حساب فرآیندهای کنترل شده برای کاهش وابسته به سن در حافظه اپیزودیک

A decrease in controlled processes has been proposed to be responsible for age-related episodic memory decline. We used the Process Dissociation Procedure, a method that attempts to estimate the contribution of controlled and automatic processes to cognitive performance, and entered both estimates in regression analyses. Results indicate that only controlled processes explained a great part of the age-related variance in a word recall task, especially when little environmental support was offered.

Automatic memory processes, also called “ecphoric” processes, allow the activation of a memory trace from a cue without intention or conscious effort, at least if there is overlap between this cue and the memory trace. On the contrary, controlled processes are not supported by environmental support and do not depend on habits (Bialystok, Craik, Klein, & Viswanathan, 2004). They are involved in the rejection of irrelevant stimuli, the orienting of attention to the relevant ones, the conscious mental manipulation of information, etc. They also allow intentional and active memory search of a past event; in the memory context, they are called “recollection” and are assumed to be independent of the automatic use of memory (Jacoby, Bishara, Hessels, & Toth, 2005). It is now well known that elderly adults, compared with younger adults, are typically disadvantaged in their performance on direct or explicit tests of memory (such as free recall, recognition, and cued recall), in which participants are asked to recollect a previous study episode (for reviews, see Craik et al., 1995 and Craik and Jennings, 1992). For the last 15 years, much work has focused on the possible causes of such a decline. One interesting hypothesis stems from the convergence of two sets of data. Studies on brain-damaged patients and studies using brain imaging, have shown that the frontal lobes play an important role in memory tasks requiring controlled processes such as initiation, planning, reorganisation, spatial and temporal contextual details retrieval or in dual task memory situations (Craik and Grady, 2002, Moscovitch, 1994a, Norman and Schacter, 1996 and Shimamura et al., 1990). Moreover, age-related physiological brain changes are particularly visible in those cerebral structures (Grady and Haxby, 1995, Martin, 1998 and Prull et al., 2000) and could therefore disrupt controlled processes in aging people (Buckner & Koutstaal, 1998). Age effects on tasks implying control and effortful strategies have indeed been frequently described. Craik, 1986 and Craik, 2002 noted that aging persons obtained lower memory performances than young adults, especially when strategic demands were high, and that giving the participants environmental support reduced these age-related differences. For instance, telling people in advance the best way to encode the to-be-presented words reduced differences in memory performance between young and elderly adults (Bäckman and Karlsson, 1986 and Bäckman and Larsson, 1992). In the same way, giving them semantic cues at retrieval reduced the gap between elderly and young people (Bäckman and Karlsson, 1986, Bäckman and Larsson, 1992 and Craik et al., 1987) because retrieval cues allowed them to recover information from memory even if specification of some retrieval cues (attributed to controlled processes) was deficient (Burgess & Shallice, 1996). As also noted by Moscovitch (1992), free recall implies strategic processes to retrieve information (since no cues are given) while cued recall can be based on more automatic processes. However, strategic processes can also be implied in cued recall tasks in order to consciously retrieve encoded information (free and cued recall tasks being explicit tasks). It is therefore difficult, with these manipulations (e.g. giving cues), to distinguish the contribution of automatic and controlled processes to memory performance. In contrast to the age-related decline in tasks of direct or explicit memory, some studies have proposed that performance of older and younger adults do not differ on indirect or implicit memory tests (e.g. Ergis et al., 1995, Light et al., 1995 and Winocur et al., 1996). Implicit tasks, such as word fragment completion, in which no conscious recall of a prior learning episode is necessary, should involve automatic processes that are not conscious, rapidly performed, difficult to control, with minimal effort or with automatic allocation of attention (Hasher and Zacks, 1979 and Schneider and Shiffrin, 1977). Although these findings suggest that implicit memory does not change with age, there has been contradictory evidence showing the presence of a small age-related difference in performance on indirect tests (La Voie & Light, 1994). Moreover, Fleischman and Gabrieli (1998) pointed out that, although 85% of the studies reviewed reported age invariance in priming, almost half of these studies showed a 10% reduction in priming magnitude in elderly participants compared with young subjects. One factor often considered to explain these diverging results is the possibility that performance on indirect tests may be contaminated by consciously controlled use of memory and