Detecting the presence of an object is a different process than identifying the object as a particular object. This difference has not been taken into account in designing experiments on the neural correlates of consciousness. We compared the electrophysiological correlates of conscious detection and identification directly by measuring ERPs while participants performed either a task only requiring the conscious detection of the stimulus or a higher-level task requiring its conscious identification. Behavioral results showed that, even if the stimulus was consciously detected, it was not necessarily identified. A posterior electrophysiological signature 200–300 msec after stimulus onset was sensitive for conscious detection but not for conscious identification, which correlated with a later widespread activity. Thus, we found behavioral and neural evidence for elementary visual experiences, which are not yet enriched with higher-level knowledge. The search for the mechanisms of consciousness should focus on the early elementary phenomenal experiences to avoid the confounding effects of higher-level processes.
When you become conscious of a visual object, does it pop into your visual consciousness as a fully formed object with a determinate identity? Or is there first a brief moment of elementary visual experience of the object during which you do not yet know its identity? In the literature on visual categorization, researchers have reached controversial conclusions such as “as soon as you know it is there, you know what it is” (Grill-Spector & Kanwisher, 2005) or that “as soon as you know an object is there, you do not necessarily know what it is” (Mack, Gauthier, Sadr, & Palmeri, 2008). At least in difficult viewing conditions, the mere detection of an object is easier than its full identification (Straube & Fahle, 2011). The standard categorization studies do not, however, reveal what the observer subjectively, consciously, experiences. They focus only on the objectively measured categorization performance, which may occur independently of awareness (Koivisto, Kastrati, & Revonsuo, 2014). In other words, it is not clear whether early during the processing the stimulus is coarsely experienced and sensed, without any finer representations that define what the object is.
The possible solution to this problem has fundamental implications for studies on consciousness, both theoretically and methodologically. The ultimate goal of the neuroscience of consciousness is to reveal the mechanisms by which the physical brain enables conscious, phenomenal experiences (Revonsuo, 2006; Koch, 2004). During the last 20 years, the neuroscience of consciousness has proceeded less ambitiously by searching for the neural correlates of consciousness (NCCs) in the visual system, brain events that co-occur with visual awareness,1 the subjective experience of seeing. Typically, the experiments on visual awareness have contrasted brain activation in response to visual stimuli that enter awareness with activation to physically identical stimuli that do not enter awareness. Functional brain imaging (Bisenius, Trapp, Neumann, & Schroeter, 2015) and electrophysiological recordings (Koivisto & Revonsuo, 2010) have revealed several spatially and temporally distinct NCCs. As a consequence, there is no theoretical agreement even on the very basic question about whether visual phenomenal experiences are generated early in the visual cortex (Koch, Massimini, Boly, & Tononi, 2016; Lamme, 2010) or later in frontoparietal areas (Dehaene, 2014).
Psychophysical evidence suggests that detection and identification are not the same process (Hillis & Brainard, 2007), and brain imaging shows that brain areas are involved differently in these processes (Straube & Fahle, 2011). The distinction between detection and identification has not been taken into account in the search for NCCs, but the operationalization of visual awareness and the task criteria have varied radically in the neuroscience of consciousness as well as in behavioral experiments from simple detection whether a stimulus was presented (e.g., Pins & ffytche, 2003) to full identification of the stimulus (e.g., Dehaene et al., 2001). The results from studies applying tasks that differ in their processing requirements have been interpreted as if the results would tell about the same phenomenon, which may be one of the major sources of the conflicting views on the nature and localization of the underlying mechanisms of consciousness. A systematic attempt to separate the activity related to aware detection, the entrance of the object into phenomenal experience by becoming aware of its qualitative features or qualia, from the activity related to higher-level awareness of the identity or conceptual properties of the object, is lacking. The importance of this distinction is underlined by a behavioral study that found that objectively measured discrimination performance and subjective awareness followed different psychophysical curves depending on whether low-level visual or high-level semantic features had to be identified in the perceptual task (Windey, Gevers, & Cleeremans, 2013). To reveal the processes that correlate with phenomenal experiences, it might be especially important to focus on the processes that correspond to aware detection of the stimulus to separate the earliest, simple phenomenal experiences from the confounding effects of higher-level knowledge involved in identification or from other further processes that are consequences of awareness (Aru, Bachmann, Singer, & Melloni, 2012; de Graaf, Hsieh, & Sack, 2012).
In this study, we compared directly the ERP correlates of visual awareness during detection and identification. ERPs are time-locked with sensory or cognitive events and provide fine-grained information about the temporal dynamics of neural processing while the observer is performing a cognitive task. In the detection task, the observers' awareness of the mere presence or absence of a low-contrast digit stimulus was measured. In the identification task,2 the observers had to identify the digit to decide whether it was smaller or larger than 5. In the contrastive analysis, the difference in ERPs between aware and unaware trials was defined as the electrophysiological correlate of awareness. Within both tasks, we followed the standard procedure in NCC studies and calibrated individually the stimulus intensity such that, in about half of the trials, the participant was aware of the task-relevant feature of the stimulus (i.e., aware of the presence of the stimulus in the detection task and aware of the identity in the identification task). Note that, when the stimulation intensity is calibrated using the identification threshold, awareness of the presence of the unidentified stimulus is still possible: The stimulus remains unaware according to the identification threshold but may be aware according to the detection threshold. A mere detection does not require as much information and as large activation of different brain areas as identification (Straube & Fahle, 2011). Therefore, we expected that the ERP correlates of visual awareness could be dissociated depending on whether awareness is operationalized according to the detection or identification threshold.
