1
Pupil dilation as a marker for neuromodulation:
How strong is the evidence for the claim that pupil dilation
reflects central noradrenaline release from the locus
coeruleus?
Literature thesis MSc Brain and Cognitive Sciences: Cognitive Neuroscience Data: May 14th, 2014
Name: Thomas Meindertsma Student number: 5810593
Supervisor: Prof. Dr. Sander Nieuwenhuis (Leiden University) Co-assessor: Dr. Tobias H. Donner
Introduction
The relation between pupil diameter and cognition was picked up a long time ago, even long before the techniques were available to objectively measure it. The link between the tonic size of the pupil and slow unspecific processes like wakefulness and arousal is the most constant factor in this research field. But even faster phasic pupil size modulations have been studied for decades already, with a big wave of publications in the 1980’s, after Hess linked pupil diameter to the pleasantness or unpleasantness of pictures presented to his subjects (Hess and Polt, 1960). Many studies linked pupil diameter to all kinds of emotional states and responses, as well as to personality traits, demographic characteristics, sexual preference and many other variables that could, in some way, be related to stimulus valence. Pupil size modulation was even studied as a marker for customers’ unconscious true likes and dislikes by the commercial industry (Loewenfeld, 1993). After about a decade the results and validity of this proliferation of studies had proven to be poor. As a result, the pupil research field grew more quiet again.
In recent years, research into the relation between the phasic modulation of pupil size and cognition has flourished again. Pupil diameter has been linked to diverse and high level cognitive processes including task-relevance or novelty of stimuli, learning and believe updating, decision making and bistable perception (Aston-Jones et al., 1997; Einhäuser et al., 2008; Preuschoff et al., 2011; Nassar et al., 2012; Eldar et al., 2013; Naber et al., 2013; De Gee et al., 2014). Technical improvements have allowed for more precise measurements. And the developments in other fields of science, especially in neurophysiology, neuroanatomy and computational neuroscience, add to possible mechanisms that facilitate the link between pupil and cognition.
One new development, the coupling between pupil diameter and central neuromodulation, has been particularly influential. The widespread effects of dopamine,
2
serotonin, noradrenaline and other modulatory neurotransmitters has been the subject of many studies in the past decades. Some of these neuromodulators, due to their similarities to the pupil in response to cognitive task modulations, have been coupled to the findings in the pupillometry field. Especially noradrenaline (NA) release from the locus coeruleus in the brainstem has been argued to relate to pupil size. Sparse evidence from intracranial recordings in animals supports this (Rajkowski et al., 1993; Aston-Jones and Cohen, 2005).
The first steps in integrating findings from these different research fields have been taken. However, an overarching explanation or model is far from established. The diversity of the research that needs to be integrated might be a cause of this lack. The link between pupil size and cognition is based mostly on recent evidence from studies using human subjects and an occasional monkey. Whereas pupil anatomy and physiology build mostly on evidence from cats, rats, monkeys, dogs and even non-mammalian species, as well as some post-mortem human brain studies. Most of this research was also performed much earlier than the cognitive pupillometry studies, providing for poor cross-disciplinary discussion. The other main piece of the puzzle, central neuromodulation, has been studied in both human and animal research, mostly in the nineties.
To summarize, although not always well integrated, both the link between pupil size and cognition and neuromodulation and cognition is supported by a relatively strong base of evidence. However, we need more evidence for the link between the pupil and neuromodulation to be able to claim any causal relation between these two and develop a reliable model of the interplay between cognition, neurophysiology and the pupil. In this review I will critically evaluate the evidence for the relation between central neuromodulation and pupil size. How well-supported is the claim that pupil size modulation reflects central release of a neuromodulator (e.g. NA)? And what does this mean for the neurocognitive processes that coincide with these modulations? I will first discuss the anatomy and physiology of the pupil itself. Subsequently, I will discuss and appreciate the evidence for the link between the locus coeruleus noradrenaline (LC-NA) system and pupil dilation. In the third part, I will look into other mechanisms that might influence pupil dilation. I will also try to elucidate the possible influence of central processes on the constriction of the pupil. Pupil constriction has so far been seen only as peripheral reflexes to light (pupillary light reflex) and object distances (accommodation reflex). However, pupil constriction has been linked to cognitive processes recently, demanding a physiological explanation.
Anatomy and physiology of the pupil
We need at least some basic knowledge of the anatomy and physiology of the pupil to be able to understand in what way cognitive processes might relate to pupil size. The muscles in the pupil and the neurons innervating them have been studied and documented thoroughly. Publications about this subject date back to the 1600’s. A complete description of that work is beyond the scope of this review and can be found in other publications (Loewenfeld, 1993; Beatty and Lucero-Wagoner, 2000).
3
The pupil’s main function is to control the amount of light entering the eye by changing its size. This is facilitated by two antagonistic smooth muscles tissues: the pupillary sphincter muscle for constriction and the radial dilator muscles for dilation. The two muscles groups are under control of different parts of the peripheral nervous system, the sphincter (constrictor) muscles are innervated by parasympathetic nerves, whereas the dilator muscles are innervated by sympathetic nerves (Beatty and Lucero-Wagoner, 2000).
The afferent connections to the constricting sphincter muscles come from the Edinger-Westphal nucleus in the brainstem. Information about luminance travels from the retina, through ganglion cells of the optic nerve and the olivary pretectal nucleus, to the Edinger-Westphal. From there efferent connections go through the oculomotor nerve to the ciliary ganglion behind the eye, where they connect to the short ciliary nerves that directly control the pupil constrictor muscles. This simple pathway facilitates the pupillary light reflex (Loewenfeld, 1993).
The accommodation reflex (or pupil near reflex), when the eye is focused on a near object, is a bit more complex, but also involves pupil constriction. Additionally, the accommodation reflex causes two other muscular actions: (i), the ciliary muscles contract, which releases the tension of the zonule fibers and thereby increase the curvature of the lens. And (ii), the medial rectus nerves contract, causing the eyes to converge (Beatty and Lucero-Wagoner, 2000). The efferent pathways are grossly the same as for the pupillary light reflex, with fibers originating in the Edinger-Westphal nucleus and innervating the corresponding muscles through the oculomotor nerve. Up to now, we know very little about the afferent pathways of the accommodation response. With conventional non-invasive pupillometry methods we are not able to distinguish between these two reflexes, or any other mechanisms constricting the pupil. We must therefore always keep in mind that any effect we see in the pupil might be accompanied by lens accommodation or vergence adjustments.
The innervation of the dilator muscles is quite different from that of the constrictor muscles. Many cortical and subcortical structures have connections to the nucleus paragigantocellularis lateralis (PGi, also often abbreviated as PGL), located in the rostral ventrolateral medulla. From there the PGi sends projections through the intermediolateral cell column of the spinal cord to the superior cervical ganglion (SCG). Axons go from the SCG, through the root of the ciliary ganglion and the long ciliary nerves to the eye, where they innervate the pupil dilator muscles (Drake et al., 2009). The afferent connections to the PGi include the medial prefrontal cortex (mPFC), orbitofrontal cortex (OFC), anterior cingulate cortex (ACC) and insular cortex, but also subcortical structures such as the amygdala and peripheral inputs from the vagal nerve (Nieuwenhuis et al., 2011). This broad variety of afferents accounts for the abundance of processes it is associated with. Most sensory stimulations, not just visual, are accompanied by pupil dilation. But internally generated cognitive processes, such as cognitive load, motor output and emotions, can also cause the pupil to dilate (Beatty and Lucero-Wagoner, 2000).
