As mentioned earlier, the three perceptual tasks were chosen as three representatives of low-level perceptual processing. But the task in Experiment 2, of looking for a possible effect of associative processing on hierarchical global–local processing, had a secondary rationale. Specifically, in another line of research, it has been suggested that there is a reciprocal link between associations and mood (Bar, 2009). Indeed, studies have found that increasing the breadth of associative processing can positively influence mood (Brunyé et al., 2013; Mason and Bar, 2012). Furthermore, it was found that positive mood was directly associated with global bias in perception, and inversely related to local bias (Basso et al., 1996; Gasper and Clore, 2002). Taken together, although we do not study mood here, given that associative processing affects mood, and mood affects global and local perception, we examine in Experiment 2 the effect of associativity directly on global and local perceptions.
The development of object recognition is widely believed to be a hierarchical process, progressing from global to local properties (Kimchi, 1992; Navon, 1977). Operationally, local processing is based on selective attention to individual elements of an object or a scene. In contrast, global processing involves integrating spatial-local elements by linking them together into a larger form of a global structure (Kimchi, 1992). That processing of the global form typically precedes the local form (Navon, 1977) was dubbed the global precedence effect. Similarly, it is argued that the global properties, conveyed mainly by low spatial frequencies, trigger top-down facilitation predictions in visual recognition (Bar et al., 2006).
In this experiment, we evaluated how associative processing (or its lack thereof) can affect global vs. local perception. It has been shown that cognitive load, operationally defined, diminishes the precedence of global information (Hoar and Linnell, 2013). Therefore, if the search for an associative link even for the non-associated pairs continues into the visual task (Global vs. Local in this case), we would expect to see less of a global precedence effect in the non-associative condition.
Materials and methods
Forty Bar-Ilan University students with normal or corrected-to-normal visual acuity and with no ocular pathology (27 females, age range 18–32, mean age = 24.05) participated in the experiment for credit toward a course requirement.
Five hundred ninety-four pictures of objects, neutral in their emotional valence, were used in 276 image pairs and 42 individual images. All objects were daily items in our lives (e.g., glass, hat, and telephone). The target display was a ‘Navon’ stimulus. The inconsistent Navon set of stimuli consisted of compound letters consisting of several numbers of lower case Ss or Hs (14 lower case Ss, 12 lower case Hs; local letters) configured to form a global H or S (‘Navon’ stimuli), respectively. The consistent Navon set of stimuli consisted of a global H formed by lower case Hs and a global S formed by lower case Hs. The global letters subtended 3 × 2 degrees of visual angle, and the smaller local letters were 0.4 × 0.35 degrees of visual angle. Each letter was displayed 54 times per block.
The experiment began with a practice phase that consisted of 12 image pairs followed by Navon stimuli. During the practice phase, the image pairs were a mixture of associated and non-associated pairs (six image pairs for each condition). The experiment was divided into three blocks (baseline, associated, and non-associated block). Each block consisted of 216 trials (determined by the performance described in the literature (Navon, 1977)). The associated and non-associated blocks also consisted of 21 catch trials which were used to verify the performance of the task (a total of 237 trials each; 216 experimental trials, 21catch trials). The baseline block did not consist of image pairs, but rather only a Navon Stimulus. Participants participated in all three blocks, one after the other. Each pair was seen no more than twice during the experiment.
Before the experiment began, participants were given instructions about the general nature of the task and were instructed to attend to the objects. No instructions were given about the nature of the images or the possible relations between them.
During the baseline block, a Navon stimulus was presented in the center of the screen for 40 ms. Participants had to judge which letter they perceived first and respond to it as quickly as possible by pressing “S” if the perception was the letter s or by pressing “H” if the perception was the letter h.
