The nature of visual illusions is hotly debated in the scientific literature, in search of a theory to explain how perceptual distortions arise upon daily interactions with the world. The present study provides the first direct test of Day’s (1989) Conflicting Cues theory to account for the Muller-Lyer illusion. Perceptual compromise was investigated, by measuring the impact of global and local processing on perceptions of size, as modulated by Navon stimuli. Following exposure to global, local or neutral cues, participants adjusted the length of a line to match the length of an illusory stimulus, in a series of varying trials.
However, the error rate for global and local groups did not significantly differ from the baseline condition, thus failing to support Day’s (1989) theory, and casting doubt on the usefulness of Navon stimuli in the current context. Methodological flaws and directions for future research are discussed, reaching the conclusion that multiple theories may be necessary to account for the different perceptual mechanisms responsible for the Muller-Lyer illusion.
The impact of global and local processing on the perceived adjustment error in the Muller-Lyer illusion. A test of Day’s (1989)
Conflicting Cues Theory.
Identifying the mechanisms responsible for visual illusions, facilitates comparisons of perceptual accuracy and inaccuracy, and thus helps to inform an understanding of the way in which environmental stimuli come to be represented in the human psyche (Woloszyn, 2010). In the pursuit of such knowledge, controversy continues to reign regarding the Muller- Lyer Illusion. Despite the equal length of the left and right line in Figure 1, confluxion describes the overestimation of the right line due to the fins-out arrangement and the underestimation of the left line due to the fins-in arrangement. By manipulating both the angle and length of the diagonal lines, Dewar (1967) confirmed the nature of the illusion.
Operating independently of each other, angle and length were linearly related to the size of the illusion exhibiting negative and positive relationships, respectively. Early theorist Gregory (1967), attributed this effect to the misapplication of the size constancy principle, where the incorporation of depth information serves an adaptive function in three dimensional settings. However in two dimensional settings the fins-out arrangement becomes associated with an inside corner and thus greater distance. Conversely the fins-in arrangement becomes associated with a protruding outside corner and thus a shorter distance, in all producing the illusion (Gregory, 1967).
Figure 1. The Original Muller-Lyer Figure.
This provided one of the first quantifiable theories that could be subjected to controlled laboratory conditions (Pressey, 1970). Nevertheless this theory has been widely criticized (Woloszyn, 2010). The common theme of these critiques encapsulates the notion that a comprehensive theory must possess the explanatory power to account for various forms of the illusion. Contrary to Gregory (1967), in the absence of any cues for depth perception, the illusion persists for variants of the Muller-Lyer figure (Day, 1989). Indeed, Delucia and Hochberg (1991) have demonstrated the illusion holds for three dimensional figures where there is no conflicting depth information. Given this, other researchers have deemphasized the importance of the shape of the retinal image and processing of the retina in general when accounting for the illusion (Restle & Decker, 1977).
Alternatively, Assimilation theory (Pressey, 1970) explains the illusion in terms of tendency to underestimate the largest stimulus aspect, corresponding to line length of the fins-in arrangement and overestimate the smallest stimulus aspect, corresponding to the line length of the fins-out arrangement. However this notion of regression towards the mean, has not withstood experimental manipulation of the fins-out Muller-Lyer figure (Day, 1989). Confusion theory on the other hand (Sekuler & Erlebacher, 1971), stipulates that the illusion is determined by the inter-tip distance between the arrowheads, where the larger distance of the fins-out arrangement and the smaller distance of the fins-in arrangement results in perceptual expansion and contraction, respectively .Like Assimilation theory, Confusion theory fails to accurately predict the size of the constant error for the experimental manipulation of the fins-out arrangement (Sekuler & Erlebacher, 1971).
The inability of a single theory to account for all forms of the illusion has led some researchers to conceptualize Muller-Lyer as two distinct illusions. Nevertheless, echoing the common theme of perceptual averaging, Day (1989) proposed the Conflicting Cues theory to account for Muller-Lyer as a unitary phenomenon. Herein, the conflict between two cues for size, namely the actual line length and overall length of the figure, is resolved via a compromise between local and global processing in the brain. The present study investigates the relevance of perceptual compromise to the Muller-Lyer illusion.