As any system that is governed by a pair of antagonists, pupil size modulations are always caused by the net result of constrictor and dilator muscle activity. Pupil dilation
4
can be caused by excitation of the dilator pathway, inhibition of the constrictor pathway or a combination of the two. To disentangle the influence of these factors, the ambient light intensity can be manipulated (Steinhauer et al., 2004). In a dark environment, the constrictor muscle is relatively relaxed, since the retina does not have to be protected from incoming light. Inhibition of the constrictor muscles in dark conditions would thus not result in strong pupil dilation. Excitation of the dilator muscle, on the other hand, still results in pupil dilation in dark environments. Pupil dilation under low light conditions can therefore be attributed mostly to sympathetic dilator muscle activation. When light intensity increases, so will constrictor muscle stimulation and, as a consequence, the effect of constrictor muscle inhibition. The higher the light intensity is, the stronger the potential influence of constrictor muscle inhibition becomes. Looking at pupil size modulations under various light conditions can give you an estimate of the respective role of the sympathetic dilator pathway and the parasympathetic constrictor pathway. Alternatively, local pharmacological interventions can give insight into the relative influence of the pupil constrictor and dilator muscles on pupil size modulations. It is possible to block (sympathetic) alpha-adrenergic receptors or (parasympathetic) muscarinic receptors separately with eye drops, thereby blocking the dilator muscles or constrictors muscles, respectively (Steinhauer et al., 2004). Importantly, these interventions do not have any known central effect, thereby nicely separating the neural processes and their effect on pupil size.
Of course, not every study that included pupil measurements incorporated these experimental manipulations. We must therefore be cautious when interpreting pupil size modulations to be caused by the pupil dilator or constrictor pathway.
The role of the LC-NA system in pupil size modulation
The research and discussion on the relation between pupil size modulations and cognition-related brain activity has focused mainly on the dilation of the pupil. The most prominent theory in the past years has been that pupil dilation is a marker for noradrenaline release from the locus coeruleus. The LC-NA system has been implicated in many overlapping cognitive processes. Tonic levels of LC-NA activity are linked to arousal and the amount of sympathetic peripheral output. Phasic responses of the LC-NA system are related to attention, uncertainty, surprise and the outcome of task-related decision processes (Aston-Jones and Cohen, 2005).
Different models exist that integrate the tonic and phasic characteristics of the LC-NA system. Aston-Jones and Cohen approach the subject from an evolutionary perspective: the exploration vs. exploitation trade-off. When the utility of a task or environment decreases, the baseline activity in the LC increases and subjects disengage from the task. Subjects become more receptive to new or task-unrelated features. This tonic mode of LC activity thus biases the subjects towards more exploratory behavior.
When utility is high within a task, the baseline level of LC activity decreases and subjects engage. In this mode the LC responds with a short phasic burst of activity to events that help to maximize profit or optimize task performance. In task-related terms this means for example that the LC shows a phasic increase to behaviorally relevant targets, but not to distractors in oddball tasks (Aston-Jones et al., 1997). This phasic mode biases the subjects towards exploitation (Aston-Jones and Cohen, 2005).
5
Dayan and Yu proposed a similar model, but worked with a slightly different point of view (Yu and Dayan, 2005; Dayan and Yu, 2006). In their model, the overall uncertainty about probabilistic relationships in the environment is reflected in the tonic activity of the LC. High uncertainty about the probability of upcoming events and task performance leads to high baseline activity in the LC. Whereas more confidence about the performance or expectations leads to a lower baseline activity. So when the level of uncertainty is consistent throughout a task, the baseline LC activity will be low. When it is not, the baseline activity will be high. This level of overall uncertainty about the task or environment is highly related to level of engagement in Aston-Jones and Cohen’s model. Phasic responses of the LC in Dayan and Yu’s model are caused by unexpected uncertainty within a task (Dayan and Yu, 2006). This means that the phasic LC responses are pronounced when subjects have strong predictions about upcoming events that do not match with reality. For example, when someone performs an oddball task in which targets are very rare, he or she will strongly predict the next stimulus to be a distractor, resulting in a strong phasic LC response when a target appears. Whereas if that same subject would perform an oddball task in which targets are almost as common as distractors the phasic response to targets will decrease, since the uncertainty about the task was expected (i.e. the prediction about the upcoming stimulus were not as strong).
These models fit well with findings from pupillometry studies. For example, Nassar et al. showed in a predictive inference task that the baseline pupil dilation reflected the uncertainty about the underlying mean of the noisy data that the subjects were presented with. Phasic pupil responses in their study scaled with the noisiness or distance from the predicted value of single events, i.e. the unexpected uncertainty (Nassar et al., 2012). Interestingly, pupil dilation also predicted the weighting of evidence in the subsequent trials. Larger phasic pupil dilation, coinciding with bigger prediction errors, correlated with an increase in learning rate on the next trials. These results are consistent with earlier findings that showed that baseline pupil diameter was larger before exploratory choices than exploitative choices (Jepma and Nieuwenhuis, 2011). In this study it was also found that subjects with a higher baseline pupil diameter were more prone to exploratory behavior. As would be expected from the models, baseline pupil diameter is negatively correlated with phasic pupil dilation on a trial-to-trial basis (Gilzenrat et al., 2010).
So, based on the similarities between models about the LC-NA system and the results of the studies mentioned here (and many others), there is reason to argue for a close interaction between pupil size modulations and the LC-NA system. However, it is not well known what the mechanism underlying this relation is. Measuring pupil size and LC activity simultaneously is a technically challenging and tedious job, and evidence from such research is extremely sparse. Any inferences regarding the link between pupil and LC must rely mostly on evidence from anatomy and retrograde tracing studies. So what does the anatomy and the few simultaneous recording studies tell us?
First of all, famous data from simultaneous intracranial recordings of an LC neuron and eye tracking in one monkey shows an obvious positive relation between tonic activity of a single LC neuron and pupil diameter (Rajkowski et al., 1993; Aston-Jones and Cohen, 2005). Unfortunately, these data have never been published as a full research article
6
(only as a conference abstract) and therefore remain preliminary. Furthermore, measurements of phasic LC or pupil responses were not studied in the experiment. That missing link was accounted for in a more recent experiment by Kalwani and Gold (Kalwani and Gold, 2010). However, this work was still in progress during the publication of the conference abstract and has not yet resulted in a full research article either. In the study activity from 25 LC neurons and pupil diameter was recorded during a countermanding task. Additionally, pupil diameter was recorded from several human subjects performing the same task. Preliminary results suggested that LC neurons fire around both the commitment of making and inhibiting motor responses to task events. The pupil diameter increased around the same events in both the monkey and the human subjects. Although these studies are consistent with each other and support the most prominent hypothesis about the mechanism underlying cognitively induced pupil fluctuations, this work was not completed and therefore lacks the reliability to be key evidence for or against any model or hypothesis.
Recently, a study using a different approach was published. Murphy et al. measured pupil dilation in humans while simultaneously recording BOLD fMRI from the brainstem (Murphy et al., 2014). Many other studies attempted measuring LC BOLD activity before, but measuring fMRI from the brainstem is extremely challenging and conventional techniques are not well suited for the small nuclei it consists of. The study by Murphy et al. has an advantage over these experiments for two reasons: (i) They used a combination of high-precision imaging techniques to locate the LC, including neuromelanin-sensitive imaging, a precise LC atlas and retrospective statistical control for physiological noise. (ii) They measured pupil diameter simultaneously, which in the context of this review is the variable of interest, but can also boost the reliability of the LC BOLD signal since it is hypothesized to closely relate to each other. The experiment consisted of measurements both at rest and while the subjects performed a visual oddball task. The results indicated a relationship between BOLD activity in the LC and pupil diameter during rest. Furthermore, relevant targets in the oddball task were found to evoke phasic modulation of both pupil diameter and LC activity, this had so far only been found with invasive methods in animal studies.
