We often need to infer unknown properties of objects from observable ones, just like detectives must infer guilt from observable clues and behavior. But how do inferential processes change with age? We examined young and older adults' reliance on rule-based and similarity-based processes in an inference task that can be considered either a categorization or a multiple-cue judgment task, depending on the nature of the criterion (binary vs. continuous). Both older and young adults relied on rule-based processes in the multiple-cue judgment task. In the categorization task, however, the majority of older adults relied on rule-based processes while young adults preferred similarity-based processes. Moreover, older adults who relied on rule-based processes performed poorly compared with young adults who relied on the same process, suggesting that aging is associated with deficits in applying rule-based processes.

Current memory theories generally assume that memory performance reflects both recollection and automatic influences of memory. Research on people's predictions about the likelihood of remembering recently studied information on a memory test, that is, on judgments of learning (JOLs), suggests that both magnitude and resolution of JOLs are linked to recollection. However, it has remained unresolved whether JOLs are also predictive of automatic influences of memory. This issue was addressed in 3 experiments. Using the process-dissociation procedure, we assessed the predictive accuracy of immediate and delayed JOLs (Experiment 1) and of immediate JOLs from a first and from a second study-test cycle (Experiments 2 and 3) for recollection and automatic influences. Results showed that each type of JOLs was predictive of both recollection and automatic influences. Moreover, we found that a delay between study and JOL improved the predictive accuracy of JOLs for recollection, while study-test experience improved the predictive accuracy of JOLs for both recollection and automatic influences. These findings demonstrate that JOLs predict not only recollection, but also automatic influences of memory. (PsycINFO Database Record

Abstract. A memory measurement model is presented that accounts for judgments of remembering, knowing, and guessing in old-new recognition tasks by assuming four disjoint latent memory states: recollection, familiarity, uncertainty, and rejection. This four-states model can be applied to both Tulving's (1985) remember-know procedure (RK version) and Gardiner and coworker's ( Gardiner, Java, & Richardson-Klavehn, 1996 ; Gardiner, Richardson-Klavehn, & Ramponi, 1997 ) remember-know-guess procedure (RKG version). It is shown that the RK version of the model fits remember-know data approximately as well as the one-dimensional signal detection model does. In contrast, the RKG version of the four-states model outperforms the corresponding detection model even if unequal variances for old and new items are allowed for.We show empirically that the two versions of the four-statesmodelmeasure the same state probabilities. However, the RKG version, requiring remember-know-guess judgments, provides parameter estimates with smaller standard errors and is therefore recommended for routine use.

Zusammenfassung. Manche Methoden zur Erfassung verschiedener Gedächtnisprozesse setzen voraus, dass sich die Stärke der an Rekognitionsaufgaben beteiligten bewussten Erinnerungsprozesse nicht von den an Reproduktionsaufgaben beteiligten unterscheidet. Diese Kommensurabilitätsannahme scheint zwar zweifelhaft, ist bis jetzt aber empirisch nicht überprüft worden (vgl. Clark, 1999 ). Die Autoren zeigen, dass die Verlängerung von Wortstämmen in der Wortstammergänzungsaufgabe neben Rateprozessen auch bewusste Erinnerungsprozesse fördert, während automatische Gedächtnisprozesse unbeeinflusst bleiben. Dieser Befund spricht gegen die Gültigkeit der Kommensurabilitätsannahme. Die Konsequenzen des Befunds für Theorien des Gedächtnisses werden diskutiert.

No abstract