Endogenous Control and Reward-Based Mechanisms Shape Infants’ Attention Biases to Caregiver Faces

Abstract

Infants rely on developing attention skills to identify relevant stimuli in their environments. Although caregivers are socially rewarding and a critical source of information, they are also one of many stimuli that compete for infants’ attention. Young infants preferentially hold attention on caregiver faces, but it is unknown whether they also preferentially orient to caregivers and the extent to which these attention biases reflect reward-based attention mechanisms. To address these questions, we measured 4- to 10-month-old infants’ (N = 64) frequency of orienting and duration of looking to caregiver and stranger faces within multi-item arrays. We also assessed whether infants’ attention to these faces related to individual differences in Surgency, an indirect index of reward sensitivity. Although infants did not show biased attention to caregiver versus stranger faces at the group level, infants were increasingly biased to orient to stranger faces with age. However, infants with higher Surgency scores showed more robust attention orienting and attention holding biases to caregiver faces. These effects varied based on the selective attention demands of the task, suggesting that infants’ attention biases to caregiver faces may reflect both developing attention control skills and reward-based attention mechanisms.

Developing attention skills facilitate infants’ ability to direct their attention to and maintain focus on their social partners (Leppänen, 2016). Caregivers are one of the most frequently experienced and rewarding stimuli in infants’ environments (Minagawa-Kawai et al., 2009; Sugden & Moulson, 2019), a source of physical and psychological safety (Bowlby, 1969, 1988), and a critical source of early learning input (Barker & Newman, 2004; Barry-Anwar et al., 2017; Montague & Walker-Andrews, 2002). Beginning shortly after birth, neonates preferentially look towards caregiver faces compared to stranger faces when they are presented simultaneously (Bushnell, 2001; Pascalis et al., 1995; Rigato et al., 2023). Yet caregiver faces are only one of many stimuli present in an infant’s environment and it is unclear whether these previously observed attention holding biases to caregivers extend to more complex contexts. Specifically, it is unknown whether infants detect caregiver versus stranger faces more efficiently in contexts with many items competing for attention (i.e., biased attention orienting). Furthermore, while research has established that infants’ orienting biases to unfamiliar stranger faces reflect age-related improvements in endogenous attention control skills (e.g., Frank et al., 2014; Kelly et al., 2007; Kwon et al., 2016; Quinn et al., 2002), caregivers activate neural reward circuitry in both infancy and childhood (Abrams et al., 2016; Liu et al., 2019; Minagawa-Kawai et al., 2009) and research has shown that this type of social reward, can also drive attention orienting biases in adulthood (Awh et al., 2012; Kim et al., 2021; Luck et al., 2021). However, research has not yet investigated the extent to which social reward may contribute to infants’ attention biases to caregivers over the first year. To address these questions, we directly compared 4- to 10-month-old infants’ attention orienting and attention holding biases to caregiver and stranger faces within the same visual search task. Using this approach, we examined the extent to which participant age, selective attention demands of the task, and individual differences in reward sensitivity contribute to these attention biases to caregiver versus stranger faces in infancy. We offer initial evidence that, when tested in the same task context, infants show parallel attention holding and orienting biases to caregiver faces that reflect both endogenous attention control and reward-based mechanisms.

Infants rely on developing attention skills to select meaningful information among competing stimuli in crowded environments (attention orienting) and sustain looking on a selected stimulus for detailed processing and encoding (attention holding; Cohen, 1972). Attention orienting and attention holding function together in everyday contexts (Fisher 2019) but are assessed via separate measures and reflect distinct underlying mechanisms. To evaluate attention holding, researchers have traditionally used visual paired comparison (VPC) tasks to compare duration of looking across two stimuli that appear simultaneously (e.g., Fagan, 1970; Fantz, 1956), with an attention holding bias defined by longer looking towards one stimulus (i.e., preferential looking; Cohen, 1972). Attention orienting has instead been assessed using visual search tasks, in which a target appears among multiple task-irrelevant distractors. As the number of distractors increases, individuals typically detect the target more slowly or less frequently, reflecting increased competition from the additional competing information (e.g., Gerhardstein & Rovee-Collier, 2002; Wolfe, 2015). An orienting bias towards a specific stimulus can be indexed based on relatively faster or more frequent orienting to one type of target compared to another (e.g., Hunter & Markant, 2021).

Faces are a prevalent and valuable information source in an infant’s environment and elicit both attention holding and attention orienting biases in the first year (Jakobsen et al., 2016). Newborn infants look longer to faces or face-like stimuli compared to non-faces, indicating an early attention holding bias to faces (Farroni et al., 2005; Morton & Johnson, 1991; Valenza et al., 1996). However, infants do not reliably orient to faces within more complex scenes or arrays until 6 months of age (Di Giorgio et al., 2012; Gliga et al., 2009; Gluckman & Johnson, 2013; Hunter & Markant, 2021; Jakobsen et al., 2016; Kwon et al., 2016; Prunty et al., 2020; Simpson et al., 2020; but see Simpson, Maylott, Leonard, et al., 2019). These increasingly robust orienting biases to faces reflect age-related improvements in infants’ endogenous attention control skills, which support their ability to voluntarily ignore physical salience and resolve visual competition between distractors (e.g., Di Giorgio et al., 2012; Hunter & Markant, 2021; Kwon et al., 2016; Simpson, Maylott, Leonard, et al., 2019). For example, 4-month-old infants showed an orienting bias to a face only when it was presented alongside one distractor that was relatively low in physical salience, whereas 6- and 8-month-old infants reliably oriented to a face that appeared among five distractors, regardless of their physical salience (Kwon et al., 2016). However, because attention holding biases to faces have typically been evaluated using relatively less demanding tasks, it is unclear whether developing endogenous control similarly influences infants’ attention holding biases to faces in more complex contexts.

