Illusory Recollection 1 Running Head: ILLUSORY RECOLLECTION Illusory Recollection and Dual-Process Models of Recognition Memory

Higham and Vokey (2000, Experiments 1 & 3) demonstrated that a slight increase in the display duration of a briefly presented word prior to displaying it in the clear for a recognition response increased the bias to respond “old”. In this research, 3 experiments investigated the phenomenology associated with this illusion of memory using the standard R-K procedure and a new, independent-scales methodology. Contrary to expectations based on the fluency heuristic, which predicts effects of display duration on subjective familiarity only, the results indicated that the illusion was reported as both familiarity and recollection. Furthermore, manipulations of prime duration induced reports of false recollection in all experiments. The results—in particular, the implications of illusory recollection—are discussed in terms of dual-process, fuzzy trace, two-criteria signal-detection models and attribution models of recognition memory. Illusory Recollection 3 Illusory Recollection and Dual-Process Models of Recognition Memory Dual-process theories of recognition memory (e.g., Jacoby & Dallas, 1981; Mandler, 1979, 1980) propose that hits (accepting studied targets as “old”) are based on two processes, commonly referred to as familiarity and recollection. Dual-process theory assumes that each process gives rise to a distinctive phenomenology; recollection (with or without familiarity) gives rise to the experience of reliving a past event, whereas familiarity (in the absence of recollection) provides a vague feeling of remembrance. Because participants’ phenomenology has been thought to track the processes underlying hits in recognition, new procedures have been developed to assist researchers in gaining access to participants’ mental experiences during memory tasks. One of these is the remember-know (R-K) procedure, introduced by Tulving (1985). With this procedure, as it applies to recognition memory, participants are instructed to indicate whether each test item rated “old” was “remembered” (R) or “known” (K). If the item is given an R rating, that is said to indicate that some aspect(s) of the context surrounding the item’s study episode has been retrieved. K ratings, on the other hand, are thought to indicate that the person believed that the test item was in the study list, but that the details regarding that previous encoding context of the item have not been retrieved. In dual-process theory, although both recollection (measured by R ratings) and familiarity (measured by K ratings) are assumed to underlie the hit rate, only familiarity is assumed to support false alarms (accepting nonstudied distracter items as “old”). Because false alarms, by definition, involve the acceptance of items for which there is no associated veridical study experience, recollection cannot form the basis of these errors. Dual-process theory also assumes that R judgments are an accurate measure of recollection and do not reflect variations in familiarity (e.g., Gardiner, 1988; Jacoby, 1998; Rajaram, 1993; Yonelinas & Jacoby, 1995) – an assumption that we will refer to as the recollection-remember identity (RRI) assumption. Because of the RRI assumption, a serious problem emerges: most dual-process accounts of recognition have no mechanism to account for systematic variation in false R ratings. In fact, such errors are usually just considered to be noise. For example, in a recent comprehensive review of dual-process theory, Yonelinas (2002) estimated recollection by subtracting R false alarms from R hits. This assumes that the observed R hit rate is Illusory Recollection 4 composed of “true” recollection plus some error (which should be removed); error, in this case, is the R false alarm rate. This simple, but crucial, gap in dual-process theory has important ramifications, and it is one of the central issues in the research reported in this article. Dual-process Models and False Recollection Probably because of the assumption that R judgments can only arise from mechanisms applied to or initiated by old items, false R judgments are usually not the focus of R-K research motivated from a dual-process theoretical viewpoint. To illustrate, consider first the fact that analyses of R data in published research have concentrated principally on the R hit rate (although see Donaldson, 1996; Gardiner, Richardson-Klavehn, & Ramponi, 1997; Hicks & Marsh, 1999; Hirshman & Henzler, 1998; Hirshman & Master, 1997; Inoue & Bellezza, 1998; Strack & Förster, 1995; Xu & Bellezza, 2001, and a few others for exceptions). The R false alarm rate is assumed (and often shown) to be minimal and to reflect a negligible number of responses. As a result, analyses of false alarms, or even discrimination measures (e.g., d’ or A’), are rare. As Donaldson has stated, “Recently, there has been an increased tendency [in recognition memory