Attention Biases to Threat in Infants and Parents: Links to Parental and Infant Anxiety Dispositions

The current study focused on dwell times as a behavioral measure of infant and parent attention. All available data from parents and infants were included in the analyses and were processed in a uniform way to increase comparability. Gaze data from parents and infants were extracted using EyeLink® Data Viewer (SR Research Ltd., Mississauga, Canada) using the default velocity, and acceleration-based algorithm. The dwell times to dynamic expressions were obtained by first summing the duration of the fixations to the face area (with a minimum duration of 100 ms) within each trial, and next by averaging the duration of the fixations across trials of a given emotion category. Data from 7 infants and 3 parents were missing following this reduction. Next, infants and parents who contributed less than 10 trials were removed from the analyses (N = 10 infants and 2 parents). At this stage, dwell times were available on average for 34.32 of 40 trials (SD = 6.75, range: 10–40) from infants and for 38.77 trials (SD = 3.63, range: 14–40) from parents. Following this data reduction, the trials with neutral and sad faces were excluded from further analysis.

Statistical analyses were conducted using IBM SPSS Statistics for Mac, Version 25.0. We used multi-level regression models to account for the repeated structure of the dataset. The main outcome measure in the analysis of infants’ and parents’ attention biases was the mean dwell time per emotional expression. Emotion was initially randomized in the mixed models with infants’ and parents’ dwell times as the outcome variable, but was redundant, and therefore not randomized further in the analyses. Emotion was, therefore, included as a fixed effect with the happy faces as reference. We used maximum likelihood for estimation in all models. The intercept was randomized. The initial models included theoretically relevant interactions (specified per model below, and presented in the Supplement).

Finally, we investigated individual differences explained by parental anxiety dispositions, as well as parental dwell times on infant dwell times in models that simultaneously included the main effects of emotion, parental anxiety dispositions, and dwell times, along with all the two and three-way interactions between these predictors, N = 195. None of the tested interactions were significant in the initial model, see Table S5, (p’s > 0.153), reducing the final model (presented in Table 7) to main effects. This model revealed no significant main effect of parental anxiety, p = 0.934, whereas the significant link between parent and infant dwell times remained significant F (1, 441.72) = 4.86, β = 0.12, SE = 0.06, p = 0.028. We conclude that the relationship between parent and infant attention to threat-relevant emotions holds independent of parental anxiety dispositions. In a final step, we explored whether a link exists between parental anxiety and infant attention before controlling for parental attention. Parental anxiety did not predict infant attention, neither alone p = 0.942, nor as a function of infant age, or parent gender, p’s > 0.239. We conclude that the link between parental anxiety dispositions and infant dwell times is not significant, regardless of any concordance in attention bias in parent-infant dyads.

This study aimed to investigate the links between parental and infant attention biases to threat, along with anxiety dispositions in these same infants and parents. The current findings reveal a significant positive association between attention to threat-relevant (versus happy) facial expressions in parents and infants, and a marginally significant positive link between parental and infant anxiety dispositions. In turn, there was no significant link between an individual’s attention biases and his or her anxiety dispositions. This held for both infants and parents. Parental anxiety dispositions were also not related to their infant’s attention biases to threat. Below we discuss each of these findings in more detail within the broader framework of information processing model of intergenerational anxiety transmission (Creswell et al., 2010)

Creswell and colleagues (2010) suggest that an anxious parent’s vigilance to potentially threatening stimuli or situations in their environment may trigger or potentiate selective attention for threat in the offspring, leading to child anxiety. Accordingly, the link between parental and offspring anxiety is proposed to be at least partially accounted for by shared attention biases in the processing of threat. The current study was a first preliminary investigation of the proposed overlap in attention and anxiety dispositions of parents and infants, in a cross-sectional, correlational design, without a formal mediation analysis. In line with family studies of anxiety (e.g., Bolton et al., 2006; Lau et al., 2006; Creswell & Waite, 2015), the findings revealed a trend for an association between parental and infant anxiety dispositions: Parents who reported higher levels of anxiety and stress had infants who showed higher levels of temperamental fear and distress. A methodological limitation in the current design concerns the use of parental reports for measuring both infant and parental anxiety dispositions, as a parent’s anxiety dispositions may bias their perception of child temperament (Kelley et al., 2017). We aimed to address this limitation by using reports of infant temperament from both parents when available. Our findings suggest that the relationship between parental traits and infant risk is evident, and stable, across the first 18 months of life. This is in line with the findings capturing shared genetic substrates that influence psychosocial functioning in both generations (Bolton et al., 2006; Lau et al., 2006). Needless to say, genetically-informed studies are needed in order to disentangle the genetic and environmental contributors to this overlap in typical development (Creswell & Waite, 2015). However, even within the limitations of the current within-family cross-sectional design, the use of multiple informants helps alleviate some of these concerns. We then build on this foundation to examine concordance in a putative functional mechanism for intra- and inter-individual patterns of anxiety.

To our knowledge, this is the first study to explore the potential overlap in infant and parental attention biases to threat-relevant versus happy emotional expressions as a mechanism explaining the relationship between infants’ and parents’ anxiety in a free-viewing experiment. In line with the proposal by Creswell and colleagues (2010), the findings revealed that parents of infants who showed more attention to emotional faces also showed more attention to emotional faces. Thus, this link was not specific to threat-relevant expressions, but also extended to happy faces. Despite this overall link in the attention of parents and infants, separate analyses of parent and infants reveal differences in their relative attention to threat-relevant faces: Although both parent and infants dwelled marginally longer on fearful than happy faces, attention to angry versus happy faces differed between infants and parents: Parents spent longer time fixating on angry expressions compared to happy faces, whereas a reverse pattern was observed for infants who spent longer time looking at happy than angry faces. Thus, in contrast to parents, infant negativity bias was specific to fear, whereas an avoidant response was observed for angry expressions.

There is some earlier evidence that suggests that an attention bias for fearful faces may precede angry faces in development, which may emerge sometime between eighth month and third year of life (Leppänen et al., 2018; LoBue & DeLoache, 2010). A bias for angry faces was not observed in the current study, which may suggest that our sample did not capture the appropriate developmental window. The avoidant attention pattern observed among infants in the current study may be related to the differences in the threat value of angry and fearful expressions or infant’s differential exposure to anger and fear in everyday situations. In the case of a fearful face, the signal often points to an indirect threat, as the individual is reflecting their response to an external stimulus as yet unknown to the infant (Adams & Klerck, 2003). As such, the signal is ambiguous and more computationally taxing as the infant must deduce the source of the emotional response. In contrast, when observing an angry face, the individual is the direct source of the threat. The direct threat value of angry faces may have triggered an avoidant reaction among infants in the current study. As an alternative, we can consider the prototypical pattern of faces infants are exposed to in their daily experience. The assumption is that in adaptive and supportive environments, infants are likely to have more direct experience with positive (happy) faces relative to anger and fear. Angry faces, in particular, may be rare in a community sample, leading to the pattern of responses noted here. While these different lines of reasoning need to be tested, the core finding of the current study suggests that, within the first 18 months of life, infants and parents display parallel patterns of attention, in line with proposed mechanisms of intergenerational transmission (Creswell et al., 2010).

Despite a significant overlap between parental and infant attention, as well as a marginally significant overlap in anxiety profiles, no empirical support was found for the idea that this overlap in attention can partially or fully account for the overlap in anxiety profiles in this community sample: no significant associations were found between attention biases and anxiety dispositions, in either infants or parents. Earlier evidence from adult studies has revealed a link between attention biases to threat and anxiety in adults with clinical or subclinical levels of anxiety (Bar-Haim et al., 2007; van Bockstaele et al., 2014). In contrast to earlier studies that noted a link between parental anxiety and infant attention biases to threat in infancy (Burris et al., 2019; Forssman et al., 2014; Morales et al., 2017), we failed to note a similar relation in the current study (before and after considering parental attention biases).

The lack of significant associations between parental anxiety and attention biases in either infants or parents may be explained by the fact that most of the participating parents reported little to mild levels of anxiety dispositions in this community sample. Out of 216 parents who contributed to the current study, only 4 parents reported moderate levels of anxiety, and only 8 parents reported moderate or higher levels of stress. Thus, in the current parent sample, there was more individual variation in the generalized anxiety symptoms and associated negative affect dimension, than in the physiological arousal and fearfulness dimension of anxiety, which may characterize rather subclinical and clinical levels of anxiety displays. Overall, these rates of psychopathology in our sample seem to be lower than rates reported in community samples (Bijl et al., 1998), thus non-anxious parents may have been more likely to join this infant study. The limited number of parents with moderate or clinical levels of anxiety dispositions may have limited our ability to detect any associations were they to exist across a fuller spectrum of anxiety (Bar-Haim et al., 2007).

We also failed to note a link between infant anxiety dispositions (a composite measure of temperamental fear and distress reported by both parents) and their attention biases. This finding is not consistent with earlier evidence suggesting that infants with a more negative temperament take longer to disengage from threat (Conejero & Rueda, 2018; Nakagawa & Sukigara, 2012), although Morales and colleagues (2017) found no significant link between temperamental negativity and disengagement from threat. Here, we may point to differences in the task at hand. The prior studies used an overlap task that captured the infant’s ability, and propensity, to disengage from a face when provided with an alternate stimulus in the periphery of the visual field. Here, the infants were provided with a single, albeit dynamic, face at a time. As such, we may have captured the infant’s pattern of processing of, and interest in, the face, as opposed to comparing competing targets for the infant’s limited attentional resources. Indeed, Rothbart and Posner (2006) note that temperamental variation may differ across associated components of attention, such as vigilance and disengagement. Future studies are needed to further specify the nature of this association and to answer whether the inconsistent findings are related to child or task-related factors.

