Does feature integration affect resolution of multiple simultaneous forms of ambiguity?

Abstract

Ambiguity resolution, perceptual grouping, and feature integration all occur seamlessly and subconsciously. When multiple regions of an image share ambiguous features, perceptual grouping can yield an integrated object percept rather than one of multiple objects, each with its individual features. Here, perceptual resolution and grouping of chromatically rivalrous Necker cubes were investigated in three experiments to determine the principles that underlie these coherent percepts. The first experiment showed perceptual grouping beyond independent resolution of each cube’s color and orientation, but the second experiment did not show grouping greater than expected from separate color- and orientation-grouping processes. The third experiment found no reliable difference in grouping when two features (color and orientation) were part of the same object versus when they were distributed across separate objects. These findings fail to support a role for feature conjunctions in grouping objects with multiple ambiguous features.

1. INTRODUCTION

In Max Wertheimer’s pioneering monograph on perceptual organization, he reported standing at the window and seeing “a house, trees, sky” rather than simply the “brightnesses and nuances of color” transmitted by his eyes [1]. This perceptual grouping of parts of the visual field into collections of objects is one of the constructive processes of unconscious inference, a term coined by Helmholtz to describe the near-automatic reconstruction of the properties of objects from incomplete sensory neural representations [2].

Perceptual grouping is not, however, the only constructive process in vision. The visual system simultaneously performs other, equally important processes such as feature integration and rivalry resolution, which integrate seamlessly with perceptual grouping. As an illustration of rivalry resolution and feature integration—and how automatic they are—suppose that you are the one looking out the window. It is nearly dusk, starting to get dark outside, but the room is well-lit. You see a blue house, dark green-leaved trees, and the purple-to-orange fade of the horizon at sunset. After contemplating the organization of the various “brightnesses and nuances of color” into these discrete units (perceptual grouping), you pause to think about how the colors and shapes of these units are tied together in your perception, so that you effortlessly see a blue house and an orange sky, even though the sky is usually blue and a house could perfectly well be orange. Shifting your attention, you suddenly notice your own reflection in the window. Your reflection was there all the time, of course, just as visible in the well-lit room as the darkening scene outside. Yet you previously took no account of it, and it did not interfere with your perception of the house, trees, and sky. Along with grouping the visual field into objects, your visual system also integrated or bound the features of those objects together and resolved the monocular rivalry [3] caused by the superimposed images of the outdoors and of your reflection.

The experiments here investigate perceptual grouping, feature integration, and rivalry resolution together to better determine how they contribute to our percepts. These experiments build upon a long line of studies of these three processes, many of which have given basic insights into the nature of visual perception and consciousness [1,4-6].

Perceptual rivalry demonstrates that percepts can change without a corresponding change in a stimulus. Different types of perceptual rivalry can be induced by various types of stimuli. Dichoptic stimuli are conflicting images presented to the same location for each eye, causing percepts that switch back and forth between the stimulus presented to each eye [7]. In the laboratory, dichoptic stimuli are often viewed with a haploscope, a system of mirrors that presents one half of the display to each eye. Dichoptic stimulus presentation methods include standard binocular rivalry (SBR) in which rivaling images are presented steadily to each eye, and interocular-switch rivalry (ISR) in which rivaling images are swapped between eyes several times per second. Ambiguous figures, on the other hand, induce rivalry between alternate interpretations of an unchanging stimulus such as the Necker cube, which may be perceived to face front-upward or front-downward [8]. See Fig. 1(A) for illustrations of SBR and ISR, a haploscope, and Necker cube orientation rivalry.

Fig. 1.

Forms of rivalry and grouping. (A) Left: stimulus for standard binocular rivalry (SBR), as presented through the haploscope. Center top: stimulus for chromatic interocular-switch rivalry (CISR), in which equiluminant chromatic stimuli are swapped between eyes at 3.75 Hz. Center bottom: percepts that switch back and forth over time while viewing SBR or CISR. Right: Necker cube, percepts of which switch back and forth over time. OS = “oculus sinister” (presented to left eye); OD = “oculus dexter” (presented to right eye). (B) Perceptual grouping cues. Top row: no grouping cues. Second row: proximity cue. Third row: color cue.