Together, these studies represent an important piece of the puzzle, but more research is needed to accurately and reliably depict the (functional) relationship between pupil diameter and LC activity. Evidence for a structural relationship has to come from anatomical studies, which luckily have been conducted in much higher numbers.
Although thoroughly studied, there is no reliable evidence for a strong direct connection between the LC and the pupil. Some studies claim to have found an inhibitory connection from the LC to the Edinger-Westphal nucleus (EW), which would inhibit pupil constriction and thereby cause the pupil to dilate. For example, a retrograde tracing study in cats found projections from the LC to the EW (Breen et al., 1983). However, the validity of these studies has been questioned (Nieuwenhuis et al., 2011). The NA cells in the cat LC are intermingled with non-LC cells (Chu and Bloom, 1974). With the methods used by Breen et al. it is impossible to say whether the cells they trace to the LC are actually NA cells. Furthermore, a different histochemical tracing study showed that the EW is innervated by NA cells, but not from the LC (Levitt and Moore, 1979). Instead, this NA innervation comes from NA cell groups in the lower medulla.
A direct anatomical connection between the LC and EW is thus unlikely. But some evidence exists for an indirect connection. In both tracing and lesion studies in rats, the
7
LC has been found to project to the paraventricular nucleus (PVN) in the hypothalamus (Cunningham and Sawchenko, 1988; Gaillet et al., 1993; Palkovits et al., 1999). The PVN, in turn, was found to project to the EW in an anterograde tracing study (Geerling et al., 2010). Although only weak connections were found, and it is not clear whether these are excitatory or inhibitory, it is possible that the LC influences the EW, and thus the pupil, though this indirect route. Alternatively, the LC might be connected to the EW via a loop involving the cortex and hypothalamus. The LC has broad projections to the cortex and frontal cortical regions have been found to affect tonic activity in the EW through indirect connections passing the hypothalamus (Szabadi and Bradshaw, 1996). Both these indirect pathways are plausible, but the indirect and diffuse character of these routes is in no way matched to the strong link between pupil diameter and LC activity. It is highly unlikely that these pathways can fully account for that relation.
A different perspective was introduced by Nieuwenhuis et al. (Nieuwenhuis et al., 2011). They also conclude that direct influence of the LC on the motoric control of the pupil, or the other way around, is unlikely. Instead, they propose that both are innervated by the same upstream source: the nucleus paragigantocellularis. This nucleus in the rostral ventrolateral medulla sends strong projections down the intermediolateral cell column of the spinal cord to the superior cervical ganglion, from where the dilator muscles of the pupil (as well as other sympathetic targets) are directly innervated. Microstimulating the PGi triggers pupil dilation, pleading for a causal influence of the PGi on pupil diameter (Koss and Wang, 1972). Furthermore, the PGi was found to have strong excitatory projections to the LC in rats (Aston-Jones et al., 1986) and LC responses to sensory input can be blocked by pharmacologically disrupting the PGi-LC pathway (Ennis and Aston-Jones, 1988), suggesting a key role for the PGi in mediating phasic LC responses to sensory events (Nieuwenhuis et al., 2011).
Importantly, the PGi receives efferent connections from both cortical and subcortical areas and autonomic input from the vagal nerve. Cortical input connections include insular cortex, orbitofrontal and anterior cingulate cortex and medial prefrontal cortex, all of which are involved in cognitive processes associated with the LC-NA system and pupil dilation (e.g. task-relevance and prediction of upcoming sensory input). Additionally, the amygdala might project to the PGi by a pathway through the hypothalamus and periaqueductal grey (Nieuwenhuis et al., 2011). The influence of the vagal nerve on LC-NA activity is supported by studies that used vagal nerve stimulation (VNS), a method normally used as epilepsy treatment (De Taeye et al., 2014). In an animal model of epilepsy VNS was found to not only suppress seizures, but also to increase the hippocampal NA concentration (Raedt et al., 2011). In human subjects, VNS was found to increase the amplitude of the P3 component of the event-related potential after targets in an oddball paradigm (De Taeye et al., 2014), an hypothesized marker of phasic LC-NA activity (Nieuwenhuis et al., 2005). Taken together, the hypothesis that the PGi is the common input that causes the close relation between the LC and pupil diameter is well supported by empirical evidence. Both its efferent and afferent connections make the PGi very well suited to fulfill this role, and pharmacological interventions further support this idea.
One concern that needs to be solved, however, is that the direct connection from the PGi to the LC has been found in rats but not humans. More specifically, a tracing study in post-mortem human brainstems was unable to find any projections from the PGi to the
8
LC (Zec and Kinney, 2001). In their paper, the authors point to two possible explanations for the lack of labeling in the LC: (i) the projection to the LC might have a strict topography in the PGi, meaning that they might be missed when placing the tracer (DiI). Or (ii), the study was conducted on mid-gestational fetal post-mortem brains. It is possible that the connection does not exist in humans at that stage, but comes about at a later developmental stage. The first of these explanations is most likely, but both are plausible. Still, finding this connection in humans would greatly increase the translational validity of the anatomical connections found in animals that the PGi hypothesis builds on.
If the pupil dilator muscles and the LC are not directly connected, but only share a common input, the effects we measure in pupil size are not caused by the LC. Pupil diameter can merely serve as a reporter variable for LC activation. This should be taken into account when measuring pupil diameter, but does not necessarily have to be a problem. When pupil size is found to reliably coincide with LC activation its role as reporter variable of the LC-NA system is valid. As mentioned earlier, the evidence for parallel LC activity and pupil size modulation (from simultaneous recordings) so far is meager, so prudence is called for when interpreting pupil size modulations as LC activity.
It is important to realize that the discovery of any pathway or mechanism does not necessarily exclude the existence of other pathways and mechanisms. If the pupil muscles would be directly innervated by the LC, this would not mean that the pupil muscles could not be innervated by any other source. This notion is even more important when the proposed relationship between two signals is that of a common input, such as the suggested relation between the PGi, the LC and the pupil musculature. If this is the only relation between the LC and the pupil, activation or suppression of the LC by other inputs than the PGi would not be reflected in the pupil. Likewise, any input influencing pupil size that is not the PGi would not affect the LC-NA system.
In the next section I will discuss and evaluate possible alternatives or parallel influences on pupil size, which do not necessarily include the PGi and the LC-NA system.
Alternatives and parallel mechanisms
Alternative mechanisms to the parallel activation of the LC-NA system and the pupil by the PGi can be expected in different forms. Some central pathways that influence pupil size might not include the PGi or the LC, which make them clearly distinct from the default PGi model. Others might be connected to the PGi pathway, so activation of these mechanisms would also activate the PGi and the LC. These alternatives still have implications for the interpretation of the pupil signal though, and especially for the interpretation of neurocognitive effects that coincide with it. Pupil diameter would still be a proxy for NA release, but also for the alternative mechanism, which might be more closely related to the coinciding neural processes.
