Social Dual Task Developmental Dataset and Analyses

Project abstract: Multitasking is part of the everyday lives of both adolescents and adults. We often multitask during social interactions by simultaneously keeping track of other, non-social information. Here, we examined how keeping track of non-social information impacts the ability to navigate social interactions in adolescents and adults. Participants aged 11-17 and 22-30 years old were instructed to carry out two tasks, one social and one non-social, within each trial. The social task involved referential communication, requiring participants to use social cues to guide their decisions, which sometimes required taking a different perspective. The non-social task manipulated cognitive load by requiring participants to remember non-social information in the form of one two-digit number (low load) or three two-digit numbers (high load) presented before each social task stimulus. Participants showed performance deficits when under high cognitive load and when the social task involved taking a different perspective, and individual differences in both trait perspective taking and working memory capacity predicted performance. Overall, adolescents were less adept at multitasking than adults when under high cognitive load. These results suggest that multitasking during social interactions incurs performance deficits, and that adolescents, compared to adults, are more sensitive to the effects of cognitive load while multitasking. Data: These files include our dataset, as well as the scripts used to analyze the data. You will need to download R (http://www.r-project.org/) to use these files. Data are from 37 adolescents (11-17 years) and 30 adults (22-30 years). Ages are not provided to preserve anonymity. Participants completed an adapted version of the “Director Task” (Dumontheil, Hillebrandt, Apperly, & Blakemore, 2012) with an embedded working memory (WM) Task component. Afterwards, participants completed a verbal reverse digit-span task as a measure of WM capacity and the Interpersonal Reactivity Index questionnaire to assess individual differences in trait perspective taking (Davis, 1980). We have also included information about the errors made by individuals, as well as the Director Task stimuli. Data Analysis: We used mixed effects modelling to determine what factors best predicted multitasking performance. Accuracy was determined on a trial-by-trial basis, where a trial was considered accurate only if participants correctly performed both the Director Task and embedded WM Task. As our main interest was performance during social interactions, and not recall of non-social information, we analysed Director Task RT (correct trials only). We used the lme4 package in R (Bates, Maechler, & Bolker, 2013) to perform a linear mixed effects analysis on the relationship between our factors of interest and multitasking performance (accuracy and RT). Our factors of interest included three from the task: cognitive load (low vs. high), condition (DA vs. DP), perspective (same vs. different), and two individual traits: WM capacity and trait perspective taking (TPT). As we hypothesised an interaction between WM capacity and TPT would relate to our task, we included a combined measure of these two individual traits by calculating and summing the ratios of TPT and WM capacity (Combined Traits). We determined which factors best predicted performance for our measures of interest by testing global models including our factors of interest as fixed effects. Each model included a random intercept for each participant. Because of computational limitations, we performed a two-step procedure that involved five global models. First, all possible combinations of the variables within each of the five global models were tested using an automated model selection procedure (MuMIn1.9.0; Barton, 2013). Models were ranked using Second-order Akaike Information Criterion (AICc; Burnham & Anderson, 2002). Second, the best fitting model for each of the five global models were compared and ranked using AIC and likelihood ratio tests. The object_positions.xls file shows the object positions (slot number) for each of the 48 Director Task stimuli. For example, in a trial using picture number 1, the participant would need to select the object in slot number 3 during 3-object trials in which the Director shared the same perspective as the participant. However, the participant would need to select the object in slot number 11 during 3-object trials in which the Director had a different perspective than the participant (i.e., the Director is behind the shelf array). Finally, on 1-object trials, the participant would need to select the object in show number 6. You can use this file to do error analyses with the provided dataset to see what kind of errors participants made in the Director Task.

项目摘要：多任务处理是青少年与成年人日常生活的组成部分。我们常常在社交互动过程中同时追踪非社交类信息，以此实现多任务并行。本研究旨在探究追踪非社交信息会如何影响青少年与成年人的社交互动能力。实验招募了11至17岁的青少年以及22至30岁的成年人作为被试，每名被试需在每轮试次中同时完成两项任务：一项社交任务与一项非社交任务。社交任务为指涉性沟通任务，要求被试借助社交线索指导决策，部分情境下需要采择他人视角。非社交任务通过记忆非社交信息来操纵认知负荷：在每项社交任务刺激呈现前，向被试展示1个两位数数字（低认知负荷）或3个两位数数字（高认知负荷），要求被试记住这些数字。结果发现，当处于高认知负荷条件，且社交任务需要采择他人视角时，被试的任务表现会出现缺损；个体的特质采择能力与工作记忆（working memory, WM）容量均能预测任务表现。总体而言，在高认知负荷条件下，青少年的多任务处理能力弱于成年人。上述结果表明，社交互动过程中的多任务处理会导致任务表现缺损，且相较于成年人，青少年在多任务处理时对认知负荷的影响更为敏感。

数据集说明：本数据集包含实验原始数据与数据分析所用的脚本文件，需下载R语言（http://www.r-project.org/）方可运行相关文件。本实验共招募37名青少年（11-17岁）与30名成年人（22-30岁），未公开被试具体年龄以保护匿名性。被试需完成改编版的“指导者任务（Director Task）”（Dumontheil, Hillebrandt, Apperly, & Blakemore, 2012），该任务内嵌了工作记忆（working memory, WM）任务组件。实验结束后，被试需完成言语反向数字广度任务以测量工作记忆容量，并完成人际反应指数量表以评估特质采择能力的个体差异（Davis, 1980）。本数据集同时包含被试的错误作答记录与指导者任务的刺激材料信息。

数据分析：本研究采用混合效应建模（mixed effects modelling）以确定哪些因素能够最优预测多任务处理表现。试次准确率的计算以单试次为单位，仅当被试同时正确完成指导者任务与内嵌工作记忆任务时，该试次才被记为准确。由于本研究的核心关注点为社交互动中的任务表现，而非非社交信息的回忆，因此仅针对正确试次的指导者任务反应时进行分析。我们使用R语言的lme4包（Bates, Maechler, & Bolker, 2013），针对我们关注的因素与多任务表现（准确率与反应时）之间的关系开展线性混合效应分析。我们关注的因素包括三类任务相关变量：认知负荷（低vs.高）、任务条件（DA vs. DP）、视角采择类型（一致vs.不一致），以及两类个体特质变量：工作记忆容量与特质采择能力（trait perspective taking, TPT）。基于我们提出的工作记忆容量与特质采择能力的交互作用会与任务表现相关的假设，我们通过计算特质采择能力与工作记忆容量的比值并求和，生成了这两类个体特质的综合指标（Combined Traits）。我们构建了包含所有关注因素作为固定效应的全局模型，且每个模型均为每名被试设置了随机截距项。受限于计算资源，我们采用了包含5个全局模型的两步分析流程：第一步，借助自动化模型选择流程（MuMIn 1.9.0; Barton, 2013），对5个全局模型内的所有变量组合进行测试，并以二阶赤池信息准则（Second-order Akaike Information Criterion, AICc; Burnham & Anderson, 2002）对模型进行排序；第二步，对5个全局模型各自拟合效果最优的模型进行比较，并基于AIC与似然比检验完成最终的模型排序。
object_positions.xls文件记录了48个指导者任务刺激对应的物体位置（槽位编号）。例如，在使用图片1的试次中，若指导者与被试视角一致的3物体试次中，被试需选择槽位3中的物体；而当指导者与被试视角不一致（即指导者位于货架阵列后方）的3物体试次中，被试需选择槽位11中的物体。在单物体试次中，被试则需选择编号为6的展示位上的物体。您可借助该文件结合本数据集开展错误分析，以探究被试在指导者任务中出现的错误类型。