To see or not to see: the interaction between
information processing and cortical state to
generate consciousness
Roozendaal, D.H.M
___________________________________________________________________________ Abstract
Consciousness is a widely-used term, but actually consists of two different concepts: conscious content; the awareness of presented sensory information, and conscious state; the intrinsic neural fluctuations of the brain which determine whether conscious content is actually possible. In general, studies of consciousness only focus on conscious content, while the role of conscious state is often neglected. This review will focus on how dynamics of the brain are established, how sensory information is integrated in relation to the dynamics of the brain, and what neuromodulatory mechanisms regulate these dynamics. To the best of my knowledge, this is one of the first reviews that combines findings from different disciplines and various techniques to discuss the exact relationship between conscious state and conscious content.
___________________________________________________________________________
Student number: 10001424
MSc in Brain and Cognitive Sciences, track cognitive neuroscience Literature thesis, 12 ECTS
Supervisor: Dr. Simon van Gaal Co-assessor: Dr. Filip van Opstal 14/10/2016 – 23/12/2016
2 Table of contents
1. Introduction 3
1.1 Current theories of consciousness 3
1.2 Cortical state in relation to conscious content 4 2. What are the dynamics of cortical state during sleep? 5
2.1 Non-REM sleep and slow-wave sleep 6
2.2 REM sleep 7
2.3 Synchronized and desynchronized cortical state 8 3. How do neuromodulatory circuits influence cortical state? 9
3.1 Cholinergic ascending projections 9
3.2 The locus coeruleus – noradrenaline system 11
3.3 Regulation of pupil diameter 12
4. How does cortical state influences sensory perception? 14 4.1 Sensory processing is cortical state dependent 14 4.2 Variability in cortical responsiveness predicts sensory perception 15 4.3 Cortical desynchronization facilitates conscious content 16 4.4 Cortical synchronization attenuates conscious content 17 4.5 Pupil diameter as a surrogate measure for cortical state 18
5. Discussion 20
3
1. Introduction
If someone walks into your office, while you are highly concentrated on reading these words, it is plausible that you do not notice the person entering. What happens in the brain that generates conscious perception of these words, while leaving other environmental influences unimportant? Consciousness is a widely-used term, but actually consists of two different concepts: conscious content and state (figure 1). Conscious content refers to the awareness of presented sensory information, like in the previous example when you are aware of these words but
not of the person entering the room. Conscious state refers to the intrinsic, neural fluctuations of the brain during the course of the day and determines whether conscious content is actually possible. Patients in coma or deep sleep are considered nonconscious, since no conscious content is possible. In the course of the day, the internal dynamics of the brain fluctuate. These ongoing, spontaneous fluctuations are often referred to as cortical state. Cortical state is not necessarily determined by sensory input but rather due to interactions of external stimuli with spontaneous activity, or solely the latter (Harris & Thiele, 2011). Arousal, attention and awareness can subsequently modulate cortical state and enhance sensory perception. The question remains how this modulation
is exactly established and how this enhances conscious content. And what is different from awake, conscious state compared to deep sleep and coma, which underlies the differences in conscious content? These questions are the main focus of this review and will be discussed with different techniques and within different disciplines.
1.1 Current theories of consciousness
An influential idea of how consciousness is generated is the global workspace theory (Baars, 2005). This theory proposes that conscious perception is generated when sensory information enters a so called “global workspace”. Once in the workspace, consciousness has an integrative function in distributing sensory information to higher cortical areas (figure 2). To enter the global workspace and thus ignite consciousness, sensory information needs to have sufficient strength and top-down attention. The necessity of sufficient stimulus strength to generate conscious content can be explained by masking experiments. In a typical masking paradigm, a sensory target stimulus is briefly presented, and quickly preceded and followed by other stimuli. This masking of the target stimulus results in reduced or even imperceptible experience. Conscious perception of the target stimulus is therefore affected by reducing sensory stimulus strength with surrounding stimuli (Fahrenfort et al., 2007). A second criteria to generate conscious perception according to
Figure 1. Simplified illustration of the two underlying concepts of consciousness. Conscious content (y-axis) as a function of conscious state (x-axis) to illustrate different levels of consciousness. Coma is accompanied by the minimum level of conscious state and content, where wakefulness has the maximum of both (image from Laureys, 2005).
4
the global workspace theory, is sufficient top-down attention. Top-down attention refers to attention to a particular spatial or temporal location which selectively gates the focus of interest (Moran & Desimone, 1985). Inattentional blindness is an elegant example of the necessity for top-down attention in establishing conscious content. Mack and Rock (1998) asked subjects to report changes in a visual discrimination task while fixating at the center of the screen. Changes of substantial strength and duration occurred at different locations than attention was located. A large percentage of the subjects failed to report the unattended and unexpected changes, indicating that top-down attention is necessary for conscious content.
In addition to the global workspace theory, Lamme and Roelfsema (2000) described the neural cortical connections involved in conscious perception with the recurrent and feedforward processing model of sensory information. After stimulus presentation, sensory information enters the brain through the thalamus and is rapidly forwarded from the primary sensory cortices to higher cortical areas. Neurons in this “feed-forward sweep” remain active after forwarding information. During these latencies, feedback information from higher cortical areas propagates back along horizontal and recurrent connections and can be integrated in the activated neurons of the feed-forward sweep. This involvement of recurrent projections is thought to establish conscious content.
1.2 Cortical state in relation to conscious content
The so far described theories, the global workspace theory and the recurrent and feedforward processing model, suggest that conscious content depends on the recurrent integration of information from higher and local cortical areas, and that this integration is achieved according to an all-or-none mechanism. However, nonconscious information is still processed in the brain as evidenced by several
neuroimaging and electrophysiological studies (Kakigi et al., 2003; Kim & Blake, 2005; Kouider et al., 2014). Since these studies mainly focus on conscious content, e.g. if certain stimuli result in awareness, it is interesting to study the relation with conscious state. Perhaps several intrinsic processes cooperate and interact, preceding sensory stimuli presentation, to generate conscious content (Dehaene et al., 2014).
The role of cortical state is often neglected in studies of consciousness. Sleep is a useful starting point to investigate the underpinnings of cortical state. The dynamics of the brain fluctuate during the transition from wakefulness to sleep and within different sleep stages, even in the absence of sensory input. Cortical state during sleep is regulated by various, ascending neuromodulatory systems located in the brainstem, hypothalamus and basal forebrain, which release numerous neurotransmitters to the cortex and thalamus (Steriade et al., 1993). The relationship between neuromodulatory systems and conscious
Figure 2. Schematic overview of the global workspace theory. If sensory information has sufficient strength and top-down attention, it enters the global workspace. Once in the workspace, sensory information is forwarded into the brain and consciousness is generated (image adapted from Dehaene et al., 2003).
5
content during wakefulness can be investigated with drugs and anesthetics. These substances function as transmitter agonists, which enhance transmitter secretion, or antagonists, which block transmitter secretion. However, the exact relationship between neuromodulatory transmitters and conscious content remains largely unknown (Self et al., 2012).
