The goal of this study was to investigate whether the processing of scene congruency/incongruency (or object/context interactions) exist in the absence of awareness. Furthermore, does the privilege of animal stimuli exist in the interactions between object and background without awareness. In
Experiment 1, using a b-CFS paradigm, we observed that congruent scenes had shorter response times than incongruent scenes, but this difference was only marginally significant. Further analysis showed that the difference between congruent and incongruent scenes was significant with vehicle targets but not with animal targets. On the other hand, in
Experiment 1, the response time for animal stimuli was longer than for vehicle stimuli. This was opposite to the results found in previous studies in the conscious condition (
New et al., 2007;
Ohman, 2007;
Mormann et al., 2008;
Crouzet et al., 2012;
Yang et al., 2012;
Drewes et al., 2015). A possible explanation may be found in the search/detection strategy used by our participants; although instructed to respond as soon as they saw any part of any target (and thus without instructed preference of one or the other target category), they may have been actively looking for one of the two target types more than the other, resulting in shorter response times for that category (here: vehicles). Such behavior as a proxy strategy has been postulated before (
Drewes et al., 2015). When expecting (or actively looking for) one type of target, response times to the appearance of this stimulus category may be slightly faster (
Stein & Peelen, 2015;
Stein 7 Peelen, 2017;
Stein, Utz, & van Opstal, 2020). The same expectation may have facilitated the appearance of congruency effects for the expected target category: together with the expectation of, for example, with a vehicle one might expect a road in the background but not a forest or ocean.
Experiment 1 was a breaking-CFS paradigm, and differences in response (or break-through) time are generally considered indications of differences in unconscious/unaware processing (before breakthrough) rather than in aware processing after breakthrough. The results ran contrary to findings in previous studies in the aware condition. To replicate the previous studies (in the conscious/aware condition) with the same stimulus material and the same participants and to compare with our result in unconsciousness in the b-CFS experiment, we ran
Experiment 2. Considering the response time advantage for animal stimuli observed in the most classic studies by using 2AFC-style paradigms for object recognition (
Kirchner & Thorpe 2006;
Crouzet et al., 2012;
Drewes et al., 2015), we used a 2AFC paradigm in
Experiment 2, which was designed to be comparable to previous studies, except for the use of monocular target presentation. In the aware condition in
Experiment 2, we did observe the same results with previous studies: animal stimuli had shorter response times than vehicles and the response time of congruent scenes was faster than incongruent scenes (
Palmer, 1975;
Biederman et al., 1982;
De Graef et al. 1990;
Boyce & Pollatsek, 1992;
De Graef et al., 1992).
In the unaware (suppressed) condition in
Experiment 1, the congruency effect was overall only marginally significant, possibly because of existing (as far as statistical significance is concerned) only within the vehicle images (and even there only within the congruent ones). Although the congruency ratings (stronger in congruent vehicle images) might explain the partial effect in
Experiment 1 found only with these, we can still exclude that the unexpected overall results we found in
Experiment 1 originated from either our choice of stimuli or an unusual batch of participants, because both were the same throughout the entire study.
However, it would be hasty to conclude from these results that congruency effects with animal images do not exist without awareness: in the unconscious condition (b-CFS), participants were requested to press a single button once they saw any part of the stimulus images, which required them to perform (indiscriminate) object (or stimulus) detection. In the conscious condition (2AFC) however, participants were required to press separate buttons for animals and vehicles, which required object discrimination (recognition). Differences between the experiments may therefore originate either from difference between conscious and unconscious conditions, or from the difference in perceptual task (detection/discrimination). To address this, we designed
Experiment 3. Participants were asked to press different buttons for animals and vehicles during a CFS paradigm (named here 2AFC-CFS). In
Experiment 3, participants were required to perform object recognition in a CFS paradigm. This was intended to avoid the task difference between conscious and unconscious conditions, therefore improving comparability between conscious and unconscious conditions.
The results from
Experiment 3 showed longer response times to animal stimuli than that to vehicle stimuli. This is consistent with the results in
Experiment 1 (with suppression) but opposite with the results in in
Experiment 2 (without suppression) and earlier studies in the conscious condition. This suggests that the difference between animals and vehicles on response time in the unaware (suppressed) condition is opposite from that found in the aware condition and is not caused by the difference in task (recognition vs. detection).