In addition to developing endogenous attention control, infants’ prior experience with categories of faces can also shape their attention biases to faces. Extensive research has established that infants show stronger attention holding biases towards faces that are similar to those that they regularly experience in their environment. For example, young infants show stronger attention holding biases for familiar own-race faces compared to less familiar other-race faces (e.g., Kelly et al., 2007). Critically, this effect is reversed if an other-race individual assumes a caregiving role (e.g., through adoption; Sangrigoli et al., 2005), confirming that these attention biases reflect infants’ experiences with faces. However, recent studies did not observe race-based attention orienting biases either in infancy or childhood (Hunter & Markant, 2021, 2023a; Prunty et al., 2020), suggesting that experience with specific face categories may differentially influence attention holding and orienting biases to faces.

In addition to these attention biases to familiar face categories, research has identified experience-based attention biases that are based on specific face identities. Within the first hours of life, newborn infants look longer at their caregiver’s face compared to a stranger’s face (Bushnell, 2001; Field et al., 1984; Pascalis et al., 1995). Infants continue to preferentially look to caregivers over the first several months of life (Barrera & Mauer, 1981), reflecting a robust attention holding bias towards caregiver faces. By 6 months of age, infants may show increased preferential looking to stranger faces (e.g., Bartrip et al., 2001), but these findings have been mixed (e.g., Rigato et al., 2023). Although infant research has focused on attention holding, recent work found that by 4 years of age, children are biased to detect caregiver faces in complex arrays containing competing visual information (Fields et al., 2022; Hunter & Markant, 2023b). However, it is unknown if similar attention orienting biases to caregivers are evident in infancy. This gap limits our understanding of the extent to which infants show parallel attention holding and attention orienting biases to caregivers during the first year of life.

Furthermore, the mechanisms underlying infants’ attention biases to caregiver faces are also unclear. Attention holding biases in infancy are typically interpreted as an index of recognition memory (Aslin, 2007), whereas attention orienting biases have been more closely linked to rapid stimulus detection (Leppänen, 2016). Developmental researchers have traditionally argued that this stimulus detection can be driven by perceptual salience or endogenous control (Colombo, 2001) and classic models of selective attention development primarily emphasize a shift from initial salience-based orienting in early infancy to increasing endogenous attention control starting at 4- to 6-months (Reynolds & Romano, 2016). This shift towards increased endogenous attention control over the first year is considered a critical factor underlying the development of biased orienting to unfamiliar faces at 6 months of age (e.g., Kwon et al., 2016). However, adult research has established that attention orienting can also reflect multiple additional factors, including prior learning and experience, the individual’s state of arousal, or the emotional relevance of the stimulus (Luck et al. 2021; Awh et al. 2012; Sara & Bouret 2012). For example, adults show robust orienting biases towards multiple categories of rewarding stimuli, including social reward, independent of the physical salience or task-relevance of the stimuli (Anderson, 2016; Anderson, 2019; Anderson et al., 2011; Theeuwes & Belopolsky, 2012). Caregivers are a frequently experienced face identity (Sugden & Moulson, 2019) and also selectively activate neural reward circuitry among infants and children (Abrams et al., 2016; Liu et al., 2019; Minagawa-Kawai et al., 2009). Recent work identified an orienting bias to caregivers in childhood after controlling for both perceptual salience and task-relevance (Hunter & Markant, 2023b), suggesting that attention biases towards caregivers may reflect other factors such as familiarity or reward value.

One approach to evaluating this potential reward-based attention mechanism is to relate individual differences in infants’ reward sensitivity to their attention biases to caregiver faces. Infants show relatively consistent individual differences in temperament that reflect variability in their emotional reactivity and regulation (Rothbart & Derryberry 1981; Rothbart & Bates, 2006). Infant temperament assessments (e.g., Infant Behavior Questionnaire-R; Gartstein & Rothbart, 2003) often include a measure of positive reactivity/affect (“Surgency/Extraversion”; Shiner et al., 2012), which is classically conceptualized based on infants’ propensity to approach positive valenced stimuli, tendency to find pleasure in high-intensity stimuli, and their displays of positive affect, sociability, and high activity levels (e.g., Holmboe 2016). However, research suggests that Surgency/Extraversion may also indirectly index reward sensitivity across development. For example, Vervoort et al. (2015) found that Surgency/Extraversion scores predicted parent report of 2–18-year-old children’s reward sensitivity assessed via the The Sensitivity to Punishment and Sensitivity to Reward Questionnaire (Colder & O’Connor, 2004). Functional neuroimaging studies have also revealed a relationship between adults’ trait extraversion and individual differences in prefrontal cortical reward processing regions (see Wacker & Smillie, 2015 for review). Consistent with this, Kujawa et al. (2015) found that increased positive emotionality at 3 years of age predicted a more robust neural response to reward in middle childhood. These findings suggest that measures of Surgency/Extraversion can provide an indirect measure of reward sensitivity in infancy. Initial evidence also showed that Surgency scores predicted 6- to 10-month-old infants’ face processing and recognition to a greater extent than Orientation/Regulation scores (Rennels et al., 2020). Although this finding suggests that individual differences in positive reactivity and reward sensitivity, indexed via Surgency, relate to infants’ face processing, it is unclear whether this link extends to attention biases to caregivers specifically. Therefore, examining the relationship between Surgency and infants’ attention biases to caregiver faces may provide initial insight into the extent to which these attention biases are shaped by reward-based mechanisms.

