Effects of Transient and Nontransient Changes of Surface Feature on Object Correspondence

Abstract

Object correspondence is a fundamental problem in perception. Classic theories hold that the computation of correspondence is solely based on spatiotemporal information. Recent research showed that surface features also play an important role. However, the surface features of objects in many studies did not change throughout a trial. This study investigated the effect of feature change on object correspondence using the object-reviewing paradigm. Two moving objects underwent transient feature changes on color dimension (Experiment 1A) or a combination of three dimensions (Experiment 2A). Moreover, the objects moved behind four occluders to make the feature change nontransient (Experiments 1B and 2B). Object-specific preview benefits were reduced or eliminated when feature changes were transient, but the benefits were not affected when feature changes were nontransient. The effects of transient versus nontransient changes of surface feature in object correspondence are discussed.

Keywords

object correspondence object persistence feature change occlusion

In real-world environments, objects are moving over time and sometimes are occluded by other objects, but observers can easily recognize an object as the one appearing elsewhere a moment ago. It is a classic question how the visual system establishes correspondence between two objects (for a review, see Scholl, 2007). During the past few decades, researchers have found two categories of information that could be used to solve this correspondence problem. Spatiotemporal information, as one such category, has been proposed as the dominant or even the only factor that determines the correspondence between objects (Flombaum et al., 2009; Kahneman et al., 1992; Mitroff & Alvarez, 2007). This claim derives from the prominent object-file framework. An object file is an episodic visual representation that contains surface features of an object (Kahneman & Treisman, 1984; Kahneman et al., 1992). The object is assigned a position marker and the attached position marker goes along when the object moves. Thus, the comparison of objects appearing at different locations and times is based on the spatiotemporal information. When this information indicates that the object viewed later is the one appearing earlier, then correspondence establishes between these two appearances and they are treated as the same object (see also Pylyshyn, 1989).

Empirical evidence supporting the object-file theory comes from the object-reviewing paradigm (Kahneman et al., 1992). In the paradigm, two preview letters appear in two squares, usually located equidistant to the left and right of the fixation. After a short time, the letters disappear and the squares begin moving, either clockwise or counterclockwise, before stopping just above and below the fixation. One test letter then appears in one of the squares, and the participants are required to name the letter or judge whether this letter was one of the preview letters. It is not surprising to find faster naming latencies or response times (RTs) when the letter was one of the preview letters compared with when a new letter appears. In addition to this nonspecific preview benefit, what is more important is the observation of the object-specific preview benefit (OSPB), which refers to faster naming latencies or RTs when the letter was a previous one and it appeared in the same square as it had appeared originally, compared with when the letter was a previous one but appeared in the opposite square. Note that whether the letter appeared in the same or the opposite square can only be inferred on the basis of the spatiotemporal continuity, so the OSPB is an indicator of the establishment of object correspondence. Later studies often use symbols or novel graphs instead of letters as preview and test stimuli (e.g., Gao & Scholl, 2010; Hollingworth & Franconeri, 2009). Furthermore, two test stimuli are presented when objects stop moving, and participants are asked to decide whether these two stimuli were the same as those viewed at the beginning (e.g., Hollingworth & Franconeri, 2009; Moore et al., 2010). In this case, the two test stimuli are both the same as before (match condition), or one of them is different (no-match condition). When they are the same, they can appear in the same object (congruent-match condition) or in the different object (incongruent-match condition). The OSPB is then the RT difference between congruent-match and incongruent-match conditions. These modifications make the OSPB larger and more robust than that obtained from the original paradigm proposed by Kahneman et al. (1992).

In addition to spatiotemporal information, an object carries surface feature information as well. Whether an object’s surface feature could guide correspondence has long been debated. In their seminal article, Kahneman et al. (1992) found that naming a letter which had the same color as before was not faster than naming a different-colored letter. Similarly, Mitroff and Alvarez (2007) also showed that surface feature played no role in object persistence. In their study, the two objects were distinct on a combination of feature dimensions (e.g., color, shape), but a blank interval lasting 1,000 milliseconds was used to replace the motion period (i.e., no spatiotemporal continuity). They observed that object correspondence could not be established solely on the basis of surface feature. However, Moore et al. (2010) arrived at a different conclusion. The two objects in their research were also distinct in surface feature, but they moved in a circular path, so clear spatiotemporal continuity was available. In the last 30 milliseconds of the motion period, these two objects exchanged each other’s surface feature. If surface feature had some effect in object correspondence, Moore et al. argued the exchange of surface features would destroy object persistence and consequently reduce the OSPB. Results showed that the OSPB was not only reduced but also reversed (i.e., became negative). This suggests that the effect of surface feature can even override that of spatiotemporal continuity when these two sorts of information conflict with each other. Surface feature information can also be consulted when spatiotemporal information is ambiguous. For instance, in a series of experiments conducted by Hollingworth and Franconeri (2009), two distinct objects moved from left and right along a visible path to the center of the screen and then became occluded behind a black square. After the occluder was removed, the two objects became visible again, one above and the other below the fixation. Importantly, the final locations of the objects were randomly assigned and had nothing to do with their start locations. Despite a lack of clear spatiotemporal information, correspondence computation can rely solely on surface feature. Recently, a growing body of research suggests that surface feature indeed plays an important role in object correspondence (Gordon & Vollmer, 2010; Hein & Moore, 2012; Hollingworth et al., 2008; Kawachi et al., 2011; Makovski & Jiang, 2009a; Poth & Schneider, 2016).

