Binocular vision or stereopsis provides precise depth perception by aligning the two eyes’ views. Under the eye-specific stimulation of binocular rivalry, the mature visual system enters into a continuous fluctuation between two or more perceptual states, not yielding stereopsis. While cortical binocularity in humans seems to have a relatively abrupt onset (at around 3.5 months) during ontogeny (Braddick et al., 1980; Petrig et al., 1981), driven by experience-dependent mechanisms (Kovacs et al., 2011), little is known about the onset time of binocular rivalry and its further development. Here we review information with respect to the human development of binocular rivalry, and make an attempt to assess its maturity before puberty in a behavioral experiment. We interpret our data in the wider framework of neural adaptation.

The nature of binocular vision of human infants before the occurrence of binocular 3D perception has been debated. This issue was mainly investigated in preferential looking paradigms, employing stimuli that induce binocularly rivalry in adults. Shimojo et al. (1986) found that infants younger than 3.5 months of age preferred to look at the dichoptic (interocularly orthogonal) pattern. However, at an average age of 3.5 months, a sudden shift of preference occurred from the rivalrous pattern to the fusible stimulus. They interpret this result as a preference for a blended stimulus, resulting in a grid-like pattern, which is more complex than the monocularly projected simple lines. However, from the time by which binocular functions have further developed (3.5 months of age), the two patterns begin to oscillate, which might be aversive for infants. This would account for the shift in the preference for binocularly fusible stimuli, and would suggest that pre-stereoscopic vision blends those images that are rivalrous for adults. However, these results could not be replicated (Brown and Miracle, 2003). Nor did Brown et al. (1999) find any physiological evidence for binocular rivalry using a visually evoked potential paradigm with 5- to 15-month-old infants. They attribute their result to the immaturity of dichoptic suppression.

Even less data have accumulated so far concerning the development of rivalry following infancy. In a study that aimed to compare binocular interactions of children aged 6–14 years to normal and amblyopic adults, it was found that binocular summation decreased with age in a dichoptic visual acuity task (Vedamurthy et al., 2007). The acuity of the dominant eye did not improve significantly in children in the dichoptic viewing condition compared to the monocular condition. In this respect, the performance of children was similar to that of adults. However, they found a significant negative correlation with age in the improvement of acuity of the non-dominant eye in the dichoptic condition compared to the monocular one, indicating that developmental trends in binocular interactions are present after infancy, until at least pre-puberty.

The development of binocular rivalry was investigated in 5- to 6-year-old children (Kovacs and Eisenberg, 2005). They found that children alternated significantly more quickly than adults. Verbal reports of the subjects also indicated that children perceived a patchwork of the two images more frequently than adults. On this basis, the conclusion was drawn that the visual system of 5- to 6-year-old children is not sufficiently mature to integrate entire images spatially, thus they experience more piecemeal rivalry than adults. This is in line with their earlier findings in contour integration (Kovacs et al., 1999; Kovacs, 2000) and spatial integration (Káldy and Kovacs, 2003).

Binocular rivalry shares several features with the perception of ambiguous figures, such as the Necker cube. Common features are gamma distribution of the dominance times of each percept, the high inter-subject variability of the frequency of reversals, the significant influence of stimulus properties on reversal rates, or the fact that both can be influenced by the voluntary control of the subject (see Kornmeier and Bach, 2005).

According to the findings reviewed by Leopold and Logothetis (1999), stimulus properties, such as brightness, contrast, and spatial-frequency content can have a significant impact on the balance of dominance and suppression. In addition, high-level properties of the stimuli can also modify dominance periods in multi-stable perception. Such properties include recognizability or semantic content. For instance, if a recognizable figure is inverted, then its perceptual dominance might significantly be altered in both figure/ground stimuli and binocular rivalry.