Sixteen right-handed students (mean age = 23.4 years, range = 19–30 years, two men) with normal or corrected-to-normal vision participated. The sample size was determined on the basis of our earlier studies that had used similar methods on studying electrophysiological correlates of visual awareness of low-contrast stimuli (Koivisto & Grassini, 2016; Koivisto, Grassini, Salminen-Vaparanta, & Revonsuo, 2016; Koivisto et al., 2008). The experiment was conducted with the understanding and written consent of each participant, in accordance with the Declaration of Helsinki, and accepted by the ethics committee of the Hospital District of Southwest Finland.
The stimuli were the digits 3, 4, 6, and 7 (see Figure 1). They were selected on the basis of a pilot study (n = 5) using the digits 1–4 and 6–9 so that they were approximately equally difficult to detect or identify. The size of the digits, from the viewing distance of 150 cm, was on average 0.8° × 0.4°, and the Weber contrast in the critical low-contrast stimuli was either −0.003, −0.10, or −0.13, depending on each participant's performance in the pre-experimental calibration phase (see below). The stimuli were presented with E-Prime software (Psychology Tools, Pittsburgh, PA) on a 19-in. CRT monitor with a resolution of 1024 × 768 pixels and an 85-Hz screen refresh rate (1 refresh ≈ 12 msec). The luminance of the gray background was 19 cd/m2.
Two tasks (detection task and identification task) were performed by each participant in counterbalanced order. In both tasks, each trial began with a fixation point appearing at the center of the screen for 1750 msec (Figure 1). It was followed by the digit stimulus, or a blank screen in catch trials, for an individually calibrated duration. The exact position of the stimulus was jittered randomly so that it could be centered −0.3° to +0.3° vertically and horizontally from the fixation point. After the stimulus (or blank), the participant responded to the forced-choice task and then gave the subjective rating of awareness. In the forced-choice detection task, the participant had to decide whether the stimulus (digit) was presented, whereas in the forced-choice identification task, the participant decided whether each digit was smaller or greater than 5. The response was given by pressing one of two buttons in the front of the response pad. The response to the forced-choice task was followed by the rating of subjective awareness according to the Perceptual Awareness Scale (Ramsøy & Overgaard, 2004), which was modified for the present purpose such that the rating alternatives in both tasks were (1) “I did not see any stimulus,” (2) “I saw something (but could not identify the stimulus),” (3) “I saw the stimulus almost clearly (and could identify it),” and (4) “I saw the stimulus clearly (and could identify it).” Thus, according to the second alternative, the participant was aware of the presence of the digit but could not identify it; the contrast between ratings 1 and 2 thus is related to aware detection of the object's presence without aware identification of it. According to the third alternative, the participant could subjectively identify the digit; therefore, the contrast between the second alternative and the third alternative is related to awareness of the identity of the object, while the participant is aware of the presence of the stimulus. The definitions of the rating alternatives were made clear for the participants in the instructions and appeared also on the screen in spatial locations that paralleled the locations of the corresponding response buttons at the top of the response pad.
Because the aim of the experiment was to compare two different operationalizations of awareness, the meaning of “aware” and “unaware” depends on the task; therefore, the stimulus duration and contrast were adjusted separately for the detection and identifications tasks. Before the experimental trials of the detection task, the duration (and contrast, if needed) of the critical low-contrast stimulus was calibrated so that about half of the stimuli would not be seen (rating “nothing seen”) and awareness of the presence of the stimulus would be near the subjective threshold for awareness. Similarly, before the identification task, the duration (and contrast, if needed) was calibrated such that, in about half of the critical trials, the participant was not aware of the identity of the stimulus (ratings “nothing” or “something”) and high-level awareness was thus near the subjective identification threshold. During the calibration phases, the instructions and procedure were identical to those in the experimental detection and identification tasks, but in the first calibration block, the duration of the critical low-contrast stimulus was always 4 frame rates (47 msec), and the Weber contrast was −0.01. The stimulus duration was changed after each calibration block according to the participants' ratings of awareness. If they were unaware or aware of less than 45% or more than 55% of the low-contrast stimuli, respectively, the duration was increased or decreased with 2 frame rates for the next block (if the performance then was above 55% or below 45%, the duration was decreased or increased with 1 frame rate). However, if 2-frame duration was too easy or 8 frames were too difficult, the contrast of the stimulus was decreased or increased respectively to the next difficulty level. When the appropriate stimulus duration (and contrast) with about 50% unaware trials was found, the calibration block was repeated once again with the same stimulus intensity to verify that performance remained stable. In the actual experiment, eight participants were tested with the lowest contrast stimuli (all with duration of higher than 2 frame rates; mode = 4 frame rates), and eight participants with the middle contrast stimuli in the detection task (all with 2–4 frame rates). In the identification task, two participants were tested with the lowest contrast (4 and 8 frame rates), 13 with the middle contrast (2–5 frame rates, mode = 2), and one with the highest contrast (4 frame rates).