In this paper, I will describe how fluctuations in cortical state interact with external, sensory input to generate conscious content. I start with describing the fluctuations in cortical state during sleep and how these are associated with different levels of conscious content. The well-characterized cortical rhythms at different sleep stages provide useful information regarding cortical state during wakefulness and in understanding the neural correlates of cortical state. Second, I will explore the neurobiology of conscious state: how cortical fluctuations are established and in what fashion modulatory neurotransmitters exactly contribute to sensory integration, recurrent processing and cortical state. With these guidelines, I try to investigate the influence of cortical state on sensory information processing during daytime. I will describe studies which focused on cortical state prior to stimulus presentation. This way, both conscious content and conscious state will be taken into account when consciousness is discussed. To the best of my knowledge, this is the first review paper which integrates research from sleep, neuroimaging, animal and human studies to describe the interaction between cortical state and cortical content to generate consciousness.
2. What are the dynamics of consciousness during sleep?
On average, a person sleeps eight hours a night, in which approximately 4 – 6 sleep cycles are passed. One sleep cycle lasts around 90 – 100 minutes and generally alternates between two main stages: Rapid Eye Movement (REM) and non-REM sleep. Non-REM sleep can further be divided in four stages (non-REM 1 – 4). Each sleep stage is characterized by unique electrophysiological rhythms (figure 3). These oscillations are generated due to interactions between thalamic and cortical areas, so called thalamo-cortical neurons. The awake state and REM sleep are dominated by localized, gamma-band (~40 Hz) oscillations, whereas low-frequency oscillations spread over the entire cortex during non-REM sleep (Llinas & Ribary, 1993). Even though sleep onset is accompanied by the fading of consciousness and a sleeping person is generally considered nonconscious, external sensory stimuli are still processed in the brain (Kouider et al., 2014). The differences in conscious content during different stages of the sleep cycle, suggests that the fluctuations in cortical state somehow relate to different levels of conscious content (figure 3).
Figure 3. Sleep stages in
relation to
electrophysiological rhythms and levels of consciousness. In general, one sleep cycle alternates between non-REM and REM sleep. Each sleep stage is characterized by unique electrophysiological rhythms and levels of consciousness (image from Bryant et al., 2004).
6
2.1 Non-REM sleep and slow-wave sleep
Non-REM sleep stage 1 is often referred to as the (wake-sleep) transition zone. The rapid, high frequency rhythms seen in the awake state are replaced by synchronized, slow, large amplitude oscillations. Non-REM stage 1 is characterized by slow rolling eye movements, attenuation of alpha band oscillations (7 – 14 Hz), enhancement of theta band oscillations (3 – 7 Hz) and the fading of perceptual awareness (Steriade et al., 1993). This stage is particularly interesting in the research of cortical state, since you slip away in a nonconscious sleep, but are still able to respond to external sensory stimuli (Hennevin et al., 2006). Imaging studies showed that frontal activity is attenuated during sleep onset, indicating that the frontal cortex is the first to fall asleep and the last to wake up (Ogilvie, 2001). To investigate the role of cortical connectivity in conscious state during sleep onset, Massimini and colleagues (2005) applied TMS on the motor cortex during wakefulness and several non-REM sleep stages, while simultaneously recording EEG. They found temporal and spatial propagation of TMS-evoked responses along the thalamo-cortical pathways during wakefulness. Interestingly, during non-REM sleep response patterns were initially stronger compared to the awake state but locally constrained, i.e. activity patterns did not propagate to the prefrontal cortex. This breakdown of thalamo-cortical connectivity, or inability to integrate information along the thalamo-cortical pathways, during the first stages of non-REM sleep might underlie the fading of consciousness.
Non-REM stage 2 is the predominant sleep stage and marked by the appearance of EEG spindles (~7 – 14 Hz) and K-complexes. A typical K-complex consists of a brief negative peak followed by a brief positive peak, both with high voltage amplitude, and they occur spontaneously during the first stages of the sleep cycle. EEG spindles occur every 3 to 10 seconds in bursts of waxing and waning rhythmic activity (figure 4A). These 1 – 3 seconds oscillations are generated in the thalamus due to the interaction of inhibitory neurons of the reticular thalamic nucleus, thalamo-cortical cells, and cortical pyramidal neurons between the thalamus and cortex. The reticular thalamic nucleus is located between the thalamus and cortex and therefore able to influence information processing (figure 4B). Intracellular recordings of cats’ reticular cells showed that ~7 – 14 Hz rhythmic bursts are generated due to depolarization of the membrane potential. This short depolarization results in short bursts of action potentials (McCormick &
Bal, 1997). Through divergent inhibitory projections these bursts inhibit large populations of thalamo-cortical cells, resulting in rhythmic (~7 - 14 Hz) inhibitory post-synaptic potentials (IPSPs). Via reciprocal connections, these bursts in turn facilitate the rhythmic bursts of the reticular cells. The IPSPs also continue travelling to the cortex, where EEG spindles are generated in cortical pyramidal neurons (Steriade et
al., 1993).
Non-REM sleep stage 3 and 4 are often referred to as slow-wave sleep, since the EEG spindles and K-complexes are replaced with slow-wave (<1 Hz) and
Figure 4. The interaction between the thalamus and cortex to generate EEG spindles in non-REM sleep stage 2. A: One to
three seconds sleep spindles occur every 3 – 10 seconds in non-REM stage 2, recordings from thalamic reticular nucleus of a cat. B: Schematic overview of the interaction between the reticular thalamic nucleus, thalamo-cortical cells and cortical pyramidal neurons in the thalamus and the cortex which generate the characteristic EEG spindles (image derived from Steriade et al., 1993).
7
Figure 5. Synchronized and desynchronized state of visual cortical neurons in mice. Upper row: Whole-cell recordings of the membrane potential of cortical neurons. Bottom row: Local Field Potentials (LFP). Synchronized state (left) is characterized by slow, large amplitude oscillations, whereas desynchronized state (right) is characterized by low voltage rapid oscillations (image adapted from Lee & Dan, 2009).
delta band oscillations (1 – 4 Hz). Delta frequency oscillations are generated in the thalamus and cortical layer 2, 3, 4 and 5. In the absence of EEG spindles, cortical neurons that project to thalamic neurons generate a hyperpolarizing – depolarizing oscillation in a sustained delta band frequency (Steriade et al., 1993a). The generation of slow wave oscillations are initiated in the neocortex, evidenced by thalamic lesion and in vivo studies, and the result of a cascade of synaptic inputs (Dossi et al., 1992; Steriade et al., 1993). Slow-wave sleep is considered the deepest sleep of all, since it is most difficult stage to awaken. It is therefore accompanied by the lowest level of conscious content.