research] to use the hit rate as if it accurately measures memory” (Donaldson, 1996, p. 523). Second, most researchers attempting to manipulate R judgments have done so by varying the kind of encoding that old items undergo at study. These variations include the level of processing of the study items (Gardiner, 1988; Gardiner, Gregg, Mashru & Thaman, 2001; Rajaram, 1993), reading versus generating study words (Gardiner, 1988; Java, 1994), the attentional resources available during study (Gardiner et al., 2001; Gardiner & Parkin, 1990), the orthographic distinctiveness of the study words (Rajaram, 1998), the conceptual salience of the study words (Rajaram, 1998), the perceptual form of the study items (picture versus word: Dewhurst & Conway, 1994; Rajaram, 1993, 1996; Wagner, Gabrielli, & Verfaellie, 1997), and the number of study time presentations (Jacoby, Jones, & Dolan, 1998), to name a few. The focus on old item encoding at study when it comes to manipulating R judgments suggests that investigating false R judgments is of low priority in these lines of research. The failure on the part of dual-process models to account for false R judgments is particularly problematic with respect to the growing body of research that has used the R-K procedure, or some variant, to examine the nature of false memories (e.g., Brainerd, Reyna & Brandse, 1995; Dalla Illusory Recollection 5 Barba, 1993; Dewhurst, 2001; Higham, 1998; Holmes, Waters, & Rajaram, 1998; Lane & Zaragoza, 1995; Norman & Schacter, 1997; Payne, Elie, Blackwell & Neuschatz, 1996; Read, 1996; Roediger & McDermott, 1995; Roediger, Jacoby, & McDermott, 1996; Schacter, Verfaellie, & Anes, 1997, Whittlesea, in press). Rather than the R false alarm rate being negligible, this research has found that participants quite often rate their false memories as “remembered”, rather than merely “known”, as would be anticipated by dual-process models. A number of explanations have been advanced to explain such false R data, including inappropriate binding of episodic content to new items (Holmes et al., 1998), failures of source monitoring (Higham, 1998; Johnson, Hashtroudi, & Lindsay, 1993; Lane & Zaragoza, 1995) and phantom recollection (Brainerd, Wright, Reyna & Mojardin, 2001). Our point, however, is not that there are no mechanisms that can account for false R data. It is that dualprocess theory, as it is typically understood, cannot. To illustrate further the problem for dual-process theory, consider the following two different cases in which it has been clearly highlighted. Both cases are taken from the literature on word frequency effects in recognition. Case 1. Typically in recognition experiments, high frequency words receive lower hits, but higher false alarms, than low frequency words, an effect dubbed the mirror effect (e.g., Glanzer & Bowles, 1976; Higham & Brooks, 1997). Reder et al. (2000) applied the R-K procedure to recognition experiments in which the pre-experimental frequency of the words used as stimuli was varied. They found that both K hits and false alarms were higher for high frequency words than low frequency words. In contrast, the R hit rate was higher for low frequency words than high frequency words. They argued that their results were consistent with their quantitative Source of Activation Confusion (SAC) model – essentially a dual-process model -in which words are represented in two ways. First, there are word nodes, which are abstract representations of pre-experimental experiences with the word. Links fan out from the word nodes, with greater fan for high than low frequency words (more associations). Activating a word node, either through reading a (test) word, or through activation spreading from other nodes via links, generates K responding, and is analogous to familiarity. In addition to word nodes, there are also episode nodes in the SAC model. These Illusory Recollection 6 nodes represent the contextual details of each within-experiment study encounter with each word. Links form between episode nodes and word nodes at study so that activation can spread to and from a word node’s associated episode node. If reading a previously studied word at test causes enough activation of the episode node, an R response is generated, and recollection has occurred. The SAC model accounts well for the pattern of results that Reder et al. (2000) obtained. However, because R judgments derive from episode nodes, and because new items have no associated episode nodes, no mechanism is in place in the SAC model to account for false R judgments. Consequently, any systematic variation (from independent variables) in the R false alarm rate is not and cannot be predicted by the model. Such systematic variation was found in their Experiment 1; the rate of false R judgments was significantly lower for low frequency than high frequency words, despite false R rates being near floor. Case 2. Other research on word frequency effects in recognition, that also