Across the core analyses, we noted an absence of any age effects. Although the findings reveal higher levels of temperamental anxiety dispositions in 12 and 18-month-olds, this uptick did not moderate the links between parent-infant attention or anxiety dispositions. The longer dwell times observed in 12 and 18-month-olds may reflect the normative improvement in attentional control with age. This finding is also in line with an earlier study reporting an increase in attention (i.e., dwell time to angry faces) between 4 and 24 months of age (Pérez-Edgar et al., 2017). However, the hypothesized relations with individual and parental traits were not evident here.

Taken together, current findings provide support for an overlap between infant and parental attention biases to threat, and a more tenuous relationship between infant and parental anxiety dispositions. However, we found no empirical support for relations crossing constructs of interest (attention vs. anxiety) for either infants or parents. Thus, the suggestion we based on the model by Creswell and colleagues (2010) that concordance in anxiety dispositions would be at least partially be accounted for by attention biases to threat, does not seem to hold in this community sample. Future investigations of these associations should include clinically and sub-clinically anxious parents and their infants, and should test the model by Creswell et al. (2010) in formal mediation analyses.

The current study constitutes the first evidence for a link between parental and offspring attention biases as early as infancy. Although there was a marginally significant link between parent and child anxiety dispositions, the parents in this community sample reported low levels of anxiety, precluding strong conclusions regarding the relations predicted by the Creswell et al. (2010) model for the intergenerational transmission of anxiety.

The significant link between infant and parental attention bias is likely to be adaptive in typical development in a window marked by infants increasingly exploring the world around them under their parent’s guidance (e.g., Koterba & Iverson, 2009). Shared profiles of exploration and attention may be built on an infant’s joint attention capacity, which emerges and strengthens across the second half of the first year (Mundy et al., 2007; Scaife & Bruner, 1975). In this period infants become increasingly sensitive to their interaction partner’s direction of gaze, leading to shared attention (Beier & Spelke, 2012). Shared emotional states with the caregivers may then be experienced frequently through daily face-to-face interactions with novel objects or events. The model by Creswell et al. (2010) suggests that this normative attention mechanism, may become maladaptive and mediate early anxiety transmission in cases where parental attention is repeatedly targeted to the threatening aspects of the environment. Biased attention in the parent may then trigger parental expressions of anxiety, fueling this cycle. An earlier study including parents with or without anxiety disorders noted that parental anxiety expressions towards novel stimuli, rather than the presence of an anxiety diagnosis determine child avoidant reactions to the stimuli (Aktar et al., 2013). Although this study did not measure infant and parent gaze to the novel stimuli, parents may induce an attention bias towards novel stimuli, coupled with a subsequent avoidant behavioral tendency in these situations. A longitudinal study in a sample of clinically and sub-clinically anxious parents and their infants that includes observations of parent-infant shared attention (in everyday interactions and computerized eye-tracking tasks), along with indices of parent and child anxiety dispositions would be the ideal follow-up design to gain further insight into shared attention to threat as a mechanism in anxiety transmission.

The following limitations should be considered while interpreting the current findings. First, it is important to note that only a few of the hypothesized associations were significant among all analyses conducted for the current study. Although the findings provided evidence for a significant association between parent and infant attention bias to threat in this community sample, the hypotheses regarding the role of attention in the hypothesized links between parent anxiety and infant temperament were, by large, not supported. Future research on these associations in community and clinical samples of parents with young children is needed for a more rigorous test of these hypothesized associations.

Second, we did not experimentally manipulate parental anxiety dispositions nor infant temperament. The associations between attention biases and anxiety dispositions were cross-sectional and correlational, precluding any prospective or causal inferences on the effect of anxiety dispositions on infant or parent attention to facial expressions. Recently, the importance of adopting a developmental approach was highlighted as crucial to understanding how normative attention biases in infancy may develop into a risk factor or mechanism for anxiety, and how infant and parent characteristics may play a role in this co-evolution (Burris et al., 2019; Field & Lester, 2010). Future studies should investigate these associations in longitudinal multi-method designs to establish a timeline, which is essential to understanding the developmental trajectories of attention biases.

Third, the current study did not include direct observations of parental anxiety or infant temperament. Parental anxiety symptoms and infant negative dispositions were only indirectly measured using questionnaires, which were completed by the parents. Future studies should consider measuring infant temperament and parental anxiety using naturalistic observations along with self-report measures. Fourth, the current analyses had to exclude a younger sub-sample due to difficulties related to eye-tracking, limiting the generalizations of our findings to younger ages. Finally, although parent emotion processing was incorporated in the current design, the eye-tracking data were only available from the specific parent who visited the lab, and the analyses of parental variables were limited to that parent. Future investigations should consider incorporating both parents, when available, for a more complete picture of the hypothesized associations within the family. Indeed, a family system approach often notes that the aggregated impact of the family on child development is more than a simple sum of individual contributions (Johnson & Ray, 2016). Despite these limitations, the current study constitutes the first evidence suggesting significant concordance in attention between parents and infants, prior to, or in the absence of extreme anxiety in the absence of extreme anxiety.