In summary, attention holding and orienting are distinct processes (Leppänen, 2016) that can reflect multiple underlying mechanisms (e.g., perceptual salience, endogenous control, familiarity, reward value; Cohen, 1972; Hunter et al., 1983; Kim et al., 2021). Young infants are biased to attend to faces, with attention holding biases evident among newborns (Farroni et al., 2005; Morton & Johnson, 1991; Valenza et al., 1996) and attention orienting biases emerging by 6 months of age (Di Giorgio et al., 2012; Gliga et al., 2009; Kwon et al., 2016). These attention biases to faces are shaped by experience, as infants show robust attention holding biases to familiar face categories (e.g., own-race faces) and to their own caregiver’s face (Bushnell, 2001; Pascalis et al., 1995; Rigato et al., 2023). However, evidence for experience-based orienting biases to faces has been more mixed. Although recent research found that older children were biased to orient to their caregiver’s face (Fields et al., 2022; Hunter & Markant, 2023b), it is unknown whether infants are similarly biased to orient to caregivers. Furthermore, because attention holding and orienting biases are typically assessed using tasks with varying demands (i.e., visual paired comparison vs. multi-object arrays), the extent to which these early attention biases may develop in parallel remains unclear. Finally, the mechanisms underlying early attention biases to caregivers are not understood. Infants’ orienting to unfamiliar faces is shaped by age-related improvements in endogenous attention control (e.g., Frank et al., 2014; Kwon et al., 2016), but research has also established that multiple additional factors, including social reward, can drive biased orienting in adulthood (Awh et al., 2012; Luck et al., 2021; Kim et al., 2021). Although initial work has related individual differences in Surgency/Extraversion, an indirect measure of reward sensitivity, to face processing and recognition in infancy (Rennels et al., 2020), research has yet to investigate links between this measure and infants’ attention biases to caregivers specifically.

To address these gaps in the research, we investigated the development of attention orienting and attention holding biases to caregiver versus stranger faces among 4- to 10-month-old infants. Infants viewed 6-item arrays that contained caregiver and stranger faces appearing among multiple competing distractors. We assessed infants’ frequency/speed of orienting and duration of looking to the faces to measure both attention orienting and attention holding biases to caregiver faces, respectively. We addressed three primary questions in this study. First, we examined age-related changes in infants’ attention biases to determine whether they showed parallel attention orienting and attention holding biases to caregiver versus stranger faces over the first year. Second, we determined whether these attention biases varied based on the selective attention demands of the task. Specifically, during some trials, only one face (i.e., either the caregiver or stranger) appeared among multiple distractors and during the remaining trials, both the caregiver and stranger faces appeared simultaneously within the multi-object arrays. Presenting two face types at once is more consistent with classic visual paired comparison tasks (e.g., Bushnell, 2001; Pascalis et al., 1995; Rigato et al., 2023) and increases selective attention demands by placing faces in direct competition. We expected that these changing task demands would influence infants’ attention biases to caregiver versus stranger faces but did not have a strong prediction about the direction of the effect. For example, infants may only show attention biases to caregiver faces when they are in direct competition with stranger faces (i.e., similar to visual paired comparison tasks). Alternatively, this direct competition between the faces may be too challenging to resolve, leading to similar face detection rates. Finally, we examined links between individual differences in Surgency and infants’ attention biases to caregiver versus stranger faces to evaluate the potential role of reward-based attention mechanisms (i.e., motivational salience) underlying these attention biases. Given that Surgency is an indirect index of infants’ reward sensitivity and parents are rewarding to infants (i.e., they activate neural reward circuitry), examining these links allowed us to take an initial step towards determining whether infants’ attention biases to caregivers may reflect reward-based attention mechanisms. We predicted that infants with higher Surgency scores may show stronger attention biases to caregiver faces due to their increased tendency to approach positive or rewarding stimuli. Evidence in support of this prediction would provide initial support for the idea that social reward value may contribute to caregiver attention biases in infancy.

Method

Participants

The final sample included 64 participants (31 Females, 33 Males) between 4- and 10-months-old (M_age = 7 months, 6 days, SD = 1 month, 27 days, range = 3 months, 27 days - 10 months, 1 day). An a priori power analysis in WebPower (Zhang & Yuan, 2018; alpha= .05, power = .80) indicated that this sample size would be sufficient to detect effect sizes similar to those observed in previous studies (e.g., Hunter & Markant, 2021) and allow us to examine individual differences in Surgency. According to parent report, 67.2% of participants were White/Caucasian, 12.5% were Black/African American, 7.8% were Hispanic, 4.7% were Multiracial, 3.1% were Asian, 1.6% were American Indian/Alaskan Native, and 3.1% listed their race as Other. We tested nine additional infants but excluded them from final analyses due to fussiness (N = 6), poor eye tracking calibration/validation (N = 2), or technical issues (N = 1). Caregivers provided informed consent in accordance with university institutional review board guidelines and received a $10 gift as compensation. We tested all participants between 1/28/2020 and 7/28/2023¹.

Measures

Visual Search Task

We recorded infants’ eye movements as they viewed 6-item visual search arrays that contained caregiver and stranger faces and multiple distractor objects (Figure 1). During single target trials, only one face (either the caregiver or a stranger’s face) appeared in the array with the distractors. During dual target trials, the caregiver and stranger faces instead appeared simultaneously. We took photos of caregiver faces in the lab under similar lighting conditions against a black curtain with a black drape covering the caregivers’ shoulders. We instructed caregivers to remove any glasses/hats, maintain eye contact, and display a positive facial expression without showing teeth. To control for perceptual differences across face stimuli, we used each photo once as the caregiver face stimulus and again as the stranger face stimulus for another participant. We also matched caregiver and stranger face stimuli based on the gender and race of faces when possible². The remaining distractor stimuli were images of flowers, shoes, cars, toys, and chairs. Each 150 × 150 pixel (4.2° x 4.2°) stimulus appeared within a 200 × 200 pixel (5.6° x 5.6°) white square overlaid on a gray background.

Figure 1.

Examples of multi-item search arrays. Trials included either (A) only the caregiver’s face, (B) only the stranger’s face, or (C) both the caregiver and stranger faces appearing simultaneously.

The six stimuli appeared in a circular array with the center of each image equidistant from the center of the screen (7.8° visual angle). We created the search arrays by preselecting the locations of all images. Both face types appeared once in each possible location during single target trials and appeared an equal number of times in each location across participants during dual target trials. We selected 15 unique images per non-face distractor category and placed them in the remaining locations. Each category of distractor objects appeared an equal number of times in each location across participants and unique distractor images did not repeat within-subject. After each testing session, we used the Graph-Based Visual Saliency toolbox (Harel et al., 2006) to identify the most physically salient image in each search array. On average, a face was identified as the most salient item during 8.1% of trials (SD = 8.0%), but this did not vary for caregiver versus stranger faces (p = .69). Therefore, any potential differences in attention to face types cannot be attributed to differences in physical salience alone.