Perceptual grouping phenomena show that vision is a constructive process that makes use of the retinal image [1,2] [Fig. 1(B)]. Finally, the finding that the various features of objects can be registered—but not correctly bound together in their respective objects—shows that, while features are “registered early, automatically, and in parallel,” identification of objects and binding of their features occurs later in visual processing, at a stage that “requires focused attention” [9]. Classic feature-integration theory [4] and more recent modifications [10] highlight the importance of attention in visual perception.

Since the discoveries of perceptual rivalry, perceptual grouping, and feature integration, studies have attempted to determine how these processes fit together. Perceptual grouping of figures sharing a common rivalrous feature has been shown in experiments using numerous forms of perceptual rivalry, including SBR [11,12], chromatic interocular-switch rivalry (CISR) [13], Necker cube orientation rivalry [14,15], and other types of form rivalry [16,17]. Interocular grouping when viewing rivalrous stimuli in a patchwork configuration was an important theoretical advance. In patchwork configurations, each eye views halves of two coherent patterns [11,12]. A theory of binocular rivalry based strictly upon competition between neurons representing each eye’s entire view would predict that observers would never perceive all elements in the whole pattern as identical when viewing rivalrous stimuli in patchwork configurations, but in these experiments observers frequently saw them as identical [11-13], ruling out this model of strict eye-rivalry.

The neural processes of binocular rivalry resolution remain contentious, with several models advanced in the literature. One holds that it is mediated solely by local competition between monocular neurons in distinct retinotopic zones of rivalry, which can laterally excite and inhibit one another [18]. A competing model proposes that interocular grouping requires competition between binocular neurons sensitive to feature coherence [12]; this is also supported by resolution [12] and grouping [13] of stimuli in ISR and CISR. A hybrid multistage model has been proposed, which incorporates both lateral excitation/inhibition between monocular neurons and feedback from binocular neurons onto monocular neurons to explain rivalry resolution and interocular grouping [19].

Behavioral and neuroimaging studies have shown that resolution of different types of visual rivalry involve activity at different levels of the visual system [13,20,21]. However, visually similar features can be grouped across different types of rivalry. In an experiment in which observers viewed a disc of orthogonal gratings, displayed in one-third SBR, one-third achromatic ISR, and one-third monocular rivalry, the observers perceived the disc to be a coherent uniform pattern, oriented either entirely at 45° or 135°, more often than chance [22]. This has been taken by some as evidence that SBR, ISR, and monocular rivalry have common high-level, interpretive processing [11,12,23] and by others as evidence of a high-level process that influences all forms of multistable perception by incorporating sensory and nonsensory information [24]. A high-level process like this could act upon integrated perceptual representations of objects rather than on low-level representations of individual features.

Perceptual grouping has also been studied in conjunction with feature integration. One experiment using unambiguously moving red and green dots showed feature misbinding of color and motion direction in the periphery when the two features were bound oppositely in the periphery and center [5], and a subsequent experiment indicated that this feature misbinding was mediated by a higher-order chromatic representation rather than by a low-level cone-opponent chromatic mechanism [25]. Perceptual grouping has also been found for objects with multiple ambiguous features: when observers viewed an array of gratings with binocularly rivalrous color, grating orientation, and grating spatial frequency, which all covaried in unison, a grouped percept was seen more often than when gratings had only one ambiguous feature [26]. This could be taken to indicate a role of feature integration in grouping and resolution of rivalry, but there was an important confound: In the experimental condition with multiple ambiguous features, the features were physically bound together in the stimuli. That is, though the stimuli were displayed in the patchwork configuration, each grating had one of two feature combinations (for example, RED-AND-GRAY/THIN STRIPES/135° versus GREEN-AND-GRAY/THICK STRIPES/45°). This binding of features in the physical stimuli meant that local competition in distinct zones of rivalry [18] could have been responsible for interocular feature grouping, without a contribution from feature integration mechanisms.

