UniServe*Science News, Vol. 5, November 1996

Controlling experiments in Psychology

Steve Provost, Psychology Department, University of Newcastle, Australia.

Experimental psychologists are increasingly reliant upon general-purpose computers for the collection of data in both teaching and 'publication quality' research. A typical psychological experiment consists of the presentation of a stimulus, often visual, followed by the recording of the time taken to make some kind of response. On the face of it, the computer seems an ideal tool for this kind of situation, but there are nasty issues revolving around the accurate timing of visual events on a computer monitor and responses made on a keyboard which form potentially traps for the unwary. It often seems that the psychologist must either place very heavy reliance upon 'experts' to solve such problems, or get on with the job of learning C++. Both of these strategies are costly, and should be unnecessary in a perfect world.



Figure 1. The 'New Event' dialogue box allows you to import a PICT, control its display time, and associate it with a particular response. The 'Picture' menu above the preview lets you control the picture's size, location, and other parameters associated with the presentation.

SuperLab by Cedrus comes as close to creating this perfect world as seems possible. Originally developed in a Macintosh environment, it has now been released for Windows as well. (This review is based upon SuperLab version 1.4 running under Mac OS 7.5.) SuperLab allows the experimental psychologist with virtually no programming skills whatsoever to create relatively complex designs involving visual or auditory stimuli and responses recorded either through the keyboard or serial port, without having to worry about the accuracy of either aspect of their experiment. Visual stimuli can be locked to the vertical blank, and keyboard responses can be collected with plus or minus 1 ms accuracy.

The SuperLab interface requires only that the user can point a mouse and click. Visual stimuli are imported as PICTs, and a number of properties can be attached to them (for example their duration of display, what events will cause termination of the trial, what the correct response is, etc.). Auditory stimuli may also be incorporated as SNDs. A 'trial' is then created by connecting one or more stimuli together, for example a prime, followed by a delay, followed by an event that the subject must respond to. Then blocks of trials can be created by connecting trials together in a similar way. The order of trials can be randomised, although it is to be hoped that the latest version of SuperLab has improved the procedure for seeding the random number generator. The experiment is saved as a 'script', and is run under SuperLab (with extensions switched off, of course).

SuperLab collects responses as required, and then very politely writes them out to an Excel spreadsheet compatible file, with indications of correct responses, trial type, etc. There is a facility to attach condition labels, to make summarising the data even easier. SuperLab will also save the experimental script as an Excel spreadsheet, making error checking straightforward. The manual is brief, easy to read, and informative, but hardly necessary in most cases. Some demo scripts are included which provide a good starting point for the 'cut-and-paste' experimenter.

I have found SuperLab to be invaluable. I can get an experiment up and running very quickly, and feel comfortable that the job is being done right. What's more, I have found that students, some of whom could reasonable be described as computer phobic, have been able to create experiments with very little pre-training and almost no ongoing support. This has enabled us to complete some projects which could otherwise have fallen prey to the time constraints of contemporary academic life. Better yet, creating scripts in SuperLab has provided some individuals with just about their first positive interactions with a computer, and getting a script to work correctly has a positive effect on self esteem.

SuperLab is not the magic bullet for experimental psychology, and it is important for the user to know its limitations. For example there is no 'case' or 'if-then' structure, which means that it cannot cater for experiments which involve changing what is presented on the basis of responding. However, many experimental paradigms in Psychology do not require such flexibility, and SuperLab is an ideal product in these cases. As my students have recently observed: 'Data makes everybody smile'. Collect some, and you'll see they're right.

Stephen Provost

provost@psychology.newcastle.edu.au

Figure 2. Clicking on the diamond next to an event will associate it with a selected trial. Blocks are constructed in the same way, once trials have been formed.



Chris Eccelston's review of SuperLab can be found at:

http://www.york.ac.uk/inst/ctipsych/web/CTI/ DirTxt/reviews/superlab.htm

The latest version of SuperLab for Macintosh is 1.68, and version 1.01 for Windows.

Distributor: A free demo disk can be obtained from superlab@cedrus.com. It is hoped that demos will be downloadable from www.cedrus.com before the end of November.

Cost: AU$699 (Education) for a single license: Five- and ten-user packs are also available.