Over the course of the last several decades, research on child development has suggested that

Attention bias to threat has been conceptualized in many ways over the past several decades. Generally, it indicates an increased sensitivity to potentially threatening stimuli in the environment, and is often measured by faster detection of, longer looking to, or difficulty disengaging from, angry or fearful facial configurations when compared to happy or neutral stimuli (Burris & Oleas, 2019). Given the adaptive benefits of quickly detecting and responding to threat, researchers have suggested that attention biases to threat are evolutionarily adaptive, automatic, and part of typical development (LeDoux, 1995, Petersen and Posner, 2012, Williams et al., 1998).

Although previous research has demonstrated that attention biases for stimuli like angry or fearful facial configurations are early developing, emerging as early as 7 months of age (Kotsoni et al., 2001, Peltola et al., 2008), we know very little regarding patterns of change over time. For example, some research demonstrates a stable pattern of attention bias to threat that develops between 4 and 24 months (Morales, et al., 2017a), while other studies point to normative changes in attention biases as a function of age (Leppänen et al., 2018, Peltola et al., 2018). Further, research from the clinical literature suggests that attention biases to threat may not necessarily be part of typical development, but rather markers of risk for the development of anxiety (Bar-Haim et al., 2007, Britton et al., 2012, Burris et al., 2019b, Morales et al., 2016), particularly within the context of additional risk factors, such as temperamental fearfulness, or behavioral inhibition (Pérez-Edgar et al., 2010).

Most of the literature on temperamental risk for anxiety focuses on behavioral inhibition (BI). BI is a temperament style characterized by the tendency to experience distress in response to novel stimuli (including people, places, and objects), hypersensitivity to changes in the environment, and engagement in avoidant coping behaviors (Kagan et al., 1984). Manifestations of BI encompass several domains, including motivations (withdrawal or avoidance) or behavioral reactions (flight or freezing) and social reactions (social fear or social reticence) to novelty (Buss, 2011). For example, toddlers with BI are more likely to show characteristics of social fear in childhood (Henderson et al., 2001), such as hesitation, wariness, and fearful behavior when engaging with an unfamiliar person, which in turn may put them at higher risk for social anxiety later in life (Rapee, 2014). Approximately 10–20% of children have a behaviorally inhibited temperament, which is relatively stable throughout development. Among those who are behaviorally inhibited in childhood, 30–50% remain inhibited into adolescence and adulthood (Buss, 2011).

Importantly, the developmental literature has demonstrated a link between BI and attention biases for threat (Morales et al., 2015, Morales et al., 2017b, Pérez-Edgar et al., 2011). For example, as early as 5 years of age, children who are temperamentally shy, and are thus at increased risk for the development of social anxiety, show a heightened attention bias for social threats (i.e., angry faces) when compared to non-shy controls (LoBue & Pérez-Edgar, 2014). Further, one study reported that while children who are behaviorally inhibited at ages 2 and 3 are socially withdrawn at age 5, this relationship was only significant for children who showed a heighted bias for angry versus happy faces (Pérez-Edgar et al., 2011). A second study reported a similar relationship in adolescents (Pérez-Edgar et al., 2010). Taken together, these findings suggest that a behaviorally inhibited temperament might interact with attention to threat to predict the development of anxiety symptoms.

However, some of the studies examining the relation between attention biases to threat and a fearful temperament have provided mixed results. For example, while some research has demonstrated a significant relation between fearful temperament and an attention bias away from, or avoidance of threat (Morales et al., 2015), other research has shown the opposite, such that behavioral inhibition predicted attention biases toward threat (Pérez-Edgar et al., 2011). To complicate things further, some studies simply found no relation between the two (Broren et al., 2011), while other research has found that childhood temperamental shyness was associated with both vigilance and avoidance (Morales et al., 2017b, Poole and Schmidt, 2021) rather than one or the other.

Unfortunately, most existing research on this topic consists of cross-sectional studies with infants, or longitudinal studies with older children and adults (White et al., 2017), limiting our understanding of how attention biases to threat develop over the first few years of life and whether they are linked to risk factors for anxiety related to social fear. Longitudinal studies early in infancy are needed to capture how patterns of biased attention to threat first develop, and whether they interact with individual difference factors to confer risk for anxiety over time. Field and Lester (2010) proposed three models to conceptualize how attention biases to threat might develop early in life and interact with environmental and individual difference factors. The integral bias model suggests that attention biases for threat are normative and stable over time, with very little room for developmental change. In contrast, the moderation model posits that biased attention to threat is normative but is moderated by individual factors like infant temperament. According to this model, individuals with a behaviorally inhibited temperament might maintain or show even greater biases compared to children without BI over the course of development. Finally, the acquisition model posits that attention biases to threat are not present at birth, but instead are acquired over the course of development through experience.