Given the important role of surface feature in object correspondence, one would speculate that a feature change would disrupt object persistence. However, there coexist two lines of seemingly contradictory evidence concerning the influence of feature change on object persistence other than the object-reviewing paradigm. One line of evidence relates to the tunnel effect. In this paradigm, an object moves toward a tunnel, enters it, becomes occluded for a short interval, and then comes out of the tunnel and continues moving ahead. Sometimes its surface feature changes (e.g., from red to green) after reappearing from the tunnel. Nonetheless, participants usually treat the object as the original one which has changed its surface feature, but not as a new object, provided that the object reemerges from the tunnel at the right place and time (Flombaum & Scholl, 2006; Kawachi & Gyoba, 2006; Michotte, 1946/1963). The tunnel effect shows that the change of surface feature would not disrupt object continuity. In contrary, another line of evidence shows the opposite. Using a flash-lag paradigm, Moore and Enns (2004) found that a sudden size or color change to a moving object would disrupt the object continuity. When the object underwent a feature change, participants reported seeing two distinct objects. This suggests that the change of surface feature would break the object continuity and lead to the representation of a new object. The disruption of object continuity as a result of feature change has been observed by many studies (Demeyer et al., 2010; Kanai et al., 2009; Rauschenberger, 2003; Tas, Moore, et al., 2012).

Why do these two aspects of evidence mentioned above conflict with each other? One possible answer is that the contexts of feature change are different in these two paradigms. In the tunnel paradigm, the feature change happens behind the tunnel, and thus observers do not see how the object changes its features. It is only the final product of the feature change that is visible to participants. In the flash-lag paradigm, however, there is no occlusion when the change takes place, and observers can watch the object change its surface feature. We speculate that a transient change of surface feature would disrupt object continuity, while a nontransient change would not do so. In fact, in a later study conducted by Moore et al. (2007) which used the same flash-lag paradigm as Moore and Enns (2004), they found that the perception of a new disk caused by size change no longer emerges when the moving disk passed behind an occluder with a hole in the center. Although the hole is smaller than the disk and participants saw a size-changed disk through the hole, the occluder provided a scene-based reason to prevent the feature change from disrupting the object continuity. Also, transient and nontransient stimuli have different effects on various cognitive processes (e.g., Bahrami, 2003; Cole et al., 2011; Franconeri et al., 2005; Hollingworth et al., 2010).

Based on the literature reviewed earlier, this study adopted the object-reviewing paradigm to further investigate the role of surface feature in object correspondence. Importantly, we aimed to explore whether transient and nontransient feature changes have different effects on object correspondence. To this end, four occluders were introduced to manipulate the way of feature change of the objects (see Figure 1B). When the objects moved across the surface of the occluders, they would be never masked by the occluders. When the objects moved behind the occluders, they would be masked briefly by the occluders. In Experiment 1A, two different-colored disks which moved across the surface of the occluders would suddenly change their colors to another two new colors when they passed the midpoints of their moving paths. Experiment 1B was similar to Experiment 1A, but the disks moved behind the occluders, and thus the color change was invisible. In Experiments 2A and 2B, the two objects would change their features on a combination of three dimensions (color, shape, and topology), and the feature change was visible or invisible, respectively.

Figure 1.

(A) Icons presented in preview and test displays. (B) A preview display in Experiment 1A. Four black occluders (not drawn to scale) were presented and the objects moved in front of (Experiments 1A and 2A) or behind (Experiments 1B and 2B) the occluders.

Experiment 1A

Method

Participants

Participants in all four experiments reported in this study were university students and were paid for their participation. All participants gave written informed consent before the experiment, and the study was approved by the ethics committee at the Civil Aviation Flight University of China. All participants had normal or corrected-to-normal visual acuity and none was colorblind. Seventeen students (eight females) participated in Experiment 1A. They ranged in age from 19 to 26 years, with a mean age of 22.59 (standard deviation [SD] = 1.82).

Materials and Apparatus

Stimuli were presented on a 17-in LCD screen with a refresh rate of 60 Hz. The experiment was conducted using the PsychoPy software (Peirce, 2007). All stimuli were presented against a light gray background (66 cd/m², RGB 128, 128, 128). The objects were disks, colored in red (32 cd/m², RGB 255, 0, 0), green (80 cd/m², RGB 0, 230, 0), yellow (92 cd/m², RGB 242, 204, 10), blue (26 cd/m², RGB 0, 0, 255), purple (44 cd/m², RGB 217, 0, 255), or orange (70 cd/m², RGB 255, 128, 0). From a viewing distance of 57 cm, the disks subtended 1.5° in diameter. The initial locations of the two disks were 4.5° to the left and right of the screen center. They traveled smoothly at a speed of 7.1 deg/s along a circular path, either clockwise or counterclockwise, and finally stopped above and below the screen center. The final locations were also 4.5° away from the center. Six novel icons were created as preview and test stimuli (see Figure 1A). The icons were 1.1° and were colored in white. On each trial, two icons were randomly chosen (without replacement) to appear in the preview display. Four black occluders were presented throughout the experiment (see Figure 1B). Each occluder measured 2° wide × 4° tall and tilted 45° either to the left or right. The center of the occluder was located at the midpoint of the curve along which the disks traveled. The two disks always traveled in front of the occluders (i.e., across the surface of the occluders). Thus, when the disks underwent a color change, participants would always watch the progress of the transient color change.

