Intermittent Stimulus Presentation

The gain in temporal resolution while using this paradigm to study the

endogenous reversal process comes at the cost of presenting the stimulus intermittently. This begs the question about whether the findings from studies using this paradigm are applicable to the continuous case. This question is raised because, generally, we do not experience our visual world intermittently but rather in a continuous manner with our brain constantly resolving the instability of the visual information we receive. Moreover, some researchers (e.g. Noest et al., 2007) argue that the existing neural findings of perceptual reversals during continuous viewing of an ambiguous stimulus do not explain why the same percept reappears across trials in a paradigm presenting the ambiguous stimulus intermittently. Noest et al. (2007) recently discussed the neural differences between the reversals occurring during continuous stimulus presentation and during intermittent stimulus presentation. They suggest that the underlying mechanisms differ between the two with the reversals occurring during the intermittent case being a ‘percept choice’ whereas the ones occurring during continuous presentation being a ‘percept switch’ (Guckenheimer & Holmes, 1983). According to Noest et al. (2007), ‘percept choice’ involves complex ambiguity resolution behavior whereby the near- threshold interaction of the competing neural representations is controlled by the timing of stimulus on and off times. This suggests a sort of perceptual decision about the representation of the ambiguous stimulus that is presented intermittently (Kornmeier &

Bach, 2012). ‘Percept switch’, on the other hand, occurs after prolonged stimulus presentation when neural adaptation of the current percept representation has slowly destabilized its own attractor and neural noise will push the system to escape from what is left of the attractor to the other attractor (Noest et al., 2007). This in turn leads to the other percept representation.

Kornmeier and Bach (2012) argue however, that the variation that they applied to the original design of the Onset Paradigm with a shorter ISI and a longer presentation time than the ones used in Orbach et al. (1963, 1966) and O’Donnell et al. (1988)’s experiments may lead to perceptual reversals that involve ‘percept switch’ mechanisms and that the results observed in their studies apply to the continuous case. Orbach et al. (1963, 1966) presented a Necker Cube discontinuously, varied presentation time and ISI and found that reversal rates can be modulated as function of ISI and presentation time. The results of their studies coupled with the results from other studies (e.g. Kornmeier et al., 2007; Kornmeier & Bach, 2012, Leopold et al., 2002; Maier et al., 2003)’s studies suggest that reversal rates are modulated in a non-monotonic manner as a function of ISI, with an increase in reversal rates with ISI up to 400ms after stimulus onset and then a decrease in those rates with further ISI increase. These results show a smooth

monotonic decrease in reversal rates similar to the continuous case for shorter ISIs (see Figure 2 in Chapter 1). They also found that continuous stimulus presentation also involves short interruptions due to eye-blinks that lasts about 200ms with a mean frequency of 4s (Caffier et al., 2003).

Previous findings have shown so far that the Onset Paradigm successfully

It provides a series of EEG signatures related to endogenous reversals. This paradigm that involves presenting the stimulus intermittently is suitable for the type of analyses conducted in our experiments. These analyses involve exploring the factors that

influence the ERP signatures that have been suggested to be reversal related (the RP and the RN – Experiments 1&2) and investigating the pre-stimulus period in order to see if there is a period during which the spatio-temporal profile in that window is predictive of an upcoming perceptual reversal. These involve time-locking events to stimulus onset which in turn means that the paradigm we use in our experiments involves presenting the stimulus intermittently.

Based on the focus and aims of the research, experimenters have used the Onset Paradigm with differing stimulus presentation times and ISIs. For instance, Hesselmann et al. (2008) presented Rubin’s Face-Vase stimulus (1915) for a very short period of time (150ms) and had long ISIs (>20s). The reasons for this experimental design are explained in more detail in the next section. This type of experimental design, however, allowed them to explore the pre-stimulus period in more depth and avoided any overlap of effects from the previous stimulus. We use a similar paradigm in two of our

experiments (3&4) in order to study the pattern of activity across time (until stimulus onset) in the pre-stimulus time window and see if there is a spatio-temporal profile that is linked to perceptual reversals. As for Kornmeier and Bach (2004a, 2005), their experimental design involved using longer stimulus presentation times (800ms) and shorter ISIs (400 – 1000ms) in order to investigate ERPs that are related to perceptual reversals with the idea that reversals occur at stimulus onset. In order to investigate

whether or not these ERPs are dependent on higher level factors (e.g. task or response demands), we used a similar paradigm in our other two experiments (1&2).