vice versa (performance on direct memory tests would be contaminated by automatic processes). For instance, in a comparative study between word recognition, cued recall and stem-completion, responses to a questionnaire showed that almost half of the participants used active retrieval strategies in the completion task (Gooding, Mayes, Van Eijk, Meudell, & McDonald, 1999). To summarise, one of the problems associated with explicit and implicit memory tasks is that they do not provide a pure measure of automatic and controlled processes since performance in one task is not solely supported by one kind of process (Jacoby, Toth, & Yonelinas, 1993). As both automatic and controlled processes play a role in all cognitive tasks, a procedure that could separate them in the same task seems necessary. Jacoby, 1991 and Jacoby, 1998 developed a paradigm, the “process-dissociation procedure”, that enables one to quantify properly and separately contributions of automatic and consciously controlled memory processes in the same task (familiarity versus recollection). For example, in a word-stem version of the procedure, subjects perform two tasks: one involving inclusion, the other exclusion. In the inclusion task, subjects are asked to complete a stem with a previously studied word and, if they are unable to do so, to use the first word which comes to mind. In the exclusion task, they are asked to complete a stem with a new word that was not encountered during the earlier study phase and to avoid (exclude) old studied words. It is then possible to insert performance of the two conditions in equations representing the logical relation between automatic (A) and controlled (C) memory processes. So, in the inclusion condition, a subject may complete correctly a stem with an earlier studied word either because they consciously recollect having seen the word before (C), or because it was the first word that came to mind automatically (A), when there is a failure of conscious recollection (1 − C). Thus, Inclusion = C + A(1 − C). On the other hand, in the exclusion condition, a subject may complete incorrectly an earlier studied word only if the word comes automatically to mind (A), without any controlled recollection that the word was presented earlier (1 − C): Exclusion = A(1 − C). Using these two equations and considering the subject’s performance on the two conditions, it is then possible to derive estimated contributions of automatic and controlled memory processes. Controlled processes can be estimated by subtracting the probability of responding with a studied word in the exclusion task from the probability of responding with an old word in the inclusion task [C = Inclusion − Exclusion]. Once an estimate of controlled processes has been obtained, the contribution of automatic processes corresponds to the probability of completing a stem with the studied word in the exclusion condition divided by one minus the estimate of controlled processes [A = Exclusion/(1 − C)]. This procedure was adapted for several kinds of memory evaluations and in various psychopathological states (Hay and Jacoby, 1999, Jacoby, 1999, Jennings and Jacoby, 1993, Jennings and Jacoby, 1997, Jermann et al., 2005, Smith and Knight, 2002, Titov and Knight, 1997 and Zelazo et al., 2004) as well as for non-memory performances (Spieler, Balota, & Faust, 1996). All these studies presented converging data showing an age-related decline in controlled processes along with age constancy in automatic processes. However, it should be noted that the process-dissociation procedure is not without its critics. The most controversial aspect is the assumption that controlled and automatic processes contribute independently to memory performance (from which the formulas proposed by Jacoby are derived). Alternative hypotheses are Redundancy and Exclusivity (however, see Jacoby, Yonelinas, & Jennings, 1997 for a refutation of these hypotheses). The major argument put forward by Jacoby and colleagues in favour of the independence assumption is that several studies identified variables that produce dissociate effects on the estimates of controlled and automatic processes (e.g. Debner and Jacoby, 1994, Hay and Jacoby, 1996, Jacoby, 1991, Jacoby, 1998, Jacoby, 1999, Jacoby et al., 1993, Jennings and Jacoby, 1993, Jennings and Jacoby, 1997, Kelley and Jacoby, 2000 and Yonelinas and Jacoby, 1995). As Jacoby has noted (Jacoby, 1998 and Jacoby and Hay, 1998), the inclusion/exclusion tests instructions are critical in order for controlled and automatic processes to independently influence task performance. Another criticism to the PDP procedure is that A and C estimated in a memory task could reflect non-mnesic controlled or automatic processes, for instance inhibitory processes ( Nigg, 2001; however see Hay & Jacoby, 1999 who demonstrated that C and A evaluated with the PDP effectively reflects memory processes). Assuming that the Process Dissociation Procedure is adequate to obtain correct estimates of automatic and controlled processes, the A and C estimates can then be used in regression analyses in order to verify the hypothesis that age-related decreases in episodic memory tasks are mediated by age-related decreases in controlled processes but