2.2 REM sleep
The last stage of one typical sleep cycle is REM sleep, named after the characteristic rapid saccadic movements of the eyes, and accompanied by vivid dreaming, desynchronization of oscillations (compared to non-REM stages) and the absence of muscular activity. The transition from non-REM to REM sleep or awaking is initiated by the depolarization of the thalamo-cortical and thalamic reticular neurons, i.e. the blockage of hyperpolarizations associated with slow-wave sleep (Steriade et al., 2001). REM sleep is characterized by 40 Hz thalamo-cortical oscillations generated in the brainstem, and are similar in phase and amplitude to the awake state (Llinas & Ribary, 1993). REM sleep is an interesting level of consciousness, since electrophysiological rhythms resemble wakefulness even in the absence of external input. The vivid dreams during REM sleep have conscious features such as perception and emotion, but lack higher order characteristics such as self-reflective awareness, metacognition and abstract thinking (Hobson, 2009). REM sleep and wakefulness resemble in rapid, high-frequency (~40 Hz) rhythms, indicating similar underlying thalamo-cortical oscillatory mechanisms. Although, conscious state seems similar, conscious content is different between REM sleep and wakefulness. Llinas and Ribary (1993) suggest a hyper-attentiveness during REM sleep in which sensory stimuli are unable to reach the activation threshold to generate conscious content, as which occurs during wakefulness. Although similar brain rhythms as during wakefulness are observed, the thalamo-cortical oscillations during REM sleep do not modulate the required cortical areas to generate conscious content.
8
Figure 6. Control of REM on- and offset by neuromodulatory systems. Across time, the firing rates of neuromodulatory neurons from the brainstem and basal forebrain adapt their firing rate to regulate the on- and offset of REM sleep. In red: so called REM-on cells, cholinergic neurons increase their firing rate before the onset of REM sleep. In blue: REM-off cells, aminergic neurons increase before the onset of non-REM sleep (image adapted from Hobson & Pace-Schott, 2002).
2.3 Synchronized and desynchronized cortical state
In general, the main difference between each sleep stage is the state of thalamo-cortical activity, the frequency in which thalamo-thalamo-cortical neurons oscillate, e.g. thalamo-cortical state. Non-REM is characterized by cortical synchronization, whereas REM sleep and wakefulness are characterized by cortical desynchronization (figure 5). The activity of thalamo-cortical neurons is regulated by ascending neuromodulatory systems located in the brain stem, hypothalamus and basal forebrain. Neurons in these areas synapse directly on neurons in the thalamus and cortex and increase their firing rate in anticipation of awakening and several states of arousal (McCormick & Bal, 1997). During non-REM sleep, modulatory neurons fire intermediately relative to wakefulness, in which all neurons are activated. In REM sleep however, almost all ascending modulatory neurons completely shut down except for dopamine and acetylcholine. Cholinergic (i.e. neurons which release acetylcholine) projections depolarize thalamo-cortical neurons and therefore inhibit the rhythmic and burst firing during non-REM sleep. Thus, cholinergic neurons in the brainstem are mainly responsible for the electrophysiological characteristics during REM sleep (McCarley, 2007).
In addition, electrical stimulation of these neuromodulatory systems or direct application of acetylcholine result in depolarization of cortical pyramidal neurons followed by an increase in cortical excitability, similar to the desynchronized state seen during wakefulness or REM sleep (Steriade et al., 1993). While cholinergic systems are activated during REM sleep, these activation inputs are abolished during non-REM sleep. The synchronized thalamo-cortical oscillations seen in non-REM sleep emerge under aminergic influences, neurons in the pons of the brainstem that secrete modulatory the neurotransmitters serotonin and noradrenaline. In REM sleep however, these aminergic neurons are attenuated (figure 6). These mutually exclusive neuromodulatory systems, cholinergic neurons for REM sleep and aminergic neurons for non-REM sleep, regulate the sleep-wake switching mechanisms under influence of the circadian rhythm of the hypothalamus (Hobson & Pace-Schott, 2002).
In summary, each sleep cycle alternates between non-REM and REM stages, both characterized by unique electrophysiological rhythms. Non-REM sleep is characterized by synchronized, coherent oscillations which spread over the entire cortex by propagating along the thalamo-cortical pathways. These oscillations are generated by depolarization and hyperpolarization of thalamo-cortical neurons and regulated by aminergic neurons from the
9
Figure 7. Neuromodulatory transmitter pathways. Yellow: Locus Coeruleus (LC) releases noradrenergic transmitters; Red: Raphe Nucleus (FN) and Nucleus Accumbens (NA) regulate serotonin release; Green: Pedunculopontine nucleus (PN) in the brain stem and the Nucleus Basalis of Meynert (BN) for cholinergic neurotransmitters. HP: Hypothalamus; RE: Reticular Thalamic Nucleus; CL: Centrolateral Nucleus of the Thalamus. The ascending projections release neuromodulatory transmitters, through the thalamus (blue), in almost all cortical and subcortical areas (image derived from McCrea, 2009). brainstem. REM sleep is marked by desynchronized tonic activity, which is more locally constrained. This pattern of activity resembles wakefulness and correlates with the firing rate of cholinergic neurons. The unique oscillations of each sleep stage are associated with different levels of consciousness. REM sleep is an interesting state to investigate in the light of consciousness, since it is characterized by vivid dreaming (perception and emotion) but lacks the same conscious content as seen in wakefulness. Slow-wave sleep is considered as the deepest sleep with the lowest level of conscious content. It is therefore interesting to investigate the dynamics of the brain during sleep, since it will provide more information about the neural correlates of cortical state. A better understanding of cortical state might help to investigate the relationship between cortical state and conscious content during wakefulness.
3. How do neuromodulatory circuits influence cortical state?
Modulatory neurotransmitters have long-projecting axons to almost all cortical and subcortical areas and have the unique property of modulating their own and other transmitters’ release (Gu, 2002). Numerous neurotransmitters, such as acetylcholine, noradrenaline, serotonin, dopamine and histamine are involved in the neural signaling and regulation of cognitive processes (figure 7). Acetylcholine (ACh) and noradrenaline (NE) are implicated in almost every aspect of behavioral and physiological regulation, such as sleep, arousal, attention and sensory processing, and therefore I will only focus on these two neurotransmitter systems in relation to cortical state and sensory processing.