2.2.2.1.1. Intermittent Stimulus Presentation: Inter-stimulus interval and presentation times.

I used two different experimental paradigms with different inter-trial intervals and stimulus presentation times in order to investigate both the pre-stimulus and post- stimulus time period. In this subsection, I describe the length of the times chosen for the inter-trial intervals and stimulus presentations in both experiments and the experiments on which they are adopted from.

Post-stimulus analysis focused experiments: Short ISIs and relatively long presentation times (Experiments 1&2).

Experiments 1 and 2 used an adapted version of Kornmeier and Bach (2004)’s experiment with long presentation time (800ms) and relatively short inter-trial interval (400ms when participants don’t respond – 1000ms when participants respond). The full experimental design is detailed in Chapter 3.

As was mentioned previously, Kornmeier and Bach (2004a) adopted O’Donnell et al.’s (1988) intermittent stimulus presentation experimental paradigm and modified it. They did this by shortening their ISI in order to be as close to the continuous viewing condition as possible. In an experiment conducted by Kornmeier et al. (2002) where they presented the Necker Cube discontinuously with varying ISIs increasing the ISI from 0ms (continuous observation) to 1200ms, the researchers found that increasing the ISI from 0 to 400ms, more than doubles the reversal rates (Orbach et al., 1963). At longer ISIs these rates decrease to almost 0 (Leopold et al., 2002; Maier et al., 2003).

After any button press however, Kornmeier and Bach (2004a) extended the ISI to 1,000 ms so that stimuli were not compared across intervening manual reactions. The idea behind their design was to make percept and task separable. Moreover, it avoids motor tasks between successive stimuli. In addition to shortening the ISIs, they lengthened the presentation time, long enough to allow the full development of the P300-like positivity and short enough to keep the probability of re-reversals (the occurrence of another reversal during stimulus presentation) low. Using these timings in our experiments (1&2) would yield a sufficient number of both reversal and stable trials for our analyses. This speculation is based on the previous behavioral findings of these experiments where the percentage of reversal trials is high enough to investigate the ERP correlates underlying them (e.g. Kornmeir & Bach, 2004a, 2005; Pitts et al., 2007).

Using this experimental paradigm, Kornmeier and Bach (2004a, 2005) were able to identify two reversal related ERP components, the RN and RP. One of the questions of this thesis is around the nature and underlying mechanisms of these two components and what they can tell us about ambiguous figure perception and perceptual reversals. However, although Kornmeier and Bach (2012) suggested that the RP is a marker of the destabilization process described in Chapter 1, it is unclear what the function role of the RN is and to what extent these two components reveal reversal related processes rather than task or response demand processes. As was mentioned previously, these correlates are sensitive to several factors (stimulus used, length of ISIs and presentation times, number of trials used, etc…). The task typically used to evoke the RN and RP components is what we call the “Reversal Task”. This task involves, on each

a change detection task (e.g., Rensink, 2002; Cohen, et al., 2005) in which reversal trials contain a change to detect (relative to the last stimulus) and stable trials do not.

Moreover, the RP and RN share a similar timeframe and scalp distribution to other ERP components (e.g. N2b component shows similarities to the RN; Courchesne et al., 1975; Potts, 2004; another component is the Selection Negativity; Anllo-Vento & Hillyard, 1996).

In order to address this, we used an adapted version of Kornmeier and Bach (2004a) and Pitts et al. (2007)’s experimental paradigms and varied response and task requirements. In addition to the reversal task, we introduced a second task called ‘Identity Task’. It is a modification of the reversal task paradigm (Kornmeier & Bach, 2004a, 2005; Pitts et al., 2007) where participants were asked to identify which of the two possible interpretations best matched their interpretation of the stimulus (i.e. Experiment 1: Left-Facing or Right-Facing for Necker Lattice; Experiment 2: Faces or Vase for Face-Vase). Moreover, in order to be more time efficient we used a similar paradigm to Pitts et al. (2007) where they presented ambiguous images intermittently and instructed participants to respond after each stimulus presentation about whether or not they experienced a perceptual reversal as opposed to Kornmeier and Bach (2004a) who used a paired-stimulus paradigm in which two ambiguous figures are presented in quick succession and participants indicated after the second stimulus whether they experienced a reversal or stable trial. This allowed us to address the task and response factors using a within-subjects design with a single testing session for each subject. These experiments (1&2) are explained in more detail in Chapter 3. These experiments are designed to answer the ERP related questions of my research and to identify whether

or not there is a spatio-temporal profile that is linked to perceptual reversals in the post- stimulus period.

