Crossword puzzles as a study aid

By far one of the most common types of scientific research I see involving crossword puzzles is regarding their use in education. Uses range from elementary school (vocabulary terms), to second-language learners, to advanced topics.  A recent paper in the Academic Medical Journal of India typifies this last approach.

A study comparing speed training to crossword play

A couple recent papers by a research group by Fredric Walinsky have come out of a large study using cognitive training to see if there are benefits of a specific video-game style intervention (in the image) a year after practice.  They happened to compare their speed-training with an 'attention control' that was actually a crossword puzzle, and so I thought it would be interesting to review the results here.

 The PLoS paper from 2013 on this should be accessible to anyone, although the  JAH paper is paywalled. These both came out of the same study, but looked at different outcomes.  The PLoS paper looked at a set of cognitive tests, whereas the JAH paper looked at activities of daily living and depression

Moxley et al Study of crossword experts

When it rains it pours--In the last month or so, three studies using crossword experts have been published, including a study by Toma I blogged about earlier, a study published by my own lab, and one that just appeared by Moxley, et al.  Remarkably, each of these studies looked at  crossword players during different years of the ACPT, dating back to 2005 for the Moxley paper. In fact, I think that Tuffiash, on of the authors, appeared in the Wordplay movie.

The Moxley paper is coauthored by Anders Ericsson, who is known for noticing the 10-year 10,000-hour rule of experts that now often gets attributed to Malcolm Gladwell.  This paper was only looking partly at past experience of players in order to predict performance at the tournament.  It turns out there is indeed a relatively strong correlation between tournament performance and  past experience, although it only accounted for 19% of the total variance. There must be other factors that can explain why some people are better than others then, and one of these turned out to be how long it has been since someone took a 'break' or 'hiatus'.

Many players have taken time off from playing over their history.  As Moxley et al, write, as far back as Ebbinghaus (basically the first modern memory researcher), people have been interested in how long it takes to relearn knowledge, and how that impacts retention.  It turns out that taking a hiatus from solving negatively impacted final standings at the ACPT.  Furthermore, the time since that hiatus was also important--the more time that had passed since the hiatus, the better the performance.  Hiatus predicted another 5% of the variance, which is about 20% of the total explained variability.

One of the remarkable aspects of this study is that the reported time spent playing each week is a relatively poor predictor of their final performance.  All history-based measures together accounted for about 20% of the variance, and one of the best ones here was age (with older people doing relatively worse on in the tournament).

I have a theory about this: the best players actually do not spend a lot of time playing puzzles!  This is evident from watching the bloggers who report their puzzle times.  The best players may solve 5-10 puzzles a day, but they do it in a 30-60 minutes.  Although this is on par with the 3 hrs/week spent solving puzzles across participants that Moxley reports, it might be that the time spent solving puzzles by fans of every level is constrained by their free time, and they solve as many puzzles as they can do in that time.  Here, time would only be a poor predictor.  Moxley did not ask about how many puzzles were solved, but rather how much time.  I myself have collected this data in our most recent study, and so we may have a good answer to this question before too long.

Anyway, the upshot of this study is that the amount of time you take off of any deliberate practice can hurt you, possibly for years afterwards.

On Toma, Halpern, & Berger (2014)

Two crossword-related papers were published in the last month; one by my student that I will post on soon, and another that came out last month by Michael Toma, Diane Halpern, and Dale Berger.

The Toma study compared both crossword and scrabble experts, and looked at whether there were any substantial differences between experts in these two areas.  What is probably most interesting about this study is in the differences they did not find, rather than the differences they did find.

Summer 2014 online Crossword Study

Anybody is invited to take part in the 'cognitive crossword challenge'

http://challenge.cls.mtu.edu/WebXword4/

Feel free to forward this to anyone who you may feel would be interested. Participation in the study is voluntary and you will not be paid (but hopefully it will be fun).

We are interested in players of all levels and experiences.  There is a short background survey, followed by up to six short games, all played through a web browser (you need to use a computer with a keyboard; tablets/phones won't work). In the study, you get a clue with a partially filled in answer and have to try to solve it.  You can play as few or many of the games as you would like.  This is related to the study we ran at the ACPT in March, but if you did part of that study previously, you can take part again on-line.

Instructions:

To begin, follow the link to
http://challenge.cls.mtu.edu/WebXword4/
and click on the 'register' button.

Your identify will not be tied to your data, but the registration will let you do part of the study in one setting and come back to finish it later.

Then, once you enter in a name and email (fake names are OK), you will get a message via that email to confirm registration.  That will take you back to the website to log in with the name/password you just created.

Then, click the 'game set 1' option on the left side of the screen.  Please do the survey first, and then as many of the games (up to 6) as you have time to.  You do not have to do this all in the same setting.

Presentaton at the Society for Mathematical Psychology

I'm in Quebec at the Society for Mathematical Psychology meetin
presenting some of my research on crossword puzzle solvers.  The complete slide presentation can be found here.

A highlight of the presentation was the video below, which shows an expert and novice playing the same puzzle. I'll let you guess who is who.

 

In fact, the novice was actually sped up by a factor of 4.

Below is a brief review of the presentation.

Error detection, backtracking, and expert problem solving.

As I discussed in an earlier posting about a computer model of Sudoku, computers can solve puzzles very differently than humans do.  Because computers can be very fast and do a good job of systematically exploring a problem, you can give let them churn on a problem and exhaustively figure out the answer.  For something like Sudoku, it is easy to figure out if the answer is correct (it is just simple math, or really, it is just pattern sorting). Problems like Sudoku are sometimes called 'well-structured', meaning that you can tell if you are right or wrong.  Other puzzles, problems, and tasks are ill-structured; anything that is not well-structured.

Human and Machine Problem Solving in Sudoku (and crosswords)

A new preprint in Arxiv by Radek Pelánek takes a look at human problem solving of Sudoku, and has been picked up by various media sources (see The medium,  and slashdot)

Above is a picture of how well their new method predicts solution times across several hundred puzzles.