I’m Oliver Frost, a recent graduate from University of Westminster in London studying Cognitive Neuroscience B.Sc. and I am now working for Consolidata. To begin with I’m bridging both the technical and customer facing sides of the company as both a data engineer and account executive.
On the 4th July 2015 in a secondary school near Dublin, Ireland I broke the official world record for solving a 4x4x4 sized Rubik’s Cube blindfolded in a total time of 2:02.75, beating the previous time by approximately 8 seconds and claiming my 7th world record in total (you can see a list of the current records here to compare!)
If you are confused as to what is going on, the procedure is as follows: start the timer, memorize the positions of the stickers, don the blindfold, solve the cube by solving two pieces at a time without affecting anything else, and stop the timer when you’re done.
If I told you that the methods involved in memorizing a Rubik’s Cube and executing the solution blindfolded are not all that complex, you would understandably be a bit confused, but in reality this is true. With a bit of time, patience and some good guidance it is possible to learn to memorize and solve the 3x3x3 Rubik’s Cube blindfolded in under a week – in fact, I would go as far to say it is as easy to learn to solve one blindfolded as it is to solve one with a sighted method!
I do not have a photographic memory or even a spectacular memory – maybe a slightly better memory than the average person, but that really is all. For a glimpse inside the mind of a blindfolded ‘speedcuber’, you can read here about the methods involved and why anyone can solve a Rubik’s Cube blindfolded. These memory methods can be seen in action in the clip below, where I memorise a standard sized 3x3x3 Rubik’s Cube blindfolded in around 12-13 seconds.
But today, we will talk about data.
Rubik’s Cube research
I have recently graduated from the University of Westminster studying for a Cognitive Neuroscience B.Sc., eventually achieving a 2:1. It is safe to say that my choice of degree complemented the ‘speedcubing’ hobby in a number of different ways. For example, I was able to develop much of my memory methods using literature from the field of neuroscience and use research on behaviour to improve my performance at official competitions.
It also stuck with me right into my final year dissertation which tested the cognitive abilities of ‘speedcubers’, the collective term for competitive Rubik’s Cube solvers, to see if Rubik’s Cube ability was transferable to other cognitive skills. It mostly set out to try and establish cause and effect. Do speedcubers become better in their cognitive abilities? Or are those with higher cognitive abilities in the first place more likely to pursue the hobby?
This was also where my interest in big data and how a good understanding of statistics and data analysis is crucial within research and academia and also in a whole range of other sectors.
To try and show that the cube could actually play a role in academic research and be seen as more than just a party trick, it asked:
- Could regular practice in searching for specific colour patterns improve your overall attention and how quickly your brain processes information?
- Could practicing more advanced puzzles improve one’s ability to switch between tasks and plan more efficiently?
- Since the brain regions that are responsible for certain cognitive skills often overlap, could Rubik’s Cube ability transfer to other non-related areas such as verbal fluency and language skills?
- And could the cube provide a non-invasive treatment option for sufferers of developmental disorders such as ADHD, whose attention and cognitive functions are affected by the disorder?
The project studied 39 participants in total, collecting data relating to age, gender, occupation, education, the number of languages spoken and the amount of time spent playing games that exercise spatial memory to account for external factors as much as possible. Of these, 16 participants were active speedcubers and thus data relating to speedcubing experience and the number of events they frequently practiced was collected as well. All of the participants were subjected to a cognitive battery consisting of seven tests and the differences in scores between the groups was compared – simple enough?
The table above was taken from the final submission titled ‘The Puzzle of Brain Training – The Rubik’s Cube and Executive Function’. Here are some of the basic findings:
- In the Trail A and B tasks, where participants were asked to join a series of numbers or number-letter combinations (very much much like dot-to-dot) as fast as possible, the speedcubers outperformed the control group. They found the target locations significantly faster, they planned ahead much more effectively and made less errors, suggesting that speedcubers may possess better attention and processing abilities.
- In the Phonemic Fluency test, where participants were asked to name as many words beginning with the letter F within one minute as possible (and then again for A and S, called a FAS test), speedcubers produced significantly more correct responses than the rest of the study. This suggests that they may be able to access areas of knowledge within the brain more quickly than the average non-speedcuber and do this under pressure.
- In a Spatial Span test (a variation of a short-term memory test that you can try here) speedcubers remembered 2.32 units more on average than the rest of the sample. Perhaps they possess better memories for spatial patterns?
- However, there were no real differences in scores 3/4 of the verbal tasks, suggesting that abilities from Rubik’s Cube solving might not necessarily transfer to other areas of cognition.
So while it is a third year student’s dream to find significant results with large effect sizes in their final projects, how do we separate cause and effect in this instance? Especially when this project found that speedcubers only performed significantly better on tasks that drew parallels with Rubik’s Cube solving! The next step was to perform correlational analyses and linear regressions to determine whether a model of all the variables could be formed and utilized to predict performance. The results of these are below:
Some explanations of the confusing jumbles of numbers:
- This table contains correlational analyses of a number of different variables. The performances on scores where there was no significant difference between them and the control population were taken out.
- The analysis also considered a new variable called B-A. Essentially, previous research by Corrigan and Hinkeldey (1987) found that when you minus a persons Trail A score from their Trail B score, the final score correlates with overall intelligence. Speedcubers had a significant better score than the control on this measure, so it was implemented into the model.
- r, p and N? r is the strength of the correlation on a scale of -1 to +1, where positive values indicate that A increases when B increases while negative values portray a relationship where when A increases, B decreased (or vice versa). N is the number of participants considered in each stage of the analysis and p is whether the relationship between the two variables are statistically significant (p = <0.05 is good).
- There were strong relationships between cubing experience, the number of events practiced and the performances on the tests. But still, cause and effect?
In this case, linear regressions performed in a statistics program called SPSS (used a lot by academics) determined whether cubing experience and the number of events could predict performance on the cognitive tests, rather than rely on casual correlations. It found that cubing experience accounted for 21.7% of the variance in scores on the Trail A test, whereas the number of events practiced was the biggest predictor of performance in Trail B scores, accounting for 35.3% of the overall variance! Cubing experience accounted for 32.6% of the overall variance in the spatial span task as well, suggesting that both cubing experience and the number of events someone practices are both predictors of performance on various cognitive tests.
Finishing off my third year dissertation and realising that the data analysis was the most enjoyable part was my first clue that a move into big data might be right for me (this sentiment is still met with looks of anguish to a lot of the science graduates I know!) Something that satisfies the love of problem-solving, science and combining elements to see the bigger picture that I inherited from my degree appealed to me the most. But I am under no illusions as to what is ahead of me. Learning a coding language from scratch and the inside-outs of a successful big data company will be challenging and much more complex than the types of iterative analyses that went into my final project. But if I can contribute to big data half of what I contributed to competitive speedcubing then I hopefully can’t go far wrong.