Conscious and unconscious expectancy effects: A behavioral, scalp and intracranial electroencephalography study

Highlights

• Both supraliminal and subliminal cues can affect response times and modulate brain activity.

• Supraliminal CNV originates from temporal and mesio-frontal (Supplementary Motor Areas) neural sources.

• Subliminal CNV originates from temporal lobe sources.

Abstract

Objective

The scope of unconscious cognition stretched its limits dramatically during the last 40 years, yet most unconscious processes and representations that have been described so far are fleeting and very short-lived, whereas conscious representations can be actively maintained in working memory for a virtually unlimited period. In the present work we aimed at exploring conscious and unconscious lasting (>1 second) expectancy effects.

Methods

In a series of four experiments we engaged participants in the foreperiod paradigm while using both unmasked and masked cues that were informative about the presence/absence of an upcoming target. We recorded behavioral responses, high-density scalp EEG (Exp. 2a), and intra-cranial EEG (Exp. 2b).

Results

While conscious expectancy was associated with a large behavioral effect (~150 ms), unconscious expectancy effect was significant but much smaller (4 ms). Both conscious and unconscious expectancy Contingent Negative Variations (CNVs) originated from temporal cortices, but only the late component of conscious CNV originated from an additional source located in the vicinity of mesio-frontal areas and supplementary motor areas. Finally, only conscious expectancy was accessible to introspection.

Conclusions

Both unmasked and masked cues had an impact on response times and on brain activity.

Significance

These results support a two-stage model of the underlying mechanisms of expectancy.

Introduction

During the last decades, a large range of unconscious cognitive processes, inaccessible to conscious report, have been discovered and characterized in conscious subjects using various paradigms such as visual masking, attentional blink or inattentional blindness (Dehaene et al., 2001, Dehaene et al., 1998, Kouider and Dehaene, 2007, Mack and Rock, 1998, McCormick, 1997, Naccache, 2006, Sergent et al., 2005, Van Gaal et al., 2008). Thereby, the combination of behavioral and brain activity measurements revealed the existence of unconscious perceptual, semantic, executive, motor and even emotional cognitive representations, correlated with the activity of various cortical areas and networks. The scope of unconscious cognition rapidly stretched its limits so dramatically, that the classical search for clear differences between conscious (reportable) and unconscious (unreportable) cognitive processes turned unexpectedly to a highly challenging issue. Which are the mental operations, - if any -, that require conscious mentation? Within this scientific context, one potential candidate that is still in the running deals with time. Indeed, most unconscious active representations and cognitive processes that have been described so far are fleeting and very short-lived (usually within a few tens or hundreds of milliseconds), whereas conscious representations can be actively maintained in working memory for a virtually unlimited period.

Time-related cognitive processes that can be viewed as the launching of an internal timer mechanism are of special interest. Among these processes, the expectancy of upcoming relevant stimuli is especially noteworthy because it can be measured both using reaction times (RTs) and EEG signals. Indeed, a rich literature demonstrated that expectancy is associated with the Contingent Negative Variation (CNV) ERP component (Macar and Vidal, 2004, Pfeuty et al., 2005, Walter et al., 1964). Typically, in the foreperiod paradigm a warning stimulus (S1) is followed by an imperative stimulus (S2). Once the contingency between S1 and S2 is learned after a few trials, the presentation of S1 triggers an expectancy towards S2, visible as a slow negative potential gradually increasing until the presentation of S2. This CNV then terminates abruptly right after the appearance of S2 (Walter et al., 1964). Importantly, the CNV temporally bridges S1 and S2 presentation over several seconds. Very few works investigated the relations prevailing between consciousness and expectancy (Capa et al., 2013, Faugeras et al., 2012, Hamon et al., 1994, Sergent et al., 2017, Yasuda et al., 2011).

A recent study reported that CNVs elicited consciously could be modulated by masked monetary reward cues (Capa et al., 2013) presented at the beginning of long runs of trials. Unconscious monetary cues seemed to induce changes in motivation, resulting in modifications of the CNV amplitudes. These results were interpreted as a long-lasting influence of monetary cues on motivation thus resulting in increased CNVs. Thus, unconscious monetary cues probably influenced the CNV indirectly. To our knowledge, no study has ever directly tried to probe the triggering of a CNV by subliminal stimuli or to directly influence an ongoing CNV with masked cues.

The CNV seems to be affected by levels of arousal. Indeed, subjects who had an undisturbed night of sleep show stronger CNVs than participants who were disturbed during their sleep (Yamamoto et al., 1984). These results suggest that arousal might be a key requirement to elicit a CNV. Hamon and colleagues (1994) tested the possibility of initiating anticipatory attention during sleep. They observed CNVs in awake subjects, and during rapid eye movement sleep (REM sleep), but noticed that the CNV was absent in deep sleep. Yasuda et al. (2011) also found an absence of CNV in deep sleep and a decrease of CNV amplitude with sleep onset latency. All this evidence points toward the fact that consciousness might be necessary for endogenous anticipatory temporal attention.

The presence of CNV was also probed in patients with disorders of consciousness (DOC). In a study on two comatose patients, Dolce and Sannita (1973) found that after presenting paired auditory stimuli to these patients more than a hundred times, they observed a negative shift in the EEG between S1 and S2 resembling a CNV. The present work actually stems from a set of studies by our group in which we incidentally discovered the presence of a CNV in an auditory paradigm that was associated with a previously reported neural signature of conscious access. Indeed we recently tested DOC patients with the ‘local-global’ paradigm to probe brain responses to violations of auditory regularities (Faugeras et al., 2012). We observed that patients showing a CNV were more prone to detect violations of both local (intra-trial) and global (inter-trial) regularities. Moreover, a CNV could be observed both in some patients in minimally conscious state (MCS), and in some patients in vegetative state (VS). Similarly, we also observed a CNV in three out of four clinically VS patients in a more recent work (Sergent et al., 2017). Hence, some form of temporal expectancy might also occur in the absence of consciousness. Note however that the top-down modulation of CNV seemed to require a minimal level of consciousness. These questions also motivate the present work in a translational perspective. Indeed, defining the diagnostic and prognostic values of CNV signatures in DOC patients could be valuable in these very uncertain clinical conditions. However, one should be aware of the potential multiple differences between such an unconscious cognitive process that occurs in a subject who is in a conscious state, and an unconscious cognitive process of expectancy that occurs in a subject who is in a non-conscious state. In order to begin to address this difficult issue, we aimed here at exploring expectancy in conscious participants as a function of their access to the information eliciting the expectancy process.

In the present work, we characterized conscious and unconscious expectancy effects in a foreperiod paradigm, and tested the hypothesis that short-lived unconscious representations elicited by subliminal cues could still trigger long-lasting CNV-associated processes and affect behavior. To this aim, we designed a series of masked and unmasked cueing experiments in which we explored the possibility of unconsciously initiating a sustained expectancy effect. Through four complementary experiments using behavioral measures, high-density EEG and intra-cranial recordings, we compared conscious and unconscious expectancy effects, and demonstrated that: (i) consciously perceived cues elicit large behavioral expectancy effects (effect size > 150 ms) that are accessible to introspection; (ii) conscious expectancy is correlated with typical CNVs, the generators of which are located in temporal lobes (early component) and in the supplementary motor areas and related areas (late component); (iii) subliminal cues that are not consciously reported can elicit small (effect-size ~ 5 ms) behavioral effects for values of cue-stimulus onset asynchrony (SOA) superior to one second; (iv) these unconscious expectancy effects do not seem to be accessible to introspection, (v) and they are associated with a Cz-centered CNV, the generators of which seem to be confined to the temporal lobe.