Do multiple ambiguous features group together as an integrated set, or is each feature grouped independently? That is, are ambiguous objects grouped as objects with ambiguous features bound together, or are ambiguous features grouped irrespective of other features with which they appear? The experiments here test these questions in a novel way, using paired stimuli with two ambiguous features that were not physically bound in the stimuli: rivalrous Necker cube orientation and rivalrous color. These rivalrous features are visually dissimilar, so grouping across rivalrous features based on visual similarity is not expected, unlike in a previous experiment that tested grouping across monocular, binocular, and stimulus rivalry [22]. Additionally, while resolution of binocularly rivalrous color is linked to representations of different stimuli in each eye, resolution of Necker cube orientation ambiguity is linked to representations of two different interpretations of the same stimulus and so has no ties to differences in the physical stimuli. The two feature percepts thus cannot be bound together in the physical stimuli, which means that grouping by both features in this experiment—if greater than expected from chance overlap of independent grouping processes—could not be due to only monocular competition within distinct local zones of competition. This instead would implicate feature integration and/or a higher-level process governing multiple forms of rivalry. See Fig. 2 for illustration of the chromatically rivalrous Necker cubes used in this study and ways in which they may be perceptually grouped.

Fig. 2.

Chromatically rivalrous Necker cubes: (A) stimulus, and (B–D) percepts. (A) Example of rivalrous Necker cube stimuli. (B) Representation of percept with grouped color and orientation. (C) Representation of percept with grouped color but not orientation. (D) Representation of percept with grouped orientation, but not color.

The first experiment here tests whether two separate Necker cubes with chromatic rivalry are perceptually grouped. The second experiment tests whether perceptual grouping processes for rivalrous chromaticity and orientation are independent of one another. The third experiment tests a similar hypothesis with a different technique, i.e., using rivalrous chromaticity and orientation either distributed across separate objects or integrated within them. This experiment complements results from the second experiment. The first two experiments use stimuli presented in conventional and patchwork stimulus configurations, and in SBR and CISR rivalry modes, to determine whether grouping depends on eye-of-origin information and the neural level of the rivalrous representations.

2. EXPERIMENT 1: DO SEPARATE STIMULI, EACH WITH TWO AMBIGUOUS FEATURES, GROUP MORE OFTEN THAN CHANCE?

A. Apparatus and Stimuli

Stimuli were displayed on a CRT monitor (NEC MultiSync FP2141SB) driven by an Apple iMac computer. Stimuli were viewed through a haploscope [Fig. 1(A)] with a path length of 115 cm from the viewer’s eyes to the screen. A chin rest kept the observer’s head position stable, and the pair of mirrors closest to the observer could be adjusted forward and backward to aid fusion. Stimuli were bounded by a thin white rectangular box with Nonius lines. The top and left Nonius lines were displayed to the left eye and the bottom and right lines to the right. Fusion was achieved when horizontal and vertical Nonius lines were aligned. Stimuli were Necker cubes, which were displayed in orthographic projection and spanned 1.5 deg of visual angle in width and height. The lines defining a cube were approximately 12 arcmin wide and appeared either red or green. Stimulus chromaticity was defined in the MacLeod–Boynton chromaticity space, [L/(L + M), S/(L + M)] [27], using [0.71, 0.3] for “red” and [0.62, 0.3] for “green.” The MacLeod–Boynton coordinates for the background were [0.665, 1.0], metameric to equal-energy spectrum “white.” Cube lines’ luminance was set to perceptually equate to “green” at 5 cd/m², and the background’s luminance was 3.4 cd/m². Cube centroids were positioned 1.5 deg above or below the center of the fixation cross.

There were two different session types, named “view-one” and “view-two, report-two” (see “Procedure” for further details on the session types). In “view-one” sessions, a cube was presented in one of two rivalry conditions: SBR or CISR. For each rivalry type, there were trials with the cube displayed above fixation (“cube-above-fixation”) or below fixation (“cube-below-fixation”). These four configurations were each counterbalanced by chromaticity, giving a total of eight stimulus configurations in “view-one” sessions.

In “view-two, report-two” sessions, the paired cubes were presented in one of four rivalry conditions: SBR conventional, SBR patchwork, CISR conventional, or CISR patchwork. These four rivalry conditions were each counterbalanced by chromaticity, giving a total of eight stimulus configurations in “view-two, report-two” sessions.

B. Procedure

Observers completed both session types during each day of testing with their order randomized. Over four days of testing, each observer completed four repetitions of each session type. In all sessions, each trial was 70 s long. The first 10 s of each trial were discarded to avoid onset bias and to ensure that the initial stimulus in each eye did not asymmetrically adapt the two eyes [28,29]. In “view-one” sessions, observers could hold one of four buttons during each trial to report conjunctions of the perceived color and orientation of the cube (“color-and-orientation conjunction”). The four percepts that observers reported were (1) cube RED and FRONT-LEFT, (2) cube RED and FRONT-RIGHT, (3) cube GREEN and FRONT-LEFT and (4) cube GREEN and FRONT-RIGHT. When either or both features were unstable or not as stated above, no button was held. Color-and-orientation conjunction times for each of the four percepts were measured in each trial and expressed as proportions of the 60 s reporting period (“color-and-orientation conjunction proportions”).

