Study Skills

In a series of experiments involving college students, drawing pictures was found to be the best strategy for remembering lists of words.

The basic experiment involved students being given a list of simple, easily drawn words, for each of which they had 40 seconds to either draw the word, or write it out repeatedly. Following a filler task (classifying musical tones), they were given 60 seconds to then recall as many words as possible. Variations of the experiment had students draw the words repeatedly, list physical characteristics, create mental images, view pictures of the objects, or add visual details to the written letters (such as shading or other doodles).

In all variations, there was a positive drawing effect, with participants often recalling more than twice as many drawn than written words.

Importantly, the quality of the drawings didn’t seem to matter, nor did the time given, with even a very brief 4 seconds being enough. This challenges the usual explanation for drawing benefits: that it simply reflects the greater time spent with the material.

Participants were rated on their ability to form vivid mental images (measured using the VVIQ), and questioned about their drawing history. Neither of these factors had any reliable effect.

The experimental comparisons challenge various theories about why drawing is beneficial:

  • that it processes the information more deeply (when participants in the written word condition listed semantic characteristics of the word, thus processing it more deeply, the results were no better than simply writing out the word repeatedly, and drawing was still significantly better)
  • that it evokes mental imagery (when some students were told to mentally visualize the object, their recall was intermediate between the write and draw conditions)
  • that it simply reflects the fact that pictures are remembered better (when some students were shown a picture of the target word during the encoding time, their recall performance was not significantly better than that of the students writing the words)

The researchers suggest that it is a combination of factors that work together to produce a greater effect than the sum of each. These factors include mental imagery, elaboration, the motor action, and the creation of a picture. Drawing brings all these factors together to create a stronger and more integrated memory code.

[4245] Wammes JD, Meade ME, Fernandes MA. The drawing effect: Evidence for reliable and robust memory benefits in free recall. The Quarterly Journal of Experimental Psychology [Internet]. 2016 ;69(9):1752 - 1776. Available from:

I’ve spoken before about the effects of motivation on test performance. This is displayed in a fascinating study by researchers at the Educational Testing Service, who gave one of their widely-used tests (the ETS Proficiency Profile, short form, plus essay) to 757 students from three institutions: a research university, a master's institution and a community college. Here’s the good bit: students were randomly assigned to groups, each given a different consent form. In the control condition, students were told: “Your answers on the tests and the survey will be used only for research purposes and will not be disclosed to anyone except the research team.” In the “Institutional” condition, the rider was added: “However, your test scores will be averaged with all other students taking the test at your college.” While in the “Personal” condition, they were told instead: “However, your test scores may be released to faculty in your college or to potential employers to evaluate your academic ability.”

No prizes for guessing which of these was more motivating!

Students in the “personal” group performed significantly and consistently better than those in the control group at all three institutions. On the multi-choice part of the test, the personal group performed on average .41 of the standard deviation higher than the control group, and the institutional group performed on average .26 SD higher than the controls. The largest difference was .68 SD. On the essay, the largest effect size was .59 SD. (The reason for the results being reported this way is because the focus of the study was on the use of such tests to assess and compare learning gains by colleges.)

The effect is perhaps less dramatic at the individual level, with the average sophomore score on the multichoice test being 460, compared to 458 and 455, for personal, institutional, and control groups, respectively. Interestingly, this effect was greater at the senior level: 469 vs 466 vs 460. For the essay question, however, the effect was larger: 4.55 vs 4.35 vs 4.21 (sophomore); 4.75 vs 4.37 vs 4.37 (senior). (Note that these scores have been adjusted by college admission scores).

Students also reported on motivation level, and this was found to be a significant predictor of test performance, after controlling for SAT or placement scores.

Student participants had received at least one year of college, or (for community colleges) taken at least three courses.

The findings confirm recently expressed concern that students don’t put their best efforts into low-stakes tests, and that, when such tests are used to make judgments about institutional performance (how much value they add), they may well be significantly misleading, if different institutions are providing different levels of motivation.