Procedure

Participants were tested individually in a quiet, dimly lit room. Instructions were shown on the screen and explained by the experimenter. Participants were required to place the index fingers of both hands on the F and J keys of the computer keyboard. At the beginning of each trial, two different-colored disks were shown to the left and right of the screen center. After 500 milliseconds, two icons (see Figure 1A) were presented on the center of the disks for 1,200 milliseconds (preview display). Upon the disappearance of the graphs, the disks remained static for another 300 milliseconds and then began moving along a circular path for 1,000 milliseconds. The direction of movement was either clockwise or counterclockwise and was randomly assigned for each trial, with each direction on half of all trials. There were two types of trials. On the change trials, the colors of the two disks were changed to two new colors after the disks passed the midpoint of their journey. Note that the new colors were new for the entire display, not just the particular object (i.e., not just a switch of the colors of the two disks shown in the preview display). On the no-change trials, the colors of the disks remained the same throughout the entire path. The disks stopped at locations above or below the screen center, and two icons were shown on the disks (test display). On half of the trials, one of the icons was different from those viewed in the preview display, while the other was one of the icons shown before (no-match condition). On the other half trials, the two icons were both the same as those shown in the preview display (match condition). In addition, the match trials were further classified into two conditions: the icons appeared in the same disks as defined by spatiotemporal continuity (congruent-match condition) or they appeared in the opposite disks (incongruent-match condition). The test display remained on the screen until participants pressed a key. They were instructed to judge as quickly and accurately as possible whether the two icons were the same as those presented in the preview display. One group of participants was told to press the F key if the icons were the same as the original two and press the J key if one of the icons was different. For the other group of participants, the response mappings were reversed. The next trial started 1,000 milliseconds after the key press. Participants first completed 32 practice trials drawn randomly from the full design. Feedback (correct or incorrect) was given after each trial in the practice session. Then, they completed 320 experimental trials, divided by 8 blocks of 40 trials each. No feedback was given during the experimental session. Participants were asked to have a rest after finishing a block. The whole experiment lasted about 1 hour.

Design

Experiment 1A was a 2 (Feature Change: change, no-change) × 2 (match, no-match) × 2 (congruent, incongruent) repeated-measures design. Match or no-match means that the icons presented in the test display were the same as or different from those in the preview display. Congruent or incongruent was meaningful only for the match trials. It means that the icons in the test display were located in the same or different disks compared with those in the preview display. All three factors were randomly mixed within the experiment.

Results and Discussion

In all experiments reported in this study, analyses of interest were conducted over match trials. On no-match trials, a new icon was introduced in the test display, and this in general yielded slower responses compared with match trials. As no-match trials are less important for theoretical concern, we do not analyze them in the main text, although these data are presented in Table 1. For RT analyses, only trials with correct responses were included. In addition, RTs fell outside 3 SD of the cell mean were eliminated as outliers.

Table 1.

Mean Response Times (Milliseconds) and Accuracy (%) for Each Experiment.

Exp.	Feature no-change			Feature change
Exp.	No-match	Incongruent-match	Congruent-match	No-match	Incongruent-match	Congruent-match
1A	1,018 (91.80)	974 (96.32)	908 (97.94)	1,034 (92.74)	978 (96.03)	949 (97.21)
1B	1,028 (91.87)	960 (96.87)	905 (97.34)	1,059 (92.81)	989 (97.19)	932 (96.72)
2A	929 (91.68)	885 (95.67)	824 (97.83)	922 (89.11)	898 (94.67)	888 (96.67)
2B	967 (90.83)	899 (96.67)	849 (97.33)	972 (92.33)	944 (95.67)	901 (97.17)

Overall accuracy was high (95.34%) and did not differ between incongruent-match and congruent-match trials for either the no-change condition (96.32% vs. 97.94%), t(16) = 1.734, p = .102, Cohen’s d = .387, or the feature change condition (96.03% vs. 97.21%), t(16) =1.224, p = .239, Cohen’s d = .301. A repeated-measures analysis of variance (ANOVA) was conducted on the RTs with feature change (change, no-change) and congruency (congruent, incongruent) as within-subjects factors. The main effects were both significant for feature change, F(1, 16) = 6.856, p = .019, $η_{p}^{2}$ = 0.299, and for congruency, F(1, 16) = 34.225, p < .001, $η_{p}^{2}$ = 0.681. These two factors interacted significantly, F(1, 16) = 15.794, p = .001, $η_{p}^{2}$ = 0.481. The main effect of congruency showed that responses were generally faster when the graphs appeared in the same disks as before, no matter whether the colors of the disks changed or not. Note that the sameness was defined only by spatiotemporal continuity in the feature change condition whereas by both spatiotemporal and feature continuity in the no-change condition. The main effect of feature change may arise from the updating of old information or the creation of a new representation when the disks underwent sudden color changes. Responses were 23 milliseconds slower when the objects underwent a color change relative to when the objects did not change their colors.

