Loading...
Please wait, while we are loading the content...
Similar Documents
Heft 3 Process Dissociation Measurement Models : Good versus Better
| Content Provider | Semantic Scholar |
|---|---|
| Author | Erdfelder, Edgar Buchner, Axel |
| Copyright Year | 2004 |
| Abstract | .2 Zusammenfassung.3 Process Dissociation Measurement Models: Good versus Better.4 Process Dissociation Measurement Models.5 The Extended Measurement Model .6 The Dual-Process Signal-Detection Model.15 Alternative Methods to Correct for Response Bias .20 Evaluation of Process Dissociation Measurement Models.21 Confidence Ratings.21 Evaluation of the DPSDM. 22 Evaluation of the EMM. 24 Experimental Manipulations of Response Bias.32 Evaluation of the DPSDM. 33 Evaluation of the EMM. 35 Improvement of Process Dissociation Measurement Models.37 Generalizations of the Dual-Process Signal-Detection Model .38 The distribution-free DPSDM. . 38 The correlated-processes signal-detection model. 43 Generalizations of the Extended Measurement Model.46 The multiple-groups two-high threshold EMM. 52 The single-group two-high threshold EMM. 53 Discussion .56 References.60 Appendix .63 Proof of the Identifiability of the DPSDM.63 Authors' Notes.64 Page 2 Process Dissociation Measurement Models Abstract Several methods to account for response bias in the process dissociation procedure have recently been proposed. A. P. Yonelinas and L. L. Jacoby (1995b) favor a dual-process signal-detection model (DPSDM) and claim that threshold-based models such as the extended measurement model (EMM) suggested by A. Buchner, E. Erdfelder, and B. Vaterrodt-Plünnecke (1995 should be rejected because threshold models are inconsistent with nonlinear receiver operating characteristics (ROCs) as obtained from confidence ratings. Their claim is shown to be incorrect. An EMM variant for confidence ratings is developed which accounts perfectly for nonlinear ROCs. It is demonstrated that, in contrast, the DPSDM cannot fit the ROC data of at least two of the three experiments reported by A. P. Yonelinas (1994). Further, it is argued that experimental manipulations of response biases result in much more thorough tests of process dissociation measurement models than confidence ratings. We close by suggesting a new two-high threshold extended measurement model which fits the Buchner et al. data better than both the EMM and the DPSDM.Several methods to account for response bias in the process dissociation procedure have recently been proposed. A. P. Yonelinas and L. L. Jacoby (1995b) favor a dual-process signal-detection model (DPSDM) and claim that threshold-based models such as the extended measurement model (EMM) suggested by A. Buchner, E. Erdfelder, and B. Vaterrodt-Plünnecke (1995 should be rejected because threshold models are inconsistent with nonlinear receiver operating characteristics (ROCs) as obtained from confidence ratings. Their claim is shown to be incorrect. An EMM variant for confidence ratings is developed which accounts perfectly for nonlinear ROCs. It is demonstrated that, in contrast, the DPSDM cannot fit the ROC data of at least two of the three experiments reported by A. P. Yonelinas (1994). Further, it is argued that experimental manipulations of response biases result in much more thorough tests of process dissociation measurement models than confidence ratings. We close by suggesting a new two-high threshold extended measurement model which fits the Buchner et al. data better than both the EMM and the DPSDM. Process Dissociation Measurement Models Page 3 Zusammenfassung In der neueren Literatur zur Prozeß-Dissoziations-Prozedur werden unterschiedliche Vorschläge zur Berücksichtigung von Antworttendenzen gemacht. A. P. Yonelinas und L. L. Jacoby (1995b) favorisieren ein Dual-Process-Signal-DetectionModell (DPSDM) und kritisieren High-Threshold-Modelle wie z.B. das ExtendedMeasurement-Modell (EMM) von A. Buchner, E. Erdfelder und B. VaterrodtPlünnecke (1995), weil derartige Modelle mit nichtlinearen Receiver Operating Characteristic Curves (ROCs) wie sie für Konfidenzratings beobachtet wurden unvereinbar seien. Es wird gezeigt, daß diese Behauptung falsch ist. Eine Variante des EMM für Konfidenzratings, die nichtlineare ROCs fehlerfrei erklären kann, wird vorgestellt. Es wird gleichzeitig nachgewiesen, daß das DPSDM mit den Daten von mindesten zweien der drei Experimente von A. P. Yonelinas (1994) nicht vereinbar ist. Weiterhin wird die These vertreten, daß experimentelle Manipulationen von Antworttendenzen strengere Tests von Prozeß-Dissoziations-Meßmodellen erlauben als Konfidenzratings. Im abschließenden Teil der vorliegenden Arbeit wird ein TwoHigh-Threshold EMM vorgestellt, das mit den Buchner et al.