On a personal level, of course, the findings may be taken as further confirmation of the importance of non-academic factors in academic achievement. Something looked at more directly in the next study.

Motivation, study habits—not IQ—determine growth in math achievement

Data from a large German longitudinal study assessing math ability in adolescents found that, although intelligence was strongly linked to students' math achievement, this was only in the initial development of competence. The significant predictors of growth in math achievement, however, were motivation and study skills.

Specifically (and excitingly for me, since it supports some of my recurring themes!), at the end of Grade 5, perceived control was a significant positive predictor for growth, and surface learning strategies were a significant negative predictor. ‘Perceived control’ reflects the student’s belief that their grades are under their control, that their efforts matter. ‘Surface learning strategies’ reflect the use of rote memorization/rehearsal strategies rather than ones that encourage understanding. (This is not to say, of course, that these strategies don’t have their place — but they need to be used appropriately).

At the end of Grade 7, however, a slightly different pattern emerged, with intrinsic motivation and deep learning strategies the significant positive predictors of growth, while perceived control and surface learning strategies were no longer significant.

In other words, while intelligence didn’t predict growth at either point, the particular motivational and strategy variables that affected growth were different at different points in time, reflecting, presumably, developmental changes and/or changes in academic demands.

Note that this is not to say that intelligence doesn’t affect math achievement! It is, indeed, a strong predictor — but through its effect on getting the student off to a good start (lifting the starting point) rather than having an ongoing benefit.

There was, sadly but unfortunately consistent with other research, an overall decline in motivation from grade 5 to 7. There was also a smaller decline in strategy use (any strategy! — presumably reflecting the declining motivation).

It’s also worth noting that (also sadly but unsurprisingly) the difference between school types increased over time, with those in the higher track schools making more progress than those in the lowest track.

The last point I want to emphasize is that extrinsic motivation only affected initial levels, not growth. The idea that extrinsic motivation (e.g., wanting good grades) is of only short-term benefit, while intrinsic motivation (e.g., being interested in the subject) is far more durable, is one I have made before, and one that all parents and teachers should pay attention to.

The study involved 3,520 students, following them from grades 5 to 10. The math achievement test was given at the end of each grade, while intelligence and self-reported motivation and strategy use were assessed at the end of grades 5 and 7. Intelligence was assessed using the nonverbal reasoning subtest of Thorndike’s Cognitive Abilities Test (German version). The 42 schools in the study were spread among the three school types: lower-track (Hauptschule), intermediate-track (Realschule), and higher-track (Gymnasium). These school types differ in entrance standards and academic demands.

A study into how well students understand specific diagrams reminds us that, while pictures may be worth 1000 words, even small details can make a significant difference to how informative they are.

The study focused on variously formatted cladograms (also known as phylogenetic trees) that are commonly used in high school and college biology textbooks. Such diagrams are hierarchically branching, and are typically used to show the evolutionary history of taxa.

Nineteen college students (most of whom were women), who were majoring in biology, were shown cladograms in sequential pairs and asked whether the second cladogram (a diagonal one) depicted relationships that were the same or different as those depicted in the first cladogram (a rectangular one). Taxa were represented by single letters, which were either in forward or reverse alphabetical order. Each set (diagonal and rectangular) had four variants: up to the right (UR) with forward letters; UR with reverse letters; down to the right (DR), forward letters; DR, reverse. Six topologies were used, creating 24 cladograms in each set. Eye-tracking showed how the students studied the diagrams.

The order of the letters turned out not to matter, but the way the diagrams were oriented made a significant difference to how well students understood them.

In line with our training in reading (left to right), and regardless of orientation, students scanned the diagrams from left to right. The main line of the cladogram (the “backbone”) also provided a strong visual cue to the direction of scanning (upward or downward). In conjunction with the left-right bias, this meant that UR cladograms were processed from bottom to top, while DR cladograms were processed from top to bottom.