Our focus was to compare the OSPBs under the feature change and the no-change conditions. The OSPB was computed by subtracting the mean RT for congruent-match trials from the mean RT for incongruent-match trials. As can be seen in Figure 2A, a significant OSPB was observed for the no-change condition (974 milliseconds vs. 908 milliseconds, for incongruent-match and congruent-match trials, respectively), t(16) = 7.368, p < .001, Cohen’s d = .459, and for the feature change condition (978 milliseconds vs. 949 milliseconds), t(16) = 2.994, p = .009, Cohen’s d = .203. The effect sizes of OSPBs were 66 milliseconds and 29 milliseconds for the no-change and the feature change conditions, respectively, with the latter significantly smaller than the former, t(16) = 3.974, p = .001, Cohen’s d = .733. Results suggest that although both OSPBs under feature change and no-change conditions were robust, the former was reliably smaller than the latter. The result shows the role of surface feature for object correspondence. When objects undergo a feature change, only one category of information (i.e., spatiotemporal information) can be resorted for the correspondence operation. Therefore, the benefits are not as large as when two categories of information can both be utilized. These results are in line with the research by Moore et al. (2010), indicating the role of surface features in object persistence. In addition, while Moore et al. found negative OSPBs when two objects exchanged their surface features, our study found smaller but positive OSPBs when objects changed their surface features. This may result from the differences between feature switch and feature change. It seems that a feature switch may have more impact than a feature change, thus producing OSPBs with different magnitudes.

Figure 2.

Mean RTs for Incongruent and Congruent Trials as a Function of Feature Change Condition in (A) Experiment 1A (No-Change vs. Transient Change) and (B) Experiment 1B (No-Change vs. Nontransient Change). Error bars represent the within-subject standard error of the mean (Morey, 2008). **p < .01. ***p < .001. RT = response time.

Experiment 1B

The results of Experiment 1A show that feature change would decrease the OSPB. However, the feature change was an abrupt transient that served as a low-level visual input. What would happen if feature change is rendered nontransient? Recall the tunnel effect that a feature-changed object after coming out of the tunnel can be treated as the same object which enters the tunnel. In such case, the nontransient feature change seems to induce the visual system to fill in the gap amodally (Burke, 1952; Michotte, 1946/1963). Following this consideration, we predict that the nontransient feature change would not reduce the OSPB. To make feature change nontransient, the disks in Experiment 1B would move behind the occluders, with different colors before and after passing these occluders.

Method

Participants

Sixteen new participants (seven females) completed Experiment 1B. They ranged in age from 19 to 26 years, with a mean age of 22.31 (SD = 1.96).

Materials, Apparatus, Procedure, and Design

These were identical to those in Experiment 1A except that the disks were moving behind the occluders. Specifically, they started to move from left and right along a circular path, each passed behind an occluder and became hidden briefly, and then reappeared from the occluder and continued moving. The moving period lasted for 1,000 milliseconds, as in Experiment 1A. On the no-change trials, the colors of the disks moving into and coming out of an occluder remained the same. On the change trials, the disks changed to two new colors that were not initially presented in the preview display after passing the occluders. However, what was different from Experiment 1A was that the change of color was invisible here, whereas the color change was clearly visible in Experiment 1A.

Results and Discussion

Overall accuracy was high (95.47%) and did not differ between incongruent-match and congruent-match trials for either the no-change condition (96.87% vs. 97.34%), t(15) = .469, p = .646, Cohen’s d = .202, or the feature change condition (97.19% vs. 96.72%), t(15) = .588, p = .566, Cohen’s d = .254. A repeated-measures ANOVA was conducted on the RTs with feature change (change, no-change) and congruency (congruent, incongruent) as within-subjects factors. The main effects were both significant for feature change, F(1, 15) = 8.358, p = .011, $η_{p}^{2}$ = 0.358, and for congruency, F(1, 15) = 29.299, p < .001, $η_{p}^{2}$ = 0.661. Responses were 27 milliseconds slower when the objects underwent a color change relative to when the objects did not change their colors. However, these two factors did not interact significantly, F(1, 15) = 0.023, p = .880, $η_{p}^{2}$ = 0.002. As shown in Figure 2B, a significant OSPB was observed for the no-change condition (960 milliseconds vs. 905 milliseconds, for incongruent-match and congruent-match trials, respectively), t(15) = 4.188, p < .001, Cohen’s d = .382, and for the feature change condition (989 milliseconds vs. 932 milliseconds), t(15) = 4.439, p < .001, Cohen’s d = .380. The effect sizes of OSPBs were 55 milliseconds and 57 milliseconds for the no-change and the feature change conditions, respectively.

We further combined the data obtained from Experiments 1A and 1B and conducted an ANOVA on the OSPBs, with feature change as the within-subjects factor and experiment as the between-subjects factor. Feature change approached significant, F(1, 31) = 3.662, p = .065, $η_{p}^{2}$ = 0.115. In addition, the interaction was significant, F(1, 31) = 4.760, p = .037, $η_{p}^{2}$ = 0.133. Simple effect analysis showed that when surface feature did not change, the OSPBs were the same no matter whether the objects moved in front of or behind the occluders, t(31) = .710, p = .483. When surface feature underwent a change, however, the OSPB was larger when the change was nontransient compared with when the change was transient, t(31) = 1.749, p = .045, Cohen’s d = .454.