Both tasks were conducted in two blocks of stimuli, separated by brief resting periods. Half of the participants performed the detection task first (calibration + two experimental blocks), followed by the identification task (calibration + two experimental blocks); half of the participants performed the tasks in the reversed order. Each experimental stimulus block consisted of 80 critical trials, 26 catch trials, and 26 control trials with a longer stimulus duration (10 refresh frames). Each calibration block included 24 critical trials, 8 catch trials, and 8 control trials.
EEG Recording and Data Analysis
EEG was recorded with Ag–AgCl sintered ring electrodes attached to recording cap (EASYCAP GmbH, Germany). The scalp electrodes were placed on international 10–20 system sites Fp1, Fp2, F3, F4, F7, F8, Fz, P3, P4, Pz, C3, C4, Cz, T3, T4, T5, T6, O1, and O2. The reference electrode was on the nose, and the ground electrode was placed in front of Fz. An electrode below the left eye was used for monitoring vertical eye movements and blinks, and an electrode 1.5 cm to the left of the left eye was used for monitoring horizontal eye movements. EEG was amplified (SynAmps, Mesa, AZ) using a bandpass of 0.05–100 Hz, with a sampling rate of 500 Hz. A 50-Hz notch filter was used. The impedance was kept below 5 kΩ. Brain Vision Analyzer (Brain Products GmbH, Germany) was used for offline EEG analyses. The activity between −100 and 0 msec before the onset of visual stimulus was used as baseline. Eye movements were corrected with the Gratton and Coles algorithm (Gratton, Coles, & Donchin, 1983). Trials with artifacts (>70 μV) in any of the electrodes were rejected offline (14% of trials). The data were filtered with 0.1-Hz high-pass and 30-Hz low-pass filters.
In the detection task, we averaged the ERPs separately for trials in which the observers reported “seeing something” (detected) and trials with “seeing nothing” (undetected). In the identification task, the trials with “seeing almost clearly” (identified) and “seeing something” (unidentified) ratings were averaged. Thus, in the detection task, the contrast between aware and unaware trials is related to the awareness of the presence of the stimulus whose identity is not represented in awareness (detected vs. undetected), whereas in the identification task, the contrast between aware and unaware trials is related to awareness of the identity of the stimulus whose presence is represented in awareness (identified vs. unidentified).
The data from four participants were eliminated from the main ERP analyses because after artifact rejection their data did not contain enough trials for computing the ERPs for each condition (at least 20 per stimulus type). On average, each ERP wave in the remaining group (n = 12) was based on 60 trials (SD = 25).
On the basis of previous electrophysiological studies that have used low-contrast stimuli at the limen of visual awareness (Koivisto & Grassini, 2016; Koivisto et al., 2008, 2016; Pins & ffytche, 2003), we expected that the electrophysiological differences between aware and unaware trials should be manifested as an enhancement of negativity of the amplitudes in aware trials relative to unaware ones, onsetting around 200 msec, in occipital and posterior temporal electrodes (this difference has been named as visual awareness negativity, VAN; Koivisto & Revonsuo, 2003). In addition, we expected an enhancement of P3 potential in aware trials compared with unaware ones, peaking around 450 msec over parietal electrode locations (the difference has been named as late positivity, LP; Koivisto & Revonsuo, 2003). Visual inspection of the grand-averaged ERPs confirmed these expectations. Therefore, we analyzed the average amplitudes of ERPs in the time windows of 200–300 and 400–500 msec in occipital (O1, O2), posterior temporal (T5, T6), and parietal (P3, P4) electrodes.
In the detection task, the participants made 46.3% hits (correct “yes” responses; 95% CI [38.1, 54.7]), in the critical stimulus-present trials and 93.8% (95% CI [89.7, 97.9]) correct rejections (correct “no” responses) in the stimulus-absent (catch) trials. In the identification task, 73.9% (95% CI [66.0, 81.8]) of the critical stimuli were correctly responded to. The control stimulus, with a longer duration, was correctly responded to with “yes” responses in 85.4% (95% CI [77.0, 93.8]) of trials in the detection task and with correct judgments of whether the digit was smaller or greater than 5 in 87.9% (95% CI [79.9, 96.0]) of trials in the identification task.
The distribution of the ratings of awareness is presented in Figure 2. In 39.8% (95% CI [32.2, 47.5]) of the critical stimulus-present trials in the detection task, the observers reported having seen something, and in 54.8% (95% CI [47.6, 62.1]) of trials, they rated having seen no stimulus at all. The difference in ERPs between these stimulus types (ratings “something” vs. “nothing”) was defined as the correlate of conscious detection of the presence of the stimulus. In the identification task, the observers rated having seen the stimulus almost clearly (and having been able to identify it) in 31.4% (95% CI [20.8, 42.1]) of the critical trials and having seen something (without being able to identify the stimulus) in 38.2% (95% CI [27.2, 49.2]) of the trials. The differences in ERPs to these stimulus types (“almost clear” vs. “something”) were regarded as the correlates for aware identification. In the catch trials without any stimulus, the observers usually rated having seen nothing (detection task: M = 95.1%, 95% CI [91.4, 98.7]; identification task: M = 90.8%, 95% CI [81.3, 100.0]).