makes the RRI assumption characteristic of dual-process theory, has revealed larger effects on false R judgments. For example, Joordens and Hockley (2000, Experiments 1A & 1B) limited recollection by extending the retention interval between study and test in a standard recognition experiment using the R-K procedure. They predicted that the hit rate portion of the mirror effect would reduce, in response to the retention interval increase, because it was based primarily on recollection. Furthermore, the effect of retention interval on hits should be reflected primarily in R judgments because of the RRI assumption. In contrast, they predicted that the false alarm portion of the mirror effect, because it was based primarily on familiarity, would remain intact across manipulations of retention interval, and would be reflected primarily in K judgments. Joordens and Hockley (2000) made very similar assumptions in their predictions as Reder at al. (2000); that is, they assumed that R judgments reflect recollection, which supports hits, not false alarms. However, their independent variables had systematic effects on false R judgments; across two experiments, both low and high frequency words showed more than double the rate of R false alarms in the delay condition than in the immediate condition. Additionally, across four different comparisons, high frequency words received more R false alarms than low frequency words. These effects cannot be explained by Joordens and Hockley’s dual-process model because of the absence Illusory Recollection 7 of any mechanism to account for systematic variation in R false alarms. Fuzzy Trace Theory and Phantom Recollection Fuzzy trace theory (FTT) also implies a dual-process model of recognition. In early versions of the theory (e.g., Brainerd et al., 1995; Reyna & Brainerd, 1995), encountering a stimulus (at study) was thought to produce two representations in memory: a verbatim representation, which contained literal details of the original encoding incident, and a gist representation, which contained semantic content. At test, old items caused retrieval of both verbatim and gist representations, whereas new items resulted in the retrieval of only gist representations. Old-new recognition was based on the extent to which the retrieved representation was similar to the test item’s surface form and/or gist. In terms of the R-K distinction, R judgments resulted from retrieval of verbatim representations, whereas gist representations drove K judgments. This version of FTT greatly resembles Reder et al.’s SAC model: SAC’s word nodes are analogous to FTT’s gist representations and SAC’s episode nodes are analogous to FTT’s verbatim representations. Thus, early versions of fuzzy trace theory suffered the same problem that SAC currently does: both make the RRI assumption and both assume that recollection only reflects activation of a (veridical) representation that contains contextual detail. Recently, however, Brainerd, et al. (2001; see also Brainerd, Reyna & Mojardin, 1999) have explicitly recognized the problem that early versions of FFT had with accounting for R false alarms. Accordingly, they added a third parameter (process) to their model to account for such effects, called “phantom recollection”. Phantom recollection is argued to arise if a new item’s semantic gist was strongly instantiated at study and if the new item is an excellent retrieval cue for that gist memory (Brainerd et al., 2001). Such a mechanism is useful for explaining semantic false R effects, such as those seen in the Deese-Roediger-McDermott paradigm in which a critical lure’s meaning is primed by the presentation of several related targets at study (e.g., Deese, 1959; Dewhurst, 2001; Roediger & McDermott, 1995). However, other effects on false R judgments, such as Joordens and Hockley’s (2000) retention interval effect, and, indeed, our own effects involving display duration, are not easily explained in terms of variations in the strength of semantic gist and/or the efficacy of lures to access gist representations. We return to our results, and possible explanations of them with respect to FTT and phantom recollection, in the General Discussion. Illusory Recollection 8 Overview of the Experiments The procedure we used in our experiments was based on one we developed elsewhere (Higham & Vokey, 2000; see also Watkins & Gibson, 1988). In those experiments, participants first studied a long list of words, and later were administered a recognition memory test. However, prior to the presentation of each test item (target) for a recognition judgment, the same item (prime) was presented very briefly and then masked with ampersands (e.g., the target “table” was primed with a brief presentation of “table”). Participants were asked to try to identify the prime and then to make a recognition response to the target subsequently presented in the clear. Unbeknownst to participants, we varied the duration of the prime