These results with those from Experiment 1A demonstrate different effects of transient and nontransient feature change on object correspondence. Nontransient feature change behind occluders would not break object continuity by means of amodal integration (Burke, 1952; Michotte, 1946/1963), whereas an abrupt transient feature change will indeed disrupt object continuity to some extent. In fact, some studies have found that transient changes of surface feature would destroy object continuity and cause the establishment of the representation of a new object (Moore & Enns, 2004; Moore et al., 2007). If transient feature changes induce the representation of a new object, however, there remains a question for the present results. That is, the perception of new disks caused by transient feature change should lead to the creation of new object files, which should contain no information about the graphs, because the graphs are only related to the original disks. Consequently, the visual system should treat the congruent and incongruent conditions in an unbiased manner because the congruent and incongruent labels only have a sense with respect to the original disks but have nothing to do with the feature-changed new disks. If the congruent trials were not in an advantageous status, there should be no OSPB in the feature change condition. This was quite different from the results of Experiment 1A. The OSPB in Experiment 1A was reliable in the feature change condition. How could we explain this? A possible explanation is that the feature change may be not drastic enough in Experiment 1A: Only one feature dimension (i.e., color) was changed. The shape, size, and many other attributes of the disks were kept the same. Thus, the new objects were somewhat similar to the original ones in a sense. In addition, the new objects shared the same spatiotemporal property as the old ones because the new disks emerged at the exact position where the old disks disappeared and continued moving along their paths. In a word, the new disks were somewhat similar to the previous ones with respect to surface feature and spatiotemporal properties. These similarities may induce old and new objects to share some information (e.g., the graphs and their locations). All these considerations make it possible to find a reliable OSPB even when objects change their surface features. Suppose if the objects undergo more drastic feature changes, they would be much dissimilar to the old ones. One way of introducing more drastic feature change is to have objects change their properties on a combination of feature dimensions. It is potential for feature changes on a combination of dimensions to further reduce similarities between new and old objects, and as a result decrease the OSPB more, or even eliminate the benefit (Feldman & Tremoulet, 2006). We test this hypothesis in Experiment 2A.

Experiment 2A

In this experiment, the objects would undergo feature changes on a combination of three dimensions. Specifically, the objects had different colors and shapes, and their topological properties were also different. That is, they were solid or hollow. When the objects changed their features, their colors, shapes, and the topological properties all changed to a new attribute. If multidimensional feature changes had more effect in generating the representation of a new object, the OSPB would decreased more, or even totally disappeared.

Method

Participants

Fifteen new students (six females) participated in Experiment 2A. They ranged in age from 19 to 27 years, with a mean age of 22.20 (SD = 2.19).

Materials, Apparatus, Procedure, and Design

These were the same as those in Experiment 1A except mentioned later. The shapes of the objects were disks, squares, pentagons, or hexagons (see Figure 3). Each shape measured approximately 1.6° × 1.6°. The objects had different topological features. They were solid or hollow. When they were solid, the whole object was filled with a color. When they were hollow, only the contour (0.2°) of the object was filled with a color, while the inside part was transparent and had the same color as the background. Thus, the object looked as if having a big hole in it when it was hollow. For instance, the hollow disk was actually a ring. At the beginning of each trial, two different shapes with different colors and same topological properties were presented. On half of all trials, the two objects were both solid, and on the other half trials, they were both hollow. The topological properties of the objects were randomly assigned on each trial. On the no-change trials, the shape, color, and the topological property of each object remained the same during the motion. On the feature change trials, these two objects changed to another two shapes in another two colors and also changed the topological properties after passing the midpoint of the moving path. That is, if the original objects were both solid, they both transformed into hollow shapes, or vice versa. The reason why we did not use a hollow shape and a solid shape as the original objects and then change the topological properties after the midpoint is that such a manipulation may cause the perception of a switch of two objects when they exchange each other’s topological properties. This perception may not be interpreted as the emergence of two new objects but mere a switch of locations of two old objects. The exchange of properties has the potential to generate a negative OSPB (e.g., Moore et al., 2010).

Results and Discussion

Overall accuracy was high (94.27%) and differed between incongruent-match and congruent-match trials for both the no-change condition (95.67% vs. 97.83%), t(14) = 2.982, p = .010, Cohen’s d = .578, and the feature change condition (94.67% vs. 96.67%), t(14) = 2.347, p = .034, Cohen’s d = .612, suggesting preview benefits on accuracy. A repeated-measures ANOVA was conducted on the RTs with feature change (change and no-change) and congruency (congruent and incongruent) as within-subjects factors. The main effects were both significant for feature change, F(1, 14) = 21.045, p < .001, $η_{p}^{2}$ = 0.601, and for congruency, F(1, 14) = 16.540, p = .001, $η_{p}^{2}$ = 0.542. Responses were 38 milliseconds slower when the objects underwent a feature change relative to when the objects did not change their properties. In addition, these two factors interacted significantly, F(1, 14) = 16.933, p = .001, $η_{p}^{2}$ = 0.547. As illustrated in Figure 4A, a significant OSPB of 61 milliseconds was observed for the no-change condition (885 milliseconds vs. 824 milliseconds, for incongruent-match and congruent-match trials, respectively), t(14) = 4.694, p < .001, Cohen’s d = .423. For the feature change condition, however, the RT difference between incongruent trials (898 milliseconds) and congruent trials (888 milliseconds) was only 10 milliseconds and was not significant, t(14) = 1.265, p = .226, Cohen’s d = .067.

Figure 3.

Examples of Objects Used in Experiments 2A and 2B. The first row was solid objects, and the second row was hollow objects. The hollow part of the object had the same color as the background. The color of the objects was red, yellow, green, blue, purple, or orange.

Figure 4.

Mean RTs for Incongruent and Congruent Trials as a Function of Feature Change Condition in (A) Experiment 2A (No-Change vs. Transient Change) and (B) Experiment 2B (No-Change vs. Nontransient Change). Error bars represent the within-subject standard error of the mean (Morey, 2008). ***p < .001. RT = response time.

The disappearance of the OSPB in the feature change condition indicates that transient feature changes from multidimension could create the representation of a new object. Although the new object still shares some spatiotemporal property of the old one, this information can be overridden by the drastic feature change. In Experiment 2B, we explored whether nontransient multidimensional feature change could also disrupt the OSPB.