We used an EyeLink 1000 Plus eye tracker (SR Research, Mississauga, Ontario, Canada) to record infant eye movements at 500 Hz throughout the task. Infants sat in a highchair or on their parent’s lap³ 60 cm away from a 17-inch display (1280 × 1024 resolution). Infants could not see their caregiver’s face or the experimenter’s face while they completed the eye tracking task. All participants first completed a 3-point calibration and validation protocol (average deviation = 0.67°, SD = 0.40°), after which each trial began with the presentation of an animated audiovisual stimulus to center participants’ gaze. The visual search array appeared when participants looked at the central attention-getting stimulus (> 500 ms) and remained visible for 5000 ms. Infants completed six trials of each trial type (single target-caregiver, single target-stranger, dual target), resulting in a total of 18 trials. Half of the participants completed all dual target trials first and the remaining participants completed all single target trials first. Participants viewed the arrays in randomized order within these two sets of trial types.

Infant Behavior Questionnaire.

Parents completed the Infant Behavior Questionnaire-Revised (IBQ-R; Gartstein & Rothbart, 2003) online via REDCap (Harris et al., 2019) within 3 days prior to their infant’s scheduled test session. This questionnaire is validated to measure 3- to 12-month-old infant temperament from a series of 91 multiple-choice questions. The questions asked parents to rate the frequency of discrete behaviors displayed by their child over the previous 2 weeks across multiple contexts (e.g., “When given a new toy, how often did the baby get very excited about getting it?”). Caregivers rated each item on a 7-point Likert scale from 1 (never) to 7 (always). Previous research has demonstrated the reliability of the IBQ-R, with reported Cronbach alpha coefficients ranging from 0.59 to 0.92 for the Surgency/Extraversion scale (Castro Dias et al., 2021; Gartstein & Rothbart, 2003; Parade & Leerkes, 2008). We observed similar reliability for the Surgency/Extraversion scale within the current sample (α = .69).

Data Processing

Visual Search Task.

We used the DataViewer software (SR Research, Mississauga, Ontario, Canada) to create 280 × 280 pixel (7.8° x 7.8°) interest areas surrounding each image in the search array. We identified fixations using EyeLink’s fixation parsing algorithm and removed any fixations less than 100 ms in duration. On average, eye tracking data were unavailable for 0.22 trials per subject (SD = 0.65 trials) due to data loss. We also excluded any trials in which the participant was not looking at the center of the screen at the start of the trial (M = 0.66 trials per subject, SD = 1.07 trials) or those in which the participant did not look at any of the interest areas during the trial (M = 0.17 trials per subject, SD = 0.38 trials). The proportion of dropped trials was unrelated to trial type, age, or Surgency scores (p’s > .6).

We assessed attention orienting based on initial orienting frequency, which may reflect a reflexive response to faces (e.g., DiGiorgio et al., 2012; Kwon et al., 2016), as well as overall orienting frequency, which may reflect a broader view of how infants detect faces over time (e.g., Hunter & Markant, 2021; Jakobsen et al., 2016). We computed initial orienting frequency based on the proportion of trials in which a face was the first item that the infant looked at in the search array and overall orienting frequency based on the proportion of trials in which infants looked at a face at any point during the trial. We also computed orienting speed based on the response time of infants’ first look at a face. Finally, we assessed attention holding based on the total duration of looking to the interest areas containing a face. We computed all variables separately for each face type (caregiver and stranger faces) within each trial type (single and dual target trials). We included all infants for analyses of initial and overall orienting frequency (N = 64). However, we computed orienting speed and duration of looking based only on trials in which infants looked at a face. One infant did not look at any stranger faces during dual target trials and was excluded from analyses of orienting speed and look duration (N = 63).

Infant Behavior Questionnaire.

The IBQ-R yields three temperament factors, including Surgency/Extraversion, Negative Affect, and Orientation/Regulation. However, we focused all analyses on Surgency/Extraversion scores, consistent with our a priori hypotheses described earlier. We computed Surgency scores by averaging the subscales that load most heavily on that factor, including the approach, high intensity pleasure, vocal reactivity, smiling and laughter, activity level, and perceptual sensitivity scales (Gartstein & Rothbart, 2003). Consistent with prior research (e.g., Gartstein & Rothbart, 2003), preliminary analyses indicated that participant sex was unrelated to Surgency ratings (p = .81) but older infants were more likely to be rated higher in Surgency, r(64) = .52, p < .001. .

Results

We assessed infants’ attention orienting based on their initial orienting frequency, overall orienting frequency, and speed of orienting to the caregiver and stranger faces, while attention holding was based on their duration of looking to these faces. We examined each of these variables using repeated measures analyses of covariances (ANCOVAs) with face type (caregiver vs. stranger) and trial type (single vs. dual target) as within-subjects factors and age and Surgency scores included as continuous centered covariates⁴. We examined age and Surgency scores as continuous variables in all ANCOVA models and follow-up analyses. To interpret the observed interactions between these continuous covariates and the face type variable, we computed bias scores based on the difference in attention to the caregiver face versus the stranger face (e.g., initial orienting to caregiver - initial orienting to stranger). Positive values indicated a bias to the caregiver face, whereas negative values indicated a bias to the stranger face. Initial analyses revealed no reliable differences in overall orienting frequency or looking time to faces in general (i.e., collapsed across caregiver and stranger faces) based on participant sex (p’s > .2). There was a significant effect of participant sex on initial orienting frequency (M_Female = .50, SD = .11; M_Male = .43, SD = .14), t(62) = 2.20, p = .031, d = 0.56, and a marginally significant effect of sex on orienting speed, (M_Female = 790.67 ms, SD = 217.38; M_Male = 921.72 ms, SD = 311.02), t(57.38) = 1.96, p = .054, d = 0.49. However, the ANCOVA results did not change regardless of whether we included participant sex in the model. We therefore report the ANCOVAs without participant sex in the model, as we did not have any a priori predictions about this variable. ANCOVA results for measures of attention orienting (i.e., initial orienting frequency, overall orienting frequency, orienting speed) and attention holding (i.e., duration of looking) are presented in Tables 1 and 2 and summarized below.

