On the Automaticity of Familiarity in Short-term Recognition: A Test of the Dual-Process Assumption with the PRP Paradigm.

Klaus Oberauer

doi:10.5334/joc.21

Abstract

Dual-process models of recognition often assume that one retrieval process, generating a familiarity signal, is automatic, whereas the other, recollection, is controlled. Four experiments are presented to test for automaticity of familiarity in a short-term recognition task. The experiments use the Psychological Refractory Period (PRP) paradigm to assess whether familiarity requires central processing capacity. Task 1 was an oral tone-classification task. Task 2 was a local-recognition task, in which participants decided whether a probe matched a particular item in the memory set, identified by its screen location. Intrusion probes, matching an item of the memory set in a different location, were slower and more difficult to reject than new probes. The size of this intrusion cost reflects the influence of familiarity on recognition. In all four experiments the size of the intrusion cost was additive with the stimulus-onset-asynchrony (SOA) of Task 1 and Task 2, demonstrating that extraction of familiarity requires central capacity. In addition, Experiment 2 showed additive effects of memory set size and serial position with SOA, confirming that recollection, too, requires central capacity. Experiments 3A and 3B compared a condition including new probes to one including only positive and intrusion probes; in the latter condition the familiarity signal was completely uninformative. Participants showed some ability to reduce the influence of familiarity when it was completely uninformative, but only when they were explicitly told to do so (Experiment 3B). To conclude, by one criterion familiarity is a controlled process: It demands central processing capacity. It might also be controlled by another criterion: People can intentionally reduce the influence of familiarity on recognition decision, but they fail to do so spontaneously even when it would be advantageous. All raw data are available on the Open Science Framework: osf.io/7pr72.

Highlights

Dual-process models of recognition assume that recognition decisions draw on information from two processes, referred to as familiarity and recollection (Diana, Reder, Arndt, & Park, 2006; Jacoby, 1991; Mandler, 1980; Oberauer, 2008; Yonelinas, 2002)
The additive effects of SOA and probe type support the assumption that both familiarity and recollection rely on central processing capacity, so that they have to wait until the central processing stage of Task 1 has finished
Larger memory sets increase response times, and this effect is in large part due to the longer duration of retrieval (Shepherdson, Oberauer, & Souza, 2017)

Summary

Introduction

Dual-process models of recognition assume that recognition decisions draw on information from two processes, referred to as familiarity and recollection (Diana, Reder, Arndt, & Park, 2006; Jacoby, 1991; Mandler, 1980; Oberauer, 2008; Yonelinas, 2002). Recollection refers to memory of the recent occurrence of a stimulus in a particular context It is described as remembering the episode in which the stimulus has previously occurred, including some of its context. Dual-process theories of recognition vary in the degree to which they assume qualitative differences between familiarity and recollection as retrieval processes. Often referred to as dual-process theories, these theories could be interpreted as assuming a single retrieval process that extracts evidence from two kinds of information in memory: The strength of the probe’s representation in memory – yielding the familiarity signal – and the links of the probe’s representation to elements of the relevant study context – yielding the recollection signal. Two key attributes are: (1) Automatic processes proceed without being intended, and as such, can operate against an intended mental process or action – for instance, word reading in the Stroop task (MacLeod, 1991). (2) Automatic processes are usually assumed to operate without capacity limits, such that they can run in parallel with other processes without suffering performance costs

Objectives

Methods

Results

Conclusion