3.1 Cholinergic ascending projections
Cholinergic neurons are located in the basal forebrain and brainstem. Neurons in the basal forebrain secrete acetylcholine in the cerebral cortex via long-range, ascending projections, whereas neurons in the brainstem only project to the thalamus (figure 7). Lesions of the basal forebrain and the selective removal of cortical cholinergic projections evidence the involvement of acetylcholine in attentional processes (Hasselmo & Sarter, 2011). More specifically, changes in the level of cortical acetylcholine correlate with changes
10
in attentional performance (Parikh et al., 2007). Synaptic acetylcholine secretion in the medial prefrontal cortex of rats was measured during a cue-detection task. Attended cues were associated with increased cholinergic secretion, prior to stimulus presentation. In contrast, unattended or missed cues were not preceded by changes in cholinergic activity. The “phasic” release of acetylcholine, thus, mediates cognitive attentional processes. This hypothesis is supported by evidence from direct stimulation of cholinergic neurons in the visual cortex of mice (Pinto et al., 2013). During a go/no-go target detection task, activation of cholinergic neurons prior to stimulus presentation resulted in enhanced target detection, defined by the ratio of hits to false alarms. Thus, an increase of acetylcholine secretion in the visual cortex, preceding stimulus onset improved visual discrimination. Additionally, the authors tested whether this increased activity also enhanced visual encoding of the stimuli. They found that cholinergic activation significantly increased firing rates of visual cortical neurons and reduced trial-by-trial variability. These results indicate a positive correlation between pre-stimulus cholinergic activity and enhanced sensory information processing.
Interestingly, cholinergic activity primarily influences the maintenance of attention, with less effects on initiating or shifting attention (Furey et al., 2008). Cholinergic modulation during a selective attention task was investigated by pharmacological application in humans. Two double-exposure, i.e. morphed, pictures of houses and faces were presented, each in one hemifield. Participants were cued to attend to either faces or houses, and asked to respond if the two pictures of the cued category were the same. Reaction time at the first trial after the cued category was switched (from faces to houses or vice versa), indicated the attentional performance of shifting attention. Whereas later (than the first) trials indicated attentional performance of sustained attention. Cholinergic enhancement significantly increased overall reaction time, but specifically of later trials after the switch of cued category. Conversely, cholinergic reduction selectively decreased reaction times at trials later than the first trial. These findings suggest the involvement of acetylcholine in maintaining selective attention, rather than shifting attention.
As described in the previous chapter, cholinergic neurons regulate the onset of REM sleep. Direct activation of cholinergic neurons in the basal forebrain of mice during wakefulness, resemble the cortical desynchronization as seen in REM sleep (Pinto et al., 2013). Increased secretion of acetylcholine reduced low-frequency (1 – 5 Hz) power and increased high-frequency (60 – 100 Hz) power. Conversely, inactivation of cholinergic neurons significantly increased low-frequency oscillations and reduced spontaneous firing rates. Lee and colleagues (2005) measured extracellular recordings of cortical cholinergic neurons in sleeping-waking rats. Across the entire sleep-wake cycle, bursts of cholinergic firing activity were positively correlated with gamma power and negatively correlated with delta power. Thus, cholinergic activity plays a facilitating role in cortical desynchronization.
In addition, cholinergic activity in the thalamus regulate the spontaneous firing of thalamo-cortical neurons (Hirata & Castro-Alamancos, 2010). Application of a cholinergic agonist in the somatosensory thalamus enhanced thalamo-cortical oscillations. The increase of acetylcholine secretion in the barrel cortex of rats resulted in local cortical desynchronization. The authors provide evidence for the control of local cortical state by thalamo-cortical firing. In conclusion, cholinergic neurons in the basal forebrain, facilitate cortical desynchronization and therefore enhance sensory information processing.
11
3.2 The locus coeruleus – noradrenaline system
The locus coeruleus in the pons of the brain stem is the only origin of noradrenergic, ascending pathways and often referred to as the LC-NE system (figure 7). Activity of LC-NE neurons is highly correlated with the sleep-waking cycle. Firing rates are high during the awake state, low during non-REM and SWS sleep, and nearly silent during REM sleep. LC activity may be the primary factor that differentiate wakefulness from REM sleep (Aston-Jones & Cohen, 2005). Whereas cholinergic activity increases spontaneous thalamo-cortical oscillations, noradrenaline decreases spontaneous firing (Gu, 2002). Direct application of noradrenaline in the rats’ thalamus significantly reduced spontaneous activity to 0 Hz (Hirata & Castro-Alamancos, 2010). Stimulation of the rats’ whiskers during these elevated levels of noradrenaline resulted in increased evoked responses in the thalamo-cortical neurons, compared to baseline. These results indicate that noradrenaline enhances the signal-to-noise ratio (SNR) between sensory information (signal) and ongoing, spontaneous oscillations (noise). Turetsky and colleagues (2002) support this finding by pharmacological manipulation of noradrenergic secretion in humans, while measuring EEG during an auditory detection task. A typical EEG component that represents stimulus evaluation is the stimulus evoked P300 peak. Inhibition of noradrenaline resulted in decreased P300 amplitude, whereas enhancement of noradrenaline increased the P300 amplitude. The elevated levels of noradrenaline also decreased task performance, e.g. the accuracy on correctly detected stimuli. The authors proposed a noradrenergically modulating mechanism that tunes, i.e. reduces the variability in stimulus-evoked responses, cortical responsiveness. Thus, the LC-NE system increases the signal-to-noise ratio of sensory information and therefore facilitates behavioral performance.
Activity of the LC typically exhibits two states: phasic and tonic firing activity. Phasic mode to bursts of firing, selectively for target stimuli, over a short (milliseconds) scale of time and tonic firing refers to elevated levels of ongoing LC activity compared to baseline. LC activity highly correlates with behavioral performance. Aston-Jones et al. (1994) proposed a positive correlation between behavioral performance and LC firing rate. During an auditory oddball task, noradrenergic neurons in the LC were extracellularly recorded. Phasic firing prior to response onset correlated with enhanced behavioral performance, i.e. correctly detected targets and few errors. Conversely, periods of attenuated LC firing were associated with poor performance, i.e. increased false alarm rate.
LC firing activity is not strictly coupled with sensory information but rather is associated with optimization of behavioral performance according to an inverted U-pattern (Yerkes & Dodson, 1908). The optimization of behavioral performance can be described with the exploration – exploitation trade-off (Aston-Jones & Cohen, 2005). Exploitation refers to the optimization of the current task-performance as long as utility is high, for example when an animal is hungry and the reward is food. Whereas new seeking behavior when utility decreases refers to exploration. The authors investigated the role of monkeys’ LC firing activity in this trade-off with a target reversal experiment. The target in this task becomes the distractor and the distractor becomes the target. Correct target detection was rewarded. Recordings of LC activity revealed that if a new target was presented, phasic firing decreased over time while tonic firing increased. This pattern persisted until the target reversal elicited phasic firing again. It seems, thus for optimal behavior, that LC neurons fire tonically during focused attention, and respond phasically during task-relevant target stimuli. The authors show that the relationship between tonic firing of LC neurons and behavioral performance follow an inverted U-pattern (figure 8). Moderate tonic firing with
12
Figure 8. Inverted U-pattern describes the relationship between behavioral performance and tonic LC firing activity. Optimal behavior is associated with moderate levels of tonic firing and target-evoked phasic responses. Poor performance is associated with low tonic firing (inattentiveness and non-alert) and high tonic firing (distracted and hyper-attentiveness). Image retrieved from Aston-Jones & Cohen, 2005. phasic LC responses to target stimuli predict optimal performance. Poor performance is associated with low tonic firing result because animals are non-alert and inattentive, and high levels of tonic firing because animals are scanning (continuously exploring) and hyper- attentive (McGinley et al, 2015).
