THE HUMAN AMYGDALA IS INVOLVED IN GENERAL BEHAVIORAL RELEVANCE DETECTION: EVIDENCE FROM AN EVENT-RELATED FUNCTIONAL MAGNETIC RESONANCE IMAGING Go-NoGo TASK

Abstract

The amygdala is classically regarded as a detector of potential threat and as a critical component of the neural circuitry mediating conditioned fear responses. However, it has been reported that the human amygdala responds to multiple expressions of emotions as well as emotionally neutral stimuli of a novel, uncertain or ambiguous nature. Thus, it has been proposed that the function of the amygdala may be of a more general art, i.e. as a detector of behaviorally relevant stimuli [Sander D, Grafman J, Zalla T (2003) The human amygdala: an evolved system for relevance detection. Rev Neurosci 14:303–316]. To investigate this putative function of the amygdala, we used event related functional magnetic resonance imaging (fMRI) and a modified Go-NoGo task composed of behaviorally relevant and irrelevant letter and number stimuli. Analyses revealed bilateral amygdala activation in response to letter stimuli that were behaviorally relevant as compared with letters with less behavioral relevance. Similar results were obtained for relatively infrequent NoGo relevant stimuli as compared with more frequent Go stimuli. Our findings support a role for the human amygdala in general detection of behaviorally relevant stimuli.

Classically, the amygdala has been regarded as a critical component of the neural circuitry mediating conditioned fear responses as well as sensitivity to intrinsically salient emotional stimuli (LeDoux, 2000; Davis and Whalen, 2001; Ohman and Mineka, 2001). Several research groups have shown that the amygdala contributes to acquisition and expression of conditioned fear responses, both in animals (LeDoux, 2000) and humans (LaBar et al., 1998). In monkeys, bilateral damage to the amygdala was suggested to cause reduction in the fear-inducing potency of predators as well as cause deficits in complex social interactions (Amaral, 2002). Compared with healthy controls, people with selective amygdala lesions were shown to have difficulties recognizing fearful facial expressions (Adolphs et al., 1994), and tended to judge unknown individuals as more trustworthy and approachable (Adolphs et al., 1998). Further, patients with amygdala lesions showed blunted emotional responses during fear conditioning (Buchel and Dolan, 2000), a phenomenon also noted in Alzheimer's patients who have amygdala atrophy (Hamann et al., 2002).

During the past several years the classical view of the amygdala as a threat detector/fear-module has been expanded following multiple reports linking the amygdala to non-fearful emotional facial expressions including happiness (Breiter et al., 1996; Morris et al., 1996; Canli et al., 2002), sadness (Blair et al., 1999), contempt (Sambataro et al., 2006) and anger (Fitzgerald et al., 2006). Further, stimuli that are experienced as arousing or motivating can elicit amygdala responses (Zald, 2003). Also, socially relevant stimuli without explicit emotional valence like novel, neutral faces (Schwartz et al., 2003a,b; Wright and Liu, 2006), untrustworthy faces (Winston et al., 2002; Engell et al., 2007), facial attractiveness (Winston et al., 2007) and eye-gaze (Adams et al., 2003) as well as unpredictable auditory stimuli elicit amygdala responses (Herry et al., 2007).

Together these findings indicate a more general function for the amygdala than in the classical view, and one suggested hypotheses is that the amygdala is an evolved system for relevance detection (Sander et al., 2003).

Detection of relevance is important in order to survive in an environment with multiple simultaneously-occurring sensory stimuli. According to Sander et al. (2003), “an event is relevant for an organism if it can significantly influence (positively or negatively) the attainment of his or her goals, or the maintenance of his or her own well-being.” Their hypothesis integrates the diverse findings from imaging, human lesion and animal studies, but has yet to be directly tested in humans.

The aim of the current study was to explore the role of the human amygdala in general relevance detection using functional magnetic resonance imaging (fMRI) and a modified Go-NoGo task composed of letter and number stimuli. We hypothesized that although simple letter and number stimuli used in Go-NoGo tasks lack emotional valence, manipulating the behavioral relevance of these stimuli so that some stimuli are more behaviorally relevant than others, will be reflected in the pattern of amygdala activation.