In “view-two, report-two” sessions, observers held one of four buttons in each trial to report when both cubes appeared to be both the same color and the same orientation (“color-and-orientation identity”). The four percepts that observers reported were (1) both cubes RED and FRONT-LEFT, (2) both cubes RED and FRONT-RIGHT, (3) both cubes GREEN and FRONT-LEFT, and (4) both cubes GREEN and FRONT-RIGHT. When either or both features were mixed or not stable, no button was held. Color-and-orientation identity times for each of the four percepts were measured in each trial and expressed as proportions of the 60 s reporting period (“color-and-orientation identity proportions”).

Each session was comprised of two blocks of trials, with a block consisting of one repetition of each stimulus configuration for that session type. Both sessions had 16 total trials (eight trials per block × two blocks per session).

C. Null Hypothesis: Independence Prediction

For each rivalry configuration in each session type, the measured percepts’ proportions were averaged across the four trials in a session (two blocks × two counterbalances per block). This gave four values for each rivalry configuration within each session type, which were used for statistical comparisons. In “view-one” trials, the four averaged color-and-orientation conjunction proportions were calculated separately for cube-above trials and cube-below trials. For each rivalry type, independence predictions were calculated by multiplying the averaged color-and-orientation conjunction proportion for cube-above-fixation by the proportion for cube-below-fixation (e.g., cube-above-fixation average for RED + FRONT − LEFT × cube − below − fixation average for RED + FRONT − LEFT). Summing over the four color-and-orientation percepts gave the expected proportion of reporting time in “view-two, report-two” trials that both cubes would be seen as identical in both color and orientation, assuming the percepts of the top and bottom cubes’ color and orientation were resolved independently. All proportions were arcsine-transformed for statistical computations [30].

D. Observers

In total, five observers (three women and two men; mean age 26 years) participated in this experiment. One observer was author RL; all other observers were naïve as to the purpose and design of the study. All observers were color normal, as assessed by Ishihara pseudoisochromatic plates and by Rayleigh matches on a Neitz OT-II anomaloscope; observers also passed stereo vision testing with the titmus stereo test (to 80 arcsec). All observers completed heterochromatic flicker photometry (HFP) to equate the luminance of the two chromaticities (“red” and “green”) used in the experiment. HFP used a 1.6° foveally fixated disc viewed binocularly, which alternated between two chromaticities at 12.5 Hz. Observers adjusted the luminance of one chromaticity to minimize perceived flicker. All observers gave informed written consent according to the policy of the University of Chicago’s Institutional Review Board.

E. Results and Discussion

The hypothesis that two cubes’ color and orientation would appear the same more often than chance was tested by comparing color-and-orientation identity proportions from “view-two, report-two” trials against independence predictions from above-fixation and below-fixation “view-one” trials. These comparisons showed that measured view-two color-and-orientation identity proportions were significantly greater than independence predictions under most viewing conditions. Under SBR conventional and CISR conventional conditions, all five observers showed significantly greater color-and-orientation identity proportions than independence (i.e., chance). Under the CISR patchwork condition, three of five observers showed significantly greater color-and-orientation identity proportions than chance predictions. With patchwork SBR, however, only one observer showed a significantly greater color-and-orientation identity proportion than the chance prediction. Comparisons for each observer of measured color-and-orientation identity proportions under the four viewing conditions to corresponding independence predictions are shown in Fig. 3.

Fig. 3.

Results from five observers. The vertical axis is the proportion of the 60 s reporting period that the two cubes were expected or perceived to be the same color and orientation. Red bars (“Pred.”) indicate proportions expected if perceptual resolution of the two cubes above and below fixation were independent. Blue bars indicate measured proportions for conventional (“Conv.”) and patchwork (“Patch.”) stimulus presentations. Bottom right box: Representations of each percept that observers reported during experimental trials. All contrasts are pairwise with Bonferroni correction for four comparisons per observer. “SBR” = standard binocular rivalry. “CISR” = chromatic interocular-switch rivalry. NS: p > 0.05; * : p < 0.05. Error bars ± 1 SEM.