not in automatic processes. However, many studies based on regression analyses use raw scores obtained in tasks that are multi-compound. For instance, to evaluate how speed processing can explain age-related decrease in memory, speed is defined as time to solve tasks such as Digit-Symbol or perceptive comparisons. Time measures are then entered in regression analyses to evaluate how they are able to reduce age effect on an independent memory task ( Bryan et al., 1999, Earles et al., 1997, Park et al., 1996, Salthouse et al., 1996 and Sliwinski and Buschke, 1997). Unfortunately, the tasks used to evaluate processing speed imply other variables in addition to the strict speed component, in other words, they are not “pure”, for instance working memory plays a role in the Digit-Symbol task and strategic processes are involved in perceptive tasks. Moreover, it is often difficult to determine if the observed relationships are due to one, several or all factors involved in the tasks. As noted by Salthouse, Toth, Hancock, and Woodard (1997), it is important to have tasks providing a pure evaluation of the suspected factors. They studied the influence of several factors on age-related decrease (from 18 years to 78 years) in a word learning task. These factors were perceptual speed, executive functions evaluated with Verbal Fluency tests and the Trail Making Test, spatial abilities, and controlled and automatic processes engaged in a stem-completion task and in a spatial attention task. In this spatial task, left or right directed arrows appeared on the left or on the right side of a screen. Location on the screen and arrow’s direction did or did not correspond (both were left, both were right, or one was left and the other was right). Subjects had to respond to the arrow’s direction only and ignore its position on the screen. The stem-completion task was similar to the ones used by Jacoby and colleagues (Jacoby, 1998 and Jacoby et al., 1993). Perceptual speed was the best explanatory factor of the age-related decline in the word learning task. C obtained with the spatial attention task also accounted for a great part of the age-related variance (78%). However, C obtained with the PDP on the stem-completion task only explained 18% of this variance. This result was rather unexpected since the material involved is this task was closer to the words to be learned than was the spatial task. Moreover, C for stem-completion did not significantly correlate with age, 3 which is in contradiction with results frequently described (e.g. Hay and Jacoby, 1999 and Jacoby, 1999). These results could be due to the fact that, in the stem-completion task, word encoding was incidental. This encoding procedure, which differs from the procedure generally used in studies of stem completion, could render the task more difficult. Indeed, C estimates were relatively low (respectively 31% and 24% for young and old participants). In this context, the major aim of our study was to re-evaluate how age effect on controlled processes could explain age-related memory decrease in “classical” memory tasks such as learning words lists. To obtain a more valid measure and to avoid floor effects (as suspected in Salthouse et al., 1997’ stem completion task), we modified somewhat the procedure proposed by Salthouse et al. Automatic and controlled processes were also evaluated with the Process Dissociation Procedure, as this procedure appears to give pure measures of C and A processes (in contrast with procedures where a variable is directly evaluated by the score obtained in the task), but we proposed an intentional encoding of the words ( Jacoby et al., 1993). As noted, since retrieval and encoding are not independent ( Craik, 2002), the fact of enhancing strategic memory processes thanks to intentional encoding should increase performance and controlled processes at retrieval, avoiding both floor performances and too small C values. Using regression analyses, we then correlated C and A estimates derived from the PDP task with memory scores obtained on “independent” words learning tasks (items in the PDP task and in the learning lists were different). In our study, the learning words lists varied in modality (they were orally or visually presented), in speed presentation (either fast or slow) and in support given at encoding (by using a list where words were organised in semantic clusters) or at retrieval (by providing semantic cues at retrieval). We manipulated these variables in order to evaluate the C involvement in several memory tasks: we expected that the C decrease would be a better explanatory factor than A changes in tasks requiring controlled and effortful processes (for instance when no support is given, or with faster presentation), but not necessarily in tasks where support is giving at encoding or, at least, in retrieval, given that support is supposed to reduce the intervention of controlled processes ( Craik, 2002). Therefore, variables manipulated in the word recall tasks should be susceptible to modify the contribution of controlled processes and particularly the possibility of C to explain age effects on memory performance ( Bäckman and Karlsson, 1986, Bäckman and Larsson, 1992 and Craik et al., 1987).