Table 1.

Analysis of covariance results for attention orienting measures

Predictor	Initial Orienting Frequency				Overall Orienting Frequency				Orienting Speed
	df	F	p	ηp2	df	F	p	ηp2	df	F	p	ηp2
Face Type	1,61	0.04	.85	< .01	1,61	0.06	.82	< .01	1,60	0.01	.91	< .01
Trial Type	1,61	69.09	< .001***	.53	1,61	31.58	< .001**	.34	1,60	14.38	< .001***	.19
Age	1,61	5.56	.02*	.08	1,61	11.73	.001**	.16	1,60	1.17	.28	.02
Surgency	1,61	1.22	.28	.02	1,61	2.39	.13	.04	1,60	0.02	.89	< .01
Face Type x Trial Type	1,61	0.01	.92	< .01	1,61	0.17	.69	< .01	1,60	0.03	.87	< .01
Face Type x Age	1,61	9.96	.002**	.14	1,61	4.93	.03*	.08	1,60	3.12	.08†	.05
Trial Type x Age	1,61	0.02	.90	< .01	1,61	4.90	.03*	.07	1,60	0.32	.58	< .01
Face Type x Surgency	1,61	11.21	.001**	.16	1,61	9.53	.003**	.14	1,60	0.92	.34	.02
Trial Type x Surgency	1,61	0.14	.71	< .01	1,61	0.25	.62	< .01	1,60	1.06	.31	.02
Face Type x Trial Type x Age	1,61	5.64	.02*	.09	1,61	1.49	.23	.02	1,60	2.24	.14	.04
Face Type x Trial Type x Surgency	1,61	14.16	< .001***	.19	1,61	9.25	.003**	.13	1,60	0.54	.47	.01

Note.

^†p < .10,

^*p < .05,

^**p < .01,

^***p < .001

Table 2.

Analysis of covariance results for attention holding measure

Predictor	Duration of Looking
	df	F	p	ηp2
Face Type	1,60	0.76	.39	.01
Trial Type	1,60	15.30	< .001***	.20
Age	1,60	10.07	.002**	.14
Surgency	1,60	0.42	.52	.01
Face Type x Trial Type	1,60	0.01	.91	< .01
Face Type x Age	1,60	0.12	.73	< .01
Trial Type x Age	1,60	0.01	.92	< .01
Face Type x Surgency	1,60	4.39	.04*	.07
Trial Type x Surgency	1,60	0.78	.38	.01
Face Type x Trial Type x Age	1,60	0.20	.66	< .01
Face Type x Trial Type x Surgency	1,60	4.99	.03*	.08

Note.

^*p < .05,

^**p < .01,

^***p < .01

Attention Orienting

Initial Orienting Frequency

Infants showed reliable initial orienting to the faces, collapsed across face type, at above chance levels during both single target trials (chance = .17; M = .55, SD = .19), t(63) = 16.07, p < .001, d = 2.01, and dual target trials (chance = .33; M = .38, SD = .11), t(63) = 3.48, p = .001, d = 0.43. ANCOVA results indicated a main effect of trial type on frequency of initial orienting to the faces, F(1,61) = 69.09, p < .001, η_p² = .53, reflecting overall increased initial orienting to a face during single target trials (i.e., one face in the array) compared to dual target trials (i.e., both faces in the array; Figure 2A). Infants were also overall more likely to initially orient to faces with age, F(1,61) = 5.56, p = .022, η_p² = .08. However, this main effect of age was further moderated by a significant face type x age interaction, F(1,61) = 9.96, p = .002, η_p² = .14, and a significant face type x trial type x age interaction, F(1,61) = 5.64, p = .021, η_p² = .09. Follow-up analyses showed that the face type x age interaction was not reliable during single target trials (p = .272) but was significant during dual target trials, F(1,61) = 13.98, p < .001, η_p² = .19. To further examine this interaction, we computed an initial orienting bias score for dual target trials (i.e., initial orienting to caregiver - initial orienting to stranger). Positive values indicated an initial orienting bias to the caregiver face, whereas negative values indicated an initial orienting bias to the stranger face. When controlling for Surgency, age was negatively correlated with initial orienting bias scores during dual target trials, r(61) = −.53, p < .001 (Figure 2B). These results indicate an age-related decrease in initial orienting to the caregiver versus stranger when both faces appeared in the array.

Figure 2.

Effects of trial type, age, and surgency on infants’ initial orienting to faces. (A) Infants were more likely to initially look at a face during single target trials (one face in the array) compared to dual target trials (both faces in the array). Error bars represent SEM. (B) Partial regression plot illustrating the relation between participant age and initial orienting bias to caregiver vs. stranger faces during dual target trials, when controlling for Surgency scores. Positive bias scores indicated increased initial orienting to the caregiver face. (C) Partial regression plot illustrating the relation between Surgency scores and initial orienting bias to caregiver vs. stranger faces during dual target trials, when controlling for participant age.

Results also indicated a significant face type x Surgency interaction, F(1,61) = 11.21, p = .001, η_p² = .16, which was further moderated by a significant face type x trial type x Surgency interaction, F(1,61) = 14.16, p < .001, η_p² = .19. Similar to the face type x age interaction, follow up analyses showed that the face type x Surgency interaction was reliable during dual target trials, F(1,61) = 21.97, p < .001, η_p² =.27, but not during single target trials (p = .714). However, the direction of this effect differed for age versus Surgency. When controlling for age, Surgency scores were instead positively correlated with initial orienting bias scores, r(61) = .52, p < .001 (Figure 2C), indicating that infants with higher Surgency scores were more likely to initially orient to their caregiver’s face relative to the stranger face when both faces appeared in the array.

