Lesson 7: Visual constancies and illusions
Analyse Müller-Lyer illusion data
SkillsAnalysing perception dataConstancies and illusions
Worked example
Eight students each adjusted a comparison line to match the 100 mm shaft of a Müller-Lyer figure, once with fins outward and once with fins inward. Half did the fins-outward version first and half the fins-inward version first.
| Student | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 |
|---|---|---|---|---|---|---|---|---|
| Fins outward (mm) | 116 | 121 | 112 | 119 | 124 | 114 | 118 | 120 |
| Fins inward (mm) | 93 | 90 | 95 | 91 | 88 | 94 | 92 | 89 |
(a) Calculate the mean and standard deviation of each condition, to 2 decimal places. (b) Calculate the mean illusion strength for the fins-outward condition. (c) A paired t-test gave , . Draw a conclusion. (d) Identify the design and one control it used.
Step 1: Means
Fins outward: mm. Fins inward: mm.
Step 2: Standard deviations
Fins outward: the squared deviations sum to 106, so mm. Fins inward: they sum to 42, so mm.
Divide by , the sample standard deviation.
Step 3: Illusion strength
: on average the shaft with outward fins was seen as 18% longer than it is.
Step 4: Conclusion from the t-test
, which is less than 0.05, so the difference is statistically significant. Reject the null hypothesis: fin direction affected the perceived length of the shaft, with outward fins seen as longer (mean 118.00 mm against 91.50 mm).
Step 5: Design and control
Every student did both conditions, so it is a repeated-measures design (hence the paired t-test). Half did each order first: counterbalancing, which controls order effects such as practice.
Answer
(a) Fins outward 118.00 mm (SD 3.89); fins inward 91.50 mm (SD 2.45). (b) 18%. (c) The difference is significant (p < 0.001 < 0.05): outward fins make the shaft look longer. (d) Repeated measures, with counterbalancing.
Try a similar one
A fresh question of the same kind, checked as you answer, with hints and a full solution.
Method
Mean, then the sample SD (divide by n − 1).
Illusion strength = (matched − true) ÷ true × 100.
Compare p with 0.05, then state the conclusion in terms of the variables.
Watch out for
Using the population SD (dividing by 8): 3.64 mm instead of 3.89 mm.
Concluding "the hypothesis is proven". A significant result supports it; a type I error is still possible.