Several attentional and learning processes have been found to play a major role in pathological anxiety (i.e., anxiety disorders). Recent research suggests that attentional control (AC) and fear extinction learning feature prominently among such processes (e.g., Bar-Haim et al., 2007; Cisler and Koster, 2010; Eysenck and Derakshan, 2011; Milad and Quirk, 2012; Heeren et al., 2013; VanElzakker et al., 2014; Duits et al., 2015; Hadwin et al., 2016). In this study, we explore the possible association between these two processes.

Attentional control is a construct that defines our ability to regulate attention allocation, including our ability to maintain sustained attention, ignore distractors, and shift attention between tasks (Derryberry and Reed, 2002). Deficient AC has been found to characterize both clinical (Olatunji et al., 2011) and subclinical anxiety (e.g., Fajkowska and Derryberry, 2010; Sportel et al., 2013). Such deficits may also account for the attention bias to threat commonly observed in anxious individuals (Hadwin et al., 2016). Moreover, such deficits are associated with reduced ability to regulate emotion (Fajkowska and Derryberry, 2010; Armstrong et al., 2011; Tortella-Feliu et al., 2014; Hsu et al., 2015; Morillas-Romero et al., 2015; O’Bryan et al., 2017). AC can be assessed under emotional conditions (emotional AC, e.g., Barry et al., 2013) or neutral conditions (non-emotional AC, e.g., Derryberry and Reed, 2002; Pacheco-Unguetti et al., 2011; Richey et al., 2016).

Fear extinction learning refers to the decrease in fear following non-reinforced exposure to a feared conditioned stimulus (CS) and is typically investigated in humans within a differential fear learning paradigm preceded by a conditioning (i.e., acquisition) phase. The test of how fear extinction learning is retrieved after re-exposure to the extinguished CS is usually called extinction recall. Deficient fear extinction learning (Duits et al., 2015) or extinction recall (Graham and Milad, 2011; Milad and Quirk, 2012) could be a marker of anxiety disorders (e.g., Graham and Milad, 2011). Importantly, fear extinction is a form of emotion regulation (Hartley and Phelps, 2010) and there is evidence that similar neurobiological mechanisms (i.e., hipoactivity of the prefrontal cortex) may be involved in fear extinction, reduced emotion regulation capabilities, and low AC in non-emotional conditions (Bishop, 2007, 2008, 2009; Hartley and Phelps, 2010; Milad and Quirk, 2012; Ochsner et al., 2012; Shiba et al., 2016; Ball et al., 2017). Therefore, individual differences in AC under non-emotional conditions may be associated with individual differences in fear extinction, although this has not been investigated so far, as far as we are aware.

Moreover, fear extinction learning procedures are considered experimental models for exposure therapy (Craske et al., 2014: Pittig et al., 2016) and both fear extinction learning (e.g., Waters and Pine, 2016; Ball et al., 2017; Forcadell et al., 2017) and attentional functioning (e.g., Barry et al., 2015a) may be associated with the outcomes of exposure therapy, and constitute putative targets for improving such outcomes (Bar-Haim, 2010; Craske et al., 2012, 2014; Barry et al., 2015b; Mogg and Bradley, 2016; Pittig et al., 2016). A better understanding of the association between AC and fear extinction may therefore have important therapeutic implications.

The role of attention in fear learning has been a topic of research for years (e.g., Mackintosh, 1975; Wagner, 1981; Le Pelley et al., 2016). Most studies have focused on how attention allocation changes during or after acquisition (e.g., Beaver et al., 2005; Koster et al., 2005) or extinction (e.g., Robbins, 1990; Van Damme et al., 2006; Barry et al., 2016b) affect the magnitude of learning. Moreover, attentional biases to threat in anxiety could reflect a much broader dysregulation of AC (Bishop, 2009; Moriya and Tanno, 2009; Pacheco-Unguetti et al., 2011). The use of non-emotional stimuli allows to isolate potential general attention deficits beyond those observed when individuals face emotional materials (see further below). A few recent studies have focused on how baseline individual differences in attention predict the magnitude or gradient (“speed”) of fear extinction learning (Waters and Kershaw, 2015; Barry et al., 2016a, 2017).

The study by Waters and Kershaw (2015) belongs to the research tradition that focuses on analyzing how attentional bias to threat-related information is associated with increased anxiety (valence-specific models) (for a review see Heeren et al., 2013; Hadwin et al., 2016). Waters and Kershaw (2015) found that clinically anxious children who showed attention to threat in a visual probe task exhibited greater fear extinction learning than those who avoided threat.

The studies by Barry and colleagues (Barry et al., 2016a, 2017) represent a second research tradition that has explored how deficient AC is associated with anxiety vulnerability, and more specifically with cognitive and inhibitory control impairments observed in anxious individuals. In two separate studies in healthy participants, these authors investigated how emotional AC, as measured by self-report (Barry et al., 2013), was associated with fear extinction learning (Barry et al., 2016a, 2017). In the first study, participants were confronted with a perceptually similar stimulus presented after extinction of the original CS, and it was observed that higher emotional AC was associated with faster extinction learning and greater return of fear (Barry et al., 2016a). In the second study, during extinction learning participants were confronted with a similar stimulus as during acquisition, and were instructed to attend toward the common features between the acquisition and extinction stimuli, or toward the unique features of the extinction stimulus. For participants who, during extinction, were instructed to attend toward the unique features of the extinction stimulus, lower emotional AC tended to be associated with a greater return of fear (Barry et al., 2017). The authors suggested that those with low emotional AC may have been unable to shift attention to other features of the extinction stimulus, which may have facilitated the return of fear when confronted with a perceptually similar stimulus.

In the studies mentioned above, individual differences in attention functioning were investigated using emotional conditions. We share the view of Heeren et al. (2015b, p. 136) that the focus on emotional materials has “neglected the empirical exploration of basic attentional deficits from non-emotional material,” and precludes the assessment of general attentional abilities that may be relevant to several clinical phenomena (see also Snyder et al., 2015). Furthermore, in the two studies investigating the association between AC and extinction learning, AC was assessed by self-report (Barry et al., 2016a, 2017). The use of performance-based tasks (see below) may provide important information on the role of different attentional networks beyond general AC (see Heeren et al., 2015a; Heeren and McNally, 2016).

According to the attention system model (Posner and Petersen, 1990; Posner and Rothbart, 2007), attention consists of three major networks, which can be assessed separately: executive control, orienting, and alerting (see Posner et al., 2007 for a review). The executive control network is specialized in conflict resolution and voluntary control of attention while resisting distraction by other competing stimuli. While the executive control network has traditionally been equated to AC, some authors have recently expanded the definition of AC to include the orienting and alerting networks (e.g., Heeren and McNally, 2016). The orienting network is involved in attention engagement to new stimuli and attention disengagement from the current focus. Finally, the alerting network is devoted to maintaining adequate sensitivity to perceive and process stimuli. The functioning of these three attentional networks when facing non-emotional cues has been related to anxiety and emotion regulation. For example, reduced efficiency of the executive control and orienting networks has been associated with high trait and clinical anxiety (Moriya and Tanno, 2009; Pacheco-Unguetti et al., 2010; Pacheco-Unguetti et al., 2011; Heeren et al., 2015b; Heeren and McNally, 2016), and faster spontaneous emotion regulation (Morillas-Romero et al., 2015). Finally, increased efficiency of the alerting network has been associated with state anxiety (Pacheco-Unguetti et al., 2010).

In this present study, we explore whether individual differences in non-emotional AC (defined as a multifaceted construct including executive control, orienting, and alerting attentional networks), are associated with individual differences in fear extinction (learning and recall) in a sample of subclinical phobic participants (individuals with moderate to strong fear of spiders). The use of subclinical samples is a valid strategy for studying anxiety-related processes, can be generalized to individuals with an anxiety diagnosis (Stopa and Clark, 2001; Abramowitz et al., 2014) and also has some advantages (e.g., avoid comorbidity, medications or the impact from previous treatments) compared to clinical samples. Furthermore, previous studies exploring the association between attentional bias to threat and fear extinction (Waters and Kershaw, 2015) and between fear learning and treatment outcome (Waters and Pine, 2016) included children with specific phobias, but to the best of our knowledge, ours is the first study exploring the role of non-emotional attention and its association with fear extinction in (subclinical) adult phobic individuals using “truly” non-emotional stimuli.

We used self-report and performance-based measures of AC under non-emotional conditions. Fear extinction was assessed using three different measures: Unconditioned Stimulus (US) expectancies, Skin Conductance Response (SCR), and Fear-Potentiated Startle (FPS). Given the well-established association between trait anxiety and AC (e.g., Pacheco-Unguetti et al., 2010, 2011; Sportel et al., 2011), and between trait anxiety and fear extinction (Sehlmeyer et al., 2011; Gazendam et al., 2013; Haaker et al., 2015), we tested the magnitude of these associations after controlling for trait anxiety.

In this paper, we found that individual differences in self-reported non-emotional AC are not significantly associated with fear extinction learning and recall. However, we did find that the two components of orienting network functioning are significantly negatively associated with fear extinction learning and recall beyond that accounted for by trait anxiety. Specifically, participants with enhanced orienting costs (i.e., difficulties in disengaging attention from cues not relevant for the task) showed faster extinction learning, while those with enhanced orienting benefits (i.e., facilitated orientation by valid cues) exhibited faster extinction recall as measured by FPS and US expectancies, respectively.

The lack of a significant association between self-reported non-emotional AC and fear extinction is at odds with the findings of Barry et al. (2016a), who reported that higher self-reported emotional AC was associated with faster fear extinction learning. In our calculation of the extinction gradients, we considered the difference between CS+ and CS- in each trial; however, when we used the same criteria as Barry and colleagues (Barry et al., 2016a,b) (i.e., considering only the response to the CS+) the results did not change (see Supplementary Table S3). The main difference between Barry’s studies and ours is that we measured AC in non-emotional conditions whereas Barry et al. (2016a, 2017) measured emotional AC. This underlines the importance of the distinction between emotional and non-emotional AC.

Our results on orienting network functioning are not easily comparable to previous research because, to our knowledge, this is first study to analyze how individual differences in the functioning of attentional networks under non-emotional conditions are associated with individual differences in fear extinction learning and recall. Our results on orienting costs resemble those reported of Waters and Kershaw (2015) using threat-related stimuli. While Waters and Kershaw (2015) found that higher allocation to threat was a predictor of higher extinction learning, we found that lower capacity for disengaging attention from irrelevant cues was a predictor of faster extinction learning. It may be that both processes tap into a similar construct that becomes apparent using both threat-related and non-emotional conditions. Consistent with this, Heeren et al. (2015b) stated that specific impairment of the orienting network, such as difficulty in disengaging attention from task-irrelevant distractors, may be a mechanism underlying attentional bias. Then, during fear extinction learning we could assume that people with greater orienting costs with non-emotional stimuli also display higher attentional allocation to the CS+ that speeds up the extinction learning process.

However, this interpretation conflicts with data showing that higher self-reported emotional AC is positively associated with faster extinction learning, as shown by Barry et al. (2016a). Nevertheless, in the study by Barry et al. (2017), a different stimulus was presented during extinction learning and during conditioning. The authors interpreted that participants with high AC were better able to shift attention from the common, threatening features of the original CS+ to the distinct features of a different but similar CS+, therefore “speeding up” the extinction learning process.

The fact that orienting costs and orienting benefits showed a different association with extinction learning and recall also suggests that distinct attentional capabilities may have different relationships with various fear learning processes. This is also consistent with previous studies showing that extinction learning and extinction recall are independent processes (Phelps et al., 2004; Milad et al., 2007; Quirk and Mueller, 2008).

In a previous study, Morillas-Romero et al. (2015) reported that, under non-emotional conditions, orienting network functioning was not related to spontaneous emotion regulation, but with explicit emotion regulation styles. To our knowledge, ours is the first report showing that individual differences in orienting network functioning under non-emotional conditions are related to a form of spontaneous emotion regulation (i.e., fear extinction). Our results underline the potential prominence of the orienting component of attention in anxiety and related processes, as shown recently by Heeren et al. (2015b), Heeren and McNally (2016) in adults with social anxiety disorder.

We also explored the contribution of the alerting network to fear extinction, and found that enhanced alerting was associated with faster fear extinction recall. Previous studies (Tortella-Feliu et al., 2014) have shown that the alerting network is related to self-reported emotion regulation strategies (i.e., enhanced alerting predicts a higher probability of suppressing distressing cognitions). Therefore, there seems to be a positive relationship between alertness and emotional regulation, and future research will need to investigate this relationship further.

Our results could have several methodological and clinical implications. From a methodological perspective, our data indicate that fear extinction gradients can provide a better understanding of individual differences in fear learning (and, more generally, a broader view of fear learning processes) than those offered by “static” fear indices, as previously observed by Barry et al. (2016a, 2017). They support that the empirical exploration of individual differences in non-emotional AC can be relevant for several clinical phenomena, as recently emphasized by Heeren et al. (2015b), and shown here for extinction learning and recall. Furthermore, our results highlight the utility of performance-based measures of attentional network functioning beyond general and self-reported AC.