Significantly greater color-and-orientation identity proportions than predicted by independence showed that resolution of rivalrous color and Necker cube orientation were not independent for the cubes above and below fixation. Identity proportions from patchwork SBR trials were the major exception to this trend, as only one of five observers saw identical percepts significantly more often than chance under the patchwork SBR condition. For CISR, however, three of five observers saw identical percepts significantly more often than chance under the patchwork CISR condition. These findings—good evidence for grouping under conventional but not patchwork SBR and for both conventional and patchwork CISR—are consistent with the multistage processing account of rivalry resolution. When stimuli were presented with patchwork SBR, chromatic resolution was affected in part by monocularly driven neurons representing the entire view of each eye, so color grouping was lower than with conventional SBR. When stimuli were presented with patchwork CISR, for which chromatic resolution was driven by binocularly driven neurons, color grouping was high. This is consistent with measurements of color grouping of 16 simultaneously viewed chromatically rivalrous discs [13].

While the results of Experiment 1 showed that resolution of rivalrous color and orientation for Necker cubes above and below fixation was not independent—that is, there was perceptual grouping—the possibility that grouping mechanisms for each feature (color and orientation) acted independently remained an open question. Experiment 2 set out to test this.

3. EXPERIMENT 2: IS GROUPING WITH TWO AMBIGUOUS FEATURES GREATER THAN EXPECTED FROM GROUPING BY EACH FEATURE INDEPENDENTLY?

A. Apparatus and Stimuli

All stimuli were viewed through the haploscope. Two Necker cubes (one above fixation and one below) were presented to corresponding locations in the two eyes in all trials. Stimuli were presented in four configurations: SBR conventional, SBR patchwork, CISR conventional, and CISR patchwork. Each rivalry configuration was counterbalanced by color, to give a total of eight stimulus configurations in each session type (see “Procedure” for further details on the session types). Specifications for the Necker cube sizes and chromaticities and, for the background, were the same as in the previous experiment.

B. Procedure

The experiment was broken into three session types. In each session type, observers reported whether both cubes appeared identical in color and/or orientation during a series of 70 s trials. As in the first experiment, no button-hold data were collected during the first 10 s of each trial. In the first session type, “report-color-and-orientation-identity,” observers held buttons to indicate when the two Necker cubes displayed in chromatic rivalry appeared to be both the same color and same orientation. In the second session type, “report-color-identity,” observers held buttons to indicate when two Necker cubes displayed in chromatic rivalry appeared to be the same color, irrespective of perceived orientation. In the third session type, “report-orientation-identity,” observers held buttons to indicate when two Necker cubes displayed in chromatic rivalry appeared to be the same orientation, irrespective of perceived color. All sessions used identical paired chromatically rivalrous Necker cube stimuli.

Observers completed nine experimental sessions in total: one of the three session types on each of the nine days they participated. These nine sessions were split into three separately randomized session blocks, each consisting of one repetition of each session type. Each session was comprised of five trial blocks, each consisting of one repetition of each trial type with each possible stimulus configuration (as detailed in “Stimuli”), presented in random order. The three session types each had 40 trials (five blocks of trials × eight trials per block). For each session type, total reported identity time for each trial was recorded, expressed as a proportion of the 60 s reporting period and arcsine-transformed for statistical computations.

C. Null Hypothesis: Independence Prediction

For each rivalry configuration in each session type, identity proportions were averaged across the 10 trials (five blocks × two counterbalances per block) in a session. This gave three values for each rivalry type within each session type—one from each day of testing—which were used for statistical purposes. For each rivalry type, independence (chance) predictions were calculated by multiplying color-identity proportions from “report-color-identity” trials by orientation-identity proportions from “report-orientation-identity” trials. Proportions were arcsine-transformed for statistical computations.

D. Observers

In total, six observers (four women and two men; mean age 24 years) participated in Experiment 2. All observers were naïve as to the purpose and design of the experiment. All observers passed the same assessments, completed heterochromatic flicker photometry, and gave informed written consent as in the previous experiment.