Alternative pathways can be roughly grouped in four categories: (i) the inputs to the PGi (and not NA release from the LC) cause the cognitive and neurophysiological effects that are linked to pupil size, (ii) the PGi connects to other systems beside the LC-NA
9 system that cause these cognitive
and neurophysiological effects, (iii) there are different mechanisms that affect the pupil dilator muscles, or (iv) there are central mechanisms that influence the pupil constrictor muscles (figure 1). The first and second category leave the link between the pupil and the LC-NA system intact, but have important implications for the (lack of) exclusivity of that link. They open the door for parallel links to other neural mechanisms that possibly play a more prominent role in pupillary control and the coinciding cognitive processes than the LC-NA system.
Some prominent alternatives that fall into one or more of these four categories will be discussed below. The likelihood of the four options playing a part in pupil physiology will be evaluated in the discussion.
Dopamine and the pupil
As a neuromodulator system, the LC-NA system has a temporally specific but spatially global effect on cortical processing. This makes it well suited as a mechanism to explain the global modulation of cortical processing (such as the P3 component of the event-related potential found in EEG studies, Nieuwenhuis et al., 2011) that coincides with pupil dilation. However, there are other neuromodulator systems that have similar characteristics, in particular the midbrain dopamine (DA) system.
The chemical properties of DA are very similar to those of NA (DA is actually the direct precursor of NA): both are catecholamines that were originally also studied as such (Servan-Schreiber et al., 1990). Dopamine is released from the ventral tegmental area (VTA) and the substantia nigra (SN), both located in the midbrain. Similar to the LC-NA system, the midbrain DA system is reciprocally connected to frontal cortical areas and the limbic system and sends widespread projections to the cortex (Krichmar, 2008). Given all these similarities: how likely is it that the DA system also has a similar relation to pupil size?
Apart from the similar connectivity and chemical structure, the role that the DA system plays in cognition also overlaps with that of the LC-NA system and the pupil. Intracranial recordings in monkeys showed that midbrain dopamine neurons do not only respond to reward, but also to cues that were predictive of a reward. More specifically, when a
10
reward was fully predicted by a cue, the DA neurons no longer responded to the cue at all and instead only fired after the onset of the cue (Schultz et al., 1997). In other words, DA system activity reflects the prediction error about reward. Follow-up experiments showed that this DA response affects learning and scale with the level of temporal uncertainty about the reward (Hollerman and Schultz, 1998; Fiorillo et al., 2008), much like the LC-NA system behaves with respect to behaviorally relevant stimuli (Aston-Jones and Cohen, 2005).
Some studies measuring pupil responses to rewarding stimuli in humans have been performed. For example, in one study subjects were given rewards when they were lucky to make a ‘right’ decision to a random question (the content of the decision was either the timing of the button press or the digit they chose) (Einhäuser et al., 2010). In other studies, subjects won or lost money in an auditory gambling task (Preuschoff et al., 2011) or an Iowa gambling task (Lavín et al., 2014). In yet another study, subjects were given the choice between exploiting a source of reward or exploring new ones (Gilzenrat et al., 2010). All these studies showed that pupil size is modulated by rewards or the possibility of future rewards. However, all authors show that the pupil response might not be so much to the reward itself, but rather the uncertainty or surprise about the reward. This, combined with other studies linking the LC-NA system to the pupil, leads them to argue that it is not DA release but NA release that is involved in reward uncertainty and pupil size modulation.
Two points of criticism weaken their conclusion. First, the exclusion of DA influence on the pupil only holds when there is no physiological relationship between the pupil and midbrain dopamine neurons, which is far from established. Second, the physiological relationship between reward uncertainty and the LC-NA system is not well studied (Preuschoff et al., 2011), so strong claims about this should be avoided. DA neurons, on the other hand, are known to respond to the uncertainty about upcoming rewards in monkeys in a very similar context to the experiments mentioned above (Fiorillo et al., 2003), leaving the door wide open to the influence of DA on pupil size.
Publications about pupil responses in dopamine-related tasks such as reward prediction or temporal difference learning are rare. One fMRI study, that aimed at elucidating the brain structures in reward-based learning, measured pupil size an ‘objective index of learning’ (O’Doherty et al., 2003). The published analyses on these pupil data were not very elaborate, but show that pupil dilation was significantly higher after conditioned stimuli signaling reward compared to stimuli not coupled to a reward. This pupil size modulation could still be caused by NA release triggered by the arousal that the reward brings about, but a role of dopamine in this process can certainly not be excluded. Anatomical studies would be the most straightforward option to test this possibility.
It is not easy to reliably distinguish dopaminergic cell fibers from NA fibers. Not only are dopamine and NA chemically similar, dopamine is also a precursor for NA synthesis. To trace dopaminergic innervation in parts of the brain where NA cells are abundant it is necessary to discriminate between actual dopaminergic cells and NA cells that contain dopamine as a precursor. Kitahama et al. performed dopamine immunohistochemistry with a highly specific dopamine antiserum to study dopamine innervation in the rat brainstem (Kitahama et al., 2000). To exclude the possibility that they measured dopamine as a precursor in NA cells, they compared their results to a separate experiment in which they used a specific histochemical tracer for NA (namely, an inhibitor of dopamine-beta-hydroxylase, the enzyme that transforms DA into NA). Among
11
the many areas that were found to be innervated by dopaminergic cells were two structures that are of particular interest for this review: the LC and the PGi. The histochemical substances in this study were not injected locally; instead they were injected in the ascending aorta during anesthesia. This means that, based on these results alone, not much can be said about the origin of the dopaminergic fibers found in the LC and the PGi. However, other studies that used retrograde tracing can complement these findings. In a study were horseradish peroxidase was injected into the rat PGi, the substantia nigra pars reticulata turned up in the retrograde targets (Andrezik et al., 1981). This finding was not replicated in a later study , in which the authors suggested the precise location of the target cells was not in the substantia nigra, but the oculomotor nucleus (Van Bockstaele et al., 1993). In yet another study, cholera toxin B subunit was injected in rat LC. Among the targets, dopamine cells in the midbrain were found (Luppi et al., 1995).
This histochemical evidence suggest that the midbrain dopamine system might influence both the LC directly, and the PGi, through which both the LC and the pupil will be influenced. It is not sufficient to support a link between dopamine and the pupil. More anatomical research and especially simultaneous recordings of the pupil and midbrain dopamine neurons should be performed. Nevertheless, this research combined with the modest body of pupil reward studies definitely does not speak against a role of dopamine in pupil diameter control.
Serotonin and the pupil
Serotonin is another neuromodulator with widespread effects in the central nervous system. The serotonin system is associated with many and diverse big themes in cognitive neuroscience, including depression, obsession, anxiety, behavioral inhibition and punishment. It has so far been proven to be hard to pinpoint how serotonin is involved in these processes. Besides it’s relatively poorly understood and ill-defined role in cognition, its physiological mechanisms are also much less well known than those of noradrenaline and dopamine. Measuring rapid serotonin level fluctuations is challenging because good markers are hardly available and the absolute concentrations of serotonin in the brain are typically much lower than for example dopamine levels (Dayan and Huys, 2009). The serotonin system is usually not incorporated in models or hypotheses about pupil diameter modulations, so why would we study the relation that serotonin has to pupil diameter?