When we compared the response times between
Experiment 1 and
Experiment 3, we found the response time for detection (
Experiment 1) was generally shorter than recognition (
Experiment 3; see
Table 1). This could be explained by the 2AFC-CFS task actually including two processing steps: detection and recognition, with recognition requiring additional time after the initial stimulus detection. We do note, however, that the time difference between the b-CFS and the 2AFC-CFS condition is on average in the order of 80 ms, which is faster than the results obtained in previous CFS studies (see introduction, at least 120ms but typically more than 200 ms). However, the processing requirements between the b-CFS and 2AFC-CFS most likely overlap substantially, so we do not believe that a simple addition of the response times of the b-CFS paradigm and those of a conventional 2AFC-paradigm would be a good model of our 2AFC-CFS Experiment. Although further investigation would be required to disentangle the contributions of CFS and 2AFC to the outcome of
Experiment 3, we interpret the overall longer response times in the 2AFC-CFS experiment to be an indication that the congruency effects found likely do not emerge at the lowest levels of perception but emerge later in the processing chain, when the observer is at least partially aware of the stimulus. Effects of congruency might then emerge in the unconscious condition to a weaker extent, in our case only surfacing with the vehicle, but not the animal stimuli in
Experiment 1—perhaps comparable to the earlier stages considered by
Leroy et al. (2020) (early processing or memory matching)—whereas a certain degree of conscious access would be required for the effect to develop fully, comparable to the later stages of (
Leroy et al., 2020). It would also indicate that the processing involved in our CFS experiments is different than what would be expected from (
Grill-Spector & Kanwisher, 2005): In their study, Grill-Spector and Kanwisher used forward/backward masking rather than monocular masking (CFS), which is likely to introduce a different kind of suppression.
An alternative explanation for the results, particularly in
Experiment 1, would be differences in the overall congruency of the stimuli. Vehicle images were indeed judged as being more congruent overall, both for the congruent and for the incongruent samples (yet with a significant difference between), which may have led to a stronger effect with the congruent images. On the other hand, the difference between the ratings of the congruent and incongruent stimuli was larger with the animal images, which one would have expected to lead to a stronger effect. Furthermore, this effect did not re-appear in
Experiments 2 and
3, which would also suggest it to be tied to the novelty of the images. In sum, it is unclear why the higher congruency rating with the vehicle images would have led to an effect, while the larger difference between congruency ratings with the animal images did not.
During the 2AFC experiments, accuracy ratings are very high throughout. This indicates that our participants had no trouble performing the recognition task. Values near the ceiling can in principle obscure differences between conditions; in our case, however, all main effects were significant despite the high overall values.
It should be noted that the assignment of the response keys in the 2AFC conditions was fixed (animals on the left arrow), and because most participants were right-dominant, this would have put stimulus images in the view of the left eye. This may in principle have biased responses in favor of animals. However, only in
Experiment 2 did we find an animal advantage; more importantly, this advantage should have affected all three experiments in the same direction and certainly cannot explain the opposite results (vehicles faster than animals) in
Experiments 1 and
3.
Stimuli were selected from natural scenes, without editing the images by pasting objects on backgrounds. This eliminated possible artefacts from cut and paste or similar but prevented us from creating a perfectly balanced stimulus set. It is therefore possible that certain differences between the animal and vehicle images influenced the results reported here. However, because the images were the same in all three experiments, this cannot be the cause for the main differences found between the CFS, 2AFC, and CFS/2AFC paradigms.
Finally, considering the order of the experiments, because all participants completed the experiments in the same order, a certain amount of learning- or training-based improvement is to be expected—both from training in the experimental condition (e.g., in CFS) but also from the fact that the images are no longer novel to the participants after they passed
Experiment 1. This may have improved overall response times and hit ratios in
Experiments 2 and
3, potentially affecting direct comparisons of experiments. However, because of the task differences between the experiments, results were not intended to be compared across experiments directly, by design. The analysis was based on the relative differences within experiments. Thus keeping the same order of experiments for all participants did not endanger the validity of the analysis and reduced within-experiment result variability compared to a randomized approach. The remaining concern would be the fact that animal stimuli in
Experiment 1 resulted in slower response times, yet in faster response times in
Experiment 2. Seen alone, we would not immediately be able to exclude this to be the result of a training effect. However, if this was the case, then the RTs to animal stimuli should have remained faster in
Experiment 3, as there appears to be no obvious reason why a training effect such as this, if it existed, would reverse itself with even more training. Ultimately, we thus conclude that any training effect that may exist in our study did not affect the reported results in a significant way.