MEASUREMENT OF EYE SIZE ILLUSION 167 upward to downward or from downward to upward. Because the Eyelash 5 condition was not combined with all levels of eyeliner, and its PSE did not differ from the PSEs of the Eyelash 2–4 conditions, Eyelash 5 was excluded from further analysis. A two- way repeated-measures analysis of variance (ANOVA) with eyelash makeup and eye- liner as the independent factors indicated that both main effects were statistically signifi cant, F3, 63 = 91.33, p 0.001 and F4, 84 = 5.05, p = 0.001, respectively. In addition, the interaction between the two factors was also statistically signifi cant, F12, 252 = 9.79, p 0.001. The multiple comparisons between the eyelash makeup conditions revealed that the perceived eye size without eyelash makeup (Eyelash 1) was smaller than those with eyelash makeup (Eyelash 2–4), p 0.05. However, there was no statistically signifi cant difference between the conditions in which eyelash makeup was present (i.e., Eyelash 2–4), indicating that thin eyelash makeup was just as effective as thick eyelash makeup. The mean of perceived eye size with eyelash makeup (excluding the condition without eyelash makeup) was approximately 106.1% (113% in area). The simple main effects revealed that eyeliner affected the perceived eye size especially when eyelash makeup was not applied, F4, 336 = 25.04, p 0.001. Multiple comparisons between stimuli with eyeliner but without eyelash makeup revealed that the perceived eye size tended to increase as the eyeliner became thicker, although there was no statisti- cally signifi cant difference between the stimuli of brown eyeliner on both eyelids (Eye- liner 3) and black eyeliner on both eyelids (Eyeliner 4). The perceived eye size of the stimulus Eyeliner 5, Eyelash 1 (i.e., the thickest eyeliner without eyelash makeup) was approximately 104.8% (110% in area). Eyeliner and/or eyelash makeup induced the size illusion eyes with makeup appeared larger than they really were. The increase in perceived eye size with eyelash makeup was equivalent to approximately 6% of the original eye size. Interestingly, within the range of eyelash makeup in this experiment, any degree of eyelash makeup had the same effect on eye size perception. On the other hand, the infl uence of eyeliner was clear only when eyelash makeup was absent. Eyeliner did not make the eyes appear any larger in the pres- ence of eyelash makeup. These indicate that, as far as eye size illusion is concerned, eye- liner and eyelash makeup do not function additively. EXPERIMENT 2 In this experiment, we psychophysically examined whether another major component of eye makeup, namely eye shadow, also causes eye size illusion. In addition, we tried to overcome some of the limitations of Experiment 1, which used a facial image of only one individual as the basis for all the standard stimuli. This might limit the generalizability of the conclusions in Experiment 1 because the results might depend on particular facial features of the model. To overcome this problem in Experiment 2, we used several models whose faces were fairly distinct from one another. We also increased the number of ob- servers to more than 100 so that the results would better refl ect the general public’s perception. In Experiment 2, only female observers were involved because most cosmetics are purchased and used by females. Therefore, from a marketing perspective, investigat- ing how females perceive faces with makeup is more useful and cost-effective than inves- tigating male perceptions.

JOURNAL OF COSMETIC SCIENCE 168 METHOD PARTICIPANTS A total of 104 females (mean age 35.1 years, S.D. = 8.5) voluntarily participated in this experiment. They received a reward for their participation. All reported normal or corrected- to-normal visual acuity and normal color vision. All participants were unaware of the purpose of the experiment and were tested individually. STIMULI AND APPARATUS The experiment was conducted using a computer with custom software. The stimuli were presented on a 24.1-inch LCD screen (NANAO FlexScan SX2461W-BK, EIZO Corpo- ration, Ishikawa, Japan). Although we did not secure the observing position with an apparatus, the viewing distance remained constant at approximately 75 cm. The stimuli were photographs of six Japanese females with/without eye shadow. The ap- plied eye shadow was Maquillage True Eye Shadow® manufactured by Shiseido Co., Ltd. (Tokyo, Japan) This product is a multicolor palette eye shadow composed of fi ve carefully balanced shades including “line color” for the edge of eyelids and “main color” for broader area around the eyes, and so on. The general method of creating the stimuli was the same as that in Experiment 1. The dimensions of the stimuli were 744 pixels wide (16.3° in visual angle) and 1052 pixels high (22.9°). The mean dimensions (across the six models) of the face itself were approximately 479 (S.D. = 19.6) pixels wide (10.6°) at the cheek- bone level and approximately 807 (S.D. = 26.7) pixels high (17.7°) from the top of the head to the tip of the chin. The stimuli were color images with a gray background. The standard stimuli were facial images with/without eye shadow (Figure 5). Because the levels of eye shadow were binary (with or without) for the six models, the total number of standard stimuli was 12. The comparison stimuli were images of individual faces, without any eye makeup, whose eye size was sequentially changed from 90% to 110% of the original eye size (i.e., 100%) in steps of 2% both horizontally and vertically (Figure 6). These comparison stimuli were made for each model. Note that the original eye size of one model might differ from that of another model thus the percentage indicates the relative size within models. PROCEDURE The same procedure as that of Experiment 1 was used, except for the following. For coun- ter balancing, half of the participants were instructed to choose the stimulus whose eyes appeared smaller, and the other half were instructed to choose the stimulus whose eyes ap- peared larger. For each standard stimulus, there were two initially ascending staircases and two initially descending staircases hence, the experiment consisted of 48 concurrent staircases of trials, randomly interleaved. In all the trials, the model in the presented comparison stimulus was always the same person as in the standard stimulus. The eye size of the comparison stimulus for the fi rst trial of each staircase was selected from either 90% (ascending series) or 110% (descending series).