Stimulus Delivery on the Web : What Can Be Presented when Calibration isn ’ t Possible

At the heart of most psychological experiments is the stimulus. It is the feature of an experiment over which the researcher has most control. However, as a concept, the stimulus has generally received very little attention for such a central feature of our experiments. There are some notable exceptions (Gibson, 1960), but generally, it appears that psychology has a tendency to assume that the stimulus is sufficiently robust that any error attributable to the stimulus or variations in the stimulus does not affect the outcome of the experiment. Such a position is contradicted by the relatively small but vital literature on the need to calibrate video terminals and other methods of stimulus presentation (e.g., Brainard, 1998; Krantz, 2000; Metha, Vingrys & Badcock, 1993; Olds, Cowan, & Jolicoeur, 1999). There are many circumstances that require careful calibration so that the stimulus does not become a confound in an experiment. Let me give one simple example that will be amplified below. One common assumption is that the screen is the same over the entire surface. That is the stimulus will be identical regardless of the position on the screen. The luminance, the size, the color, all of the physical attributes of the stimulus are identical regardless of position. However, the luminance is not identical across the screen surface (Cook, Sample, & Weinreb, 1993; Hu & Klein, 1994). Krantz (2000) found a variation in the luminance of over 20% on a CRT and over 40% on a LCD. This variation will affect the brightness and contrast of the stimulus depending on location. Both of these variations can affect the behavioral responses of the participants (Krantz, Silverstein, & Yeh, 1992; Snyder, 1988). This need for careful monitor calibration of our stimulus presentation device is in direct conflict with the diffuse nature of the Web. Calibration requires knowing what hardware one is using. With the Web, one cannot be certain what type of equipment the person is using at the other end of the connection. While some information such as browser, operating system and even resolution and number of colors can be