Pre-stimulus analysis focused experiments: Identity Task only, short presentation time and long ISIs (Experiments 3&4).

In order to investigate the pre-stimulus period, we used an experimental paradigm that is adapted from Hesselmann et al. (2008)’s experiment. A brief, sparse, uncued presentation paradigm was employed with unpredictable onsets. Ambiguous stimuli were presented for a short duration of 150ms followed by a 100ms mask of white noise. The long inter-stimulus intervals (ISIs) were drawn from a truncated Poisson distribution with a minimum of 3s and a maximum of 14s. After each stimulus presentation participants were required to report their perception. A detailed description of this experimental paradigm is written in Chapter 5.

Due to the brief presentations, only one percept per trial was possible. Their percept was established through participants’ manual responses. Long and sparse ITIs were chosen in order to explore the pre-stimulus period, to maximize the

unpredictability of stimulus onset, and therefore minimizing volitional control of perception, obtaining independent responses in successive trials, and minimizing the occurrence of a re-reversal. In previous psychophysical piloting, these settings were found to lead to roughly similar percentages of faces and vase percepts and to prevent perceptual switching within single trials. The purpose of this thesis is to explore the underlying mechanisms of endogenous perceptual reversals. One of the aims is to explore the activity in the pre-stimulus period with regards to endogenous perceptual reversals and ambiguous figure perception. Therefore, we want to minimize the

influence of task related activity on the effects measured. This means minimizing the influence of exogenous high-level factors, such as volitional control, on reversal rates is necessary. This is done via the unpredictability of stimulus onset. As was mentioned previously, volitional control has a modulatory effect on perceptual reversals (see section 1.2.2.1. in Chapter 1), rendering these reversals exogenously induced instead of purely endogenous reversals. In addition to that, seeing as we are exploring the activity in the pre-stimulus period with regards to ambiguous figures, we need to avoid any carry-over effects between successive trials. In order to ensure that, independent responses in successive trials are important.

Previously, activity in the pre-stimulus period has been baseline corrected and discarded as unexplained variance when investigating the neural correlates underlying trial-by-trial response variations. However, a growing body of literature has suggested that these variations are linked to spontaneous ongoing fluctuations in the pre-stimulus period (e.g. Coste et al., 2011; Sadaghiani et al., 2010). As was mentioned previously, using a paradigm with long ITIs and short presentation time, Hesselmann et al. (2008) found that prestimulus activity in the FFA was higher when subjects subsequently perceived faces instead of the vase. They suggested that this finding indicates that endogenous variations in activity in the prestimulus time period biased subsequent percept. These findings suggest that ongoing slow activity fluctuations have an impact on how we make up our mind during subsequent perceptual inferences from sensory input.

Ronconi et al. (2017) conducted an experiment using EEG and a different type of analysis (e.g. pattern classification defined in the next section) with a similar paradigm

(short presentation times and long ISIs) to study the pre-stimulus activity linked to bistable/multistable perception. They suggest that their findings confirm the notion of the presence of ongoing fluctuations in the pre-stimulus period that biases subsequent perception. They found that there are frequency modulations in the pre-stimulus period that bias subsequent interpretation of the ambiguous stimulus. These modulations occur early on in the pre-stimulus window (~500ms before stimulus onset). They found two significant time windows that are predictive of the upcoming percept in the pre-stimulus period that differed between the two types of ambiguous stimuli they used. Other studies have found frequency modulations that occur at different time windows during the pre- stimulus period as well. For instance, Ehm et al. (2011) found gamma modulations on reversal trials 200ms before stimulus onset. Muller et al. (2005) found changes in EEG activity on reversal trials 300ms before onset. These variations could be due to the different ambiguous stimuli used. However, the common theme in these findings is that they seem to suggest a presence of an unknown intermediate mechanism at later and perhaps even earlier times which are not identified through the analyses conducted and parameters included in these analyses.

Based on these findings, the experimental paradigm used in Experiments 3 (Necker Lattice) and 4 (Face-Vase) is designed in order to investigate a large pre- stimulus window and not the activity occurring during stimulus onset or after it. This would allow to further explore a wider time window than has been investigated in previous studies. It may be that activity is predictive for a long period before reversals rather than just at the moment of presentation.