Holistic Processing of Faces: Bridging Paradigms
Jennifer J.
Richler, Isabel Gauthier, Michael J. Wenger & Thomas J. Palmeri
Several sequential matching composite face paradigms are used to assess “holistic processing” of faces. In the Selective Attention paradigm, participants decide whether one face part (e.g., top) is the same as a previously seen face part. This judgment is affected by whether the irrelevant part is also the same or different. This failure of selective attention implies holistic processing. However, we show that this task alone cannot distinguish between perceptual or decisional sources of holism. This distinction can be addressed by the Complete Identification paradigm, where both face parts are judged same or different, allowing analyses based on General Recognition Theory (GRT; Ashby & Townsend, 1986). Previous research has found evidence for decisional, but not perceptual, holism (Wenger & Ingvalson, 2002). We use a different paradigm, Sequential Responses, to relate these two paradigms. This paradigm produced the same results as the Selective Attention and Complete Identification paradigms. Moreover, disruptions of holistic processing due to misalignment corresponded with significant changes in the decisional components of GRT. This suggests a significant decisional component of holistic face processing in the composite face task.
Supplementary Information
In this paper we used Multidimensional Signal Detection Analysis (MSDA) to analyze data with respect to General Recognition Theory. Because these analyses are quite complex, in the paper we only focused on the important results. For each experiment we conducted MSDA at the group level to draw qualitative conclusions regarding the General Recognition Theory constructs, Perceptual Separability (PS), Decisional Separability (DS) and Perceptual Independence (PI). We also conducted MSDA on individual participant data to extract marginal d’ and marginal c values, which index PS and DS respectively, which were then submitted to statistical analysis to provide a quantitative measure of perceptual vs. decisional holism.
Here we provide all of the confusion matrices which were analyzed using MSDA2 software (Kadlec, 1995, 1999) as well as the complete output files from this program.
All the confusion matrices have the following form:
Rows are stimuli (top-same/bot-same, top-diff/bot-same, top-same/bot-diff, top-diff/bot-diff)
Columns are responses (top-same/bot-same, top-diff/bot-same, top-same/bot-diff, top-diff/bot-diff)
Experiment 1
In Experiment 1a we compared performance on the Selective Attention Task with performance in the Sequential Responses Task. We also analyzed data from the Sequential Responses Task using MSDA. Links to the aggregate confusion matrix and MSDA output are provided below. For the individual analyses, links are provided to the folders which contain all the confusion matrices and output files for each participant. These are labeled according to subject number.
The aggregate confusion matrix is available here:
http://www.psy.vanderbilt.edu/faculty/palmeri/holistic2007/Experiment_1a/Aggregate_Confusion_Matrix.txt
The aggregate MSDA output is available here:
http://www.psy.vanderbilt.edu/faculty/palmeri/holistic2007/Experiment_1a/Aggregate_MSDA_output.txt
Individual subject confusion matrices are available here:
Individual subject MSDA output files are available here:
In Experiment 1b we compared performance on the Complete Identification Task with performance in the Sequential Responses Task. Data from both tasks were analyzed using MSDA. Links to the aggregate confusion matrices and MSDA output for each task are provided below. For the individual analyses, links are provided to the folders which contain all the confusion matrices and output files for each participant. These are labeled according to subject number.
Complete Identification Task
The aggregate confusion matrix is available here:
The aggregate MSDA output is available here:
Individual subject confusion matrices are available here:
Individual subject MSDA output files are available here:
Sequential Responses Task
The aggregate confusion matrix is available here:
The aggregate MSDA output is available here:
Individual subject confusion matrices are available here:
Individual subject MSDA output files are available here:
Experiment 2
In Experiment 2 we investigated the effects of misaligning the test stimulus on performance in the Sequential Responses Task and the Complete Identification Task. For each task, MSDA was conducted at each level of alignment. Links to folders containing the aggregate confusion matrices and aggregate MSDA outputs for each alignment condition are provided below. These files are named according to alignment condition (A – aligned, M – misaligned, VM – very misaligned). The individual subject files are named according to subject number and alignment condition.
Sequential Responses Task
The aggregate confusion matrices for each alignment condition are available here:
The aggregate MSDA outputs for each alignment condition are available here:
Individual subject confusion matrices for each alignment condition are available here:
Individual subject MSDA output files for each alignment condition are available here:
Complete Identification Task
The aggregate confusion matrices for each alignment condition are available here:
The aggregate MSDA outputs for each alignment condition are available here:
Individual subject confusion matrices for each alignment condition are available here:
Individual subject MSDA output files for each alignment condition are available here: