UC San Diego
A signal-detection-based investigation into the nature of recognition memory
- Author(s): Mickes, Laura Beth
- et al.
Recognition memory is the ability to consciously appreciate that an item or event was previously presented or experienced. Signal detection theory has long provided one influential interpretation of recognition memory, and numerous investigations conducted over the last 50 years have sought to clarify the particulars of this account. Analyzing receiver operating characteristic (ROC) data can distinguish between two versions of signal detection theory, specifically, the equal and unequal variance models. The equal variance signal detection model is intuitively appealing, but the unequal variance signal detection model usually provides a better fit of the ROC data. Chapter 1 describes two experiments that provide a novel test of the unequal variance assumption. This new method of analysis required subjects to directly rate their memory strength on a fine-grained scale, and then the mean and standard deviations of the target and lure ratings were directly computed. Results from the new method support the unequal variance signal detection model. Though the unequal variance signal detection theory of recognition memory provides a useful way to conceptualize recognition, there is another long-standing theory of recognition known as dual process theory that seems to contradict it. This theory holds that two processes (familiarity and recollection) contribute to recognition decisions. A critical point of contention between standard dual process models and signal detection theory concerns the nature of the recollection process, specifically, whether it is continuous or categorical. Dual-process theories generally assume that recollection is categorical, but signal detection theory requires it to be continuous. Chapters 2 and 3 provide direct evidence that recollection is a continuous process. In Chapter 2, two versions of a source memory experiment were conducted. The continuous view of recollection was supported because the relationship between confidence and accuracy on this recollection-based task was graded. The results detailed in Chapter 3 further validate recollection as a continuous process. The method involved an associative recognition test, which purportedly tests recollection in the absence of familiarity. In this task, word pairs were studied and then at test, the pairs were either intact or rearranged. When the word pairs were strengthened, we observed the typical result of an increasingly curvilinear ROC. Evidence from various procedures converged to suggest that recollection is a continuous process. The three chapters support the unequal variance signal detection theory of recognition memory and the idea that two continuous processes aggregate to yield a continuous memory strength variable