Put like that, the results are less surprising. Diagonal cladograms going up to the right were significantly harder for students to match to the rectangular format (63% correct vs 70% for cladograms going down to the right).

Moreover, this was true even for experts. Of the two biology professors included in the study, one showed the same pattern as the students in terms of accuracy, while the other managed the translations accurately enough, but took significantly longer to interpret the UR diagrams than the DR ones.

Unfortunately, the upward orientation is the more widely used (82% of diagonal cladograms in a survey of 27 high school & college biology textbooks; diagonal cladograms comprised 72% of all diagrams).

The findings suggest that teachers need to teach their students to go against their own natural inclinations, and regardless of orientation, scan the tree in a downward direction. This strategy applies to rectangular cladograms as well as diagonal ones.

It’s worth emphasizing another aspect of these findings: even the best type of diagonal cladogram was only translated at a relatively poor level of accuracy. Previous research has suggested that the diagonal cladogram is significantly harder to understand than the rectangular format. Note that the only difference between them is the orientation.

All this highlights two points:

Even apparently minor aspects of a diagram can make a significant difference to how easily it’s understood.

Teachers shouldn’t assume that students ‘naturally’ know how to read a diagram.

Novick, L., Stull, A. T., & Catley, K. M. (2012). Reading Phylogenetic Trees: The Effects of Tree Orientation and Text Processing on Comprehension. BioScience, 62(8), 757–764. doi:10.1525/bio.2012.62.8.8

Catley, K., & Novick, L. (2008). Seeing the wood for the trees: An analysis of evolutionary diagrams in biology textbooks. BioScience, 58(10), 976–987. Retrieved from

Trying to learn two different things one after another is challenging. Almost always some of the information from the first topic or task gets lost. Why does this happen? A new study suggests the problem occurs when the two information-sets interact, and demonstrates that disrupting that interaction prevents interference. (The study is a little complicated, but bear with me, or skip to the bottom for my conclusions.)

In the study, young adults learned two memory tasks back-to-back: a list of words, and a finger-tapping motor skills task. Immediately afterwards, they received either sham stimulation or real transcranial magnetic stimulation to the dorsolateral prefrontal cortex or the primary motor cortex. Twelve hours later the same day, they were re-tested.

As expected from previous research, word recall (being the first-learned task) declined in the control condition (sham stimulation), and this decline correlated with initial skill in the motor task. That is, the better they were at the second task, the more they forgot from the first task. This same pattern occurred among those whose motor cortex had been stimulated. However, there was no significant decrease in word recall for those who had received TMS to the dorsolateral prefrontal cortex.

Learning of the motor skill didn't differ between the three groups, indicating that this effect wasn't due to a disruption of the second task. Rather, it seems that the two tasks were interacting, and TMS to the DLPFC disrupted that interaction. This hypothesis was supported when the motor learning task was replaced by a motor performance task, which shouldn’t interfere with the word-learning task (the motor performance task was almost identical to the motor learning task except that it didn’t have a repeating sequence that could be learned). In this situation, TMS to the DLPFC produced a decrease in word recall (as it did in the other conditions, and as it would after a word-learning task without any other task following).

In the second set of experiments, the order of the motor and word tasks was reversed. Similar results occurred, with this time stimulation to the motor cortex being the effective intervention. In this case, there was a significant increase in motor skill on re-testing — which is what normally happens when a motor skill is learned on its own, without interference from another task (see my blog post on Mempowered for more on this). The word-learning task was then replaced with a vowel-counting task, which produced a non-significant trend toward a decrease in motor skill learning when TMS was applied to the motor cortex.

The effect of TMS depends on the activity in the region at the time of application. In this case, TMS was applied to the primary motor cortex and the DLPFC in the right hemisphere, because the right hemisphere is thought to be involved in integrating different types of information. The timing of the stimulation was critical: not during learning, and long before testing. The timing was designed to maximize any effects on interference between the two tasks.

