Geriatric Research Education and Clinical Center (GRECC), VA Boston Healthcare System, Boston, MA 02130, USA.
Cognition. 2013 Jan;126(1):87-100. doi: 10.1016/j.cognition.2012.09.004. Epub 2012 Oct 16.
Although holistic processing is thought to underlie normal face recognition ability, widely discrepant reports have recently emerged about this link in an individual differences context. Progress in this domain may have been impeded by the widespread use of subtraction scores, which lack validity due to their contamination with control condition variance. Regressing, rather than subtracting, a control condition from a condition of interest corrects this validity problem by statistically removing all control condition variance, thereby producing a specific measure that is uncorrelated with the control measure. Using 43 participants, we measured the relationships amongst the Cambridge Face Memory Test (CFMT) and two holistic processing measures, the composite task (CT) and the part-whole task (PW). For the holistic processing measures (CT and PW), we contrasted the results for regressing vs. subtracting the control conditions (parts for PW; misaligned congruency effect for CT) from the conditions of interest (wholes for PW; aligned congruency effect for CT). The regression-based holistic processing measures correlated with each other and with CFMT, supporting the idea of a unitary holistic processing mechanism that is involved in skilled face recognition. Subtraction scores yielded weaker correlations, especially for the PW. Together, the regression-based holistic processing measures predicted more than twice the amount of variance in CFMT (R(2)=.21) than their respective subtraction measures (R(2)=.10). We conclude that holistic processing is robustly linked to skilled face recognition. In addition to confirming this theoretically significant link, these results provide a case in point for the inappropriateness of subtraction scores when requiring a specific individual differences measure that removes the variance of a control task.
尽管整体加工被认为是正常面孔识别能力的基础，但最近在个体差异背景下，关于这一联系的报告却大相径庭。这一领域的进展可能受到减法分数广泛应用的阻碍，由于其与对照条件方差的混合，这些分数缺乏有效性。通过从感兴趣的条件中减去控制条件，而不是减去控制条件，可以通过统计上消除所有控制条件方差来纠正这个有效性问题，从而产生一个与控制测量无关的特定测量值。我们使用 43 名参与者，测量了剑桥面孔记忆测试（CFMT）与两种整体加工测量（复合任务（CT）和部分-整体任务（PW））之间的关系。对于整体加工测量（CT 和 PW），我们对比了从感兴趣的条件（PW 的整体；CT 的对齐一致效应）减去控制条件（PW 的部分；CT 的非对齐一致效应）的结果，与从控制条件减去控制条件（PW 的部分；CT 的非对齐一致效应）的结果。基于回归的整体加工测量值彼此相关，与 CFMT 相关，支持了一种整体加工机制在熟练面孔识别中起作用的观点。减法分数的相关性较弱，尤其是对于 PW。基于回归的整体加工测量值与各自的减法测量值相比，对 CFMT 的预测方差高出两倍多（R²=.21）。我们得出的结论是，整体加工与熟练的面孔识别密切相关。除了证实这一具有理论意义的联系外，这些结果还提供了一个恰当的案例，说明了在需要一个特定的个体差异测量值来消除控制任务的方差时，减法分数的不适当性。