EXPERIMENTAL PROCEDURES

Subjects

Twenty-five subjects (14 women) aged 32.8±9.5 years participated in this study. All of our subjects had participated in an earlier blood oxygen-level dependent (BOLD) signal fMRI study, to which they were randomly selected from the Norwegian citizen registration (Statistics Norway). Participants reported no history of neurological or psychiatric disorder, were not on any medication, and all participants provided written informed consent before participating in the study. All participants received an honorarium (500 NOK) for completion of the protocol. The study was conducted at Ullevål University Hospital, Norway, and approved by the Regional Committee for Medical Research Ethics and the Norwegian Data Inspectorate.

Experimental paradigm

A modified Go-NoGo paradigm was developed with emotionally neutral stimuli (see Table 1). Purple letters or numbers were presented serially on a black screen. Each stimulus was displayed for 1 s with a jittered intertrial interval (ITI) lasting 3.5±1 s. During the ITI, a fixation cross appeared on the screen. The task was to press a response button with the index finger of one hand whenever a letter appeared on the screen, except for the letter “t.” For the “t” stimulus participants were instructed to respond with the index finger of the other hand. For example, participants responded with a right index finger button press for every letter except for “t,” to which they responded with their left index finger. Hence, the letter “t” stimulus was instructed to be particularly relevant compared with other letters, which were associated with a pre-potent finger response. The use of hands was counterbalanced across subjects. Numbers interspersed in the sequence demanded no button pressing. As numbers required an inhibition of response, and therefore a behavioral shift compared with standard letters, they also gained behavioral relevance.

Table 1.

Paradigm rationale

Condition		Description
Background letters (e.g. “s”)		Background (instructed to respond with the index finger of one hand for every letter except “t”)
Letter “t”		Salient+Behaviorally Relevant (instructed to respond with the index finger of the other hand)
Letter “r”		Salient+Behaviorally Irrelevant (not instructed about the color change. Hence the response was the same as for background letters)
Numbers (e.g. “5”)		Salient+Behaviorally Relevant (instructed not to press the button)

To study the response to salient stimuli per se, a green “r” was interspersed in the sequence of stimuli. The letter “r” was made green for two reasons. First, it was perceptually salient compared with the other stimuli, and by using this in the contrasts, we wanted to eliminate the salience factor that by instructions are given both to letter “t” and to some degree numbers. Secondly, the color change of “r” was irrelevant (same behavior), hence we could compare relevant information (letter “t” required a hand shift) with irrelevant information (the color change of “r”). A total of 80 trials (20 “t,” 20 “r,” 20 other letters, 20 numbers) were conducted in a randomized event-related design. Subjects responded to the stimulus during the stimulus presentation or the subsequent ITI. The total duration of the experiment was 5.8 min. The paradigm rationale is demonstrated in Table 1.

Apparatus

E-prime software (Psychology Software Tools, Inc; Pittsburgh, PA, USA) controlled the stimulus presentations and the stimuli were presented using VisualSystem (Nordic NeuroLab, Bergen, Norway). Finger responses were collected using Response Grip (Nordic NeuroLab).

Image acquisition

MRI scans were acquired on a 1.5T scanner (Siemens Magnetom Sonata, Siemens Medical Solutions, Erlangen, Germany) supplied with a standard head coil. In one session, 145 volumes (30 contiguous axial slices, each slice spanning 4 mm) covering the whole brain were acquired using an echo planar imaging (EPI) BOLD sequence (repitition time (TR)=2400 ms, echo time (TE)=40 ms, field of view (FOV) 200×200 mm, flip angle=90°, matrix size 64×64). Magnetization prepared rapid acquisition gradient echo (MPRAGE) T1-weighted anatomical images were acquired for spatial normalization and co-registration of the functional data (TR=2000 ms, TE=3.9 ms, FOV 256×256 mm, flip angle=7°, matrix size 128×128).