Taken together with previous studies on the association between both fear extinction (e.g., Waters and Pine, 2016; Ball et al., 2017) and attentional functioning (e.g., Barry et al., 2015a) and exposure-based interventions, our results suggest that differences in the orienting network functioning may explain differences in exposure therapy outcomes, a question that deserves to be explored in future studies.

Our result may also have implications for other forms of psychological interventions, especially attention training. Attention training is a generic term that refers to repetitive “practice in conflict-related tasks, working memory tasks or others tasks involving executive control mechanisms” (Tang and Posner, 2009, p. 222) that requires directed effortful attention control. Importantly, most of these interventions train attention using non-emotional materials, as is the case of the attention training technique for anxiety disorders (Fergus and Bardeen, 2016; Knowles et al., 2016). However, in the field of anxiety disorders the most common attention intervention is attention bias modification (ABM) training (Bar-Haim, 2010). Positive results on the use of ABM, mainly as an add-on to cognitive behavior therapy, have been reported in several anxiety disorders (Linetzky et al., 2015). However, results on the efficacy of these interventions are inconsistent (Mogoaşe et al., 2014; Kuckertz and Amir, 2015), and some authors have called for ABM procedures to be improved (Mogg and Bradley, 2016). Notably, although these procedures are intended to train participants to disengage attention from threatening information, several studies have found that any active attentional training procedure will reduce anxiety symptoms, including procedures that use only neutral stimuli (Klumpp and Amir, 2010; Heeren et al., 2015c). Heeren et al. (2015c) also reported that several attention training procedures improve the executive control and alerting components of AC, but not the orienting one, in socially anxious patients. Therefore we propose that attentional training with neutral stimuli should be further explored as a way to improve AC, especially by directly manipulating the orienting network, as proposed by Heeren et al. (2015b). This in turn could be related to increased effectiveness of exposure.

Our study has several strengths. It is the first to explore the association between non-emotional AC and fear learning. Moreover, we measured fear learning using three different fear measures, controlled for some variables that are known to affect fear learning (menstrual cycle, time of the day), and used both magnitude and gradient-based measures. Also, our AC variables included self-reported and performance-based measures. We included analyses and results for the whole sample and only for those participants with successful conditioning and extinction learning. Finally, while previous studies focused on extinction learning, we also studied extinction recall.

We also note some limitations. Participants in our study were individuals with moderate to strong fear of spiders, which also involves disgust, not only fear (Cisler et al., 2009), and this may affect some of the processes investigated. Second, in the paradigm employed here conditioning and extinction learning occurred consecutively with no time for longer consolidation of fear conditioning. Third, some of our analyses including only participants with successful conditioning and extinction learning (those involving SCR and FPS measures and magnitude-based indices) were based on relatively small samples (less than a half of the whole sample for some analyses). However, the main findings from these analyses are fully consistent with those obtained using the whole sample (n = 50), which is a relatively large one for fear learning psychophysiological experiments using a 2-day procedure, compared to many previous studies (e.g., Milad et al., 2005; Rabinak et al., 2013, 2014; Garfinkel et al., 2014; Pineles et al., 2016). We computed post-hoc power (two-tailed) for our significant bivariate correlations using sample size, the effect size, and the alpha error (0.05). We had 53–63% (moderate) power to detect a significant correlation between the orienting network functioning and extinction learning and recall. This indicates that Type II error probability was still relatively high in our study, and therefore a replication with larger samples is warranted to ensure the generalizability of our findings. Finally, most significant findings could not be replicated across different fear measures, although this is consistent with many previous fear-learning studies (e.g., Sevenster et al., 2014).

Despite these limitations, we consider that these results contribute to a better understanding of how non-emotional AC is related to fear extinction learning and recall. The most important theoretical contribution is that attentional biases to threat in anxiety could reflect a much broader dysregulation of AC observed in face of non-emotional material.

In summary, in this exploratory study we showed that orienting network functioning is related to fear extinction learning and recall over and above trait anxiety. To the best of our knowledge, this is the first study that links non-emotional AC to fear extinction. An important avenue for future research is to explore the association between AC and anxiety treatments (i.e., exposure therapy).

This study was carried out in accordance with the recommendations of Clinical Research Ethics Committee of Institut Hospital del Mar d’Investigacions Mèdiques, with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the Clinical Research Ethics Committee.

EF, DT-R, MF, and MT-F contributed to the development of the study hypothesis, and to the contribution of the study and experiment designs. EF, DT-R, and DT prepared the stimuli, and prepared data and software for analysis. EF, DT-R, and MT-F performed the data analysis. All authors drafted the manuscript, discussed the results, implications, and literature, and approved the final version of the manuscript for submission.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.