ERPs as a Function of Conscious Detection and Conscious Identification
We performed Task (2: detection vs. identification) × Awareness (2) × Area (3: occipital, temporal, parietal) repeated-measures ANOVAs on the mean amplitudes, separately for the time windows of 200–300 and 400–500 msec. Greenhouse–Geisser corrections were applied when the sphericity assumption was not met. The ERPs for different stimulus types are presented in Figure 3. Figure 4 shows the electrophysiological correlates of awareness more directly in difference waves, which were calculated by subtracting the ERPs to unaware stimuli from those to aware ones. Visual inspection of the ERPs and the differences between aware and unaware conditions suggested that there was first an enhanced negativity (N2 potential) over occipital and posterior temporal areas, followed by a more widespread enhancement of positive amplitudes (P3 potential), which was the strongest in parietal electrodes.
200–300 msec (N2)
In the 200–300 msec time window, awareness had a significant influence on the amplitudes, F(1, 11) = 5.15, p = .044, ηp2 = 0.319, indicating that amplitudes in aware trials were more negative than those in unaware trials. Awareness interacted with Area, F(2, 11) = 9.83, p = .001, ηp2 = 0.472, showing that the negative difference between aware and unaware trials was observed over occipital areas, F(1, 11) = 8.16, p = .016, ηp2 = 0.426, mean parameter estimate = −2.39, 95% CI [−4.23, −0.548], and temporal areas, F(1, 11) = 14.21, p = .003, ηp2 = 0.564, mean parameter estimate = −2.56, 95% CI [−4.05, −1.06], but not over parietal areas, F(1, 11) = 3.35, p = .094, ηp2 = 0.234, mean parameter estimate = −1.180, 95% CI [−2.597, 0.238]. Most importantly, the Awareness × Task interaction was significant, F(1, 11) = 7.09, p = .022, ηp2 = 0.392, which suggests that the electrophysiological correlates depended on whether the detection or identification task was performed. The enhanced negativity to aware stimuli, compared with unaware stimuli, appeared only in the detection task, F(1, 11) = 8.62, p = .014, ηp2 = 0.439, mean parameter estimate = −2.042, 95% CI [−3.57, −0.51], but not in the identification task, F(1, 12) = 0.36, p = .562, ηp2 = 0.031, mean parameter estimate = −0.307, 95% CI [−1.438, 0.824]. In other words, detected stimuli elicited larger negativity than undetected stimuli, whereas there was no difference between ERPs to identified and unidentified stimuli. Thus, only the detection task was sensitive to awareness in this time window. In the detection task, awareness referred to aware detection of the stimulus' presence (“something seen” vs. “nothing seen”), whereas in the higher-level identification task, awareness referred to awareness of the identity (“almost clearly seen [but identified]” vs. “something [unidentified] seen”).
Although awareness of identity was not reflected in the ERPs in the 200–300 msec range, the amplitudes in response to “something seen” trials in the identification task were more negative than those to “nothing seen” trials in the detection task, F(1, 11) = 8.07, p = .016, ηp2 = 0.423, mean parameter estimate = −1.88, 95% CI [−3.34, −0.42], suggesting that awareness of the presence of the stimulus was associated with enhanced negativity also in the high-level task. The point here is that, if awareness would have been operationalized only as the ERP difference between identified and unidentified stimuli, we would have missed the negative enhancement (VAN) as a correlate of visual awareness.
400–500 msec (P3)
In this time window, a large-scale enhancement of positive amplitudes (P3) in response to aware stimuli, compared with unaware stimuli, was observed. The main effect for Awareness was significant, F(1, 11) = 24.20, p < .001, ηp2 = 0.687, showing that aware stimuli elicited larger positivity than unaware stimuli. This difference was modified by the Awareness × Area interaction, F(2, 22) = 14.20, p < .001, ηp2 = 0.563, which indicated that, although the difference between aware and unaware trials was the largest over the parietal lobe, F(1, 11) = 31.10, p < .001, ηp2 = 0.739, mean parameter estimate = 3.79, 95% CI [2.29, 5.28], it was present also over the occipital lobes, F(1, 11) = 15.08, p = .003, ηp2 = 0.578, mean parameter estimate = 2.57, 95% CI [1.11, 4.01], and the temporal lobes, F(1, 11) = 24.48, p < .001, ηp2 = 0.690, mean parameter estimate = 2.82, 95% CI [1.57, 4.08]. The Awareness × Task interaction was not statistically significant, F(1, 11) = 0.46, p = .513, ηp2 = 0.040, suggesting that the task did not influence the ERP difference between aware and unaware trials. However, the three-way interaction between Awareness, Task, and Area was significant, F(1.30, 14.35) = 4.29, p = .048, ηp2 = 0.281, but further analyses separately for each area did not find any source for the interaction as they did not reveal any effect for the task on the amplitudes as a function of awareness (parietal: F(1, 11) = 1.32, p = .276, ηp2 = 0.107; occipital: F(1, 11) = 0.33, p = .580, ηp2 = 0.028; temporal: F(1, 11) = 0.02, p = .893, ηp2 = 0.002). Thus, in the 400–500 msec range, awareness correlated with a wide enhancement of positivity over large brain areas, and this effect was observed both in the detection task, F(1, 11) = 12.31, p = .005, ηp2 = 0.528, mean parameter estimate = 3.43, 95% CI [1.28, 5.58], and in the identification task, F(1, 11) = 16.51, p = .002, ηp2 = 0.600, mean parameter estimate = 2.68, 95% CI [1.23, 4.13].