Experiment 2B

In Experiment 1, we found that color change would disrupt object continuity and reduce the OSPB to some extent, and the effect of color change vanished when occluders were introduced to make the color change nontransient. In Experiment 2A, multidimensional feature changes eliminated the OSPB, showing a greater effect than single-dimensional feature change. As multidimensional feature changes are more effective in disrupting old objects and establishing new ones, it is interesting to ask whether the effect of nontransient feature change would still survive when the change involves a combination of feature dimensions. Experiment 2B was designed to explore this issue.

Method

Participants

Fifteen new students (nine females) participated in Experiment 2B. They ranged in age from 18 to 26 years, with a mean age of 21.87 (SD = 2.03).

Materials, Apparatus, Procedure, and Design

These were the same as those in Experiment 2A except that the objects here were moving behind the occluders. As in Experiment 1B, when the objects passed the occluders, they would be hidden so that the multidimensional feature change was nontransient.

Results and Discussion

Overall accuracy was high (95.01%) and did not differ between incongruent-match and congruent-match trials for either the no-change condition (96.67% vs. 97.33%), t(14) = 1.075, p = .301, Cohen’s d = .311, or the feature change condition (95.67% vs. 97.17%), t(14) = 1.790, p = .095, Cohen’s d = .356. A repeated-measures ANOVA was conducted on the RTs with feature change (change, no-change) and congruency (congruent, incongruent) as within-subjects factors. The main effects were both significant for feature change, F(1, 14) = 10.326, p = .006, $η_{p}^{2}$ = 0.424, and for congruency, F(1, 14) = 41.110, p < .001, $η_{p}^{2}$ = 0.745. Responses were 48 milliseconds slower when the objects underwent a feature change relative to when the objects did not change their properties. These two factors did not interact significantly, F(1, 14) = .375, p = .550, $η_{p}^{2}$ = 0.026. As depicted in Figure 4B, a significant OSPB was observed for the no-change condition (899 milliseconds vs. 849 milliseconds, for incongruent-match and congruent-match trials, respectively), t(14) = 5.550, p < .001, Cohen’s d = .338, and for the feature change condition (944 milliseconds vs. 901 milliseconds), t(14) = 4.427, p < .001, Cohen’s d = .273. The effect sizes of OSPBs were 50 milliseconds and 43 milliseconds for the no-change and the feature change conditions, respectively.

We further combined the data obtained from Experiments 2A and 2B and conducted an ANOVA on the OSPBs, with feature change as the within-subjects factor and experiment as the between-subjects factor. In addition to a main effect of feature change, F(1, 28) = 11.454, p = .002, $η_{p}^{2}$ = 0.291, the interaction of these two factors was also significant, F(1, 28) = 6.420, p = .017, $η_{p}^{2}$ = 0.187. Simple effect analysis showed that when surface feature did not change, the OSPBs were the same no matter whether the objects moved in front of or behind the occluders, t(28) = .638, p = .529, Cohen’s d = .132. When surface feature underwent a change, however, whether the change was transient or nontransient had different effects, t(28) = 2.692, p = .012, Cohen’s d = .552. These results demonstrate that the occluders are a powerful tool which can tolerate huge feature changes from a combination of feature dimensions.

General Discussion

In a series of experiments in which four occluders were introduced into the object-reviewing paradigm, we explored the role of surface feature in object correspondence. Our results provide further evidence to the claim that object correspondence can be established not only by spatiotemporal information but also on the basis of surface feature. The role of surface feature is reflected in the fact that feature change will more or less disrupt object continuity and correspondence. More importantly, the effect of surface feature depends on whether feature change is transient or nontransient. Specifically, transient feature change will disrupt object continuity, whereas nontransient feature change will not do so.

Although spatiotemporal information is claimed to be dominant in determining object correspondence in various domains such as object-file framework, multiple object tracking, apparent motion, and tunnel effect, surface feature has been proven to also guide the operation of object correspondence in many studies (Feldman & Tremoulet, 2006; Hein & Moore, 2012; Hollingworth & Franconeri, 2009; Moore et al., 2010; Richard et al., 2008; Tas, Dodd, et al., 2012). Makovski and Jiang (2009a, 2009b) explored the effect of feature distinctness on multiple object tracking and found that the tracking performance improved when objects had distinct features on a surface feature dimension compared with when they were all identical. In addition, object correspondence across saccadic eye movements could also rely on feature information (Poth et al., 2015; Poth & Schneider, 2016). It often happens that our eyes fail to land on the correct target object and then we have to redirect our gaze to the target. Surface feature has been demonstrated to affect the correction process (Hollingworth et al., 2008; Richard et al., 2008). Apparent motion has long been regarded as a phenomenon which can only be perceived under appropriate spatial and time constraints and surface feature plays little role in motion correspondence (but see, e.g., Green, 1989). Recently, however, evidence in favor of surface feature correspondence in apparent motion has been found. For instance, whether Ternus motion (Ternus, 1926)—a classic apparent motion—is perceived as group motion or element motion could depend on how specific objects are corresponded on the basis of their surface features (Hein & Cavanagh, 2012; Hein & Moore, 2012, 2014; Hein & Schütz, 2019; Hsu et al., 2015; Petersik & Rice, 2008; Silva & Petersik, 2015; Stepper et al., 2019).

