Infant looking behaviors measured during visual assessment paradigms may be more reliable predictors of long-term cognitive outcomes than standard measures such as the Bayley Scales of Infant Development typically used in environmental epidemiology. Infrared eye tracking technology offers an innovative approach to automate collection and processing of looking behavior data, making it possible to efficiently assess large numbers of infants. The goals of this study were to characterize infant looking behavior measures including side preference, fixation duration, and novelty preference using eye tracking and an automated version of an established visual recognition memory paradigm that includes both human faces and geometric figures as stimuli. An ancillary goal was to assess the feasibility of obtaining a precise measure of looking to the eye region of faces from the eye-tracking data. In this study, 309 7.5-month-old infants from a prospective birth cohort were assessed using a visual recognition memory (VRM) paradigm. Infrared eye tracking was used to record looking time as infants were shown nine blocks of trials with a pair of identical faces or shapes followed by two trials in which the familiar stimulus was paired with a novel one. Infants were assessed in one of four conditions: in conditions A and B, stimulus set 1 were the familiar stimuli and set 2 were novel; in conditions C and D, set 2 were familiar and set 1 novel. The novel stimuli were presented on the right first in conditions A and C and on the left first in conditions B and D. We observed a significant right side preference, which has not been reported before (57% of looking time spent looking at right side stimulus, p-value < 0.0001). Infants showed a preference for the novel stimuli similar to that published in prior studies (57–60% of looking time spent looking at the novel stimulus, p-value < 0.0001), as well as average fixation durations similar to previous studies. Infants also showed a strong preference for the eyes versus the rest of the face (p-value < 0.0001). Novelty preference was significantly higher when set 2 stimuli were novel (p-value < 0.0001), suggesting a preference among infants for set 2 stimuli compared to set 1 stimuli. The pattern of novelty preference across trials was significantly different between infants who saw the novel stimuli on the left first and those who saw them on the right first (p-value < 0.0001) but the overall mean novelty preference was not significantly different between these groups. There were also significant differences in average fixation duration and eyes preference measures across stimuli (p-values < 0.05). These findings show that VRM assessment can be automated for use in large-scale epidemiological studies using infrared eye tracking with looking behavior measure results similar to those obtained with standard non-automated methods, and that side and stimulus preferences are important modifiers of looking behavior that are critical to consider in this type of assessment.

与环境流行病学中通常使用的贝利婴儿发育量表等标准措施相比，在视觉评估范式中测量的婴儿外观行为可能是长期认知结果的更可靠预测指标。红外眼动追踪技术提供了一种创新方法来自动收集和处理外观行为数据，从而可以有效地评估大量婴儿。本研究的目标是使用眼动追踪和已建立的视觉识别记忆范式的自动版本（包括人脸和几何图形作为刺激）来表征婴儿的外观行为测量，包括侧向偏好、注视持续时间和新奇偏好。一个辅助目标是评估从眼动追踪数据获得精确测量面部眼部区域的可行性。在这项研究中，使用视觉识别记忆 (VRM) 范式对来自前瞻性出生队列的 309 名 7.5 个月大的婴儿进行了评估。红外线眼动追踪被用来记录注视时间，因为婴儿被展示了九个具有一对相同面孔或形状的试验，然后是两个试验，其中熟悉的刺激与新的刺激配对。婴儿在以下四种条件之一进行评估：在条件 A 和 B 中，刺激组 1 是熟悉的刺激，组 2 是新的；在条件 C 和 D 中，第 2 组是熟悉的，第 1 组是新颖的。新刺激在条件 A 和 C 中首先出现在右侧，在条件 B 和 D 中首先出现在左侧。我们观察到了显着的右侧偏好，这在之前从未报道过（57% 的观看时间花在观看右侧刺激上，p 值​​ < 0.0001）。婴儿对新刺激的偏好与先前研究中发表的相似（57-60% 的观看时间花在观看新刺激上，p 值​​ < 0.0001），以及与之前研究相似的平均注视持续时间。与面部其他部分相比，婴儿还表现出对眼睛的强烈偏好（p 值 < 0.0001）。当第 2 组刺激是新颖的（p 值 < 0.0001）时，新奇偏好显着更高，这表明与第 1 组刺激相比，婴儿更喜欢第 2 组刺激。试验中的新奇偏好模式在先看到左侧新刺激的婴儿和先看到右侧的新刺激的婴儿之间有显着差异（p 值 < 0.0001），但这些组之间的总体平均新奇偏好没有显着差异. 跨刺激的平均注视持续时间和眼睛偏好测量也存在显着差异（p 值 < 0.05）。这些发现表明，VRM 评估可以自动化用于大规模流行病学研究，使用红外眼动追踪，其观看行为测量结果类似于使用标准非自动化方法获得的结果，并且侧面和刺激偏好是观看行为的重要修饰符在此类评估中至关重要。0001），但总体平均新颖性偏好在这些组之间没有显着差异。跨刺激的平均注视持续时间和眼睛偏好测量也存在显着差异（p 值 < 0.05）。这些发现表明，VRM 评估可以自动化用于大规模流行病学研究，使用红外眼动追踪，其观看行为测量结果类似于使用标准非自动化方法获得的结果，并且侧面和刺激偏好是观看行为的重要修饰符在此类评估中至关重要。0001），但总体平均新颖性偏好在这些组之间没有显着差异。跨刺激的平均注视持续时间和眼睛偏好测量也存在显着差异（p 值 < 0.05）。这些发现表明，VRM 评估可以自动化用于大规模流行病学研究，使用红外眼动追踪，其观看行为测量结果类似于使用标准非自动化方法获得的结果，并且侧面和刺激偏好是观看行为的重要修饰符在此类评估中至关重要。