Overall Orienting Frequency

Infants’ overall frequency of orienting to caregiver and stranger faces largely mirrored the effects observed for initial orienting frequency. Overall frequency of orienting to the faces was at above chance levels during both single target trials (chance = 0.17; M = .71, SD = .19), t(63) = 23.00, p < .001, d = 2.88, and dual target trials (chance = 0.33; M = .59, SD = .17), t(63) = 12.29, p’s < .001, d = 1.54. ANCOVA results indicated a main effect of trial type on overall frequency of orienting to faces, F(1,61) = 31.58, p < .001, η_p² = .34. Infants were overall more likely to orient to a face during single target trials compared to dual target trials (Figure 3A). Infants were also overall more likely to orient to faces with age, F(1,61) = 11.73, p = .001, η_p² = .16. Similar to initial orienting frequency, this main effect of age on overall orienting frequency was moderated by a face type x age interaction, F(1,61) = 4.93, p = .03, η_p² = .08. However, unlike initial orienting frequency, the three-way face type x trial type x age interaction was not significant (p = .226). To further examine the face type x age interaction, we computed an overall orienting bias score based on the difference in overall orienting frequency to caregiver versus stranger faces. Because the three-way interaction was not significant, we computed this bias score collapsed across both single and dual target trials (i.e., orienting to caregiver - overall orienting to stranger). When controlling for Surgency, age was negatively correlated with overall orienting bias scores, r(61) = −.27, p = .03 (Figure 3B). Thus, similar to the effect observed for initial orienting, these results indicated an age-related decrease in orienting to caregiver versus stranger faces across trial types. Finally, unlike the results for initial orienting, we also found a trial type x age interaction, F(1,61) = 4.90, p = .03, η_p² = .07, that reflected an age-related increase in overall orienting to faces, regardless of face type, during dual target trials, r(61) = .51, p < .001, but no reliable change in orienting to faces during single target trials (p = .09).

Figure 3.

Effects of trial type, age, and surgency on infants’ overall orienting to faces. (A) Infants were more likely to look at a face during single target trials (one face in the array) compared to dual target trials (both faces in the array). Error bars represent SEM. (B) Partial regression plot illustrating the relation between participant age and overall orienting bias to caregiver vs. stranger faces across trial types when controlling for Surgency scores. Positive bias scores indicated increased overall orienting to the caregiver face. (C) Partial regression plot illustrating the relation between Surgency scores and overall orienting bias to caregiver vs. stranger faces during dual target trials, when controlling for participant age.

Similar to initial orienting frequency, the results for overall orienting frequency also indicated a significant face type x Surgency interaction, F(1,61) = 9.53, p = .003, η_p² = .14, which was further moderated by a significant face type x trial type x Surgency interaction, F(1,61) = 9.25, p = .003, η_p² = .13. Follow-up analyses indicated that the face type x Surgency interaction was reliable during dual target trials, F(1,61) = 18.82, p < .001, η_p² = .24, but not during single target trials (p = .811). To further examine this interaction, we computed an orienting bias score as above (i.e., overall orienting to caregiver - overall orienting to stranger) that was specific to dual target trials. When controlling for age, Surgency scores were positively correlated with overall orienting bias scores during dual target trials, r(61) = .49, p < .001 (Figure 3C), indicating that infants with higher Surgency scores were more likely to orient to their caregiver’s face relative to the stranger’s face when both faces appeared together in the array.

Orienting Speed

Infants’ overall speed of orienting to the faces depended on trial type, F(1,60) = 14.38, p < .001, η_p² = .19. When collapsed across face types, infants looked to faces faster during single target trials (M = 751.98, SD = 328.12) compared to dual target trials (M = 974.03, SD = 394.56), consistent with effects observed for initial and overall orienting frequency. The overall pattern of results for face type, age, and Surgency mirrored the effects that we observed for initial and overall orienting frequency but these effects did not reach significance for orienting speed (p’s > .08).

Attention Holding

Duration of Looking

ANCOVA results indicated a main effect of trial type on infant’s duration of looking to faces, F(1,60) = 15.30, p < .001, η_p² = .20, such that infants spent overall more time looking at the faces during single target trials (M = 1402.52 ms, SD = 462.22 ms) compared to dual target trials (M_Dual = 1175.52 ms, SD = 398.32; Figure 4A). Infants also showed overall increased looking to faces with age, F(1,60) = 10.07, p = .002, η_p² = .14. However, this overall effect of age did not interact with trial type or face type (p’s > .60).

Figure 4.

Effects of trial type and surgency on infants’ attention holding to faces. (A) Infants spent more time looking at faces during single target trials (one face in the array) compared to dual target trials (two faces in the array). Error bars represent SEM. (B) Partial regression plot illustrating the relation between Surgency scores and preferential looking to caregiver vs. stranger faces during single target trials when controlling for participant age.

Results again showed a significant face type x Surgency interaction F(1,60) = 4.39, p = .04, η_p² = .07, which was further moderated by a significant face type x trial type x Surgency interaction, F(1,60) = 4.99, p = .029, η_p² = .08. However, in contrast to the orienting measures described above, follow-up analyses showed that the face type x Surgency interaction was significant during single target trials, F(1,60) = 8.67, p = .005, η_p² = .12, but not during dual target trials (p = .733). To further examine this interaction, we computed a preferential looking score based on the difference in duration of looking to the caregiver face versus the stranger face (i.e., duration of looking to caregiver - duration of looking to stranger) during single target trials only. When controlling for age, Surgency scores were positively correlated with preferential looking scores during single target trials, r(61) = .35, p = .005 (Figure 4B), indicating that infants with higher Surgency scores showed increased preferential looking to caregiver versus stranger faces when only one face appeared in the array.