Analysis

The images were visually inspected for signal dropout in the amygdala region, as this area is prone to magnetic susceptibility. One subject had signal dropout in the area of the amygdala, and was therefore excluded. All functional MRI volumes were preprocessed and analyzed using SPM2 (http://www.fil.ion.ucl.ac.uk/spm). All volumes were realigned to the first volume (Friston et al., 1995) and the anatomical image was co-registered to the mean functional image to ensure that they were aligned. All subjects moved less than 2.5 mm in any direction during the scan. The images were spatially normalized (Friston et al., 1995), resampled at 2×2×2 mm and smoothed using a 6 mm full width-half maximum (FWHM) isotropic kernel. Data were high-pass filtered using a cutoff value of 128 s. The model was built by convolving stick functions for the onsets of four different event types with a canonical hemodynamic response function. The four event types were “t,” “r,” other letters and numbers. The individual contrast images were entered into second-level random effects model. Small volume corrections for multiple comparisons were applied for our regions of interest (ROIs) (bilateral amygdala) based on anatomically defined ROIs using the SPM Anatomy toolbox developed by Eickhoff et al. (2005).

RESULTS

Behavioral results

In order to be included in the analyses, accuracy in each condition had to be above 90%. Six subjects were discarded due to behavioral performances not fulfilling these criteria. One of these had signal dropout in the amygdala and thus, 19 subjects were included in the analyses. We chose such stringent inclusion criteria to eliminate subjects' most likely misunderstanding the instructions or not paying attention to the task. The 90% cutoff allowed subjects to make a few coincidental mistakes, but excluded subjects systematically responding in a wrong way. The mean response time for each condition is shown in Table 2.

Table 2.

Mean Response time for each condition

Mean	t	r	Other letters	Numbers
Mean(ms)±S.D.(ms)	654±82	620±94	665±110	—

t vs. r, P<0.05; r vs. other letters, P<0.01; t vs. other letters, n.s.

Imaging results

We compared whether the behaviorally relevant letter “t” yielded significantly greater amygdala activation than the letter “r,” for which the color change was of no behavioral relevance. As seen in the Table 3 and Fig. 1, there were significant bilateral amygdala activations for the contrast letter “t”>letter “r.” Further, there was no significant amygdala activation for the contrast letter “r”>other letters.

Table 3.

BOLD fMRI responses in amygdala for our contrasts of interest

Contrast	Peak coordinates (MNI)	Peak Z	P value
t>r
L. amygdala	−20, −10, −8	3.52	<0.05
R. amygdala	26, −2, −12	3.25	<0.05
r>Other letters
L. amygdala	—	—	—
R. amygdala	—	—	—
Numbers>r
L. amygdala	−16, −8, −10	2.69	0.004a
R. amygdala	24, −4, −16	1.97	n.s.
t>Other letters
L. amygdala	—	—	—
R. amygdala	26, −4, −10	2.65	0.004a

Data are thresholded at P<0.05 (uncorrected) and small volume corrected for our ROI.

^aUncorrected.

Abbreviations: MNI, Montreal Neurological Institute; L, left; R, right.

Fig. 1.

Left amygdala (A, B) and right amygdala (C, D) responses obtained for the contrast letter “t”>letter “r.” (A, C) Statistical parametric maps (SPM) showing the location of the peak voxel in left amygdala (x=−20, y=−10, z=−8) and right amygdala (x=26, y=−2, z=−12) respectively, for the given contrast. For illustrative reasons, the images were thresholded at P<0.001, uncorrected. The colors refer to t-values as coded in the bar to the right of the images. (B, D) Beta-values for the conditions t and r in the same peak voxel in left amygdala and right amygdala, respectively.

To further explore the ability of behaviorally relevant stimuli to recruit the amygdala, we conducted the contrast numbers>letter “r.” In our paradigm, numbers signaled NoGo and a shift in prevailing behavioral responses (no button press). Thus, numbers are also behaviorally relevant. The contrast numbers>letter “r” yielded left amygdala activation that fell short of statistical significance after correction for multiple comparisons (Table 3). We also looked at the contrast letter “t”>other letters, revealing a trend in right amygdala (Table 3). However, the other letters did not contain any irrelevant component (like the color change of “r”), and hence this contrast was of lesser interest.