presentation; primes in the long duration condition were presented for 30 ms longer than primes in the short duration condition. For the brief durations we used in our prior research, we found that both old and new targets following long duration primes were more likely to be rated “old” than were targets following short duration primes. Consequently, given that we used similar durations in the experiments reported here, we expected a similar illusory memory effect of prime duration on recognition. However, our main concern here was with the effect of prime duration on false alarms and, in particular, with how the associated phenomenology would be rated by participants. Is the effect on new items manifested principally on familiarity ratings and K judgments as anticipated by dual-process models, or can we demonstrate similar effects on recollection ratings and R judgments? The second purpose of the current research was to introduce a new methodology for investigating participants’ phenomenology during recognition. In Experiments 2A and 2B, instead of using the standard R-K procedure, we asked participants to make two independent memory judgments for each item. After being taught the distinction between familiarity and recollection by adapting instructions that are now standard in the R-K literature, participants were asked to indicate how familiar the item was and, separately, to indicate how much the item was recollected. A different 4-point scale was used to make each judgment (1 = low; 4 = high, for both scales). As we discuss subsequently, this methodology avoids a number of undesirable side-effects of the R-K procedure, such as a necessary arithmetic dependence between the R and K rates, and its inability to isolate confident correct rejections. Illusory Recollection 9 Experiment 1 Other research has been conducted that has used prime manipulations similar to ours. For example, Rajaram (1993) replicated an illusion of memory first demonstrated by Jacoby and Whitehouse (1989). The illusion occurs when participants increase their bias to respond “old” to a recognition stimulus that follows a brief presentation of itself versus a brief presentation of some other, unrelated word. Rajaram found that when the R-K procedure was used to measure participants’ phenomenology, this illusion of memory was manifested in K judgments. Rajaram maintained that her results supported Jacoby and Whitehouse’s claim that the perceptual fluency of the target was enhanced by the matching prime compared to the mismatching prime. This enhanced fluency resulted in a feeling of familiarity which, in turn, enhanced the likelihood of a K judgment. The Jacoby-Whitehouse illusion investigated by Rajaram (1993) is not the same as the one reported in Higham and Vokey (2000) because her comparison was between prime-target match and mismatch trials, not between different prime display durations for prime-target match trials. However, to explain the latter Higham and Vokey effect, it could similarly be argued that prime duration enhances the fluency with which the subsequent matching target is processed and that participants experience this fluency as familiarity, which then increases their bias to respond “old”. From a fluency perspective, then, any effect of prime duration should be manifested in familiarity (K) judgments. However, we argued (Higham & Vokey, 2000) that the memory illusion we observed was due to use of an identification heuristic, and not target processing fluency; that is, participants partially based their recognition decisions on their ability to identify primes, using successful identification as indicator that the item was probably old. Participants may rely on identification performance because old items are generally identified better than new items (perceptual priming), so such performance is in fact predictive of prior presentation. However, the memory illusion arises when participants fail to consider (consciously or otherwise) other sources of identification ability (such as the slight, but effective, difference in prime duration) and over-attribute identification success to prior presentation. For current purposes, if this memory illusion is, indeed, based on prime identification success and not target processing fluency, it is not clear what the phenomenology associated with the memory illusion should be, or how participants would translate that experience into memory judgments. To find out, in Illusory Recollection 10 Experiment 1, we induced the memory illusion using a prime display duration manipulation at test, and investigated participants’ phenomenology using standard R-K methodology. Method Participants Participants were 40 undergraduate psychology students who participated in return for course credit. Participants were tested individually. Apparatus All displays were controlled by Pentium PCs using the default 25 x 80, white-on-black MSDOS text mode with 640 x 400 pixels displays. The vertical refresh