Our study demonstrates the use of faces in central visual field as a viable tool for studying masking. Using a backward mask, we obtained peak masking at a non zero SOA (type B masking). Using a trailing mask, we obtained masking greater than that with a common onset common offset mask (SOA 0ms), demonstrating that the trailing portion of the mask was responsible for this effect. Our model, which was based on data from common onset masking with various portions of the trailing mask removed, as well as from backward making data, suggests that there is a specific period during the time course of visual processing, when the presence of a mask can exert itself. Our model also provides a framework within which backward and common onset masking operate through the same mechanisms.
There are, however, some important limitations to this model. For one, our model does not take into account offset transients, which have been shown to be associated with threshold elevations, at least in the context of masking by light (Crawford, 1947). The account given by Macknik and Livingstone (1998), who used bars as targets and mask, suggests that the disinhibitory rebound associated with a sharp luminance decrement of the mask (offset transient) is responsible for inhibiting the target. As our model does not take offset transients into account, it could be underestimating the masking effect of the offset transients introduced when a trail is interrupted. However, data from Experiment 2c shows that a series of pulsed masks (which introduced multiple transients) showed no increase in masking compared to an uninterrupted trail, a finding that held across a wide range of pulse frequencies. Nevertheless, it would be interesting to follow up this work to explore the potential role of transients in our paradigm (e.g. Sackur, 2011; Tapia, Breitmeyer, & Jacob, 2011).
Our model also predicts that backward masking will peak at an SOA of around 100ms, whereas our data suggest a peak quite a bit earlier (our baseline SOA runs in Experiment 3 showed a peak SOA of 58.3ms for a few participants). This discrepancy could be due to any number of the following reasons. First, our model was based on data from methodologically different experiments. While the common onset data was derived from the same observers, using the same method of calculating threshold elevations, the backward masking data in Experiment 1 was derived from a different set of observers, using a different method of calculating threshold elevations. Indeed, when combining the two data sets into a unified set of elevation thresholds, the peak SOA conditions from Experiment 1 showed slightly greater masking than the Full Trail condition from Experiment 2, while within Experiment 1, the peak SOA thresholds and Full Trail thresholds were virtually identical. Second, our model is a rather simple one, in that it posits a single tsa hdac across the entire trail duration, while in reality, there may be a number of distinct mechanisms each of which has its own integration window (see Breitmeyer and Ogmen (2006), p. 50). Third, our model only takes into account the magnocellular response of the mask, and does not model the influence of mask modulated parvocellular activity upon target visibility. It should also be noted that our model predicts unusually sharp backward masking functions (see Fig. 14), which is due to the narrow width of S(t).
Another important issue is that it is challenging to determine whether the increased masking found with a common onset trailing mask, relative to a common onset common offset mask, is due to the trailing portion’s ability to interfere with target processing for an extended period of time, or whether it is simply due to the extended trail being temporally integrated into a (perceptually) higher contrast mask. In other words, increasing mask duration may be equivalent to simply using a higher contrast mask that offsets with the target offset, within a limited integration window (Bloch’s Law). Increasing mask contrast relative to that of the target, while keeping mask and target durations fixed, has been shown to increase masking at an SOA of 0ms (Stewart & Purcell, 1974; Weisstein, 1972). Similarly, increasing mask duration, while keeping mask and target contrast fixed, also results in increased masking at an SOA of 0ms (Breitmeyer, 1978). Thus, it is difficult to say whether the increased masking we found with a common onset trailing mask is due to stronger sustained-on-sustained inhibition, or whether it is due to the trailing mask having more time to interfere with target processing (our model only takes into account the latter possibility). Indeed, this same question could be asked of the results of Bischof and Di Lollo (1995) study, which showed an increase in masking in central visual field as a function of mask duration. There is, however, some interesting data that suggest that the duration of a centrally presented contour mask is capable of driving masking independently of its (luminance x time) energy. Di Lollo, von Mühlenen, Enns, and Bridgeman (2004) showed that while masking remained fairly constant across brightness matched targets of varying durations (increasing target duration or increasing target luminance reduced masking substantially), the same was not true when it came to mask energy. That is, increasing the duration of a brightness matched mask produced a very similar masking function to that when the mask duration was increased with a fixed luminance, suggesting that mask duration is capable of driving masking. While this wasn’t common onset masking, the mask manipulations were done with a target duration of 10ms, and an ISI of 0ms, which is very close to a common onset paradigm. It should be noted, however, that in this study, the mask was a contour mask, whereas in our experiments, the mask was a full face. Because there is more contour overlap with the target in a full face mask, there may be a larger component of intrachannel sustained masking, or masking due to integration (luminance channels of the mask and target being shared due to spatial overlap), relative to a simple contour mask. As such, mask energy (luminance×time) may have a larger role to play with our stimuli. One way to further address this issue would be to more rigorously sample masking as a function of mask duration with different types of masking stimuli, as we are currently investigating in our lab.