The present experiments provide additional evidence for the role of transient versus nontransient surface feature changes in object correspondence. When the change is transient, the low-level visual input acts as a strong signal for the visual system to update information or even generate a new object representation. The elimination of the OSPB indicates that no correspondence is established between the original and the new objects. If two objects no longer persist to be the same objects appearing a moment ago, there is no need to correspond any one of the later objects to be any one of the earlier objects. In fact, any one of them now is not any one of the objects viewed before. Therefore, the transient change tends to discard the representation of the old object and establish a new representation, particularly when the object suffers a drastic feature change.

In contrast, when the progress of feature change is invisible and the change can only be inferred, the higher level representation of feature change will not be interpreted as a break of object continuity. The insensitivity to nontransient feature change may arise from amodal integration that can compensate for the invisible period and tolerate even enormous feature change. Different influences of transient and nontransient changes on performance are reported in other research. Cole et al. (2011) found that while transient luminance change of a square attracts attention, nontransient luminance change does not capture attention. Hollingworth et al. (2010) reported that a new object would not attract attention and was not easier to be located than old objects when its onset was masked briefly. Bahrami (2003) observed that color changes to multiple moving objects are difficult to notice when they were hidden by mud splashes. Using a different paradigm, results of the present experiments add to this line of evidence.

Surface features of objects are constantly changing over time. Their features are registered in some kind of episodic representation, such as object files. When objects change their features, new information is extracted and updated to replace the old information. However, the visual system has its limit of tolerance for these feature changes. Large and transient feature changes to objects may not lead to the updating of information but to the creation of new entities (Moore & Enns, 2004; Moore et al., 2007; Rauschenberger, 2003). In the present experiments, two kinds of feature change were employed: surface features changed on only one feature dimension (i.e., color) in Experiment 1 and changed on three dimensions (i.e., color, shape, topology) in Experiment 2. The first kind of change only reduced but not eliminated the OSPB. It is not clear from this result whether color change induced an update of information or a creation of the representation of a new object. It seems that both accounts can explain the result. On the one hand, if the color-changed object was interpreted as the same object as the one before change, the correspondence could be established by spatiotemporal information. It thus resulted in a smaller OSPB compared with color-unchanged condition in which both spatiotemporal and surface feature information could be consulted. On the other hand, the visual system could also interpret the color-changed object as a new entity, and a reliable but smaller OSPB could still be obtained. The OSPB derives from the assumption that the new object could share the spatiotemporal information of the old one, and this spatiotemporal information could be consulted to establish correspondence. At this time we cannot distinguish one account from the other, and more studies are desirable to disentangle the problem. Nonetheless, whichever is true does not hinder us from claiming that feature change can indeed affect object persistence and correspondence.

In summary, the present work demonstrates the role of surface feature in object correspondence by showing that objects undergoing transient feature changes incline to disrupt object continuity. As a result, they tend to be perceived as new entities, particularly when the feature change is drastic. However, object persistence will not be disrupted by nontransient feature changes, no matter whether the change is moderate or drastic.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by Foundation of the Institute of Aviation Human Factors and Ergonomics of Civil Aviation Flight University of China (No. JG2019-13).

ORCID iD

Hao Jiang

References

Bahrami

(2003). Object property encoding and change blindness in multiple object tracking. Visual Cognition, 10, 949–963.

Burke

(1952). On the tunnel effect. Quarterly Journal of Experimental Psychology, 4, 121–138.

Cole

G. G.

Kuhn

Skarratt

P. A.

(2011). Non-transient luminance changes do not capture attention. Attention, Perception, & Psychophysics, 73, 1407–1421.

Demeyer

De Graef

Wagemans

Verfaillie

(2010). Object form discontinuity facilitates displacement discrimination across saccades. Journal of Vision, 10, 17.

Feldman

Tremoulet

P. D.

(2006). Individuation of visual objects over time. Cognition, 99, 131–165.

Flombaum

J. I.

Scholl

B. J.

(2006). A temporal same-object advantage in the tunnel effect: Facilitated change detection for persisting objects. Journal of Experimental Psychology: Human Perception and Performance, 32, 840–853.

Flombaum

J. I.

Scholl

B. J.

Santos

L. R.

(2009). Spatiotemporal priority as a fundamental principle of object persistence. In Hood

Santos

(Eds.), The origins of object knowledge (pp. 135–164). Oxford University Press.

Franconeri

S. L.

Hollingworth

Simons

D. J.

(2005). Do new objects capture attention? Psychological Science, 16, 275–281.

Gao

Scholl

B. J.

(2010). Are objects required for object-files? Roles of segmentation and spatiotemporal continuity in computing object persistence. Visual Cognition, 18, 82–109.

10.

Gordon

R. D.

Vollmer

S. D.

(2010). Episodic representation of diagnostic and nondiagnostic object color. Visual Cognition, 18, 728–750.

11.

Green

(1989). Color correspondence in apparent motion. Perception & Psychophysics, 45, 15–20.

12.

Hein

Cavanagh

(2012). Motion correspondence in the Ternus display shows feature bias in spatiotopic coordinates. Journal of Vision, 12, 1–14.

13.

Hein

Moore

C. M.

(2012). Spatio-temporal priority revisited: The role of feature identity and similarity for object correspondence in apparent motion. Journal of Experimental Psychology: Human Perception and Performance, 38, 975–988.

14.

Hein

Moore

C. M.

(2014). Evidence for scene-based motion correspondence. Attention, Perception, & Psychophysics, 76, 793–804.

15.