Voluntary control is another modifying factor of multi-stable perception. The influence of voluntary control of the subject was found to be stronger in the case of ambiguous figures than for binocularly presented rivalrous stimuli (van Ee et al., 2005). Taddei-Ferretti et al. (2008) also point out that the rivalry between the two possible percepts of an ambiguous figure is less automatic than the competition between two different images presented binocularly. An additional common feature of binocular rivalry and ambiguous figures that Taddei-Ferretti et al. (2008) mention is that both are influenced by eye movements (Ellis and Stark, 1978; Sabrin and Kertesz, 1980). Leopold and Logothetis (1999) consider exclusivity, inevitability, and randomness as the three most fundamental common features of multi-stable perception including binocular rivalry and ambiguous figures. Exclusivity means that only one percept is present at one time, while inevitability implies that “perception can never become ‘locked’ onto a single solution” (p. 261): perceptual hypothesizes are constantly changing concerning the presented stimuli. These attributes are characteristics of both binocular rivalry and ambiguous figures.

Reese and Ford’s (1962) pioneering study intended to investigate developmental aspects of ambiguous figure perception. Nursery-school children were shown a series of six pictures of either animals or human faces. Their task was to name each. After that, they were asked to state an expectancy about the next picture. The result was that when they were shown the Bugelski rat-man ambiguous figure, it was easier for them to provide the “animal” interpretation than the “human face” interpretation, which means that the animal interpretation was easier to prime by the previously shown pictures. This might be considered evidence of stimulus-dependency even at such an early age. However, 3- to 5-year-old children also show significant performance differences even in this short age range. Doherty and Wimmer (2005) found that 3-year-old children cannot even report both interpretations of such ambiguous images as the duck–rabbit or the man–mouse figures. However, 4-year-old children can easily interpret the ambiguous figures in both ways. Nonetheless, spontaneous reversals occurred only at the age of 5. The conclusion of this study is that understanding that the perception of the same physical image might reverse is not sufficient for spontaneous reversals to occur.

The foregoing review of the literature indicates that binocular rivalry in childhood has been poorly investigated in the past, and the case is similar concerning the broader sense of bistable or multi-stable perception. Some studies focused on changes in binocular rivalry during adulthood. In these studies, it was found that domination times became longer with age. Jalavisto (1964) for instance, investigated binocular oscillations in the age range of 40–93. It was found that the frequency of oscillation decreased with age in a regular manner, and a total lack of change became prevalent in the oldest age classes. In a more recent study, similar results were obtained (Ukai et al., 2003), in which the alternation rates in three age-groups were compared: 20–34, 35–49, and 50–64-year-old subjects were investigated. In line with the results of Jalavisto (1964), they found a prolongation in alternation time as a function of age. Information is still scarce with respect to infancy, and there is a complete lack of data on the development of binocular rivalry beyond the first 5–6 years of life in childhood. In this study, we are attempting to fill this gap by investigating the developmental trends in binocular rivalry in pre-puberty. We employ a classic behavioral paradigm with orthogonal gratings, and introduce novel statistical measures to analyze the data. These novel measures were developed by Pastukhov and Braun (2011), and they provide a sensitive tool to estimate the impact of the history of perceptual dominance on future percept durations. The Pastukhov and Braun (2011) method used here reveals a significant correlation between past perceptual history and future dominance duration, which does not become evident with conventional measures such as sequential correlations of dominance durations (Fox and Herrmann, 1967; Borsellino et al., 1972; Walker, 1975; Lehky, 1995). Assessment of the “cumulative history” and its characteristic time-constant helps us to take a look at the adaptive states of the visual system under multi-stable perception, and brings us closer to establishing a possible developmental scenario for binocular rivalry.

The t-test yielded a marginally significant difference in average dominance times (T_dom) between 9-year-olds and adults (for means and t-values see Figure 3A). Each percept tends to persist for a longer period in adults than in 9-year-old children (Figure 3B). There was no significant difference between 9- and 12-year-olds, and between 12-year-olds and adults. However, the developmental trend in Figure 3B seems to be clear: dominance times increase with age. The same tendency was found earlier in 5- to 6-year-olds as compared with adults (Kovacs and Eisenberg, 2005).