Correlates of Aware Detection as a Function of Task
In the identification task, the amplitudes in the trials in which the participants subjectively identified the stimulus (“almost clearly seen”) did not differ in the 200–300 msec time window from those in which they did not subjectively identify it (“something seen”), whereas both of these trial types were associated with enhanced negativity in relation to unaware (“nothing seen”) trials in the detection task. This suggests that also the correlate of aware detection (enhanced negativity at 200–300 msec) was present when the stimulus was processed up to high level in the identification task. However, it remains to be shown directly, by contrasting “something seen” and “nothing seen” trials, that the enhanced negativity appears as a correlate of aware detection in the identification task, similarly as it appears in the detection task. We calibrated before the experimental trials the difficulty of the identification task for each participant such that they would subjectively be aware of the identity of the stimulus in about 50% of the critical trials. Because there was variability in the distribution of the ratings of awareness (Figure 2), particularly in the identification task, seven participants had a sufficient number of trials (>20) for computing ERPs in the identification task also for “nothing seen” trials. This allowed us to compare the correlates of aware detection (“something” vs. “nothing”) between the tasks (Figure 5). Five of the participants were from the group used in the above analyses, and two were participants who were not included in those analyses as they lacked sufficient number of trials in the “almost clear” ratings.
We conducted an Area (3) × Task (2) × Awareness (2: “detected” vs. “undetected”) ANOVA on mean amplitudes in the 200–300 msec time window. It revealed an Area × Awareness interaction, F(2, 12) = 13.89, p = .001, ηp2 = 0.698, suggesting that the effect of awareness depended on the area. Although awareness did not modify amplitudes in the parietal electrodes, F(1, 6) = 0.20, p = .663, ηp2 = 0.034, mean parameter estimate = −0.15, 95% CI [−1.61, 1.58], aware trials elicited larger negativity than unaware trials in occipital electrodes, F(1, 6) = 7.31, p = .035, ηp2 = 0.549, mean parameter estimate = −1.74, 95% CI [−3.32, −0.17], and in temporal electrodes, F(1, 6) = 14.36, p = .009, ηp2 = 0.705, mean parameter estimate = −2.24, 95% CI [−3.69, −0.793]. Awareness did not interact with Task, F(1, 6) = 0.76, p = .417, ηp2 = 0.112, and the Area × Task × Awareness interaction was not statistically significant, F(1, 12) = 0.59, p = .567, ηp2 = 0.090. Thus, the enhanced negativity reflecting aware detection of the stimulus was similar in both tasks.
Corresponding analysis on the 400–500 msec time window (P3) showed that aware stimuli were associated with larger positivity than unaware trials, F(1, 6) = 15.52, p = .008, ηp2 = 0.721, and this difference depended on the area, F(1, 6) = 35.61, p < .001, ηp2 = 0.856: The largest effect was measured in parietal electrodes. The most interesting finding was that the enhancement of positivity in aware trails, compared with unaware ones, was influenced by the task, F(1, 6) = 7.08, p = .038, ηp2 = 0.541. There was a large effect for Awareness in the detection task, F(1, 6) = 19.61, p = .004, ηp2 = 0.766, mean parameter estimate = 5.16, 95% CI [2.31, 8.01], and a smaller one in the identification task, F(1, 6) = 4.99, p = .067, ηp2 = 0.456, mean parameter estimate = 2.24, 95% CI [−0.21, 4.70]. In other words, when the forced-choice task was to identify the digit, the correlate of detection awareness reduced in the P3 range, compared with the detection task in which detecting the presence of the stimulus was directly relevant for the task. This finding is in line with the proposal that the P3 reflects processing of the attended-to features that are relevant for performing the task (Pitts, Metzler, & Hillyard, 2014; Pitts, Padwal, Fennelly, Martínez, & Hillyard, 2014; Koivisto & Revonsuo, 2008b). However, one should note that, in all the trials, the forced-choice response was followed by the subjective rating task, and therefore, also the information in relation to the rating categories in the rating scale was task relevant. This may explain why the difference in P3 between undetected (“nothing” seen) and detected (“something” seen) trials did not completely disappear in the identification task.