Hein

Schütz

A. C.

(2019). Competition between color and luminance in motion correspondence. Attention, Perception, & Psychophysics, 81, 310–322.

16.

Hollingworth

Franconeri

S. L.

(2009). Object correspondence across brief occlusion is established on the basis of both spatiotemporal and surface feature cues. Cognition, 113, 150–166.

17.

Hollingworth

Richard

A. M.

Luck

S. J.

(2008). Understanding the function of visual short-term memory: Transsaccadic memory, object correspondence, and gaze correction. Journal of Experimental Psychology: General, 137, 163–181.

18.

Hollingworth

Simons

D. J.

Franconeri

S. L.

(2010). New objects do not capture attention without a sensory transient. Attention, Perception, & Psychophysics, 72, 1298–1310.

19.

Hsu

P., T.

Taylor

J. E.

Pratt

(2015). Frogs jump forward: Semantic knowledge influences the perception of element motion in the Ternus display. Perception, 44, 779–789.

20.

Kahneman

Treisman

(1984). Changing views of attention and automaticity. In Parasuraman

Davies

D. R.

(Eds.), Varieties of attention (pp. 29–61). Academic Press.

21.

Kahneman

Treisman

Gibbs

B. J.

(1992). The reviewing of object files: Object-specific integration of information. Cognitive Psychology, 24, 175–219.

22.

Kanai

Carlson

T. A.

Verstraten

F. A.

Walsh

(2009). Perceived timing of new objects and feature changes. Journal of Vision, 9, 5.

23.

Kawachi

Gyoba

(2006). A new response-time measure of object persistence in the tunnel effect. Acta Psychologica, 123, 73–90.

24.

Kawachi

Kawabe

Gyoba

(2011). Stream/bounce event perception reveals a temporal limit of motion correspondence based on surface feature over space and time. i-Perception, 2, 428–439.

25.

Makovski

Jiang

Y. V.

(2009a). Feature binding in attentive tracking of distinct objects. Visual Cognition, 17, 180–194.

26.

Makovski

Jiang

Y. V.

(2009b). The role of visual working memory in attentive tracking of unique objects. Journal of Experimental Psychology: Human Perception and Performance, 35, 1687–1697.

27.

Michotte

(1963). The perception of causality (T. R. Miles & E. Miles, Trans.). Methuen. (Original work published 1946)

28.

Mitroff

S. R.

Alvarez

G. A.

(2007). Space and time, not surface features, guide object persistence. Psychonomic Bulletin & Review, 14, 1199–1204.

29.

Moore

C. M.

Enns

J. T.

(2004). Object updating and the flash-lag effect. Psychological Science, 15, 866–871.

30.

Moore

C. M.

Mordkoff

J. T.

Enns

J. T.

(2007). The path of least persistence: Object status mediates visual updating. Vision Research, 47, 1624–1630.

31.

Moore

C. M.

Stephens

Hein

(2010). Features, as well as space and time, guide object persistence. Psychonomic Bulletin & Review, 17, 731–736.

32.

Morey

R. D.

(2008). Confidence intervals from normalized data: A correction to Cousineau (2005). Tutorials in Quantitative Methods for Psychology, 4, 61–64.

33.

Peirce

J. W.

(2007). PsychoPy—Psychophysics software in Python. Journal of Neuroscience Methods, 162, 8–13.

34.

Petersik

J. T.

Rice

C. M.

(2008). Spatial correspondence and relation correspondence: Grouping factors that influence perception of the Ternus display. Perception, 37, 725–739.

35.

Poth

C. H.

Herwig

Schneider

W. X.

(2015). Breaking object correspondence across saccadic eye movements deteriorates object recognition. Frontiers in Systems Neuroscience, 9, 176.

36.

Poth

C. H.

Schneider

W. X.

(2016). Breaking object correspondence across saccades impairs object recognition: The role of color and luminance. Journal of Vision, 16, 1.

37.

Pylyshyn

(1989). The role of location indexes in spatial perception: A sketch of the FINST spatial-index model. Cognition, 32, 65–97.

38.

Rauschenberger

(2003). When something old becomes something new: Spatiotemporal object continuity and attentional capture. Journal of Experimental Psychology: Human Perception and Performance, 29, 600–615.

39.

Richard

A. M.

Luck

S. J.

Hollingworth

(2008). Establishing object correspondence across eye movements: Flexible use of spatiotemporal and surface feature information. Cognition, 109, 66–88.

40.

Scholl

B. J.

(2007). Object persistence in philosophy and psychology. Mind & Language, 22, 563–591.

41.

Silva

M. R.

Petersik

J. T.

(2015). Flicker-defined forms in the Ternus display. Perception, 44, 315–327.

42.

Stepper

M. Y.

Rolke

Hein

(2019). How voluntary spatial attention influences feature biases in object correspondence. Attention, Perception, & Psychophysics. Advance online publication. https://doi.org/10.3758/s13414-019-01801-9

43.

Tas

A. C.

Dodd

M. D.

Hollingworth

(2012). The role of surface feature continuity in object-based inhibition of return. Visual Cognition, 20, 29–47.

44.

Tas

A. C.

Moore

C. M.

Hollingworth

(2012). An object-mediated updating account of insensitivity to transsaccadic change. Journal of Vision, 12, 18.

45.

Ternus

(1926). Experimentelle Untersuchungen über phänomenale Identität. Psychologische Forschung, 7, 81–136. (Reprinted from W. D. Ellis (Ed.), A sourcebook of gestalt psychology, 1950. Humanities Press.)