-Daten besser als das EMM und das DPSDM vereinbar ist. Page 4 Process Dissociation Measurement Models Process Dissociation Measurement Models: Good versus Better The process dissociation measurement model originally proposed by Jacoby (1991) aims at measuring the contributions of controlled (“conscious”) and automatic (“unconscious”) cognitive processes to task performance without accounting for possible effects of response biases. As a consequence, measures of controlled and automatic processes may be contaminated by response biases. Although this problem is not new and several correction methods have been proposed by various authors (cf. Jacoby, Toth & Yonelinas, 1993; Komatsu, Graf & Uttl, 1995; Roediger & McDermott, 1994) , it was not until recently that the problem was addressed by new measurement models that can account for simultaneous effects of controlled processes, automatic processes, and response biases on task performance. Buchner, Erdfelder, and Vaterrodt-Plünnecke (1995) have developed an extended measurement model (EMM) which is based on threshold theory (cf. Krantz, 1969; Luce, 1963a) whereas Yonelinas, Regehr, and Jacoby (in press) suggested a dualprocess signal-detection model (DPSDM) using the framework of signal-detection theory (cf. Green & Swets, 1966) . In their interesting and stimulating comment on Buchner et al. (1995), Yonelinas and Jacoby (1995b) argued that their DPSDM was superior to the EMM because the former, but not the latter, could account for nonlinear receiver-operating characteristics (ROCs) as obtained from confidence ratings (cf. Yonelinas, 1994) . They also argued that the experimental data used to validate the EMM by Buchner et al. (1995) were inappropriate, because the experimental manipulations might have affected response bias and memory processes simultaneously. This article aims at refuting both claims. We begin by describing the EMM as applied to the process dissociation procedure using yes-no recognition judgments, and we discuss its relation to the measurement model variant suggested by Jacoby (1991). Next, we describe the DPSDM and show that it has some methodological disadvantages compared to the EMM. Nevertheless, both the EMM and the DPSDM seem to be superior to alternative methods of response bias correction that have been suggested in the literature. Process Dissociation Measurement Models Page 5 We will then present an appropriate extension of the EMM to confidence rating scales in order to show that nonlinear ROCs do not falsify the threshold concept underlying the EMM. Moreover, we maintain that the experimental data provided by Buchner et al. (1995) are appropriate validation hurdles that must be cleared by an adequate process dissociation measurement model. We agree with Yonelinas and Jacoby (1995b) that both the EMM and DPSDM fit these data quite well but not perfectly. We differ from Yonelinas and Jacoby (1995b) in that we attribute the less-than-perfect fit to limitations of both models. We close by discussing generalizations of the EMM and the DPSDM that might help to overcome these limitations. A generalization of the DPSDM that does not require the normal distribution assumption is developed, and we show that this generalized version and, hence, the DPSDM as a submodel of it does not fit the data of Yonelinas' (1994) Experiments 1 and 2 at least. Also, a correlated processes signal detection model is developed that does not require the assumption of stochastic independence of controlled and automatic processes presumed by the DPSDM. Finally, we suggest two variants of a new two-high threshold extended measurement model and show that each of them fits the Buchner et al. (1995) data even better than both the EMM and the DPSDM. Process Dissociation |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | https://psydok.psycharchives.de/jspui/bitstream/20.500.11780/163/1/BoBe1995_3.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