Next, we analyzed the mean amplitudes to unaware critical stimulus-present trials (“nothing seen” rating in the detection task) and to stimulus-absent catch trials with “nothing seen” ratings in the detection and identification tasks (n = 12; Figure 6). As the rated awareness was constant across these three stimulus types, we should observe a difference between stimulus-present and stimulus-absent trials if the amplitudes were sensitive to unconscious processing of the stimulus.
The Type (3: unaware critical stimulus-present trials, catch trials in the detection task, catch trials in the identification task) × Area (3: occipital, temporal, parietal) ANOVA in the 200–300 msec range (N2) did not find a main effect for Type, F(2, 22) = 1.03, p = .379, ηp2 = 0.084. The Type × Area interaction was statistically significant, F(2, 22) = 9.44, p < .001, ηp2 = 0.462, but further analyses did not reveal any statistically significant effect for Type over the parietal areas, F(2, 22) = 0.54, p = .588, ηp2 = 0.047, temporal areas, F(2, 22) = 0.83, p = .451, ηp2 = 0.070, or occipital areas, F(2, 22) = 0.78, p = .471, ηp2 = 0.066. Thus, the source of the interaction remained unclear. Figure 6 reveals, however, that the mean amplitudes in unaware stimulus-present trials tend to be more positive than those in the no-stimulus trials in the 200–300 msec range, suggesting that unconscious processing in this time range is not manifested as enhanced negativity.
The comparison of the stimulus types in the 400–500 msec time window (P3) did not find any main effect for Type, F(2, 22) = 0.61, p = .521, ηp2 = 0.053, or interaction between Type and Area, F(2, 22) = 0.66, p = .626, ηp2 = 0.056. Thus, in line with earlier findings with low-contrast stimuli, the amplitudes in the studied time ranges were not sensitive to unconscious processing (Koivisto & Grassini, 2016; Koivisto et al., 2008).
We studied whether the electrophysiological correlates of visual awareness differ between visual detection and identification. The intensity of the stimuli was calibrated at the threshold of awareness in both tasks so that about half of the stimuli were consciously detected or identified in the detection and identification tasks, respectively.
The behavioral results from the detection task are in line with the existence of a low-level phenomenal experience, which does not contain sufficient information for identification. In 55% of the critical trials, the observers did not report seeing anything; in 40%, they saw something; and only in 5% of the trials did they subjectively report having seen the stimulus so clearly that they were able to identify it. Thus, the observers were in 95% of the trials either not at all aware of the presence of the object or they were only aware of “something” that they could not consciously identify, suggesting that, when the representation of the stimulus crossed the subjective threshold of awareness, the object could not necessarily be identified. Consistent with our casual impressions that sometimes we see something without knowing what it is, we conclude that aware detection is possible without aware identification; simple phenomenal experiences do emerge as such, but they are not sufficient for aware identification of the stimulus. The electrophysiological recordings allowed us to track online the time courses of processing during the trials that led to aware detection and identification.
The electrophysiological results revealed two different components that correlated with awareness. They were the enhanced posterior negativity in the N2 time window, 200–300 msec after the stimulus onset, and the widespread amplification of P3 around 400 msec. The amplitude differences between aware and unaware conditions have been called VAN in the earlier N1/N2 time window and LP in the P3 range, to separate the components that are related to awareness from the overall amplitudes of N1/N2 and P3. Both components have been observed in earlier studies on visual awareness with different manipulations of awareness (for a review, see Koivisto & Revonsuo, 2010) and across different types of stimuli (Rutiku, Aru, & Bachmann, 2016).
The electrophysiological results dissociated the correlates of aware detection from those of aware identification. Aware detection correlated with the posterior negative difference (VAN) 200–300 msec after stimulus onset between “seeing nothing” (unaware) and “seeing something” (aware). However, it is not entirely clear what the participants saw when they indicated on the Perceptual Awareness Scale (Ramsøy & Overgaard, 2004) that they saw “something.” This rating might correspond to awareness of oriented edges, some kind of figure-ground segregation, or change in luminance. In any case, VAN seems to reflect awareness of low-level visual properties. The amplitudes in this latency range were not sensitive to the difference that was related to aware identification (“seeing something” vs. “seeing almost clearly [what the digit was]”). Thus, the elementary phenomenal experiences of “seeing something,” without awareness of the higher properties of the stimulus, had a unique correlate (VAN) that was not present at the identification threshold during 200–300 msec. Only the later large enhancement of positivity around 400 msec (P3) correlated with being aware of the identity. The low-level awareness may correspond to phenomenal consciousness (Revonsuo, 2006; Block, 2001), the emergence or mere presence of visual sensations (“qualia”), which do not require any further cognitive operations such as categorization, naming, and so forth. In aware identification, which was indexed by P3/LP, the content is cognitively categorized, named, or labeled, by applying concepts and cognitive operations to it. These processes are cognitive and conceptual, hence abstract in relation to stimulus modality and not purely visual–phenomenal, that is, the crucial qualitative difference between what can be called phenomenal consciousness (required by mere detection) and reflective/access consciousness (required by identification; Revonsuo, 2006; Block, 2001) or between “phenomenology” and “cognitive accessibility” (Block, 2007). Thus, “detection” in our study corresponds to seeing and knowing that one is seeing something, that is, to phenomenal consciousness, and “identification” and ability to cognitively compare the “seen” contents correspond to how Block (2007) characterizes access consciousness.