During the associated and non-associated blocks, two images were presented consecutively in the center of the screen (54 × 54 pixels). A Navon stimulus was subsequently also presented in the center of the screen. Each trial consisted of seven presentations (Fig. 3): a fixation cross (500 ms), an object image (300 ms), another fixation cross (250 ms), a second object image (300 ms), a fixation cross (250 ms), a Navon stimulus (40 ms), and a question mark for a response. The pairs of objects could be either associated or non-associated according to the characteristics of the block. Navon stimuli (Navon, 1977) have consisted of a global letter formed by the configuration of local letters (e.g., a global S composed of lower case Hs or a global H composed of lower case Hs). Participants had to judge, in their opinion, which letter they perceived first and respond to it as quickly as possible by pressing “S” if the perception was the letter s (global or local) or by pressing “H” if the perception was the letter h (global or local).
To ensure participants attend to the objects, we added 21 catch trials in each block. In catch trials, two identical images presented consecutively in the center of the screen. After the Navon stimuli were presented, participants were required to ignore it and to press the “Space”. All stimuli were presented randomly. The order of the blocks (baseline, associated, and non-associated) was randomized and counterbalanced across participants. No feedback was given because there was no correct or incorrect response in this experiment. The time elapsed from the stimulus onset to response was computed on-line for each trial and stored as the RT and response. If a participant committed more than 50% errors in the catch trials, the participant was omitted from the analysis.
The data from the experiment were analyzed to determine whether there were differences in the number of times that the participant chose the global property between the blocks (baseline, associated, and non-associated).
The test score was calculated by the number of times that the participant chose the global property. Processing precedence was calculated by the number of times the participant chose the global property relative to the total number of attempts. To test the mean differences in scores between blocks, we conducted a repeated measurement ANOVA with baseline, associated and non-associated blocks as the independent variable. This analysis revealed a main effect in the number of times that the participant chose the global property between conditions (baseline block = 79.14%; associated block = 80.88%; non-associated blocks = 69.86%; F(2, 78) = 7.811, p < 0.001, η2 = 0.167, observed power = 0.944, BF10 = 34.92). The mean numbers of times that the participant chose the global property are plotted as a function of the blocks (Fig. 4a). Pairwise comparisons analysis (using Bonferroni correction) revealed a significant difference between the baseline and the non-associated block (p < 0.05, Cohen’s d = 0.398, CI [0.0655, 18.499], BF10 = 2.742). Also, a significant difference between the associated and the non-associated block was found (p < 0.005, Cohen’s d = 0.547, CI [3.046, 18.990], BF10 = 23.674). However, like in the previous experiment, there was no significant difference between the baseline and the associated block (p = 1, Cohen’s d = 0.153, CI [−6.234, 2.762], BF10 = 0.263), and therefore our subsequent analyses included only the associated and the non-associated blocks. The results showed that the number of times that participants chose the global property in the non-associated block was lower than in the associated block (Fig. 4a). Additionally, we conducted a paired t-test for analysis of RT between the associated and the non-associated block. The analysis revealed that the RT was significantly longer for a global property in the non-associated block than for the associated block (associated block = 447.13 ms; non-associated block = 500.50 ms; two-tailed t(39) = 2.234, p < 0.05, Cohen’s d = 0.535, CI [5.037, 101.70], BF10 = 1.564, Fig. 4b).
Like in the previous experiment, and following the same reason, we tested whether the order of the blocks affected the results. We calculated the differences in choosing the global property in each block for each participant and contrasted the results based on whether the starting block was the associated or the non-associated block. A paired t-test revealed no significant difference in choosing the global property between the blocks (starting block: associated block =12.40%; non-associated block = 9.63%; two-tailed t(19) = 0.411, p = 0.685, Cohen’s d = 0.092, CI [−11.35, 16.91], BF10 = 0.251, Fig. 4c), meaning, like we showed in the previous experiment, the starting block did not affect the results.
These results are similar in vein to what we found in the previous experiment, whereby a lack of an associative link results in subsequently degraded performance, presumably because of the cognitive load that a continued search for an associative link pose. In the third experiment, we tested for a possible effect of lack of associativity on yet another aspect of low-level visual perception.