E. Results and Discussion

The hypothesis that grouping by rivalrous color and grouping by Necker cube orientation occur independently was tested by comparing color-and-orientation identity proportions (from “report-color-and-orientation-identity” trials) to independence (chance) predictions from separate “report-color-identity” and “report-orientation-identity” trials. Using the eight conditions, four planned simple orthogonal contrasts—each one between the measured and predicted proportions for one of the four viewing configurations—were tested, without Bonferroni correction for the four contrasts. These comparisons generally showed no significant differences between the color-and-orientation identity proportions and the chance predictions. Of 24 tests (four tests/observer × six observers), 20 failed to show a significant difference between color-and-orientation identity and independence. This is despite the fact that no correction for multiple comparisons was applied to the contrasts for each observer, which makes every test more powerful than with Bonferroni correction. Moreover, no consistent pattern of differences was seen among observers, with eight of the 20 nonsignificant differences opposite in sign to the prediction of a larger proportion for “report-color-and-orientation identity” trials. Also, the four significant differences were not consistent across observers, as Obs. NA showed grouping above chance in the SBR conventional and CISR conventional conditions, while Obs. CH showed significant grouping in the SBR patchwork condition, and Obs. AR showed significant grouping in the CISR patchwork condition. Overall, there is no evidence to reject the null hypothesis that grouping by rivalrous color and grouping by Necker cube orientation occur independently. Comparisons of color-and-orientation identity proportions to chance predictions for all observers and rivalry conditions are shown in Fig. 4.

Fig. 4.

Results for six observers. The vertical axis is the proportion of the 60 s reporting period that both cubes were perceived to have the same color and orientation. Red bars (“Pred.”) indicate proportions expected if grouping by color and by orientation were independent. Blue bars (“Meas.”) indicate measured proportions. Bottom box: Representations of each percept that observers reported during experimental trials. “SBR” = standard binocular rivalry. “CISR” = chromatic interocular-switch rivalry. “Conv.” = conventional stimulus configuration. “Patch.” = patchwork stimulus configuration. NS : p > 0.05; * : p < 0.05; ** : p < 0.01. Error bars ± 1 SEM.

This experiment fails to support the hypothesis that grouping by color and by orientation acts synergistically [24]. There is, however, a potential weakness in the experimental design that might artificially inflate the independence predictions. Instructing the observers to report grouping by one feature, when two features are perceived, could bias reported grouping toward that feature. In the experiment reported here, this theoretically could have increased the reported proportions in “report-color-identity” and “report-orientation-identity” trials, thereby raising the calculated independence predictions. To rule this out, a third experiment was performed in which observers reported grouping by both color and orientation in every trial.

4. EXPERIMENT 3: IS GROUPING FACILITATED IF TWO AMBIGUOUS FEATURES ARE PART OF THE SAME OBJECT?

A. Stimuli

All stimuli were viewed through the haploscope. Cubes in both session types were the same size as in the previous two experiments. In one session type, “cubes-and-discs,” the Necker cubes were achromatic (luminance 5 cd/m²) and thus not rivalrous in chromaticity. On the opposite side of the fixation cross were two 1.5° diameter discs presented in red-green patchwork CISR. In the other session type, “chromatic cubes,” the Necker cubes were presented in red-green patchwork CISR, and no discs were presented. In both trial types, the chromatically rivalrous figures, whether cubes or discs, had the same chromaticity and luminance specifications as in the previous experiments. Both session types had four stimulus configurations, given by counterbalancing the arrangement of colors, and whether the cubes were presented to the left or right of fixation (in “cubes-and-discs” trials, this also determined whether the discs were presented to the right or left of fixation).

B. Procedure

Observers completed two sessions on separate days, in random order. Each session was comprised of 10 blocks of 70 s trials, with no button-hold data collected during the first 10 s of each trial as in the previous experiments. In each trial, observers used up to two simultaneous button holds to indicate when the two chromatically rivalrous figures (either discs or Necker cubes) appeared the same color, when the two Necker cubes appeared the same orientation, or both. Total times of identical color and orientation were recorded and expressed as a proportion of each 60 s reporting period. Each block consisted of one repetition each of all four stimulus configurations, presented in random order. In total, each session was comprised of 40 trials (10 blocks × four trials per block).

For each session type, identity proportions were averaged over the four counterbalances within a block of trials. This gave 10 values for each session type, which were used for comparisons between session types. All values were arcsine-transformed for statistical calculations.