The effect in this case mimics that of sleep — sleeping between tasks reduces interference between them. It’s suggested that both TMS and sleep reduce interference by reducing the communication between the prefrontal cortex and the mediotemporal lobe (of which the hippocampus is a part).

Here’s the problem: we're consolidating one set of memories while encoding another. So, we can do both at the same time, but as with any multitasking, one task is going to be done better than the other. Unsurprisingly, encoding appears to have priority over consolidation.

So something needs to regulate the activity of these two concurrent processes. Maybe something looks for commonalities between two actions occurring at the same time — this is, after all, what we’re programmed to do: we link things that occur together in space and time. So why shouldn’t that occur at this level too? Something’s just happened, and now something else is happening, and chances are they’re connected. So something in our brain works on that.

If the two events/sets of information are connected, that’s a good thing. If they’re not, we get interference, and loss of data.

So when we apply TMS to the prefrontal cortex, that integrating processor is perhaps disrupted.

The situation may be a little different where the motor task is followed by the word-list, because motor skill consolidation (during wakefulness at least) may not depend on the hippocampus (although declarative encoding does). However, the primary motor cortex may act as a bridge between motor skills and declarative memories (think of how we gesture when we explain something), and so it may this region that provides a place where the two types of information can interact (and thus interfere with each other).

In other words, the important thing appears to be whether consolidation of the first task occurs in a region where the two sets of information can interact. If it does, and assuming you don’t want the two information-sets to interact, then you want to disrupt that interaction.

Applying TMS is not, of course, a practical strategy for most of us! But the findings do suggest an approach to reducing interference. Sleep is one way, and even brief 20-minute naps have been shown to help learning. An intriguing speculation (I just throw this out) is that meditation might act similarly (rather like a sorbet between courses, clearing the palate).

Failing a way to disrupt the interaction, you might take this as a warning that it’s best to give your brain time to consolidate one lot of information before embarking on an unrelated set — even if it's in what appears to be a completely unrelated domain. This is particularly so as we get older, because consolidation appears to take longer as we age. For children, on the other hand, this is not such a worry. (See my blog post on Mempowered for more on this.)

[2338] Cohen DA, Robertson EM. Preventing interference between different memory tasks. Nat Neurosci [Internet]. 2011 ;14(8):953 - 955. Available from:

We know active learning is better than passive learning, but for the first time a study gives us some idea of how that works. Participants in the imaging study were asked to memorize an array of objects and their exact locations in a grid on a computer screen. Only one object was visible at a time. Those in the "active study” group used a computer mouse to guide the window revealing the objects, while those in the “passive study” group watched a replay of the window movements recorded in a previous trial by an active subject. They were then tested by having to place the items in their correct positions. After a trial, the active and passive subjects switched roles and repeated the task with a new array of objects.

The active learners learned the task significantly better than the passive learners. Better spatial recall correlated with higher and better coordinated activity in the hippocampus, dorsolateral prefrontal cortex, and cerebellum, while better item recognition correlated with higher activity in the inferior parietal lobe, parahippocampal cortex and hippocampus.

The critical role of the hippocampus was supported when the experiment was replicated with those who had damage to this region — for them, there was no benefit in actively controlling the viewing window.

This is something of a surprise to researchers. Although the hippocampus plays a crucial role in memory, it has been thought of as a passive participant in the learning process. This finding suggests that it is actually part of an active network that controls behavior dynamically.

No big surprise, surely: a new study has found that computers do not magically improve students’ study skills — they tend to study online material using the same techniques they would use with traditional texts. Which means, it appears, poor strategies.

More interestingly, the study found that undergraduates who used a method called SOAR (Selecting key lesson ideas, Organizing information with comparative charts and illustrations, Associating ideas to create meaningful connections, and Regulating learning through practice) 29 to 63% more on tests of the material compared to those who mindlessly over-copied long passages verbatim, took incomplete or linear notes, built lengthy outlines that make it difficult to connect related information, and relied on memory drills like re-reading text or recopying notes.