3.3 Regulation of pupil diameter
The LC also indirectly regulates pupil diameter, even though anatomical evidence lacks. Direct LC neuronal recordings in monkeys, which performed in an auditory discrimination task, revealed a striking correlation between pupil diameter and LC firing rate (Rajkowski et al., 1993). Tonic firing was accompanied by large baseline pupil diameter and consequently pupil constriction, whereas phasic firing was accompanied by small baseline pupil diameter and pupil dilation (figure 9). Thus, fluctuations in LC firing activity were highly correlated with the physical changes in pupil diameter, and suggest that pupillary changes can be used as an index of LC activity.
Murphy and colleagues (2014) were the first who provided direct functional evidence, in humans, for a positive correlation between pupil diameter and BOLD activity of the LC. FMRI recordings during rest revealed a positive correlation between pupil diameter and LC BOLD activity. In addition during an auditory oddball task, the specific part of the LC that correlated with pupil diameter elicited a greater BOLD response to target stimuli compared to irrelevant stimuli. This relationship is similar to the phasic firing in association with pupil dilation found in monkeys.
13
Figure 9. Relationship between Locus Coeruleus (LC) activity and pupil diameter. Tonic firing of LC neurons is highly correlated with pupillary diameter, measured in monkeys. Increased LC firing rate correlated with pupil dilation, whereas pupil constriction was associated with decreased LC firing activity (image from Gilzenrat et al., 2010).
Behavioral performance in humans also correlates with LC activity, indexed by pupillary changes (Gilzenrat et al., 2010). Baseline pupil diameter was recorded before the start of each trial, during an auditory discrimination task. Small baseline pupil diameter predicted enhanced target detection, whereas large baseline pupils were associated with poor performance. In addition, pupil dilation increased as a function of task difficulty. In engagement-promoting blocks small baseline pupils and large dilations were observed, whereas disengagement-promoting blocks were accompanied by large baseline pupil diameter and small dilations. These findings correspond to LC phasic firing and LC tonic firing respectively, as observed in animals (Rajkowski et al., 1993; Aston-Jones & Cohen, 2005).
Recently, a tight link between tonic levels of noradrenaline, indexed by pupillary changes, and cortical representations in humans was found (Warren et al., 2016). Pharmacological application of a noradrenaline transporter blocker, that increased synaptic levels of noradrenaline, enhanced neural encoding of sensory information. Trial-by-trial variability in cortical responses to various stimuli reduced, indicating the enhanced precision of perceptual encoding mediated by increased levels of noradrenaline. Thus, pharmacological increased levels of noradrenaline in humans enhance the signal-to-noise ratio of sensory information processing.
Since the recent evidence for a strong correlation between pupil diameter and the LC, and the mechanisms underlying both, not many studies have been conducted yet in the investigation of cortical state with the use of pupillary changes. However, recently Reimer et
al., (2016) studied the relationship between cholinergic and noradrenergic cortical activity,
an index for cortical state, in relation to spontaneous alternations in pupil diameter. Cholinergic and noradrenergic activity in the visual cortex of mice highly correlated with pupil fluctuations. Interestingly, during high levels of arousal (i.e. locomotion) only acetylcholine was correlated with pupil diameter, while small and rapid micro saccades represented levels of noradrenaline. The authors evidenced that rapid variations in cortical state, measured by cholinergic and noradrenergic activity, can be tightly tracked with changes in pupil diameter. These findings provide a foundation for the use of pupillary changes in humans to investigate spontaneous, cortical state.
It seems thus that fluctuations in cortical state are regulated by a cascade of various interacting and modulating systems. The link between neuromodulatory systems and the establishment of conscious state is not yet fully understood, due to the lack of reliable non-invasive cortical state recordings in humans. However, several lines of research in different
14
domains of neuroscience contribute together to a better understanding in the underlying correlates of consciousness.
4. How does cortical state influences sensory perception?
The activity of the cortex is determined by cortical state. As described in the first chapter, the most noticeable fluctuations in cortical state occur during sleep. A classical view declares that cortical state during wakefulness is a function of the sleep cycle, in which the brain alternates between synchronized, coherent and desynchronized, tonic activity (McCormick & Bal, 1997). However, these two states seem to be the extremes of a continuum of fluctuating and varying spontaneous activity (Harris & Thiele, 2011). Dehaene
et al., (2005) aimed to investigate the fluctuations of cortical state with a simplified
computational model of one thalamo-cortical pathway. They were therefore one of the first studies who attempted to describe the role of spontaneous oscillations, i.e. conscious state, in relation to conscious content. Their model showed that a minimal level of neuromodulatory input, indexed by direct depolarization and hyperpolarization of thalamo-cortical neurons, elicited conscious state. This state of thalamo-thalamo-cortical resonance was characterized by bursts of high-frequency band oscillations, similar as seen during wakefulness and REM sleep, and indicated the possibility of conscious content. The influence of neuromodulatory input on the ignition threshold was mediated by at least two mechanisms: spontaneous oscillators in the thalamo-cortical pathways and the facilitation of sensory stimuli by spontaneous activation. They concluded that conscious state, shaped by ongoing spontaneous oscillations, determined the likelihood of conscious content. This model is simplified and minimally resembles actual thalamo-cortical networks, though it points out the involvement of cortical state in consciousness. However, it remains debated how exactly cortical state interacts with sensory information.
4.1 Sensory processing is cortical state dependent
Conscious content, as defined by the global workspace theory, depends on long-range integration of sensory information to higher cortical areas. This integration is established by an optimal balance between functional differentiation and integration of information in thalamo-cortical circuits (Tononi & Edelman, 1998). Differentiation refers to the multiple, specialized areas of the thalamo-cortical system involved in conscious content. Whereas integration refers to the rapid and efficient interaction between these areas. This balance is often referred to as brain complexity and appears to be a useful tool to measure conscious state. Casali and colleagues (2013) introduced such reliable measure, named the perturbational complexity index (PCI). PCI refers to response of the cortex, as measured with EEG, to the perturbation of transcranial magnetic stimulation (TMS). The integration and differentiation of the stimulus along the thalamo-cortical pathways contains information about the complexity of the brain’s neural processes and appears to accurately discriminate wakefulness, different stages of sleep, coma, vegetative state, different states of anesthesia and brain-injured patients. PCI, thus, pinpoints the importance of cortical state in the integration and differentiation of information in the thalamo-cortical network to generate conscious content.