Although the contrast between identified and unidentified trials in the identification task did not reveal the enhanced posterior negativity at 200–300 msec (VAN), it was observed in the identification task when the “nothing seen” and “something seen” trials were compared in a subgroup of participants (n = 7) who produced a sufficient number of such ratings for analyzing both stimulus types. In other words, the correlate of aware detection appeared also in the identification task. Within the same subgroup, the late positive difference (LP) in P3 between detected and undetected stimuli was larger in the detection task than in the identification task (Figure 6), suggesting that the later component depends on the relevance of the information to the task. Thus, the later component reflects further aware processing of the attended-to features that are relevant for performing the task (Pitts, Metzler, et al., 2014; Pitts, Padwal, et al., 2014; Koivisto & Revonsuo, 2008b), rather than the earliest stage of phenomenal experience. As a whole, the electrophysiological results suggest, consistent with the coarse-to-fine framework (Campana & Tallon-Baudry, 2013) and Bachmann's (2000) idea of phenomenal microgenesis, that awareness unfolds as a function of time from coarse phenomenal sensation or gist to a full-blown awareness of the object as a specific token of a certain category of objects.
One should note that, in the identification task, the stronger cortical activity, in relation to unaware detection, was marked by N200 in both aware and unaware identification conditions. Hence, VAN correlated with consciousness only when “something” and “nothing” were compared, whereas the enhancement of P3 (i.e., LP) was correlated with awareness both in the detection and identification tasks. A counterargument for the account that the earlier negative component (VAN) reflects awareness would state that it is not directly related to the neural processes that constitute awareness, but it reflects preconscious processes that enable the stimulus to reach awareness (Lamy, Salti, & Bar-Haim, 2009; but see Koivisto & Grassini, 2016). In those trials in which the preconscious processes work inefficiently, the stimuli fail to reach awareness and do not produce this component. The late positive enhancement of P3 (i.e., LP) thus would be the proper correlate of awareness. However, if this alternative account were true, the comparison of the unaware trials (“nothing seen”) and catch trials without any stimulus (“nothing seen”) should have revealed enhanced negativity that is related to preconscious processing of the physical difference in the 200–300 msec time window, but no such activity appeared in the contrast between unaware stimulus-present trials and stimulus-absent trials. In fact, an interaction pointed toward the opposite direction (weaker N200 in stimulus-present trials than in stimulus-absent trials), which speaks against the explanation that the lack of difference was simply due to lack of power. Thus, the view that VAN represents a preconscious stage whereas LP represents the proper correlate of visual awareness seems a post hoc argument that is hard to defend in the face of the converging evidence suggesting that LP is a postperceptual component that follows the onset of aware perception. Railo, Revonsuo, & Koivisto (2015) used TMS-EEG and found in a single-trial analysis that VAN preceded participants' reports of awareness, whereas LP overlapped or followed the reports. Koivisto et al. (2016) used go/no-go versions of reporting awareness or unawareness of a low-contrast stimulus and found that the method of reporting subjective awareness influenced the amplitude of P3/LP. They observed also a negative-going shift in go trials over the motor cortex during the P3 time window. This later effect suggests that motor response preparation (related to reporting subjective awareness) occurred during the P3 time range. In addition, P3/LP is decreased in response to stimuli that enter awareness but are not task relevant, which was replicated also in this study, and this decrease can be observed even when the stimuli are clearly visible (Pitts, Metzler, et al., 2014; Pitts, Padwal, et al., 2014; Koivisto & Revonsuo, 2008b). Moreover, LP is reduced or eliminated when the to-be-recognized (aware) stimuli are repeated (Koivisto & Revonsuo, 2008a) or when they can be expected on the basis of a previous trial (Melloni, Schwiedrzik, Müller, Rodriguez, & Singer, 2011), suggesting that LP is related to updating of working memory rather than to awareness. Even if one would accept that LP is a postperceptual component that correlates with higher processes that occur after the onset of awareness, it does not necessarily follow that VAN would be the direct correlate of awareness. The account that VAN is a preconscious prerequisite of awareness still remains a possibility. However, in that case, it remains open whether there exists any genuine ERP correlate for visual awareness at all.