Discussion

The current study investigated the development of attention orienting and attention holding biases to caregiver versus stranger based on 4- to 10-month-old infants’ frequency and speed of orienting and duration of looking to these faces when they appeared in multi-item search arrays. Past research examining infants’ attention holding biases to faces has primarily used visual paired comparison tasks (two faces appearing in direct competition; e.g., Fagan, 1970; Fantz, 1956), whereas orienting biases have typically been assessed using visual search tasks (a face appearing among multiple competing distractors; e.g., Kwon et al., 2016). However, the varying selective attention demands of these different tasks have made it challenging to compare the development of these attention biases to faces in infancy. In the current study, we directly compared infants’ attention orienting and attention holding to caregiver and stranger faces within a single search task but manipulated selective attention demands by varying the number of face targets within the arrays. Specifically, during single target trials only one face competed with the distractors, whereas during dual target trials both the caregiver and a stranger’s face appeared in direct competition. Using this design, we examined age-related changes in infants’ attention biases to caregiver versus stranger faces and determined whether these attention biases varied based on task demands and individual differences in Surgency, an indirect measure of reward sensitivity. Although infants showed no overall attention biases based on face identity at the group level, participant age, task context, and individual differences in Surgency related to both attention orienting and attention holding biases to caregivers. Overall, these results provide initial evidence that infants’ attention to caregiver faces may reflect both reward-based mechanisms as well as the selective attention demands of the task.

Prior research has established that infants show attention holding biases to faces or face-like stimuli beginning shortly after birth (Farroni et al., 2005; Morton & Johnson, 1991; Valenza et al., 1996) but do not consistently orient to faces in more complex arrays until 6 months of age (e.g., Di Giorgio et al., 2012; Gliga et al., 2009; Kwon et al., 2016). This relatively slower development of attention orienting biases to faces reflects the importance of improving endogenous attention control skills for infants’ ability to resolve visual competition between faces and surrounding distractors (e.g., Di Giorgio et al., 2012; Hunter & Markant, 2021; Kwon et al., 2016; Simpson, Maylott, Leonard, et al., 2019). Consistent with these past findings, we found that infants showed increasingly robust attention holding and attention orienting biases to faces with age. The current results also extend prior work by demonstrating that infants showed stronger orienting biases when only one face appeared in the array, even though the probability of detecting a face was higher when both faces appeared in the array. The decreased competition between one face and surrounding distractor objects during single target trials may have facilitated infants’ attention to faces, whereas the increased competition between two faces during dual target trials may have been more challenging to resolve. These age- and task-based effects are broadly consistent with evidence that infants’ attention biases to unfamiliar faces in complex arrays emerge over the first year as improving endogenous attention control facilitates more efficient selection of faces (Di Giorgio et al., 2012; Gliga et al., 2009; Gluckman & Johnson, 2013; Hunter & Markant, 2021; Jakobsen et al., 2016; Kwon et al., 2016; Prunty et al., 2020; Simpson et al., 2020; Simpson, Maylott, Leonard, et al., 2019). However, because the overall size of the arrays was consistent across trial types in the current study, it is not clear how other distractor characteristics (e.g., number, physical salience; c.f., Kwon et al., 2016) may interact with these effects. In addition, although the multi-object arrays used in the current study contained more visual stimuli than classic visual paired comparison tasks, they were still far from the complex and dynamic environments that infants experience daily. Additional work examining infants’ attention to faces in more naturalistic settings (e.g., using scenes or video stimuli) will be needed to better understand how both social and non-social features of complex environments (e.g., amount of clutter or social information) interact with age-related improvements in endogenous attention control to shape infants’ attention biases to faces in general.

In addition to these attention biases to faces in general, we also examined changes in infants’ attention to caregiver versus stranger faces over the first year. Past work examining attention biases to caregiver versus stranger faces in infancy has solely focused on attention holding in the context of visual paired comparison tasks. This research has yielded mixed results; some studies reported preferential looking to caregivers as late as 6 months of age (e.g., Rigato et al., 2023) while others observed increased preferential looking to stranger faces from 1- to 5 months of age (Bartrip et al., 2001). Results of the current study indicated age-related decreases in attention orienting to caregiver versus stranger faces, indicating that infants were increasingly biased towards stranger faces, but no difference in attention holding to these faces with age. Although we observed an age-related increase in orienting to strangers, rather than attention holding, this result broadly converges with past work to suggest an overall shift towards increased attention to strangers over the first year. This effect of age was especially pronounced during dual target trials, in which caregiver and stranger faces appeared in direct competition, suggesting that infants’ ability to engage endogenous selective attention skills may contribute to this shifting attention bias over the first year. However, this age-related shift towards increased attention to strangers may also reflect changes related to the developing infant-caregiver attachment relationship. Attachment formation involves balancing behaviors that ensure proximity to caregivers with those that facilitate exploration and learning from novelty (Bowlby, 1969, 1988). Although one might expect that attachment formation may promote attention to caregivers, behaviors that reflect the formation of the infant-caregiver attachment (e.g., proximity-seeking, distress to separation, exploration) have been related to reduced attention holding to caregiver versus stranger faces (Kungl et al., 2017; Swingler et al., 2007, 2010). Thus, attachment formation may instead facilitate attention to stranger faces as infants increasingly explore novel stimuli while relying on the caregiver as a secure base (Kungl et al., 2017). The current results identifying increased attention to stranger faces over the first year are consistent with this idea, but additional research will be needed to more directly link the infant-caregiver relationship to changes in infants’ attention biases to caregivers vs. strangers over the first year. For example, infants’ attention to caregiver faces may be shaped by their experiences with motivationally salient caregiver behaviors (e.g., infant-directed speech, positive emotional facial expressions, affectionate touch, contingent responding; McFarland et al., 2020; Minagawa-Kawai et al., 2009; Nencheva et al., 2021; Simpson, Maylott, Lazo, et al., 2019). Although we did not have adequate statistical power to fully explore how parental touch influenced infants’ attention in the current study, results were similar when we limited analyses to the infants who had no direct contact with their parent during the task, suggesting that the results reflect infants’ cumulative experience with their caregiver rather than the immediate task context. Future work may also consider how broader social contexts may interact with the infant-caregiver relationship to shape developing attention biases. For example, the size or density of infants’ social networks may influence how often infants must allocate their attention towards their caregiver versus strangers in daily life. These social dynamics may have varied among infants in the current study because some infants were tested prior to the COVID-19 pandemic. Although we were unable to directly assess this possibility, future research can incorporate social network analyses (Burke et al., 2023) to better understand how these dynamics may influence infants’ attention biases to faces over the first year.