A valid reason to do this is that pharmacological manipulation of serotonin levels or serotonin receptor availability has an effect on pupil diameter. In what way this happens, however, remains a question. In mice and rats, an agonist (8-OH-DPAT) of a subtype of serotonin receptor (5-HT1A) causes the pupil to dilate (Prow et al., 1996; Yu et al., 2004),
while in humans different agonists (buspirone and lesopitron) of the same serotonin receptor subtype were found to constrict the pupil (Phillips et al., 1999). This discrepancy might be caused by the use of different agonists that have different effects on other receptors than the 5-HT1A–receptor. For example, 8-OH-DPAT, besides a
5-HT1A-agonist, is also an adrenergic antagonist (Prow et al., 1996). However, this
antagonism is thought to be too weak to have a substantial effect (Yu et al., 2004). Furthermore, this possibility does not corroborate with the hypothesized link between the
12
LC-NA system and the pupil dilator muscles: if 8-OH-DPAT would antagonize NA, the pupil would constrict and not dilate.
Taken together, this pharmacological evidence does not clarify the role of serotonin in pupil diameter modulation, but given the complicated relation between serotonin and the pupil it seems unlikely that either the constrictor or dilator muscles are directly innervated by serotonergic neurons. Instead, the serotonergic system could influence pupil size indirectly by working on one or more of the important neural hubs in the pupil circuitry.
There is anatomical evidence that supports this. Like the other neuromodulatory brainstem systems, the serotonin system has widespread ascending projections though the whole brain. But the serotonin system is not as clearly organized as the noradrenaline or dopamine systems. Serotonin is released in the central nervous system from seven or eight different structures in the brainstem and midbrain, together forming the raphe nuclei. Some of these nuclei have turned up in tracing studies concerning structures involved in pupil size modulation. Most importantly, in the previously mentioned tracing study in human post-mortem brainstems by Zec and Kinney, evidence was found for an anatomical connection between the PGi and two caudal raphe nuclei (Zec and Kinney, 2001). The experimenters injected the labeling substance DlI into the PGi, resulting in both labeled cells and fibers in the two nuclei (nucleus raphe obscurus and pallidus). In another study that was mentioned before, injecting a retrograde tracer (cholera-toxin B subunit) was injected in rat LCs (Luppi et al., 1995). The PGi was most prominently targeted, but cells in the dorsal and median raphe nuclei were also among the targets. It could be that the connections between the LC and the raphe nuclei go through the PGi and reflect the same type of connection that Zec and Kinney found, but because different raphe nuclei were targeted in the two studies it is more likely that they correspond to separate pathways. In any case, both these studies suggest that serotonin affects the pupil indirectly, by activating the PGi and (indirectly) the LC-NA system.
This indirect role of serotonin would mean that pupil size modulation reflects LC-NA activity more closely then serotonin release. This is also used as an argument that pupil size modulation, although influenced by serotonin, can still be interpreted as a proxy for NA release (Einhäuser et al., 2008). However, restraint is called for interpreting brain activity that coincides with pupil size modulation as an effect of NA release. Serotonin has similarly widespread effects on cortical processing and might be the factor that (partly) drives the LC-NA system, the pupil and processes in the cortex.
Central influence on pupil constriction
So far, I focused primarily on mechanisms that result in pupil dilation by stimulating the dilator muscles. As mentioned before, the literature about central effects on pupil diameter usually also focuses solely on dilation. Active pupil constriction is only studied in the context of brainstem reflex loops, namely the pupillary light reflex and the accommodation reflex. It is not unlikely, however, that pupil constriction is also influenced by cognitive processes. For example, it has been found that dilation could be partly attributed to active inhibition of the constrictor pathway in response to cognitive demand (Steinhauer et al., 2004). Excitation of this centrally originating pathway would thus lead to active pupil constriction.
13
While many pupil studies do not focus on constriction and might even have missed possible constriction effects in their data, some studies have reported these active constriction effects. For example, Naber et al. found that pupil constriction was larger during presentation of novel compared to familiar natural scenes and that the amount of this constriction is predictive of the memory performance on the matching stimulus (Naber et al., 2013). The authors propose that pupil constriction can function as marker for stimulus novelty. In a different experiment subjects were presented with photographs of natural scenes containing the sun (Naber and Nakayama, 2013). Pupil constriction to these stimuli was found to be higher compared to (i) the same image, but inverted, (ii) a photograph of a natural scene not containing the sun, but with similar low-level image statistics such as luminance level, and (iii) cartoon images containing the sun. The authors argue that the increased constriction to upright photographs containing the sun is due the ‘sun effect’, meaning that a picture of the sun is perceived as brighter than it actually is. This effect is suggested to be less strong or lacking in the control conditions because (i) higher-order processing of the stimulus was disrupted by inverting the image, (ii) the image lacked the sun, or (iii) a less realistic stimulus was used (Naber and Nakayama, 2013). In both studies a role of higher-order central processing is proposed to influence pupil constriction.
The studies by Naber et al. do not propose a strong neurophysiological explanation for the constriction phenomena they reported. The task used in the memory experiment is known to involve a network including the medial temporal lobe, ventrolateral prefrontal cortex and the dorsal posterior parietal cortex (Naber et al., 2013). For the ‘sun effect’ experiment a neurophysiological mechanism is even harder to point out. Turning the image upside down resulted in the largest attenuation of pupil constriction. Other experiments found that inverted images of faces lead to a decrease in fusiform face area activity, hinting at a role of higher order visual areas in this process (Naber and Nakayama, 2013).
Although direct evidence for any of these suggestions is lacking, intracranial stimulation studies in monkeys provide indirect evidence for the involvement of some regions involved in the explicit memory task. Jampel estimated pupil size changes (among other measures) following faradic stimulation and lesioning of different cortical areas (Jampel, 1960). Stimulation of several locations in the medial temporal lobe was found to constrict the pupil. This could include the area involved in the memory task used by Naber et al. (Naber et al., 2013), but the antiquated techniques used by Jampel and the big gap between human memory research that linked the medial temporal lobe to memory and monkey research make it weak evidence at best.
Stimulation and lesioning of other areas that were found to induce constriction or accommodation in Jampel’s experiments included frontal and occipital areas (specifically the primary visual cortex and area MT) (Jampel, 1960). Stimulating area MT resulted in all three characteristics of the accommodation response (pupil constriction, curvature of the lens and convergence of the eyes) and lesioning this area in monkeys was found to defect detection of nearby objects (Ungerleider et al., 1984). No direct anatomical pathway between area MT and the muscles of the eye is known, but these results suggest that area MT is in some way related to the accommodation response.
Another possible link between cortical processing and pupil constriction was found in an anterograde autoradiography study in monkeys (Hartmann-von Monakow et al., 1979). In this experiment a tracer was injected into Brodmann areas 4 and 6 (primary motor
14
cortex and premotor cortex, respectively). Among the targets of the primary motor cortex injections the experimenters discerned the Edinger-Westphal nucleus, the primary relay hub for innervation of the pupil constrictor muscle. Although the precise pathway remains unknown, this finding implies an anatomical link between premotor cortex and the pupil constrictor muscle. One year later, one of the authors published another tracing study in monkeys, in which they were able to look in more detail at the EW nucleus (Akert et al., 1980). In this study the brain area that was previously identified as the EW nucleus could be clearly separated into two functionally separate parts, one innervating the pupil constrictor muscles (the actual EW nucleus) and the other consisting of descending fibers that are possibly part of the periaqueductal gray. The target cells that were found in the earlier study by Hartmann-von Monakow et al. were situated in this latter part. This falsifies the previous evidence and makes a direct anatomical connection between primary motor cortex and the pupil constrictor muscles highly unlikely.