The study involved students first reporting on their strategies for dealing with computer-based texts, then creating study materials from an online text. Different groups were asked to (a) create notes in their own preferred format; (b) create linear notes (the S part of SOAR); (c) create graphically organized matrix notes (SO); (d) create a matrix and associations (SOA); or (e) create a matrix, associations, and practice questions (SOAR). Those using the full SOAR method did best (84% correct on testing), but the dramatic difference was between SO (37%) and SOA (72%) — pointing to the importance of connecting new material to information you already know. The S group scored an average 30%, and the controls 21%.

It’s also well worth noting that, in contradiction of self-reports made by the students at the beginning, there were no signs that students left to their own devices used any association strategies.

You may think that telling students to strive for excellence is always a good strategy, but it turns out that it is not quite as simple as that. A series of four experiments looking at how students' attitudes toward achievement influenced their performance on various tasks has found that while those with high achievement motivation did better on a task when they also were exposed to subconscious "priming" that related to winning, mastery or excellence, those with low achievement motivation did worse. Similarly, when given a choice, those with high achievement motivation were more likely to resume an interrupted task which they were told tested their verbal reasoning ability. However, those with high achievement motivation did worse on a word-search puzzle when they were told the exercise was fun. The findings point to the fact that people have different goals (e.g., achievement vs enjoyment), and that effective motivation requires this to be taken account of.

In another demonstration of the many factors that affect exam success, three experiments involving a total of 131 college students have found that seeing the letter A before an exam makes a student more likely to perform better than if he sees the letter F instead. In the first experiment, 23 undergraduates took a word-analogies test, of which half were labeled "Test Bank ID: F" in the top right corner, and half "Test Bank ID: A". The A group got an average of 11.08 of 12 answers correct, compared to 9.42 for the F group. The same pattern was confirmed in two more studies. Moreover, performance of students whose exams were labeled "Test Bank ID:J" fell between those with the A and F test papers. While hard to believe, these findings are consistent with the many findings supporting the idea of "stereotype threat" (the tendency to do less well on a test when a person fears their performance could confirm a negative stereotype about their racial or gender group).

It has been well-established that, compared to younger adults, older adults require more practice to achieve the same level of performance1. Sometimes, indeed, they may need twice as much2.

In the present study, two groups of adult subjects were given paired items to learn during multiple study-test trials. During each trial items were presented at the subject's pace. Afterwards the subjects were asked to judge how likely they were to be able to recall each item in a test.

It was found that people were very good at accurately judging the likelihood of their correct recall. Correlations between judgments and the amount of time the subjects studied the items suggested that people were monitoring their learning and using this to allocate study time.

However, older adults (with a mean age of 67) used monitoring to a lesser degree than the younger adults (with a mean age of 22), and the results suggested that part of the reason for the deficit in recall commonly found with older adults is due to this factor.


1. For a review, see Kausler, D.H. 1994. Learning and memory in normal aging. New York: Academic Press.

2. Delbecq-Derousné, J. & Beauvois, M. 1989. Memory processes and aging: A defect of automatic rather than controlled processes? Archives of Gerontology & Geriatrics, 1 (Suppl), 121-150.

Salthouse, T.A. & Dunlosky, J. 1995. Analyses of adult age differences in associative learning. Zeitschrift für Psychologie, 203, 351-360

Dunlosky, J. & Connor, L.T. (1997). Age differences in the allocation of study time account for age differences in memory performance. Memory and Cognition, 25, 691-700.

A survey of 824 undergraduate students has found that those who were evening types had lower average grades than those who were morning types.

The finding was presented at SLEEP 2008, the 22nd Annual Meeting of the Associated Professional Sleep Societies (APSS).