The c_H value is significantly higher in 9-year-olds than adults, i.e., the length of the subsequent dominance period of a particular percept shows a higher correlation with the previous dominance time ratio of the other percept in 9-year-olds than in adults (Figures 3A,C). This means that 9-year-olds and adults show a significant difference in their extent of adaptation to each percept. There was no significant difference between 9- and 12-year-olds, and between 12-year-olds and adults.

The time-constant of the build-up of the adaptation (τ_H) produced significant differences both between 9-year-olds and adults as well as 12-year-olds and adults (Figures 3A,D). There was no significant difference between 9- and 12-year-olds. The τ_H value of adults is significantly higher than that of 9- and 12-year-old children, showing that the build-up of adaptation is slower in adults.

These differences were also indicated by correlations between age-groups and the observables (see Table 1).

Our results indicate that 9-year-old children are not exactly adult-like in terms of alternation rate which is a conventional measure of binocular rivalry. Children seem to have shorter average dominance times than adults. This is consistent with an earlier study by Kovacs and Eisenberg (2005) that showed that 5- to 6-year-old children are alternating very quickly. Our findings are also in line with the results of Jalavisto (1964) and Ukai et al. (2003), who found that alternation rate decreased with age in adulthood. Although the developmental curve is not complete yet, and there are several further age-groups to be tested, it can be concluded that the development of binocular rivalry, as measured by its most salient feature, is not complete by the end of the first decade in life. That draws a conspicuously slow developmental trajectory which is not yet supported by explanatory anatomical or physiological data.

In addition, we have applied two novel measures of the effect of neural adaptation, recently suggested by Pastukhov and Braun (2011). The first such measure (c_H) was the correlation between dominance times and accumulated prior dominance history and the second (t_H) was the effective time-constant of this accumulation. To appreciate the import (and limitations) of these measures, one has to consider that perceptual reversals may have several contributing causes (Wolfe, 1984; Nawrot and Blake, 1989; Petersik, 2002; van Ee, 2009; Alais et al., 2010; Kang and Blake, 2010; Pastukhov and Braun, 2011).

Firstly, neural adaptation of the dominant representation is thought to progressively destabilize the dominant percept by both the adaptation of the dominant percept, and the recovery from adaptation of the suppressed percept. Secondly, spontaneous activity fluctuations in perceptual representations as well as external transients such as eye movements or eye blinks curtail the duration of dominance periods. Thirdly, internal transients such as shifts in attention or in other volitional processes may trigger reversals. The measure c_H is a correlative measure and estimates only the relative contribution of neural adaptation to reversal timing, that is, relative to all other possible factors. We emphasize that it should not be taken to estimate the absolute strength of neural adaptation.

Specifically, our finding that dominance durations are more correlated with prior history in children than in adults, implies simply a greater relative contribution of neural adaptation. This could either be because adaptation is more pronounced, or because other factors (e.g., neural noise, attention shifts) are less pronounced in children. Our observations that shorter dominance phase duration in children are accompanied by shorter time-constants of reconstructed neural adaptation are consistent with predictions of models of multi-stable perception (Wilson, 2007; Shpiro et al., 2009), where mean dominance duration is directly proportional to the adaptation time-constant. A related possibility is that, due to the generally shorter dominance times of children, stochastic factors such as neural noise or attention shifts simply have fewer opportunities for triggering a perceptual reversal. Voluntary control over binocular rivalry is limited (Chong et al., 2005; Hancock and Andrews, 2007) and in fact less than for other bistable displays (Meng and Tong, 2004). Rivalrous displays undergo perceptual reversals even when attention is diverted (Lee et al., 2007). Nevertheless, we cannot rule out the possibility that our findings reflect differences in attentional characteristics between children and adults.

Although our study provides the first articulate view on the human developmental trajectory of binocular rivalry, more age-groups, and the underlying factors behind the protracted developmental curve need to be further investigated.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.