Although studies that explicitly test these models in infancy are limited, very recent work has begun to describe the typical trajectory of attention to both threatening and non-threatening emotions in the first two years of life. For example, one recent study reported that between 4 and 24 months of age, infants become faster to detect angry facial configurations and display greater engagement with, and less disengagement from, angry faces compared to neutral faces. Further, a significant bias for angry or threatening facial configurations over other emotional configurations did not emerge until 24 months of age (Reider et al., 2022). This suggests that attention biases for threat develop over the first two years of life, with a clear bias for threat emerging by age 2. These data provide support for both the acquisition and moderation models (but not the integral bias model) proposed by Field and Lester (2010), such that attention biases to threat were normative and developed over time. However, because the paper focused on capturing typical development, it did not address the role of potential moderators like behavioral inhibition or social fear, which are additional known risk factors for later social anxiety (Chronis-Tuscano et al., 2009).

Here we investigated how normative patterns of attention bias interact with social fear over the first two years of life. We chose social fear over other aspects of infant temperament because social fear most closely aligns with the core components of BI, such as fear and wariness in response to unfamiliar others, and social withdrawal and avoidance behaviors (Buss, 2011, Kagan et al., 1984). Further, research has demonstrated continuity in development from temperament in infancy to BI and social fear in childhood, and later social anxiety in adolescence and adulthood (Askew et al., 2015, Pérez-Edgar and Guyer, 2014). Thus, while the literature on BI has informed our hypotheses, here we focused on a parent-report measure of social fear.

We investigated our research question among a diverse and longitudinal sample of infants who, as part of a larger study, participated in two eye-tracking tasks at 5 assessments (4, 8, 12, 18, and 24 months). Additionally, infant temperament was measured via caregiver report at each assessment. We chose two eye-tracking tasks to capture attention bias for the current investigation using metrics most evident in the literature, typically either rapid fixations to threatening stimuli (orientation) or longer looking at (or slower disengagement from) threatening stimuli—in most cases, angry facial configurations, as they are a direct signal of social threat (Leppänen et al., 2018, LoBue and DeLoache, 2010, Nakagawa and Sukigara, 2012, Peltola et al., 2018). Engagement and disengagement, in turn, is embedded within several interacting attention networks (Corbetta and Shulman, 2002, Petersen and Posner, 2012, Rothbart et al., 2011). The alerting network detects information in the visual field and is responsible for selective attention and detection of novel stimuli (Sturm & Willmes, 2001). The orienting network is responsible for the selection and prioritization of information in the environment, including disengaging, shifting, and reengaging with visual stimuli (Posner & Cohen, 1984). While these systems are functional early in life and are both assessed in the attention bias literature, little work has considered how they might uniquely impact socioemotional development.

As such, here we examined the alerting network, or rapid detection of emotional stimuli, with a Vigilance task (Fu et al., 2020), and the orienting network, or prolonged engagement with and difficulty disengaging from emotional stimuli, with a classic Overlap task (Morales et al., 2017a, Peltola et al., 2008, Vallorani et al., 2021, Vallorani et al., 2022). In the Vigilance task, infants were presented with a single stimulus—either a happy, angry, or neutral facial configuration—in one of the four corners of a screen, and we measured how quickly infants detected each target. Rapid detection was indexed by time to first fixation to threatening (i.e., angry), happy, and neutral facial configurations. In the Overlap task, infants were presented with happy, angry, or neutral facial configurations in the center of a screen, and shortly after their appearance, a checkerboard probe appeared simultaneously to the left or the right of the facial stimulus. To index engagement, we measured total looking to the facial stimuli while the probe was present, and to index disengagement, we measured total looking to the probe while the face was present. We then used these data in cross-lagged longitudinal models to discern how different components of attention to various emotional facial configurations (angry, happy, and neutral) and social fear influence one another over time.

There are several possible hypotheses. One is that there will be no relation between attention to threat and social fear over time in this young age group. Another is that attention to threat at earlier assessments will predict social fear at later assessments, suggesting that attention bias to threat is prospectively related to social fear, in line with previous work (Van Bockstaele et al., 2014). Alternatively, it is also possible that social fear at earlier assessments will predict attention bias to threat at later assessments, suggesting that social fear is prospectively related to a bias for threatening facial configurations. One final possibility is that the relation between attention to threat and social fear is reciprocal over time, such that social fear at one time point would predict attention to threat at the next time point and vice versa.