Max-Planck-Institute for Psycholinguistics

MPI Tools

A Simple Experiment Example

Here we describe at a high level of abstraction what the steps would be to prepare ad execute an experiment. For this we choose a very simple experiment where the trial timing is given by the following sequence:

  • 500 ms after the beginning of the trial, open a voice key and wait for a verbal utterance from the subject with a time out of 5 seconds
  • at the same begin time, display an image with variable content on the screen for 2 seconds
  • measure the time from stimulus onset to the reaction as reaction time
  • after a reaction give a short tone of 100 ms

The trial timing to be specified in the graphic environment would have the following layout:

 
This kind of pulse diagram has to be drawn with the help of mouse operations by the scientific "programmer" to specify the trial timing. It shows three actions which each have on- and offset-events, as well as one asynchronous event (black area in the open phase of the voice key) symbolically denoting the moment when the subject is reacting. Nevertheless, it is an event and NESU allows the user to bind actions to any event. Let us call the above trial timing "exper.tts" where "tts" stands for trial timing scheme. Let us also assume that the user finished that step correctly by generating the corresponding (invisable to the user) Smalltalk code which is called "exper.eat" where "eat" stands for event-action-table.

Assume that the experiment consists of two conditions with two trials each. In one condition the images contain animals and in the other cars. Assume also that at the beginning of the experiment it is the intention to show a very simple image with the message "start" for 5 seconds and that this simple trial timing is stored in "begin.tts" with a fixed image. Similarly at the end the text "end" has to be shown to the subject, which is covered in the file "end.tts". The trial stack file for this simple experiment would look like this:

begin
exper "dog.pla" 1
exper "ford.pla" 2
exper "lion.pla" 1
exper "bmw.pla" 2
end

The first column in each line specifies the tts to be chosen. NESU allows the user to combine many different trial timings in one experiment. The second column defines a real image file which has to be inserted in each trial. The last column defines the condition code which is written to the output file together with the reaction characteristics.

Each line in the stack file defines a trial. Only for the trials in which the "exper" timing is used is a reaction device specified in the tts-file, i.e. only for these will a result be written into the result file. Therefore, it is very straightforward to generate such a stack file with the help of efficient text processing tools.

 

The minimal hardware solution to run such an experiment with a timing accuracy for the reaction measurement of < 1 ms would be as indicated in the following figure:

A NESU-Box is connected via the printer port to a Notebook with enough memory (>= 4 MB). A microphone is connected to the NESU-Box to implement a voice key. Also headphones are coupled to output the tone generated by the NESU-Box. The screen of the laptop is used first to control the experiment and then to present the graphic stimuli to the subject.

Experiment Preparation and Statistics 

This phase is a very difficult one, since so many different steps have to be taken dependingt on the experiment and on the devices to be used. The most important aspects are briefly mentioned here.

Speech stimuli are prepared at the MPI with the help of the XWaves/ESPS package, including some home-made extensions for easy labeling and splicing of speech files. Converters can be used to generate speech files for both versions of the SpeechServer. The software-based SpeechServer accepts as a default wav-files which are standard in the PC world.

Graphic stimuli can be prepared with the help of a wide range of typical graphic programs. Finally NESU expects an optimal pla-format which is used for fast operations. There is a converter from pcx format to pla format. Currently this procedure also holds for textual stimuli. To guarantee a well-defined timing behaviour, NESU expects that textual stimuli will also be available as images. To give some degree of flexibility, NESU allows the user to integrate a number of text elements into one full text, i.e. the pages can be assembled rather than each being generated separately. A special NESU text generator is available which is intended to generate textual stimuli efficiently. It does this by allowing the user to input ASCII text and by using the mouse or automatic procedures to define the elements which are directly converted into text objects.

Of course, many operations have to be carried out to generate final stack files. Here different randomization techniques and experiment designs play a role. At the MPI a number of different tools (often UNIX scripts) are used for this. There is as yet no general program which handles all the different wishes.

The result file generated by NESU can be used to feed SPSS or Alice for statistical analysis. It contains all the information about the reaction as well as the information about the corresponding condition codes. NESU itself does not offer real statistical analysis, although it provides some simple calculations for user or experimenter feedback such as "tell the subject how fast he/she was during the last 10 trials".