Department of Psychology, Neuroscience and Behaviour, McMaster University.
Neuropsychology. 2010 Sep;24(5):637-51. doi: 10.1037/a0019803.
The Attention Network Test (ANT) is a tool used to assess the efficiency of the 3 attention networks-alerting, orienting, and executive control. The ANT has become popular in the neuropsychological literature since its first description in 2002, with some form of the task currently appearing in no less than 65 original research papers. Although several general reviews of the ANT exist, none provide an analysis of its psychometric properties.
Data from 15 unique studies were collected, resulting in a large sample (N = 1,129) of healthy individuals. Split-half reliability, variance structure, distribution shape, and independence of measurement of the 3 attention network scores were analyzed, considering both reaction time and accuracy as dependent variables.
Split-half reliabilities of reaction time based attention network scores were low for alerting (rweighted = .20, CI 95%weighted [.14, .27], Spearman-Brown r = .38) and orienting (rweighted = .32, CI 95%weighted [.26, .38], Spearman-Brown r = .55), and moderate high for executive control (rweighted = .65, CI 95%weighted [.61, .71], Spearman-Brown r = .81). Analysis of the variance structure of the ANT indicated that power to find significant effects was variable across networks and dependent on the statistical analysis being used. Both analysis of variance (significant interaction observed in 100% of 15 studies) and correlational analyses (multiple-significant inter-network correlations observed) suggest that the networks measured by the ANT are not independent.
In the collection, analysis and interpretation of any test data, psychometric properties, such as those reported here for the ANT, must be carefully considered.
注意网络测试（ANT）是一种用于评估 3 种注意网络（警觉、定向和执行控制）效率的工具。自 2002 年首次描述以来，ANT 在神经心理学文献中变得很流行，目前该任务的某种形式至少出现在 65 篇原始研究论文中。尽管有几个关于 ANT 的一般综述，但没有一个分析其心理测量特性。
收集了 15 项独特研究的数据，得到了一个大型样本（N=1129），其中包括健康个体。分析了 3 种注意网络分数的半分可靠性、方差结构、分布形状和测量的独立性，同时考虑了反应时间和准确性作为因变量。
基于反应时间的警觉（rweighted=.20，95%置信区间加权[rweighted] [.14，.27]，Spearman-Brown r=.38）和定向（rweighted=.32，95%置信区间加权[rweighted] [.26，.38]，Spearman-Brown r=.55）的注意网络分数的半分可靠性较低，而执行控制（rweighted=.65，95%置信区间加权[rweighted] [.61，.71]，Spearman-Brown r=.81）的半分可靠性中等较高。对 ANT 的方差结构分析表明，网络之间的效应显著程度的检测能力因网络而异，且取决于所使用的统计分析。方差分析（在 15 项研究中的 100%观察到显著的交互作用）和相关分析（观察到多个网络间的显著相关性）都表明，ANT 测量的网络不是独立的。
在任何测试数据的收集、分析和解释中，必须仔细考虑心理测量特性，如这里报告的 ANT 的特性。
