PAGE 8 OF 10

### Interpretation of Main Results

The result window above shows the main effects of our 2 independent variables for a hypothetical subject, that is: the main effect of type-of-trial and the main effect of (the comparison's) arrowhead. It also shows the interactive effects of these two variables. Specifically, there is a main effect of (the comparison's) arrowhead: when its arrowhead was the same as the standard's arrowhead, the effect was about 1.3%. However, when its arrowhead was different, the effect was about 11%.

Why are we interested in the difference between these 2 conditions? We know there is an illusion when the arrowheads are different, and we want to measure the effect of the illusion on the perception of relative length. In order to do this, however, we need to know how good people are at this task under a "normal," non-illusion condition (i.e., same arrowhead condition). The 1.3% results tells us that while our subject is accurate at the task when there is no illusion, the subject is not perfect. Thus the effect of the illusion probably is not fully 11% ; rather the effect is probably only 9.5% (i.e., 11%-1.5%). In other words, the same arrowhead condition is acting like a control for the "normal" errors a subject might make when trying to equate the length of two lines.

These two conditions illustrate the benefits of doing experiments: given a control condition, we can assess how other stimulus features might contribute to the Müller-Lyer illusion. For example, another experiment could keep different arrowheads on the two lines throughout the experiment while varying the angle the arrowheads make with the main line segments; in this way, the effect of the angle of the arrowheads in either enhancing or reducing the illusion could be assessed.

As for type of trial (ascending versus descending), this variable seems to have little effect: 6.5% versus 6%. Also, the effect of the comparison line having an arrowhead different from the standard does not vary much between type of trial (11.7% vs. 10.5%), and the "normal" errors our subject made in the control condition (i.e., same arrowhead condition) also did not vary much between type of trial (1.3% vs. 1.5%).

However, with another subject it is quite possible that the type of trial might interact with the type of arrowhead. For example, for whatever reason, a descending type of trial might be more difficult than an ascending type of trial. If this is so, then we might expect the effect of the Müller-Lyer illusion to be greater in a descending type of trial. On the other hand, the type of trial could have little effect in the "control" (same arrowhead) condition, but might have a big effect the "experimental" (different arrowhead) condition.