تبعیض خودآگاه و ناخودآگاه بین حافظه واقعی و کاذب

When subjects give higher confidence or memory ratings to a test word in a recognition test, do they simply raise their criterion without making better discrimination, or do they raise both criterion and true discrimination between the studied words (SW) and the lures? Given that previous studies found subjects’ false alarm responses to lures slower than to SW, and recognition latency inversely correlated with the confidence rating, can the latency difference between the lures and SW be accounted for by confidence or memory ratings? The present results showed that when subjects gave higher confidence or memory ratings, both their bias and sensitivity were raised, indicating that they could consciously distinguish the lures from the SW. However, a latency difference between true and false recognitions persisted after confidence and memory ratings were held constant, suggesting an unconscious source of discrimination between the two types of memory.

Roediger and McDermott (1995) revived and extended a memory research paradigm first developed by Deese (1959), which is now known as the Deese–Roediger–McDermott (DRM) paradigm. In this paradigm, participants are presented with a list of semantically associated words, such as bed, rest, awake, tired, dream, etc. except the thematic word sleep, i.e., the critical nonpresented word (CNPW), and are later tested on their recall and/or recognition memory of both the CNPW and the studied list words (SW). A very robust phenomenon found across numerous studies is that participants have a strong tendency to recall the CNPW or make a false positive identification of the CNPW. The false recall or recognition rate is comparable to that of an average studied word ( McDermott, 1996 and Roediger and McDermott, 1995). Subjects in the DRM paradigm experiments typically report vivid or distinctive recollections of having seen or heard “certain details” of the CNPW ( Gallo et al., 2001, Lampinen et al., 1999, Payne et al., 1996 and Roediger and McDermott, 1995). For example, they sometimes claimed which of the two or more speakers said the CNPW when the list words were presented by more than one speaker ( Lampinen et al., 1999 and Payne et al., 1996), and they sometimes felt even more confident about studying the CNPW than the words they actually studied ( Toglia, Neuschatz, & Goodwin, 1999). Because of this reason, Brainerd, Wright, Reyna, and Mojardin (2001) referred to this subjective, vivid experience as a phantom recollection, and Roediger (1996) called it a memory illusion. These findings highlight the vivid, realistic aspects of the experiences that subjects often report having, and the dramatic indistinguishability of the false memory from true memory. The present study examined the other side of the facts about the false memory.1 Namely, is there evidence that subjects may actually be able to discriminate the two types of memories both consciously and unconsciously, not in the sense that they stop making false alarms to the CNPW, but that they demonstrate knowledge, explicitly or implicitly, that the false alarm “old” and the hit old responses are different? That subjects call both the SW and CNPW “old” may not necessarily mean that their experiences with the two types of words are exactly the same. The first question asked in this study is whether subjects have any conscious awareness of the difference between their true and false memories. Although the main focus of the study is to determine if false memory can be unconsciously discriminated from true memory, any capability of discriminating the two types of memories must first be tested to see if it can be accounted for in terms of conscious processes before it can be attributed to an unconscious source or process. Jou, Matus, Aldridge, Rogers, and Zimmerman (2004) showed that overall, true recognition memory was assigned higher confidence ratings than false recognition memory. On the face of it, this seemed to be an indication of subjects’ being conscious of the distinction between the two types of memories. However, from the perspective of the signal detection theory (SDT) (Green and Swets, 1966, Macmillan and Creelman, 1991, Swets et al., 1961 and Tanner and Swets, 1954), it is only an indication of adopting different decision biases or criteria. According to SDT, the true ability of discrimination referred to as sensitivity (d’) and the stringency of the criterion (bias) used in a recognition decision can be separated. Although adopting a more stringent criterion can decrease, and adopting a less stringent criterion can increase the frequencies of hits and false alarms, the shifting of the criterion does not change the true discriminability between the signals and the noises. Sensitivity is measured as the distance (in standard deviation units) on the trace-strength dimension between the mean of the noise (new items) distribution and the mean of the signal (old items) distribution. Moving the criterion up (to the right), or down (to the left) on the strength dimension should not change the distance between the means of the two distributions, i.e., the value of sensitivity. Also, in SDT, confidence ratings are subjects’ indications of the decision criteria they use. Therefore, showing that true recognition is rated higher in confidence than false positive recognition does not unequivocally demonstrate a change in true discrimination. Hence, the issue of whether assigning higher confidence ratings to the SW than to the CNPW in the DRM paradigm indicates conscious discrimination of CNPW from SW will be revisited in this study by examining, not the confidence rating itself as in Jou et al. (2004), but the sensitivities (d’s) under different confidence ratings and memory quality judgments. Thus, the first purpose of this study is to determine whether the confidence ratings in the recognition test of the DRM paradigm reflect only a criterion shift with no change in sensitivity or indicate a criterion shift as well as a sensitivity change. Similarly, the remember/know/guess (R/K/G) judgments ( Gardiner et al., 1998 and Gardiner et al., 2002) should not have an effect on sensitivity either, if, according to some researchers ( Donaldson, 1996, Dunn, 2004 and Wixted and Stretch, 2004), these judgments are no more than criterion or bias shift in a recognition decision. Whether this basic SDT principle holds in the DRM paradigm has yet to be found out. If subjects demonstrate higher sensitivities when they give higher confidence ratings, then it means that they can consciously discriminate false memory from true memory. In other words, they “know” when their memory is more accurate and when less accurate. In that sense, they make the false alarm errors more or less “knowingly”. Note that according to the SDT, the true discrimination index d’ should not change with the increase or decrease in confidence rating. Therefore, if this basic SDT principle holds in the DRM paradigm, subjects should not show better sensitivities when they give higher confidence ratings than when they give lower confidence ratings. However, if this turns out to be not the case in the DRM paradigm, then one can argue