Cogsci // cognitive science and more

Like workers from all trades, scientists produce things. Bakers produce bread, construction workers produce buildings and such. And we scientists… well, we produce p values that are smaller than .05.

So what exactly is a p value? If a scientist wants to prove a point, she generally does so by testing a hypothesis. For example, she might hypothesize that rich people are happier than poor people. She could test this hypothesis by collecting happiness ratings from fifty rich and fifty poor people, and calculate a p value for the difference. The p value then expresses the chance that these happiness ratings would be as different as they are, or more different, if rich and poor people were really just as happy. (For a more detailed discussion, see my previous post.)

Are you still with me? Maybe not, but no matter: The important point is that a low p value means that your hypothesis is probably correct. (Actually, it means that the data is unlikely given the null hypothesis, but let’s skimp over this important detail for now.) The commonly accepted threshold is .05: If your p value is below .05, you have found something worthy of publication, otherwise you haven’t.

So there is a clear incentive for scientists to find p values that are smaller than .05. So what do yo do if you get a p value of .051? Well, you do what any sensible scientist would do: You test a few more participants, analyze the data …

Sharing science: Searchability, accessibility, and the future of academic publishing

The first journal purely dedicated to science was the Paris-based Journal des sçavans, founded in January 1665. The London-based Philosophical Transactions of the Royal Society was founded two months later. The Journal des sçavans appears to have died a fairly quiet death in 2007. But the Philosophical Transactions is still around, is still prestigious, and still publishes papers in pretty much the same way as it has done for centuries. I published a paper there myself not too long ago.

When these journals were founded, they provided an excellent platform for scientists to share their work. Science was mostly a regional affair, and very few people were wealthy enough to engage in the leisure of science. And, of course, there was no internet. In this small analog world, papers were the best form of communication.

Things change, though.

Some time ago I wrote about a dataset that I had downloaded from PubMed, which is a more-or-less comprehensive database of scientific articles. I downloaded information about only 38 journals, but even from this small selection it is clear that papers are being published at an exponentially increasing rate, and that this trend has been going on for some time.

Across these 38 journals, 14.544 articles were published in 2010. Not up until 2010, mind you: in 2010 alone. It is estimated that there are approximately 25.000 journals. Of course, not all journals publish the same number of articles. But still …

So what does this mean?

Firstly, it means that …

The Smartphone Revolution in psychological science: OpenSesame mobile

Recently, Sebastiaan managed to port the OpenSesame experiment builder to the Android operating system. Since then, I've spent some time working on adapting OpenSesame so that it can be used to create and package experiments that participants can download from the Google Play Store like any other app. When the participant has finished an experiment on his/her tablet or phone, the data are automatically sent back to the experimenter over the internet.

So far, I have designed and uploaded a simple reaction time experiment to test this idea, which I would be grateful if you could take a few minutes to download and try.

The app consists of a few questions about the phone you are using, a prototype of a new on-screen keyboard feature, and 32 trials of a basic reaction time task. If, for whatever reason, the app doesn't work as expected on your phone or tablet, please leave a comment below or on the forum.

The Smartphone Revolution

The spread of smartphones over the last few years has been phenomenal, and for the first time a huge proportion of the young adult population (who psychologists have traditionally relied upon as subjects anyway) carry powerful computers with touchscreen interfaces with them wherever they go.

Smartphones also offer a number of features not seen in the average lab computer, including an accelerometer, one or more cameras, vibration feedback, and in the future even eye-tracking technology. The potential of smartphones to revolutionise behavioural research is discussed in this review …

Setting up Kubuntu 13.04 for research

A few days ago, Kubuntu 13.04 Raring Ringtail was released. I am a Kubuntu user myself, and to celebrate this new release, I wanted to share a few tips on how to set up the perfect Kubuntu environment for neuroscientists and psychologists. Of course, the ‘perfect’ environment is different everyone, but there are a few things that almost every researcher in this field will need: An office suite, a reference manager, graphics software, statistics and analysis software, and experiment building software.

What is Kubuntu?

Kubuntu is a Linux distribution. If you’re not familiar with Linux, this may not mean much to you, so let’s start with a little background.

Some relations between various flavors of Linux.

A Linux-based operating system is a layer cake. It consists of many layers of software that can be stacked and combined in an infinite number of ways. Only the bottom layer is constant: That’s the Linux kernel, which is part of all Linux-based operating systems, including Android. On top of the kernel, there can be different layers of software. Kubuntu is essentially one specific selection of software. Other Linux distributions, such as openSUSE, have slightly different selections. Some differences are clearly visible, such as different desktop environments (i.e. the software that controls the start menu, etc.). Other differences are largely under the hood, such as different system management tools.

A Linux distribution arranges the many layers of software in such a way that you, as a user, don’t …

The black swan

Let’s consider a biologist with an interest in swan coloration. She goes on an expedition to an area where two groups of swans live, to investigate whether the two groups have different colors. The biologist takes her job very seriously, and first calibrates a photometer against two reference colors: One for the ideal black swan; one for the ideal white swan. She then measures the color (or rather luminance) of ten specimens from each group, obtaining a range of values where 0 is ideal black and 100 is ideal white:

To analyze her results, she runs an independent samples t-test on the measurements, which tells her that p = .0001. This leads her to conclude that the two groups have different colors. Just as she suspected all along:

Our biologist is probably satisfied at this point. But we are not. What exactly has she learned from this t-test and the resulting p-value? Let’s start with the basics: What exactly does p = .0001 mean? Well … it means that if the two groups were really of the same color, the chance of observing a color difference as extreme as she observed, or more extreme, is .0001. This is an odd and counter-intuitive statement. Yet it is the foundation of most research.

Personally, I have a hard time understanding statistics, and null hypothesis testing (the type of statistics that gives you p-values) in particular. And I when I finally think I have some grasp on it, the paradoxes …