Not only is the evidence for central influences on pupil constriction I mentioned so far somewhat vague and often of questionable reliability, it is all very unspecific about the functional mechanism underlying these influences as well. A possible mechanism to breach this gap is acetylcholine release. The neurons that innervate the constrictor muscles of the pupil are cholinergic (Huhtala et al., 1976; Loewenfeld, 1993). Furthermore, administration of acetylcholine antagonists scopolamine and mecamylamine dilate the pupil in healthy human subjects (Little et al., 1998), i.e. they inhibit the pupil constrictor muscles activation and thereby facilitate its antagonist: the dilator muscles. It could be that pupil constriction is under control of the cholinergic basal forebrain nuclei (Yu, 2012).
Research on if and how central processes cause pupil constriction is very rare, so more studies are needed to elucidate this mater. The pieces of the puzzle described in this section indicate that this relationship does exist; future studies on this subject promise to lead to interesting new findings.
Discussion
Research on the neural mechanisms that link cognitive processes to pupil size modulation has focused mainly on the noradrenaline system. The results from these experiments speak for a key role of this system in centrally triggered pupil dilation. Especially the ‘common input’-hypothesis proposed by Nieuwenhuis et al. is a promising development (Nieuwenhuis et al., 2011). The nucleus paragigantocellularis has both the inputs (from, among others, frontal regions involved in task monitoring and surprise) and the outputs (including connections to the LC and the pupil dilator muscles) to be a very suitable structure to coordinate the parallel modulation of LC activity and pupil size that has been found in a few animal studies (Rajkowski et al., 1993; Kalwani and Gold, 2010). Pharmacological interventions further support this (Ennis and Aston-Jones, 1988). Given this evidence it is fair to assume activity in the LC-NA system is reflected in pupil size.
Well supported as it may be, there is no reason why the LC-NA system should have a monopoly on influencing pupil size from the central nervous system. Several alternatives
15
or parallel routes still possibly exist, including the options named earlier. First of all, it might be the case that the PGi does play the role it is suggested to play by Nieuwenhuis et al., but it is not the NA release from the LC that causes the cognitive and neural effects. Instead, those effects might be brought about by a different mechanism that works both on the PGi and the cortical processes involved in the cognitive process. In other words, pupil size modulation is related to LC-NA activity, but the neurocognitive effects that coincide with it are not.
Evidence for this option comes from both the serotonin system and the dopamine system. Serotonergic innervation has not been found in the pupil, but pupil size is affected by pharmacological manipulations of the serotonin system (Prow et al., 1996; Phillips et al., 1999; Yu et al., 2004). Additionally, connections have been found between raphe nuclei, the PGi and the LC (Luppi et al., 1995; Zec and Kinney, 2001). Serotonin has a strong influence on cognition and profound and global effects on cortical processing (Lidov et al., 1980; Dayan and Huys, 2009). It is therefore not unlikely that the cognitive and neural effects that coincide with pupil size modulations, which are now often attributed to NA release in the cortex, are at least partially due to serotonin.
The dopamine system might influence the LC and the pupil in way that is similar to serotonin. Both the PGi and the LC are innervated by dopaminergic fibers (Kitahama et al., 2000). The origin of these fibers remains somewhat elusive, but there is some evidence for dopaminergic innervation of the PGi from the substantia nigra and the LC from dopaminergic neurons from different midbrain structures (Andrezik et al., 1981; Luppi et al., 1995). So in the same way as for serotonin, the dopamine system might be responsible for neurocognitive effects that are now attributed to the LC-NA system. The high similarity between the cognitive processes associated with the dopamine and noradrenaline system does not help to disentangle the physiological mechanisms.
Second, the efferent connections of the PGi might include other targets than the LC that cause the neurocognitive effects in the cortex. In the same way as the afferent connections to the PGi option described above, the pupil size would reflect LC-NA activity, but the neurocognitive effects would not. There is not much evidence that this is the case however. Besides the LC and the pupil dilator muscles, afferents of the PGi include mostly peripheral targets that control sympathetic organs such as the sweat glands and the heart and the blood vessels (Nieuwenhuis et al., 2011). It is unlikely that the efferent connectivity of the PGi would cause the pupil-linked neurocognitive effects any other way than by activation of the LC-NA system.
Third, it might be that there are pathways to the pupil dilator muscles that bypass the PGi. In some studies, direct connections from the hypothalamus to the autonomic targets that innervate the pupil dilator muscles (superior cervical ganglion) are proposed to be responsible for pupil dilation (Beatty and Lucero-Wagoner, 2000). This suggestion is based on quite early studies and is not very well specified. It has been found, in a very early study, that connections from the hypothalamus inhibit the Edinger-Westphal nucleus, which would also lead to pupil dilation (Lowenstein and Loewenfeld, 1961; Szabadi and Bradshaw, 1996). This might account for the suggestion that the pupil dilator muscles are innervated by hypothalamic structures. A more promising option for pupil dilation that bypasses the PGi is projections from the LC to the pupil. Although no reliable evidence has been found for direct connections (Nieuwenhuis et al., 2011), manipulation of LC-NA activity does affect pupil size. Phillips et al. orally administered an alpha2-adrenoreceptor agonist (clonidine) and antagonist (yohimbine) to human
16
subjects, which decrease and increase NA release from the LC, respectively (Phillips et al., 2000). Inhibition of the LC was found decrease pupil size, whereas excitation increased the pupil size. However convincing, this evidence cannot strongly plead for direct innervation of the pupil by the LC without anatomical evidence for such a link. These effects might also reflect one of previously mentioned indirect pathways between the LC and the pupil, involving the frontal cortex (Szabadi and Bradshaw, 1996) and/or hypothalamic nuclei such as the paraventricular nucleus (Palkovits et al., 1999; Geerling et al., 2010). Alternatively, the effects might be explained by two other mechanisms: (i) indirect loops involving LC projections to higher brain areas that in turn project back to the PGi, indirectly resulting in pupil size modulation. Or (ii), the adrenergic receptors in the pupil machinery itself are targeted by the pharmacological agents, meaning that the effect is peripheral, not central (Nieuwenhuis et al., 2011).
Finally, pupil constriction might reflect neurocognitive processes in a way that is not mediated by the PGi or any other pathway involved in pupil dilation. Pupil constriction has been found and sometimes reported in cognitive tasks (Naber and Nakayama, 2013; Naber et al., 2013), suggesting a role of the central nervous system in this process. Electrically stimulating and lesioning of several areas of the cortex was found to affect pupil constriction, including one area (medial temporal lobe) that is thought to be involved in one of the cognitive tasks used by Naber et al. (Jampel, 1960). The mechanism by which this happens is still very unclear, although a role of the cholinergic forebrain centers is at least supported by the fact that innervation of the pupil constrictor muscles is cholinergic and acetylcholine antagonists dilate the pupil (Loewenfeld, 1993; Little et al., 1998). Alternatively, phasic pupil constriction might reflect relaxation of the dilator muscles by inhibition of the PGi. Given the effects that acetylcholine manipulation has, it is not likely that this is the only central factor that accounts for pupil constriction. Taken together, all the evidence that is discussed so far makes a solid case for central influence on pupil diameter. The majority of the studies couple pupil size to neuromodulator systems, usually the noradrenaline system. This implies that the neural process that causes the pupil size to modulate also has widespread effects in the brain. Alternatively, pupil size could be influenced by functional processes that target the pupil musculature specifically. The only well-established pathway for pupil dilator innervation is the one involving the PGi (Nieuwenhuis et al., 2011). If this would be the only pathway to these muscles, contraction would always be accompanied by excitatory input to the LC, and therefore be closely related to neuromodulation. However, other pathways that bypass the PGi, such as the constriction pathway, do not necessarily have to be neuromodulatory. For example, the ‘sun effect’ that Naber et al. found could be interpreted as a protection reflex, caused by the perceived brightness of the stimulus, that is specifically coupled to pupil constriction and does not have to have consequences in the brain. Simultaneous recordings of the pupil and brain activity (either non-invasive measures of cortical activity or invasive brainstem recordings) could help to elucidate this matter.