that the DRM paradigm is an exceptional or limiting condition to the principles of SDT, which, in addition to being of interest to the conceptualization of false memory, is in itself an important discovery and has important theoretical implications for the SDT as well. The finding will provide the answer to this question: Do subjects respond “old” to the CNPW because they have absolutely no way of distinguishing the SW from the CNPW or because they are actually consciously accepting words that they know are questionable by simply adopting a lower decision criterion. Response speed sometimes can be indicative of implicit learning or knowledge (Lewicki et al., 1988 and Stadler, 1993). Therefore, a second question asked in this study is whether there is any evidence that false memory is unconsciously discriminable from true memory given that Jou et al. (2004) showed that the mean RT for false positive recognition was longer than for the hits. In other words, can this RT difference between the two types of memory be attributed to an unconscious source of discrimination? Jou et al. (2004) as well as some others ( Leonesio and Nelson, 1990 and Robinson et al., 1997) indicated that RT and confidence rating in recognition memory were inversely correlated, i.e., lower confidence was associated with longer RT and vice versa. Furthermore, Jou et al. (2004) found that false recognition memory was associated both with longer RTs and lower confidence ratings than true memory. Assuming that confidence rating reflects conscious experiences and given that it is correlated with RTs, is it possible that RT difference between the false memory and true memory is accessible to conscious awareness? The idea that RTs may be accessible to conscious evaluation is consistent with the notion of fluency heuristic ( Hertwig et al., 2008, Kelley and Linday, 1993 and Koriat and Ma’ayan, 2005). If confidence rating can account for the RT difference between true and false memories, then it can be argued that subjects are conscious of this RT difference between the two types of memories. If the RT difference between the hits and false alarms to the CNPW cannot be fully accounted for by confidence (i.e., if there is a significant residual in RT variance due to true versus false memories), then it can be suggested that the residual part of the RT variance that can be attributed to type of memory indicates an implicit source of discrimination between the two types of memories. Jou et al. (2004) did not address this issue of whether that RT difference between the false memory and the true memory can be attributed to the differences in confidence (a conscious experience) or to unconscious discrimination. In other words, the memory type effect (i.e., the contrast in RTs between hits and false alarms to CNPW) in Jou et al.’s (2004) study was confounded with confidence rating effect on RT. The same issue would exist when using the R/K/G judgments instead of confidence rating since RT for the R judgments was found to be shorter than the RT for the K judgments ( Dewhurst and Conway, 1994 and Dewhurst et al., 2006). In this study, the contrast in RT between the two types of memories was made under the condition in which confidence rating or conscious judgment of memory quality was held constant. Memory experiences that are given very high confidence ratings can be considered recollective experiences (Wixted, 2007). Similarly, memory experiences rated with an R judgment can be considered constituting a recollective experience (Tulving, 1985). Using a recall version of the DRM paradigm, Jou (2008) showed that the output latency of the first three or four recalled CNPW was as indistinguishably low as that of the SW, although the CNPW recalled after the first three or four positions had a significantly longer output latency than the SW recalled at these same positions. Based on this evidence, Jou (2008) suggested that the first batch of CNPW was likely generated during the encoding stage and had the phenomenological qualities of illusory recollection, indistinguishable from real recollection (they are the so-called phantom recollection, see Brainerd et al., 2001), whereas CNPW output at later serial positions might be constructed during the recall phase and had different phenomenological qualities from SW. However, the recall latency measure used in Jou’s (2008) study might not be sensitive enough (“unspeeded” and with relatively little control over the recall response process compared with a recognition test) to detect small but consistent differences in the memory retrieving speed. Therefore, the suggestion based on that finding that the CNPW output in the first three or four serial positions had the same recollection qualities as the SW might be premature. This study will contrast the RTs for the hits with those for false alarms to CNPW that were given the highest confidence rating and the R judgments (and hence supposedly both can be classified as recollective memories). This will reveal whether the illusory recollection is indeed indistinguishable from true recollection. 2. Experiment 1 Two questions were asked in Experiment 1. First, if higher confidence ratings mean merely higher criteria or response biases as defined in the SDT, then the sensitivity associated with higher confidence ratings should not be different from that associated with lower confidence ratings (Macmillan and Creelman, 1991, Swets et al., 1961 and Tanner and Swets, 1954). It is important to empirically test this hypothesis out in the DRM paradigm using the semantically associated words. If this basic principle of SDT holds in the DRM paradigm, then it means that subjects cannot filter out more CNPW by using stricter standards. If this basic signal detection principle fails to hold, that is, if higher confidence ratings are associated with higher sensitivities, and vice versa, then it means that subjects are conscious of the differences between the two types of probe words, and can adjust the decision criterion either up or down to let fewer or more CNPW pass through the “quality control” filter. The second question addressed is whether there is still an RT difference between the hits to the SW and the false alarms to the CNPW when the confidence rating is held constant. If there is no RT difference between the two responses that are given the same confidence rating, then it can be suggested that there is probably no unconscious discrimination beyond the discrimination manifested in the conscious confidence rating. In that case, it can be suggested that the illusory recollective experiences may be indeed identical to the true recollective experiences. If there is still a residual RT difference between the two types of memory given the same confidence rating, then that difference can be attributed to an implicit source of discrimination, and it can be argued that the illusory recollective experiences may not be completely identical to the true recollective experiences. In the experiment, subjects were given a 10-point and a nine-point scale for confidence rating. With this many gradations in the scale, hopefully, the fine nuances of the conscious experience can be captured in the ratings.