The role of global and local processing has previously been investigated with the use of Navon stimuli (Navon, 1977). Herein, large letters are comprised of smaller letter elements. Indeed, studies of face recognition (Macrae & Lewis, 2002; Perfect, 2003) have revealed strong and robust priming effects of Navon stimuli. However recent investigations (Lawson, 2007; Large & McMullen, 2006), have failed to replicate these findings for tasks involving judgements of inverted pictures of faces, objects or words. This raises questions of the applicability of Navon stimuli to all subsequent tasks. Indeed, effects may not translate to the following task, if it is perceived as unrelated (Lawson, 2007).
Furthermore, methodological inconsistencies in the size of the Navon stimuli, nature of the control group and motion properties of the task, may account for variable results across different study designs (Large & McMullen, 2006). Given these considerations, the current study aims to test the efficacy of Navon stimuli to modulate global and local processing in the context of the Muller-Lyer illusion. The second aim thus involves a direct test of Day’s (1989) Conflicting Cues theory. It is hypothesized that global exposure, will draw attention to the figure as a whole, exaggerate the perceived line length, and thus strengthen the illusion, as reflected by a greater adjustment error. Alternatively it is predicted that local exposure, will highlight the actual line length and thus weaken the illusion, as reflected by a smaller adjustment error.
A convenience sample of nine hundred and sixty three, Undergraduate Psychology students, from Monash University campuses, participated in the study. Participant consent was obtained. There were 225 Males and 738 Females with an overall mean age of 23.80 Years (SD= 6.75).
An experiment using a variation of the original Muller-Lyer illusion was accessed via the Online Psychology Library to investigate the adjustment error via the following URL http://opl.apa.org/Experiments/Start.aspx?EID=12 (Tew & McGraw, 2013). Each presentation consisted of an illusory stimulus with fins, presented vertically on the right, and an adjustable line without fins presented vertically on the left. Eleven displays of varying fin angle degrees were used. A continuum labelled ‘smaller to larger’ was included, with a mobile arrow above to quantify adjustment. See design section for angle and line length dimensions. A timed power point presentation of Navon stimuli was used to prime Global and Local processing. Stimuli consisted of large letters, comprised of smaller letter elements. One Hundred and twenty were used.
See Appendix A.
Participants were randomly allocated to the Global, Local or Control condition. Global and Local groups were exposed to three examples of a Navon letter. Global participants were instructed to verbalize the large letter. Local participants were instructed to verbalize the smaller letter. Control participants turned off their screen or averted their gaze for this section. One hundred and twenty Navon stimuli were presented for one second each, thus lasting two minutes in total. Commencing stage two, participants logged onto the experiment URL, using their class ID code to denote group allocation.
Using the computer mouse, participants dragged a sliding arrow to adjust the line length to match the illusory stimulus. Depending on the original set length, the adjustable line was made either longer or shorter to match the illusory stimulus. Two sets of eleven trials were conducted. The angle used was chosen at random, appearing only once within each trial. Therefore, a total of 22 trials were conducted, with each angle appearing twice.
The overall independent variable was level of processing, influenced by Navon exposure . This consisted of three levels, namely the Global, Local and Control groups. The subsidiary independent variable was fin angle, with eleven manipulations, ranging from 15 to 165 degrees, in 15 degree intervals. Fins-in ranged from 15 to 75 degrees and fins-out ranged from 105 to 165 degrees.
The overall dependant variable was the slope of the regression line, calculated by plotting adjustment errors across fin angle, thus providing a measure of illusion strength. The subsidiary dependant variable was the difference in length between the two lines, calculated by subtracting the length of the illusory stimulus from the length of the adjustable line, as measured in pixels. The illusory stimulus was set at random between 100 and 150 pixels, and the adjustable line at 90 or 160 pixels. A positive and negative difference indicated respective overestimation and underestimation of the illusion.
Data was analysed using SPSS IBM Statistics 20 (see Appendix C). A One-way independent measures ANOVA, with an alpha level of .05 was used to compare the error rate between the Global , Local and Control conditions. Scatterplots for each condition are presented below in Figures 2, 3 and 4.
Figure 2. Scatterplot for the Global Group.
Figure 3. Scatterplot for the Local Group.
Figure 4. Scatterplot for the Control Group
Table 1 presents the descriptive statistics. Table 2 presents the results of
the One-way independent measures ANOVA.
Means and Standard deviations for dependant variables
One-way Independent Measures ANOVA results