The ongoing depolarization-hyperpolarization oscillations of thalamo-cortical neurons during slow-wave sleep correlate with sensory responsiveness. Massimini and colleagues (2003) used external, electrical stimulation at the wrist at different phases of
15
ongoing slow-wave oscillations to investigate cortical responsiveness (figure 10A). Typical negative somatosensory evoked responses during wakefulness were visible around N20 and N45, whereas typical positive peaks were observed around P60 and P100. Interestingly, at these specific latencies during slow-wave sleep, a “phase-dependent variability” was observed. Increased and wakefulness-like stimulus-evoked responses at negative slopes of ongoing, spontaneous oscillations were observed, whereas little responses were evoked at positive slopes (figure 10B). This correlation between stimulus-evoked responses and cortical state indicates a cortical state-dependency for optimal sensory information processing. The authors make a link with intracellular recordings in cats and suggest that hyperpolarization of thalamo-cortical neurons correspond to a surface-negative deflection in the EEG signal. This would indicate enhanced stimulus processing during hyperpolarized periods of thalamo-cortical neurons. In other words, if cortical state is desynchronized, as a result of hyperpolarized thalamo-cortical neurons, it is likely that conscious content is possible.
4.2 Variability in cortical responsiveness predicts sensory perception
Cortical responses to external stimuli during wakefulness are variable. If the same sensory stimulus is presented multiple times, cortical responsiveness in the associated sensory area will rarely be identical. However, this variability is not just random noise. Trial-by-trial variability in cortical responsiveness can be predicted from spontaneous, ongoing oscillations. In vivo recordings in rats showed that cortical responses exhibited a nonlinear interaction between auditory stimuli and spontaneous oscillations (Curto et al., 2009). If pre-stimulus activity was in a downstate (negative slope of the ongoing oscillation), presentation of the stimulus caused an opposite, positive response. Likewise, stimulation during upstate activity (positive slope of the ongoing oscillation) cause a negative evoked response. In contrast, during desynchronized state, cortical responses were weakly modulated by the presentation of the stimulus. Thus pre-stimulus cortical activity, i.e. synchronized or desynchronized, accurately predicted cortical responsiveness. A similar relationship between spontaneous cortical activity and behavioral output was found in humans (Fox et al., 2007). Participants were instructed to press buttons during a simple visual discrimination task in the fMRI scanner. Offline segregation of hard and soft button press forces revealed a significant correlation with BOLD responses in the somatosensory
Figure 10. Phase-dependent variability between cortical responsiveness and spontaneous oscillations during sleep. A: Somatosensory stimulus were applied at positive and negative slopes of spontaneous oscillations. B: Averaged stimulus-evoked responses at peaks (left) and troughs (right) of ongoing oscillations. Increased and wakefulness-like N20 responses were observed at troughs compared to response amplitudes at peaks. Image adapted from Massimini et al., 2003.
16
cortex. A regression analysis revealed that 74% of BOLD response – behavioral output relationship was explained by spontaneous, ongoing cortical activity.
In line with these results, pre-stimulus electrophysiological activity also predicts behavioral performance. Mathewson and colleagues (2009) recorded EEG while subjects were instructed to detect targets in a masking paradigm. Detection threshold was set near threshold and stimulus parameters remained equal the entire experiment. Alpha power, measured at occipital electrodes, revealed a negative correlation with detection rate. Thus, increased preceding alpha power was associated with a poor detection rate. Additionally, increases in alpha power over time coincided with decreased task performance. The authors were the first to describe the systematic influence of pre-stimulus alpha power at behavioral performance. Spontaneous oscillations are not only defined by their power, but also by their phases. The phase of pre-stimulus oscillations predicts whether a stimulus will be perceived or not (Busch et al., 2009). Trial-by-trial analyses of oscillatory phase revealed a correlation between subject-specific detection performance and the phase of low alpha (8 – 12 Hz) and high theta (4 – 8 Hz). Specifically at ~7 Hz pre-stimulus oscillations, hits and misses could be accurately predicted solely by phases. In total, single trial phases of ongoing oscillations accounted for 16% of behavioral performance. Thus, pre-stimulus ongoing oscillations accurately predicted, either by phase or power, cortical responsiveness and therefore behavioral output.
Recently, a more specific relationship between variability in cortical responsiveness and behavioral output was found (He & Zempel, 2013). Direct cortical recordings using electrocorticography (ECoG) in humans revealed an inverted U-pattern between trial-to-trial ECoG variability and behavioral performance. Trial-to-trial ECoG variability was defined by the first component, that accounts for the largest variance, after Principal Component Analysis (PCA). During a visual discrimination task, increased or decreased ECoG activity predicted poor performance, a hit rate around change level (~50% accuracy). Whereas during trials with moderate pre-stimulus ECoG activity, hit rate was significantly enhanced (~75% accuracy).
Cortical response variability decreases through active engagement, such as task performance and attention (Von Trapp et al., 2016). Gerbils performed in an auditory discrimination task, while auditory cortical neurons were recorded. Trials that involved task-engagement were compared to trials in which the animals were disengaged from the task. Task engagement reduced the trial-by-trial variability of firing rate, compared to disengaged trials. In turn, neural sensitivity (defined by improved performance of auditory discrimination) was increased, indicating enhanced sensory perception. The above mentioned studies, thus, suggest that sensory responsiveness is highly dependent of cortical state; some states enhance sensory processing while others attenuate sensory processing.
4.3 Cortical desynchronization facilitates conscious content
Attention, arousal and awareness modulate cortical state to gate the focus of interest (Moran & Desimone, 1985). Attention refers to the ability of the brain to filter out irrelevant information and amplify relevant or important features. Neuroimaging studies showed that attention increased cortical activity in visual sensory areas, compared to unattended objects (Corbetta et al., 1990; Kastner et al., 1998). The source of top-down attention seems to be located in higher cortical areas (Kastner et al., 1999). In the absence of visual stimulation, the Frontal Eye Fields (FEF), the Supplementary Eye Fields (SEF) and the Superior Parietal Lobe (SPL) showed an increase in activity compared to baseline. This
17
sustained activity in higher cortical areas, indicates the involvement of top-down attention in sensory processing. In addition, micro-stimulation of FEF enhanced sensory responses in the visual cortex and thereby improved visual selection (Moore & Armstrong, 2003). Thus, higher areas play an important role in top-down attention and therefore enhancement of sensory perception.
In vivo recordings in animals provide direct information about neuronal fluctuations and cortical state. Studies with primates demonstrated that under the influence of attention sensory cortical neurons increase gamma band frequency oscillations (Sùper et al., 2003). Recordings of neural activity in the visual cortex during a visual discrimination task, revealed that attended stimuli in the receptive field of the recorded cortical neurons increased gamma band power compared to unattended stimuli. This local increase in gamma-band power started before and preceded during stimulus presentation, and positively correlated with stimulus detection. Fries et al. (2001) supported these findings and additionally reported reduced low-frequency (<17 Hz) synchronization during attended stimuli compared to unattended stimuli in higher visual areas. The sustained enhanced gamma band synchronization and reduced low-frequency power suggest a modulating effect of attention on cortical state to enhance sensory perception. More specifically, this modulation resembled cortical desynchronization as seen during REM sleep and wakefulness (Harris & Thiele, 2011).