Even though some central processes that affect pupil size might not be related to neuromodulation, many other are. As is quite clear by now, noradrenaline is the central player here. But how exclusive is its role? In many publications of experiments studying pupil size, the authors go quite far in interpreting the pupil size modulations they find directly and purely as NA release from the LC. Given the combination of other
17
neuromodulators that proved to influence the pupil and the scarcity of physiological evidence (from simultaneous recording or pharmacological interventions) we should still be cautious when interpreting the neural process that causes the pupil to change size. When using tasks that have been found to trigger activity in the LC (e.g. oddball target detections tasks (Aston-Jones et al., 1997)) it is maybe permissible to interpret the behavior of the pupil as NA release. But when using experimental designs that have not been strongly linked to the LC-NA system directly (e.g. bistable perception or decision making tasks, Hupé and Lorenceau, 2009; De Gee et al., 2014), the evidence that pupil size reflects NA release in those tasks is just not strong enough. Direct recordings of LC activity (either intracranially in animals or using sophisticated fMRI techniques in humans) during those tasks would strengthen the case, preferably while measuring pupil size at the same time. Especially in the pupil constrictor pathway there is room for other mechanisms (e.g. acetylcholine release from the forebrain, Yu, 2012).
Comparing the evidence from the studies discussed in this review we can divide the subject into two main questions, (i): can we reliably say that pupil diameter reflects LC-NA activity? And (ii): based on this evidence, what can we say about the neural mechanisms that cause the cognitive effects that coincide with pupil size modulation? Or more specifically, can we infer from pupil modulation that NA causes these cognitive effects? The first question has already been discussed: the evidence for a close link between pupil size and LC-NA activity is strong, although probably not exclusive.
The second question is often answered positively: if pupil dilation coincides with NA release it must be NA that causes the neurocognitive effects in the brain. This statement in not only logically invalid (an apple is a fruit, but not all fruits are apples), but there is also quite substantial evidence that other influential mechanisms affect (and are therefore linked to) pupil size. Evidence from anatomical and pharmacological studies shows that both the dopamine system and the serotonin system are linked to pupil dilation (Andrezik et al., 1981; Luppi et al., 1995; Prow et al., 1996; Phillips et al., 1999; Kitahama et al., 2000; Yu et al., 2004) and have extensive effect on the brain and cognition (Schultz, 2006; Dayan and Huys, 2009). This is further complicated by the intricate relation between all these systems: both the serotonin system and the dopamine system have been found to project to the NA system (Luppi et al., 1995; Kitahama et al., 2000; Zec and Kinney, 2001), but serotonin is also thought to connect to the dopamine system (Hervé et al., 1987). This is probably still only part of the relations that exist between these and other systems acting on the pupil. It is therefore unlikely that NA release from the LC is always the dominant factor in causing the neurocognitive effects that coincide with pupil dilation.
Pupil constriction can be interpreted as dilation inhibition, thus allowing the LC-NA link to the pupil to generalize to pupil constriction. But there also seem to be cognitive processes that actively trigger the pupil constrictor muscles. Central acetylcholine release (Loewenfeld, 1993; Little et al., 1998) and processes that target the constrictor muscle specifically (Jampel, 1960; Naber and Nakayama, 2013) seem to be more promising candidate for this than the LC-NA system. In short, pupil size (especially dilation) can be cautiously used as a reporter variable for NA release from the LC, but based on the pupil alone we should refrain from strong claims about the role of NA in neurocognitive processes.
To further extend our understanding of relation between pupil size, cognition and the neural processes that underlie them, more research is needed. Especially simultaneous
18
recording of the pupil and brain activity could help elucidate this matter. The first studies recording pupil size and intracranially from the LC have led to interesting results (Rajkowski et al., 1993; Kalwani and Gold, 2010). More of these studies, including recordings from other areas like the dopamine midbrain nuclei, are necessary. The rapid development of more sophisticated and sensitive fMRI techniques promises to enable better measurements of small neuromodulatory centers in the brainstem and the midbrain, allowing for noninvasive simultaneous recordings in humans. Other noninvasive measures of cortical processing, like EEG and MEG, might also proof to be helpful in understanding how pupil size relates to neural mechanisms. Pharmacological interventions will also be valuable. Directly manipulating levels of specific neurotransmitters will greatly benefit our understanding of their influence on pupil size and cortical processing. One more interesting and promising direction of further research on this subject is the pupil musculature. Measuring not the net pupil size, but the activity in the constrictor and the dilator muscles separately, invasively with intramuscular electrodes or non-invasively with high-definition photography, by manipulating ambient light intensity or by local pharmacology with eye drops (Steinhauer et al., 2004) could be extremely valuable in disentangling the different mechanisms and pathways that influence pupil diameter.
19 Literature
Akert K, Glicksman M a, Lang W, Grob P, Huber A (1980) The Edinger-Westphal nucleus in the monkey. A retrograde tracer study. Brain Res 184:491–498.
Andrezik JA, Chan-palay V, Palay SL (1981) The Nucleus Paragigantocellularis Lateralis in the Rat: Demonstration of Afferents by the Retrograde Transport of Horseradish Peroxidase. Anat Embryol (Berl):373–390.
Aston-Jones G, Cohen JD (2005) An integrative theory of locus coeruleus-norepinephrine function: adaptive gain and optimal performance. Annu Rev Neurosci 28:403–450.
Aston-Jones G, Ennis M, Pieribone V, Nickell W, Shipley M (1986) The brain nucleus locus coeruleus: restricted afferent control of a broad efferent network. Science (80- ) 234:734–737.
Aston-Jones, Rajkowski J, Kubiak P (1997) Conditioned responses of monkey locus coeruleus neurons anticipate acquisition of discriminative behavior in a vigilance task. 80:697–715.
Beatty J, Lucero-Wagoner B (2000) The pupillar system. In: Handbook of
Psychophysiology (Cacioppo JT, Tassinary LG, Berntson G, eds), pp.142–162. Cambridge: Cambridge University Press.
Van Bockstaele EJ, Akaoka H, Aston-Jones G (1993) Brainstem afferents to the rostral (juxtafacial) nucleus paragigantocellularis: integration of exteroceptive and
interoceptive sensory inputs in the ventral tegmentum. Brain Res 603:1–18.
Breen LA, Burde RM, Loewy AD (1983) Brainstem connections to the Edinger-Westphal nucleus of the cat: a retrograde tracer study. Brain Res 261:303–306.
Chu N-S, Bloom FE (1974) The catecholamine-containing neurons in the cat
dorsolateral pontine tegmentum: distribution of the cell bodies and some axonal projections. Brain Res 66:1–21.
Cunningham ET, Sawchenko PE (1988) Anatomical specificity of noradrenergic inputs to the paraventricular and supraoptic nuclei of the rat hypothalamus. J Comp Neurol 274:60–76.
Dayan P, Huys QJM (2009) Serotonin in affective control. Annu Rev Neurosci 32:95– 126.
Dayan P, Yu AJ (2006) Phasic norepinephrine: a neural interrupt signal for unexpected events. Network 17:335–350.