There were no significant activations outside of the amygdala in the letter “t”>letter “r” contrast after correcting for multiple comparisons across the entire brain volume. For the numbers>letter “r,” a significant cluster in parietal cortex (Montreal Neurological Institute (MNI)-coordinates x=52, y=−54, z=40, Z-value=4.99) and inferior frontal cortex, BA 44 (MNI-coordinates x=−56, y=18, z=38, Z-value=4.22) were discovered (P_{false discovery rate (FDR)}<0.05). Both of these areas have previously been linked to motor response inhibition, which is the case for the number stimuli (Rubia et al., 2001). To test if the observed effect in “t”>“r” could be due to linear habituation during the “t” or “r” condition, we did a parametric modulation with time as a parameter. We did not observe any significant linear habituation in the amygdala for any of the two conditions, though this does not exclude other possible habituation patterns.

DISCUSSION

The main finding of this study is that when made behaviorally relevant, non-emotional stimuli such as simple letters recruit the amygdala. Specifically, we found bilateral amygdala activity in the contrast letter “t”>letter “r,” where both these stimuli are salient (because of instructions or color change) compared with other letters, but only the letter “t” requires a behavioral shift. NoGo number stimuli, which were also behaviorally relevant relative to other letters, also yielded activation in the amygdala although this did not reach statistical significance after correction for multiple comparisons. Further, there was no significant amygdala activation for the contrast letter “r”>other letters, suggesting that salient information of no behavioral relevance does not elicit amygdala activation.

This pattern suggests that stimuli of behavioral relevance (i.e. letter “t” and numbers) and not just perceptual salience (i.e. letter “r”) elicit amygdala activation, providing support for a more general role of the human amygdala in detection of behaviorally relevant stimuli.

Based on its functional and anatomical connectivity, the amygdala is uniquely positioned to function in behavioral relevance detection. Most sensory modalities project to the amygdala (Price, 2003), and hence the amygdala receives information about both external stimuli and internal bodily states. Further, the amygdala has efferent connections with cortical and subcortical areas engaged in attentional and motivational processes (Holland and Gallagher, 1999; Davis and Whalen, 2001), allowing it to mediate vigilance and redirect attention as a function of shifting behavioral relevance of environmental information. Moreover, it has been suggested that the amygdala is capable of processing emotionally salient stimuli outside awareness, probably through a subcortical retino-collicular-pulvinar pathway for the visual domain (Vuilleumier et al., 2003). This pre-awareness processing may serve to preferentially guide attention to behaviorally relevant events (Merikle et al., 2001), even when incidental to current cognitive demands. All these lines of evidence indicate that the amygdala is ideally positioned to serve a role in relevance detection, receiving information about potentially relevant stimuli and current needs, perhaps even prior to awareness, and, in turn, modulating further processing of stimuli based on their importance.

Our current results are consistent with those recently reported by Schaefer et al. (2006), who demonstrated that individual differences in response speed during a three-back working memory task with non-emotional stimuli was correlated with magnitude of amygdala activity. Increased amygdala activity in this context may produce increased vigilance and facilitate selection of behaviorally relevant (the three-back matches requiring button presses) from irrelevant stimuli, leading to improved performance through downstream potentiation of behavioral control networks. Hence individuals with greater amygdala activity had faster response times. The proposal that target events (i.e. task-relevant events that require a subsequent response) elicit greater amygdala activation than non-target events (i.e. those not requiring responses) is further supported by findings of greater amygdala activity to “oddball” target stimuli (Kiehl and Liddle, 2001) as well as to Go events in an emotionally neutral Go-NoGo task (Laurens et al., 2005). Goldstein et al. (2007) created relevant stimuli using goal-relevance manipulation in an emotional linguistic Go/NoGo task. Independent of their valence, the emotional words elicited greater amygdala responses as they became goal relevant (i.e. as NoGo stimuli compared with less relevant Go stimuli).