The involvement of desynchronized cortical state in relation to the establishment of conscious content is supported by recordings of Local Field Potentials (LFP), voltage recordings of the electrical currents which flow between neurons assemblies, in anesthetized rats. Marguet and Harris (2011) investigated how pre-stimulus cortical state modulated cortical responses to continuous auditory stimuli. During synchronized state, LFPs were dominated with increased low-frequency power which support previous findings. Interestingly, the firing rate of auditory neurons could be predicted from LFP, but not from stimulus presentation. Thus, auditory stimuli had little effect on the cortical responsiveness during cortical synchronization. On the other hand, when the cortex was desynchronized, firing rate was strongly correlated with both stimulus presentation and LFP. The amplitude envelope of the continuous auditory stimuli showed increased coherence with LFP, and additionally predicted the firing rate of auditory neurons. Thus, the presentation of auditory stimuli accurately predicted cortical responsiveness during cortical desynchronization, but had little effect during cortical synchronization. The latter indicates a decoupling between the auditory stimulus and cortical state. In desynchronized state, however, sensory stimuli did predict cortical responsiveness, inferring the involvement of cortical desynchronization to enhance conscious content.
4.4 Cortical synchronization attenuates conscious content
While cortical desynchronization enhances sensory perception, cortical synchronization is hypothesized to attenuate sensory processes. Application of drugs and anesthetics, which cause a loss of consciousness, result in slow wave oscillations similar to slow wave sleep. Even though the coherence of thalamo-cortical oscillations is reduced compared to natural sleep, these results suggest that slow-wave oscillations are involved in suppressing consciousness during sleep and anesthesia (Destexhe et al., 1999). Supp and colleagues (2011) used EEG to record ongoing brain oscillations during step-wise loss of consciousness as a function of increasing drug concentration. They found an association between the loss of consciousness and the emergence of hypersynchronous cortical state in
18
the alpha frequency range (8 – 14 Hz) located in frontal areas. Long-range coherence and local alpha power were highly correlated with loss of consciousness induced by drugs and anesthetics. In contrast, they found a significant decrease in alpha power as a function of sedation in occipital areas. This study evidenced increased large-scale phase coherence and frontal alpha power during loss of consciousness, indicating long-range and local synchronization, respectively. Additionally, to investigate the cortical processing of sensory information during different stages of anesthesia, the authors applied electrical stimulation to participants’ hands. Baseline stimulus evoked responses in somatosensory cortex showed three distinct components at 0-90 milliseconds, 30-110 milliseconds and 140 – 370 milliseconds. Anesthetics specifically reduced the amplitudes of the latter components, the second and third, as a function of drug dosage. The attenuation of these late component over time correlated with the loss of consciousness. This functional inhibition indicates that recurrent processes (indicated by the stimulus evoked responses later in time) but not feedforward (earliest response) are associated with the loss of consciousness. In conclusion, this study suggest the involvement of synchronized cortical state, as a result of anesthesia, in the loss of conscious content.
From the above described results it is highly likely that attention and arousal modulate cortical state to enhance sensory processing. Specifically, thalamo-cortical neurons become hyperpolarized, with decreased low-frequency band oscillations and increased local, gamma-frequency synchronization. These characteristics resemble cortical desynchronization, as seen in REM sleep. On the other hand, studies with drugs, sleep and anesthetics describe the association between a synchronized cortical state, depolarized thalamo-cortical neurons, and the loss of consciousness. However, the described studies so far were all performed under influence of attention, during task engagement which is known to enhance sensory information processing, or without any conscious content, during anesthesia or sleep. It remains unknown what exactly happens in the brain during spontaneous, pre-stimulus oscillations and how these interact with sensory stimuli.
4.5 Pupil diameter as a surrogate measure for cortical state
Similar to variability in cortical responsiveness, different states of arousal also predict behavioral output according to an inverted U-pattern. Intermediate levels of arousal result in optimal behavioral performance, whereas low and elevated levels predict poor output. The state of arousal can be modulated by cognitive demand, such as attention, engagement and mental effort. Pupil diameter changes as a function of task difficulty, across numerous cognitive domains (Stanners et al., 1979). Intracellular recordings in mice demonstrated a strong correlation between arousal and pupillary changes (for a review see McGinley et al., 2015a). During heightened levels of arousal, indicated by locomotion, pupil diameter is increased. Conversely during periods of rest pupil diameter is constricted. Therefore, pupil diameter can be used as a physiological index for arousal.
Extracellular recordings in rodents and mice have provided useful information in the exact correlation between cortical state and arousal, as measured with pupillary changes. McGinley et al. (2015) found a similar inverted U-pattern between arousal and cortical state. Low arousal, e.g. constricted pupil diameter, was associated with low-frequency membrane potential oscillations and thus increased low-frequency power. Intermediate arousal and pupil diameter (40 – 60% of fully dilated pupils) resulted in a strong suppression of low-frequency oscillations and hyperpolarization of the membrane potential. And finally, enlarged pupil diameter and heightened arousal were associated with an overall
19
Figure 11. The correlation between fluctuations of cortical membrane potential and pupil diameter. The upper row (black line) represents the raw signal of extracellular recordings of auditory cortical neurons in rodents. The second (purple) line represents the standard deviation of the membrane potential (Vm) and the third (green) signal is the mean of the membrane potential over time. The diameter of the pupil accurately follows the raw and the mean signal of the membrane potential, indicating the strong correlation between pupillary changes and cortical state. The bottom signal (blue) represents the walking velocity. Locomotion is associated with high levels of arousal and thus with large pupil dilation and high frequency band power (image adapted from McGinley et al., 2015).
depolarization and increased high-frequency oscillations (figure 11). In line with these results, another study simultaneously measured pupillary changes and whole-cell recordings (Reimer et al., 2014). During pupil dilation, the membrane potentials of somatosensory and visual cortical neurons were hyperpolarized, low-frequency oscillations were reduced and gamma band oscillations were enhanced. Conversely during pupil constriction, low-frequency power was enhanced and local synchronization was observed. Similarly, induced arousal by application of an air puff revealed a strong positive correlation between pupil diameter and gamma band-power and negative correlation between pupil diameter and low-frequency power (Vinck et al., 2015). The authors also applied air puffs during periods during which the mouse was already aroused, indicated by continuous locomotion. Interestingly, no correlation with either gamma or low-frequency bands was found. This could be related to the hyper-attentive state proposed by Llinas and Ribary (1993), in which sensory stimuli are unable to reach the activation threshold to establish conscious content.
