As a platform for research in perception and cognition, RED supports a variety of experimental designs, including factorial designs and expanded factorial designs (see Massaro, 1998), to investigate the manner in which perceivers combine information from different knowledge sources (e.g., speech sounds and facial movements). The RED also allows researchers to design and conduct experiments in which people interact with computers to perform interesting and useful tasks. In addition, the RED provides an ideal tool for use in laboratory courses (e.g., perception, psycholinguistics) to teach students to design and conduct research. Learning to apply the scientific method promotes critical thinking, and offers students new ways to conceptualize problems and investigate the world.
We illustrate the features of the RED by describing several experiments and their implementations.
Figure 1. Expanded factorial design with three auditory syllables crossed with three visual syllables.
These 15 unique trial types are tested repeatedly for a given participant. A common strategy is to sample these 15 trial types randomly without replacement in each block of 15 trials. A given experimental session might contain between 200 and 300 trials. For our experiment, we present 16 blocks of trials per experimental session, and treat the first block as practice, giving 15 observations per condition per session. Two sessions are presented, with a 5 to 10 minute break between sessions. Thus our data analysis can treat the two sessions as a variable of interest; for example, do the results differ in some way across the two sessions?
The response alternatives must be specified for the participants. It
is possible to limit the possible alternatives to just a few choices, or
to provide essentially an open-ended set of possibilities. In this example,
we allow 9 possible responses: the three test items and the six possible
clusters resulting from their combination. These nine alternatives are
/ba/, /da/, /ga/, /bda/, /bga/, /dba/, /dga/, /gba/, and /gda/. The instructions
to the subjects are given in Table 1.
We are studying the perception of speech when you can both see and hear the speaker. In this experiment, you will two sessions of about 20 minutes, with a short break in between. On each trial you will be presented with a speech stimulus over the TV monitor and identify what is spoken. Your task will be to identify what the speaker said. On some trials, the screen will be blank and you will just be presented an auditory speech sound. On some other trials, just the picture with no sound will be presented, and on other trials, you will be presented both visual and auditory information. That is, you will both see and hear the speaker.
You will make your response using only these 3 yellow buttons: B, D, G. On some trials you might use more than one of these buttons. For example, if the speaker said "BDA" you would press B followed by D. If the speaker said "GBA" you would press "G" and then "B". If the speaker said "GA", you would just press the "G" button. If you hit a wrong button you can hit BACKSPACE to erase one button. What you type will appear on the terminal screen. After you have pushed the appropriate buttons please hit RETURN to enter your response. The next trial will start a shot time later. Altogether, there are 12 possible responses you can make.
On some trials, you might not be sure about what was said. Even if you are unsure, simply make the best decision you can. It is critical that you respond on each trial. The order of syllables is strictly random and there is no sequence of sounds that you could guess. Try to respond as soon as you arrive at a decision, avoiding any excess delay. If you don't respond within about 5 seconds, the computer will beep to remind yo u. Perhaps you have not pressed the buttons fully, so try pressing it again. We want your careful identifications of the syllables, but it is most important that we get your response for each one, OK?
In summary, on each trial you will be presented with a syllable that you will identify using one or two of these YELLOW buttons lableled B, D, and G. Any questions?
On each trial, a stimulus event is presented and the participant
selects a response. Given that some trials do not have correct answers,
accuracy feedback is not meaningful in this task. However, it is possible
to present feedback as we will illustrate in another sample experiment.
The data analysis consists of a confusion matrix of stimuli times responses. In this case, there are fifteen possible stimuli and nine possible responses. The entry in each cell represents the proportion of times each response was given to each stimulus. Of course, statistical analyses and mathematical model fitting are also possible.
In the tutorial, we describe in more detail how a simplified version of this experiment is programmed.