Studies in animals further support a relevance detection role of the amygdala. Macaques with amygdala lesions exhibit decreased fear of predators and diminished stress in novel social situations (Amaral, 2002; Emery et al., 2001), observations initially interpreted as supporting a unique role of the amygdala as danger/threat detector (Amaral, 2002). After discovering that such lesions also impair food devaluation, the amygdala's role was extended to reward-related processes (Izquierdo and Murray, 2007). However, the effects of amygdala lesions in macaques may be related to more general impairments in detecting the behavioral relevance of stimuli. As a consequence of amygdala lesions, animals may fail to identify the behavioral relevance of a predator or update the value (i.e. relevance) of a food item leading to inappropriate behaviors such as approaching a predator or continuing to consume a devalued food. Amygdala lesions may lead to similar deficits in identifying threat (Adolphs et al., 1994) and processing reward in humans (Bechara et al., 2003).

Other recent studies supporting the role of the amygdala in non-emotional evaluation of behavioral relevance include those reporting its activation by socially relevant stimuli like novel faces (Schwartz et al., 2003a,b), untrustworthy faces (Winston et al., 2002), eye-gaze shifts (Adams et al., 2003) and rising sound intensity (Bach et al., 2008). All of these stimuli represent potentially relevant cues regarding the immediate environment of an individual, and may require the reallocation of attentional resources and a behavioral change. As stated previously, through its extensive connections to both cortex and brain stem nuclei, the amygdala is ideally situated to influence attentional and behavioral responses to such stimuli. In support of this specific role, Bach et al. (2008) found that rising sound intensity (signaling relevance) facilitated the autonomic orienting reflex and phasic alertness to auditory targets, responses believed to be mediated through the amygdala (Williams et al., 2001).

According to the Sander relevance hypothesis, amygdala is necessary for detecting the relevance of a stimulus for an individual's goals or needs. Knowing that goals and needs differ between individuals, the amygdala responses to a particular stimulus are variable. This has been demonstrated both to biological needs such as food and state of satiety (O'Doherty et al., 2001; Gottfried et al., 2003) as well as in social perspectives. According to Schirmer et al. (2008), social orientation correlated with amygdala activity elicited by emotional vocalization, so that subjects interested in social exchange had significantly larger amygdala responses to emotional vocalization than those less interested. Further support of this role is given by Canli et al. (2002) who demonstrated increased amygdala responses to positive faces in extroverted compared with introverted humans, suggesting that such positive facial cues are more relevant to extroverted people.

While our current results support the role of the human amygdala in the detection of non-emotional stimuli of general behavioral relevance, there are some limitations. The experiment could resemble an oddball-task. However, in auditory oddball-tasks it is usually only target stimuli that require motor responses, not the standard stimuli. Amygdala has been shown to respond to target stimuli in auditory oddball paradigm (Kiehl and Liddle, 2001), still one cannot be certain that this response is due to the behavioral relevance of the targets or the motor response to these per se, as the other stimuli often do not require button presses. In our paradigm, we compared stimuli with different levels of relevance, similar to an oddball-task, still all stimuli in our main contrast required button-presses. Therefore, the amygdala response discovered could not be due to button presses per se. The present amygdala activation to NoGo stimuli did not reach significance. This could be due to cortical inhibitory networks being involved in response inhibition, like the anterior cingulate cortex, as these areas have been hypothesized to send inhibitory impulses to the amygdala. It is necessary to replicate the current findings in independent laboratories using similar paradigms and emotional neutral stimuli.

CONCLUSION

To conclude, the current study supports the hypothesis of amygdala being involved in relevance detection in humans. The findings indicate that simple visual stimuli can elicit amygdala responses, if they are made behaviorally relevant. Therefore, the amygdala may have a more general role in humans than traditionally suggested.

Abbreviations

BOLD

blood oxygen-level dependent

fMRI

functional magnetic resonance imaging

FOV

field of view

ITI

intertrial interval

MNI

Montreal Neurological Institute

ROI

region of interest

TE

echo time

TR

repitition time