Due to ethical and technical issues, it is difficult to measure cortical fluctuations in awake humans. The exact fluctuations of cortical state, i.e. the neuromodulatory circuits and thalamo-cortical oscillations, coupled to pupil diameter remain yet largely unknown. Several studies investigated the relationship between pupil diameter and cortical activity, but not state, with EEG or fMRI in humans. Murphy and colleagues (2011) found that pupil diameter exhibited a similar inverted U-relationship with electrophysiological responses and behavioral output. During an oddball auditory discrimination task, largest stimulus-evoked responses and optimal behavioral performance, coincided with intermediate pupil diameter. On the other hand, large and small pupil diameters preceded poor task performance. More recently, an fMRI study in which participants were sleep deprived showed that spontaneous fluctuations in pupil dilation were associated with increased activation of the resting state
20
network, i.e. thalamic and parietal areas. In contrast, spontaneous pupillary constrictions were associated with increased activation of sensory cortical areas. These results indicate that pupillary changes directly reflect the activity of brain areas involved in the regulation of arousal (Schneider et al., 2016).
In conclusion, sensory responsiveness highly depends on cortical state. Different levels of arousal modulate cortical state and therefore conscious content. Low levels of arousal, such as sleep, anesthesia and drug-induced sedation, are associated with poor performance, hyperpolarized thalamo-cortical neurons and cortical desynchronization. Intermediate levels of arousal are accompanied by optimal behavioral performance, depolarized thalamo-cortical neurons and cortical synchronization. High levels of arousal are associated with poor performance, depolarized thalamo-cortical neurons and a hyper-attentive state, in which sensory perception is difficult (Reimer et al., 2014; McGinley et al.,
2015; Vinck et al., 2015). Thus, cortical synchronization attenuates sensory perception,
whereas cortical desynchronization enhances sensory perception.
5. Discussion
This review aimed to describe the interaction between conscious content; the awareness of sensory information, and conscious state; the neural intrinsic oscillations of the cerebral cortex which determine whether conscious content is possible. Conscious state, specifically involved in consciousness, is determined by cortical state: the ongoing, spontaneous fluctuations of the brain during the course of the day. The involvement of cortical state is often neglected in consciousness research, however, recent studies have pinpointed the importance of this factor (Dehaene et al., 2005; McGinley et al., 2015).
Cortical state is a function of the sleep cycle, since the most noticeable fluctuations occur during this period of the day. Generally, the activity of the cortex alternates between synchronized and desynchronized state, during non-REM and REM sleep respectively. Synchronized state is characterized by increased low-frequency oscillations with large amplitudes, which spread across the entire cortex, and suppression of high frequency rhythms. On the other hand, cortical desynchronization characterizes REM sleep and the awake state. Locally increased high frequency power and suppression of low-frequency oscillations mark these periods (figure 5). Cortical fluctuations are the result of inhibitory and excitatory interactions between the thalamus and the cortex via so called thalamo-cortical pathways. The state of thalamo-thalamo-cortical activity is mainly regulated by cholinergic and noradrenergic neurons in the basal forebrain and brainstem, respectively. These areas secrete the modulatory neurotransmitters, acetylcholine and noradrenaline, to almost all (sub) cortical areas.
Acetylcholine and noradrenaline are involved in almost every aspect of behavioral and physiological regulation, such as arousal, attention and sensory processing. Cholinergic activity primarily influences the maintenance of selective attention, i.e. keeping cortical state in active mode to enhance sensory perception (Parikh et al., 2007; Pinto et al., 2013). This maintenance resembles cortical desynchronization as seen during non-REM sleep. Secretion of noradrenaline (NE) is regulated by the locus coeruleus (LC) in the brainstem. Noradrenaline increases the signal-to-noise ratio between sensory information (signal) and ongoing, spontaneous oscillations (noise). The LC-NE system, thus, facilitates sensory information processing (Turetsky et al., 2002). Activity of the LC typically exhibits two states: phasic and tonic mode, and strongly correlates with behavioral performance according to an
21
inverted U-pattern (Aston-Jones & Cohen, 2005). Phasic LC activity refers to bursts of firing over a short time scale, whereas tonic mode refers to elevated firing rates compared to baseline.
The locus coeruleus also regulates pupil diameter. Tonic firing of LC neurons is correlated with pupil dilation, whereas phasic firing is associated with pupil constriction (Rajkowski et al., 1993; Gilzenrat et al., 2010; Murphy et al., 2013). A tight link between noradrenergic and cholinergic activity and pupillary changes was found, indicating that pupil diameter can be taken as a surrogate measure for cortical state. Recent studies used pupillary changes as an index for cortical state, and revealed an inverted U-pattern between behavioral performance and cortical state (Reimer et al., 2014; McGinley et al., 2015; Vinck
et al., 2015). During intermediate arousal, indexed by intermediate pupil diameter and
increased gamma band power (hallmarks of cortical desynchronization), behavioral performance was optimal. On the other hand, elevated or low levels of arousal, which contain characteristics of cortical synchronization, are associated with poor performance. Thus, ongoing, spontaneous oscillations predict whether a stimulus will be perceived or not (Curto et al., 2009; Busch et al., 2009; Mathewson et al., 2009; He & Zempel, 2013). Basically, cortical synchronization attenuate conscious content (Supp et al., 2011). While cortical desynchronization, facilitated by attention and arousal, enhances conscious content (Fries et al., 2001; Sùper et al., 2003; Harris & Thiele, 2011; Marguet & Harris, 2011). From these studies can be concluded that conscious content is highly dependent on conscious state, and that numerous processes and cascades of modulatory systems contribute in establishing this relationship.
Despite the many challenges, recent techniques and novel analyses have contributed to more reliable and accurate measurements of the underlying correlates of consciousness. However, much factors remain unknown. Many of the discussed studies were performed in mice, rodents or macaques. Intra- and extra cellular recordings in animals provide valuable information about the exact fluctuations in cortical state during behavioral performance. Even though humans share numerous commonalities with animals, it remains debated how these cellular recordings can be linked to the human brain. In humans, drugs and anesthetics are often used to investigate the relationship between neuromodulatory systems and behavioral output. These pharmacologically applications manipulate neurotransmitter releases by blockage or activation of certain receptors. However, it is important to keep in mind that these substances only manipulate one receptor, or even a subunit of that receptor. The results from these studies should be interpreted with caution when discussing conscious state or content. Investigating consciousness with anesthetics or drugs only tells something about the involvement of the manipulated factor and not about a causal relationship. Additionally, it remains unknown whether blockage or activation of specific receptors does not affect other transmitter secretion.
The recent study of Reimer and colleagues (2016) provides a useful tool to non-invasively investigate fluctuations in cortical state with pupil diameter. However, this study measured micro-saccades in mice. It remains to be questioned whether it is feasible to measure these small pupillary changes in humans and if the same neuromodulatory pathways, underlying the regulation of pupil diameter, exists. In humans it is difficult to measure fluctuations in cortical state at rest, due to ethical and technical issues, but also because rest implicates no cognitive demand. Studies which measured ‘spontaneous’ activity often used pre-stimulus activity. In these cases, participants already engaged in an experiment and are anticipating on performing. As previously discussed, task-engagement is