Theoretically, the results help to integrate the current theories of visual awareness. The theories can be roughly divided into early and late theories depending on their assumptions about the onset of earliest phenomenal experiences and their localization in early or higher brain areas. In recurrent theories (e.g., Lamme, 2010), phenomenal experiences arise relatively early when the feedforward signals from the primary visual areas have traveled to higher areas and the higher and lower visual areas engage in local recurrent interactions. The relatively early onset and the occipito-temporal distribution of the negativity (VAN) related to aware detection of object's presence are consistent with this account. In late theories, for example, in the global neuronal workspace theory (Dehaene, 2014), awareness arises relatively late when the representation of the stimulus gets access to neuronal workspace, which is implemented in prefrontal and parietal circuits and makes the representation globally available for various processes such as evaluation, memory, planning, intentional action, and verbal or motor report. Interestingly, the P3/LP component has been considered as the electrophysiological marker of “global ignition” (Dehaene, 2014; Lamy et al., 2009; Del Cul, Baillet, & Dehaene, 2007), activation of the frontoparietal network without which no phenomenal experiences will ever emerge according to this theory. Our results suggest that the local recurrent processing in visual areas may be sufficient for low-level visual awareness (i.e., phenomenality), whereas awareness of higher-level properties of visual stimuli might depend on the more extensive global recurrent interactions in the neuronal workspace (reflective consciousness or cognitive access). In other words, the early theories may apply to the phenomenality itself, whereas the late theories may be about higher-level (nonvisual) cognitive processes in awareness. The transition from the first phenomenal experiences to full awareness thus evolves via the recurrent activations between visual areas to the global “ignition” of the frontoparietal global workspace. This converges with the results of a recent review (Koch et al., 2016), which concluded on the basis of different types of data (e.g., EEG, brain imaging, lesion studies) that phenomenal experiences are generated in the posterior hot zone involving temporo-parieto-occipital areas, whereas the frontoparietal network is involved in attention and task execution, monitoring, and reporting.
The account that phenomenal experiences arise early in the visual cortex does not imply that each modality would possess a microconsciousness (Zeki, 2003), with functionally specialized areas producing their own restricted type of consciousness locally. In our background theory (Revonsuo, 2006, 2010), the unity of consciousness presupposes only that all phenomenal experiences appear within a unifying spatial framework. This spatial framework is partially missing, for example, in patients experiencing unilateral spatial neglect; therefore, stimuli in the neglected field cannot be detected and cannot enter the patient's phenomenal consciousness. In healthy participants, the spatial framework is weakened by strong bilateral manipulations of spatial attention; accordingly, VAN is not observed for spatially nonattended stimuli (Koivisto & Revonsuo, 2010; Koivisto, Kainulainen, & Revonsuo, 2009).
Methodologically, our results stress the importance of defining clearly what the concept of “awareness” exactly refers to in each study and how it is operationalized in the experiments. Although visual awareness is usually defined as the “subjective experience of seeing,” it is important also to define what the term “seeing” in this definition means. The present findings show directly that a different pattern of results is obtained depending on whether “seeing” refers to aware detection of the presence of the object or to awareness of the higher-level properties of the object. If consciousness research focuses on higher-level awareness of objects, which have already entered conscious experience, it may be impossible to reveal the mechanisms that are responsible for the subjective phenomenal experiences—in such case, the baseline, against which conscious conditions are compared, already involves consciousness. Instead of uncovering the mechanisms of phenomenal subjectivity or qualia, such research reveals the mechanisms involved in pattern analysis, categorization, identification, naming, and other higher processes, which are not directly related to subjectivity and phenomenality, the central landmarks of awareness. For example, if the perceptual object is presented in a complex background, all the visual elements in the scene may be phenomenally experienced without identification of the object; thus, the “unaware” baseline already includes phenomenal awareness, and the experiment may not be sensitive to phenomenality itself but to higher-level awareness requiring pattern analysis. In such cases (e.g., Lamy et al., 2009), only the late higher-level neural correlates may be revealed. On the other hand, it is also possible to confound low- and higher-level correlates of awareness by using the identification task. For example, one may use an identification task without an accompanying measure of awareness and contrast electrophysiological responses to unidentified (“unconscious”) and identified (“conscious”) stimuli. In such case (e.g., Wilenius-Emet, Revonsuo, & Ojanen, 2004), the unconscious baseline is likely to involve also stimuli that were not even detected, and therefore, a VAN may emerge.
In conclusion, we have shown that an elementary form of phenomenality, aware detection of object's presence, has an earlier, more posterior electrophysiological correlate than aware identification of the object. The present evidence converges on the view that full perceptual awareness evolves during the first half-second from the mere phenomenal experience of the object's presence to a richer representation of the features of the object and to its full recognition (Campana & Tallon-Baudry, 2013; Bachmann, 2000). In addition, after the stimulus has been recognized, the information is propagated to medial temporal lobes where it activates concept cells and becomes associated with other sensory-independent conceptual and episodic information, giving rise to contextual or conceptual awareness, as opposed to perceptual awareness (Navajas, Rey, & Quian Quiroga, 2014). Although investigation of high-level awareness is important in itself, research that ultimately aims to attack the hard problem of revealing the neural mechanisms of phenomenal experiences (i.e., consciousness) should focus on the neural mechanisms of the earliest, most elementary experiences sometimes also called “qualia.” Such research would focus on consciousness in its purest form, without confounding it with higher cognitive processes.
This study was supported by the Academy of Finland (Project no. 269156).
Reprint requests should be sent to Mika Koivisto, Department of Psychology, University of Turku, Assistentinkatu 7, 20014 Turku, Finland, or via e-mail: email@example.com.
Terms consciousness, awareness, and phenomenal experience are used interchangeably.
We assume that, once the stimulus has been identified, deciding whether it is smaller or larger than 5 is an easy and trivial task for our participants (university students). Therefore, the task is primarily an identification task.