20
Drake RL, Vogl AW, Mitchell AWM (2009) Gray’s anatomy for students. London: Elsevier Health Sciences.
Einhäuser W, Koch C, Carter OL (2010) Pupil dilation betrays the timing of decisions. Front Hum Neurosci 4:18.
Einhäuser W, Stout J, Koch C, Carter O (2008) Pupil dilation reflects perceptual
selection and predicts subsequent stability in perceptual rivalry. Proc Natl Acad Sci U S A 105:1704–1709.
Eldar E, Cohen JD, Niv Y (2013) The effects of neural gain on attention and learning. Nat Neurosci 16:1146–1153.
Ennis M, Aston-Jones G (1988) Activation of locus coeruleus from nucleus
paragigantocellularis: a new excitatory amino acid pathway in brain. J Neurosci 8:3644–3657.
Fiorillo CD, Newsome WT, Schultz W (2008) The temporal precision of reward prediction in dopamine neurons. Nat Neurosci 11:966–973.
Fiorillo CD, Tobler PN, Schultz W (2003) Discrete coding of reward probability and uncertainty by dopamine neurons. Science 299:1898–1902.
Gaillet S, Alonso G, Le Borgne R, Barbanel G, Malaval F, Assenmacher I, Szafarczyk A (1993) Effects of Discrete Lesions in the Ventral Noradrenergic Ascending Bundle on the Corticotropic Stress Response Depend on the Site of the Lesion and on the Plasma Levels of Adrenal Steroids. Neuroendocrinology 58:408–419.
De Gee JW, Knapen T, Donner TH (2014) Decision-related pupil dilation reflects upcoming choice and individual bias. Proc Natl Acad Sci U S A 11:618–625. Geerling JC, Shin J-W, Chimenti PC, Loewy AD (2010) Paraventricular hypothalamic
nucleus: axonal projections to the brainstem. J Comp Neurol 518:1460–1499. Gilzenrat MS, Nieuwenhuis S, Jepma M, Cohen JD (2010) Pupil diameter tracks
changes in control state predicted by the adaptive gain theory of locus coeruleus function. Cogn Affect Behav Neurosci 10:252–269.
Hartmann-von Monakow K, Akert K., Künzle H. (1979) Brain Projections of Precentral and Premotor Cortex to the Red Nucleus and Other Midbrain Areas in Macaca Fascicularis. Exp Brain Res 34:91–105.
Hervé D, Pickel VM, Joh TH, Beaudet A (1987) Serotonin axon terminals in the ventral tegmental area of the rat: fine structure and synaptic input to dopaminergic neurons. Brain Res 435:71–83.
21
Hess EH, Polt JM (1960) Pupil Size as Related to Interest Value of Visual Stimuli. Science (80- ) 132:349–350.
Hollerman JR, Schultz W (1998) Dopamine neurons report an error in the temporal prediction of reward during learning. Nat Neurosci 1:304–309.
Huhtala A, Tervo T, Huikuri KT, Palkama A (1976) Effects of denervations on the acetylcholinesterase-containing and fluorescent nerves of the rat iris. Acta Ophthalmol 54:85–98.
Hupé J, Lorenceau J (2009) Pupil dynamics during bistable motion perception. 9:1–19. Jampel RS (1960) Convergence, divergence, pupillary reactions and accommodation of
the eyes from faradic stimulation of the macaque brain. J Comp Neurol 115:371– 399.
Jepma M, Nieuwenhuis S (2011) Pupil diameter predicts changes in the exploration-exploitation trade-off: evidence for the adaptive gain theory. J Cogn Neurosci 23:1587–1596.
Kalwani RM, Gold JI (2010) Temporal relationship between different modes of LC firing and pupil diameter in monkeys performing a countermanding task. In: Society for Neuroscience, pp.2–3.
Kitahama K, Nagatsu I, Geffard M, Maeda T (2000) Distribution of dopamine-immunoreactive fibers in the rat brainstem. J Chem Neuroanat 18:1–9.
Koss MC, Wang SC (1972) Brainstem loci for sympathetic activation of the nictitating membrane and pupil in the cat. Am J Physiol 222:900–905.
Krichmar JL (2008) The Neuromodulatory System: A Framework for Survival and Adaptive Behavior in a Challenging World. Adapt Behav 16:385–399.
Lavín C, San Martín R, Rosales Jubal E (2014) Pupil dilation signals uncertainty and surprise in a learning gambling task. Front Behav Neurosci 7:218.
Levitt P, Moore RY (1979) Origin and organization of brainstem catecholamine innervation in the rat. J Comp Neurol 186:505–528.
Lidov HGW, Grzanna R, Molliver ME (1980) The serotonin innervation of the cerebral cortex in the rat—an immunohistochemical analysis. Neuroscience 5:207–227. Little JT, Johnson DN, Minichiello M, Weingartner H, Sunderland T (1998) Combined
nicotinic and muscarinic blockade in elderly normal volunteers: cognitive, behavioral, and physiologic responses. Neuropsychopharmacology 19:60–69.
22
Loewenfeld IE (1993) The pupil: Anatomy, physiology, and clinical applications, 2nd ed. Oxford: Butterworth-Heinemann.
Lowenstein O, Loewenfeld IE (1961) Influence of retinal adaptation upon the pupillary reflex to light in normal man. II. Effect of adaptation to dim illumination upon pupillary reflexes elicited by bright light. Am J Ophthalmol 51:644–654.
Luppi PH, Aston-Jones G, Akaoka H, Chouvet G, Jouvet M (1995) Afferent projections to the rat locus coeruleus demonstrated by retrograde and anterograde with cholera-toxin B subunit and Phaseolis Vulsgaris leucoagglutinin. Neuroscience 65:119–160.
Murphy PR, Connell RGO, Sullivan MO, Robertson IH, Balsters JH (2014) Pupil Diameter Covaries With BOLD Activity in Human Locus Coeruleus. Hum Brain Mapp 00.
Naber M, Frässle S, Rutishauser U, Einhäuser W (2013) Pupil size signals novelty and predicts later retrieval success for declarative memories of natural scenes. J Vis 13:1–20.
Naber M, Nakayama K (2013) Pupil responses to high-level image content. J Vis 13:1– 8.
Nassar MR, Rumsey KM, Wilson RC, Parikh K, Heasly B, Gold JI (2012) Rational regulation of learning dynamics by pupil-linked arousal systems. Nat Neurosci 15:1040–1046.
Nieuwenhuis S, Aston-Jones G, Cohen JD (2005) Decision making, the P3, and the locus coeruleus-norepinephrine system. Psychol Bull 131:510–532.
Nieuwenhuis S, De Geus EJ, Aston-Jones G (2011) The anatomical and functional relationship between the P3 and autonomic components of the orienting response. Psychophysiology 48:162–175.
O’Doherty JP, Dayan P, Friston K, Critchley H, Dolan RJ (2003) Temporal difference models and reward-related learning in the human brain. Neuron 38:329–337. Palkovits M, Baffi JS, Pacak K (1999) The Role of Ascending Neuronal Pathways in
Stress-Induced Release of Noradrenaline in the Hypothalamic Paraventricular Nucleus of Rats. 11:529–539.
Phillips MA, Szabadi E, Bradshaw CM (1999) The effects of the novel anxiolytic drug lesopitron, a full and selective 5-HT1A receptor agonist, on pupil diameter and oral temperature in man: comparison with buspirone. J Psychopharmacol 13:391–397.
