doing. When students have practice answering similar kinds of questions (along with feedback) to those which might appear on tests – either of the professor’s making or the standardized varieties – we might also expect that they do better on the tasks when they counts. After all, there’s a big difference between reading a lot of books about how to paint and actually being able to create a painting that bears a resemblance to what you hoped it would look like. Similarly, answering questions about your subject matter before a test might be good at getting you to answer questions better. Simple enough. While an exceedingly-plausible sounding explanation, the extent to which active learning facilitates learning in this manner is an unknown. In the current study, as previously mentioned, active learning could involve something as brief as a few quick questions or an entire class without lecture; the duration or type of active learning wasn’t controlled for. Learning by doing seems to help, but past a certain point it might simply be overkill.

Which is good news for all you metalhead professors out there

Another potential explanation that occurs to me returns to our initial question. If we assume that many professors do not receive their jobs on the basis of their teaching ability – at least not primarily – and if increasing one’s skill at teaching isn’t often or thoroughly incentivized, then it’s quite possible that many people placed in teaching positions are not particularly outstanding when it comes to their teaching ability. If student learning is in some way tied to teaching ability (likely), then we shouldn’t necessarily expect the best learning outcomes if the teacher is the only source of information. What that might mean is that students could learn better when they are able to rely on something that isn’t their teacher to achieve that end. As the current study might hint towards, what that “something” is might not even need to be very specific; almost anything might be preferable to a teacher reading powerpoint slides which they didn’t make and are just restatements of the textbook verbatim, as seems to be popular among many instructors who use lectures currently. If some professors view teaching as more of a chore than a pleasure, we might see similar issues. Before calling the lecture itself a worse format, I would like to see more discussion of how it might be improved and whether there are specific variables that separate “good” lectures from “bad” ones. Perhaps all lectures will turn out to be equally poor, and teaching ability has nothing at all to do with student’s performance in those classes. I would just like to see that evidence before coming to any strong conclusions about their effectiveness.

References: Freeman, S., Eddy, S., McDonough, M., Smith, M., Okoroafor, N., Jordt, H., & Wenderoth, M. (2014). Active learning increases student performance in science, engineering, and mathematics. Proceedings of the National Academy of Sciences, doi: 10.1073/pnas.1319030111.