In addition to these age effects, we also found that individual differences in Surgency related to the strength of infants’ caregiver attention biases. Surgency is a measure of relatively stable individual differences in positive reactivity/affect and approach (Shiner et al., 2012) that is considered an indirect index of reward sensitivity (Kujawa et al., 2015; Vervoort et al., 2015; Wacker & Smillie, 2015). Infants with higher Surgency scores were both more likely to orient to their caregiver’s face and spent more time looking at their caregiver’s face compared to the stranger’s face. Although Surgency has previously been related to infants’ face processing more generally (Rennels et al., 2020), this is the first evidence specifically linking Surgency to infants’ attention to caregivers. Caregiver cues activate neural reward circuitry in both infancy and childhood (Abrams et al., 2016; Liu et al., 2019; Minagawa-Kawai et al., 2009; Stern et al., 2024), we therefore interpret this result as initial evidence that infants’ attention biases to caregivers may be shaped by their sensitivity to social reward value. However, because the measure of Surgency used in the current study indirectly assesses reward sensitivity, future research using other indirect and direct measures of reward processing (e.g., spontaneous or task-based eye blink rate; cortical reward network activity via fNIRS) will be needed to confirm this link between reward processing and infants’ attention to caregivers. Furthermore, the present study used caregiver faces to investigate reward-based attention mechanisms based on evidence that caregiver cues activate neural reward circuitry in both infancy and childhood (Abrams et al., 2016; Liu et al., 2019; Minagawa-Kawai et al., 2009; Stern et al., 2024). Additional work can investigate the role of reward in early attention development more broadly by examining infants’ attention orienting and attention holding biases to other classes of rewarding stimuli. However, this line of future work is currently limited by insufficient evidence regarding the range of stimuli that activate neural reward circuitry in infancy. Animal research has identified distinct neural circuitry underlying hedonic pleasure (i.e., “liking”) versus incentivizing reward (i.e., “wanting”; Berridge et al., 2009). Although infants appear to “like” specific stimuli by demonstrating visual preferences (e.g., complex patterns, Fantz, 1956), establishing that these stimuli are processed as reward at the neural level will facilitate efforts to understand the development of reward-based attention biases in infancy.

Finally, these links between Surgency and infants’ attention biases to caregiver faces also varied based on the task demands. When only one face appeared in the array, infants with higher Surgency scores detected both face types at similar frequencies but looked longer at caregiver faces. Thus, when there was relatively less competition between one face and the surrounding distractors, infants with higher Surgency scores readily detected all faces but preferred looking at their caregiver’s face. In contrast, when the caregiver and stranger faces appeared in direct competition with each other during dual target trials, infants with higher Surgency scores preferentially oriented to the caregiver face but did not look longer at the caregiver face. This result suggests that infants who were more sensitive to reward were more biased to quickly detect their caregiver when it was more difficult to resolve competition among the faces in the array. Overall, these results suggest that attention holding biases may be more robust when surrounding competition is relatively low, whereas attention orienting biases may emerge when competition is increased. Older children similarly showed stronger orienting biases to caregivers when selective attention demands of the task increased due to the presence of more distractors in a visual search task (Hunter & Markant, 2023b). The current findings converge with this prior work to demonstrate that task context can impact attention biases to caregiver faces in both infancy and childhood.

Overall, the current study extends past work by identifying similar patterns in attention holding and orienting biases to caregiver faces during the first year. Although we did not observe these attention biases at the group level, we found that they varied based on age, the extent to which the caregiver and stranger faces appeared in direct competition within the arrays, and individual differences in infants’ reward sensitivity. These results provide initial evidence suggesting that both endogenous attention control and reward-based mechanisms may contribute to infants’ attention biases to caregivers. This conclusion is consistent with adult literature indicating that selective attention can be mediated by multiple factors beyond perceptual salience and task-relevance, including motivational salience (Awh et al., 2012; Kim et al., 2021; Luck et al., 2021). Motivational salience is broadly defined as the learned value of a stimulus that can reflect both rewarding as well as negative associations (Kim et al., 2021). Past work has shown that infants are biased to orient towards stimuli that are motivationally salient due to threat cues (e.g., predators and angry faces; LoBue et al., 2017) and the current study extends this work by offering initial evidence that infants’ attention biases may also be linked to reward-based mechanisms. Although developmental models have classically emphasized the role of perceptual salience and endogenous control in early selective attention (e.g., Colombo, 2001; Reynolds & Romano, 2016), these findings suggest that these models can be expanded to consider additional factors, including motivational salience, that may drive attention biases during infancy.

In conclusion, the current study replicates past work demonstrating that infants are biased to detect and maintain attention on faces that appear in multi-object arrays. These attention biases to faces became more robust with age and depended on the extent to which the faces were in direct competition, suggesting that developing attention control skills may contribute to infants’ attention biases over the first year. Furthermore, unlike past research that has used different task designs to examine attention holding and attention orienting to faces, we found that infants showed largely parallel attention holding and orienting biases to caregiver faces when tested in the same task context. Although infants did not show attention biases based on face identity at the group level, participant age, task context, and individual differences in Surgency related to both attention orienting and attention holding biases to caregiver versus stranger faces. These findings suggest that both endogenous attention control and reward-based mechanisms shape infants’ ability to gather and process meaningful social information from their complex visual worlds.