C. Observers

In total, five observers (all women; mean age 26 years) participated in Experiment 3. All observers were naïve as to the purpose and design of the experiment and passed the same assessments, completed heterochromatic flicker photometry, and gave informed written consent as in the first experiment.

D. Results and Discussion

The hypothesis that there is greater grouping from two ambiguous features when the ambiguous features are integrated components of a single object, compared with when they are distributed across different objects, was tested by comparing the total time of reported identical color and orientation between the “chromatic cubes” and the “cubes-and-discs” session types. For four out of five observers, this difference was not significant (separate comparisons for each observer; p > 0.1). For Obs. BB, the total identity proportion was significantly lower in the “chromatic cubes” condition (p < 0.001), contrary to the idea that grouping by two ambiguous features is greater when the features are part of the same object. Comparisons of identical color-and-orientation proportions in the “chromatic cubes” and “cubes-and-discs” sessions are shown in Fig. 5. These results do not support the hypothesis that grouping by ambiguous color and Necker cube orientation is greater when the two features are integrated within a single object; therefore, they are consistent with the findings in Experiment 2.

Fig. 5.

Results for five observers. The vertical axis is the proportion of the 60 s reporting period with identical color-and-orientation percepts when the ambiguous features were in one object or two. Red bars indicate proportions from “cubes-and-discs” trials. Blue bars indicate proportions from “chromatic cubes” trials. Bottom: Representations of the four percepts reported in each of the two trial types. NS : p > 0.1; ***: p < 0.001. Error bars ± 1 SEM.

Note that, if resolution of the “chromatic cubes” causes a general reduction in grouping by color and/or by orientation compared with the resolution of two separate pairs of objects (due to, say, different spatial frequency characteristics of the cubes from the discs), this theoretically could mask greater grouping by color and orientation when both features are within a single object. Post-hoc comparisons of grouping time for each feature were made between the “chromatic cubes” and “cubes-and-discs” conditions, and they show this is not a concern. For color-only grouping, only two observers showed significantly different total grouping under “chromatic cubes” and “cubes-and-discs” conditions, (Obs. BB was greater in the “cubes-and-discs” condition, p < 0.001; Obs. JA was greater in the “chromatic cubes” condition, p < 0.05). For orientation grouping, Obs. BB alone showed a significant difference, with greater grouping in the “cubes-and-discs” condition (p < 0.001). These results are not in accord with reduced grouping overall for the “chromatic cubes” condition.

5. GENERAL DISCUSSION

The results showed perceptual grouping when two objects shared two different types of ambiguous features—here, chromatic rivalry and Necker cube orientation ambiguity. There was no evidence, however, that grouping by color and orientation was affected by higher-level processes acting on individual feature-grouping mechanisms or on feature-integrated neural representations of objects.

In the first experiment, perceptual grouping was strong with SBR conventional, CISR conventional, and CISR patchwork viewing conditions but was reduced in the SBR patchwork condition. In the SBR patchwork condition, the two cubes presented to each single eye were different and had unvarying chromaticities for the entire duration of each trial. Failing to find grouping by color in this condition is consistent with a simple model of eye-rivalry for disambiguation of rivalrous chromaticity; however, grouping was not abolished for all observers in SBR patchwork. This interocular grouping could not be due to simple eye-rivalry and likely reflects the influence of distinct zones of monocular rivalry [18] and/or feedback [19].

In the first two experiments, grouping was significantly above chance for most observers when stimuli were displayed in CISR conventional and patchwork conditions. This finding is consistent with a recent study showing grouping by color when discs were displayed in CISR conventional and patchwork conditions and further indicates that binocular neurons play a role in the resolution and grouping of color in chromatic interocular-switch rivalry [13] (though cf. [31] for a possible monocular contribution to resolution of interocular-switch rivalry).

In the second experiment, no evidence was found to support the hypothesis of color-grouping and orientation-grouping synergy when viewing a pair of chromatically rivalrous Necker cubes. Relatedly, in the third experiment there was no evidence to support the hypothesis that grouping by both color and orientation was enhanced when rivalrous chromaticity and Necker cube orientation ambiguity were integrated features of a single object compared with when they were distributed across two separate objects. Taken together, these findings are consistent with independent action of color-grouping and orientation-grouping, at least for Necker cube orientation. Thus, no evidence here supported perceptual resolution of rivalrous chromaticity and Necker cube orientation governed by a neural representation that follows feature integration.