rate was 70 Hz. All training items were displayed in lower case, centred vertically and horizontally on the screen. Design and Materials Study phase. A pool of 148 five letter English words of medium lexical frequency (M = 50 occurrences/million; SD = 52) was used. The pool was divided into two lists of 74 items that were balanced in terms of lexical frequency. Each participant was shown words from only one of the lists during the training phase. Words from the exposed list acted as “old” words on the recognition test, whereas words from the unexposed list acted as “new” words. Assignment of lists to the “old” and “new” conditions was counterbalanced across participants. Test phase. All 148 words from both training lists were presented in the test phase of the experiment. Each test trial consisted of (a) a fixation stimulus (*) presented for 1000 ms, (b) presentation of the prime for either short or long duration, (c) presentation of a mask (&&&&&) that remained in view until the space bar was depressed, and (d) presentation of the target (which always matched the prime) until a memory judgment was made. Half the words on the test list were presented for short duration (6-34 ms) whereas the other half were presented for long duration (3664 ms) and this factor was crossed with prior presentation (old versus new). Durations of 6-34 ms and 36-64 ms were achieved by programming durations of 20 ms and 50 ms, but taking into account the refresh rate of the monitors we were using. In Experiments 1 and 2A, we used 70 Hz monitors, which means that the displays was refreshed every 14.3 ms. Depending upon the location of the raster gun Illusory Recollection 11 when the video memory was changed by the program, some amount of time, from 0 to 14.3 ms, is added or subtracted from the programmed duration. Hence, participants would experience an almost perfectly uniform range of prime display durations from 6 to 64 ms, but with displays 6-34 ms defined as “short” and displays 36-64 ms defined as “long”. Thus, there were four experimental conditions; old-short, old-long, new-short and new-long. Assignment of words to the long and short duration conditions was also counterbalanced across participants. Procedure Study phase. Participants were instructed to try to memorise the word list that would be presented to them on the computer screen by reading each word aloud. To reinforce the instructions, a message to this effect appeared on the computer screen for participants to read after they had received the instructions verbally. After participants indicated that they understood the instructions, they pressed the space bar and each word from the 74 word training list was presented once to each participant, one at a time, for 2 s each in a random order. When all 74 words had been presented, a message was displayed that indicated that the training phase had ended, and to please wait for instructions. Test phase. Participants were informed that the test phase consisted of a series of trials. Each trial began with the presentation of a fixation stimulus (*) which indicated where the rest of the events in the trial would occur, and participants were instructed to look at it while it was in view. Following the fixation stimulus, a word was briefly presented and then masked with ampersands (&&&&&). These ampersands remained in view until participants tried to identify the word by writing it down on paper that was provided, guessing if necessary. To ensure that participants’ identification attempts applied to the appropriate test trial, the numbers 1 to 148 were printed on the paper participants used for prime identification and the trial number (1-148) was displayed on the computer monitor for each test trial. After attempting to identify the prime, participants were instructed to press the space bar to reveal the test item. Memory judgments were also made using paper; next to each word written as an identification attempt were two columns. After viewing the target item in the clear, participants were instructed to insert either an “O” (for “old”) or “N” (for “new”) next to the item in the first column and then either an “R” (for “remember”) or “K” (for “know”) in the second column for those Illusory Recollection 12 items rated “old”. The particular instructions used to describe the R-K procedure to participants were taken from Rajaram (1993), and are shown in Appendix A. Results and Discussion In all experiments, the recognition ratings were analysed first (both raw likelihoods and signal detection measures of discrimination and bias), followed up with an analysis of the identification performance. Our primary statistical tool was Analysis of Variance (ANOVA). An alpha level of .05 was adopted for all analyses and all significant effects from the ANOVAs are reported. However, for the sake of brevity, nonsignificant effects are not, unless they are considered important for exposition. Recognition ratings. The mean likelihoods of “old,” R and K ratings are shown in Table 1. The recognition, R ratings and K ratings were submitted to three separate 2 (prior presentation: old-new) x 2 (duration: short-long) within-subjects ANOVAs. All ANOVAs revealed main effects of prior presentation [recognition: old = .72, new = .29, F(1, 39) = 707.39, MSE = .010; R: old = .40, new = .07, F(1, 39) = 361.36, MSE = .012; K: old = .32, new = .22, F(1, 39) = 27.52, MSE = .014]. Additionally, a significant effect of duration was revealed for the analysis of the R judgments, F(1, 39) = 4.26, MSE = .004; items preceded with primes presented for long duration were judged to be “remembered” more often (.25) than items preceded by primes presented for short duration (.23). Given the importance of establishing that duration specifically influenced false recollection, the new items were analysed separately. A within-subjects one-way ANOVA indicated that there were significantly more recognition false alarms in the long duration condition (.32) than in the short duration condition (.26), F(1, 39) = 5.01, MSE = .012. More important, lengthening the prime display duration reliably induced false R judgments; although the effect was small, new items preceded by a matching prime displayed for long duration were significantly more likely to receive false R judgments (.09) than were new items preceded by a short duration prime (.06), F(1, 39) = 7.29, MSE = .002 (see Table 1). The recognition data were further analysed by determining nonparametric indices of old-new discrimination (A’, Grier, 1971) and response bias (B”D, Donaldson, 1992). These indices are shown in the top panel of Table 2. Six one-way, within-subjects ANOVAs were completed in total. All tested Illusory Recollection 13 the effect of prime display duration on recognition, K and R judgments, with the first three testing A’ and the latter three testing B”D. Only the effect of duration on B”D for R judgments was significant, F(1, 39) = 6.97, MSE = .011. Although generally quite conservative, response bias for R judgments was significantly more liberal in the long duration condition (.82) than in the short duration condition (.89). No significant effects were found for discrimination. Identification performance. The mean likelihoods of identifying primes in Experiment 1, as a function of prior presentation and duration, are shown in Table 3. A 2 (prior presentation: old-new) x 2 (duration: short-long) within-subjects ANOVA revealed main effects of prior presentation, F(1, 39) = 84.14, MSE = .004, and duration, F(1, 39) = 316.57, MSE = .035. Old primes were more likely to be identified (.64) than new items (.55) (perceptual priming; Jacoby & Dallas, 1981) and primes presented for long duration were more likely to be identified (.86) than primes presented for short duration (.33). Overall, the main results of Experiment 1 were the following: (1) Longer prime display duration significantly increased the likelihood of R judgments, and significantly reduced bias, relative to shorter prime duration. (2) For new items, the longer prime display duration significantly increased the likelihood of both recognition false alarms (“old” responses), and the associated subset of R false alarms, relative to the shorter prime duration. In contrast, the K false alarm rate was not significantly affected by the prime display duration manipulation. This pattern of results suggests that participants sometimes experience the illusion of memory brought about by varying prime duration as something they would rather attribute to recollection than to familiarity. (3) Old items were identified better than new items (perceptual priming; Jacoby & Dallas, 1981), EXPERIMENT 2 A serious problem with the standard R-K procedure is that the effect of prime duration on R judgments that we observed in Experiment 1 could be due to the fact that any effect on recognition must be reflected in R, K, or both. Indeed, inclusion of a third “guess” category in an R-K paradigm has been shown to change the pattern of R-K results obtained (e.g., Gardiner et al., 1997), Illusory Recollection 14 suggesting that the constraints imposed by the either-or nature of the standard R-K procedure are problematic. Although, strictly speaking, R judgments are not forced to increase if “old” responding increases (i.e., the effect could be manifested in K judgments), certainly an experimental effect on R would be more impressive if it were free to vary independently of old-new recognition decisions. Consequently, instead of using the standard R-K procedure in Experiment 2, we introduced an independent scales methodology to investigate our participants’ reported phenomenology during the recognition task. Participants were not asked for an overt recognition response, but instead were asked to rate how much each test item was “familiar” and separately, how much each item was “recollected” using two, separate 4-point rating scales. The distinction between recollection and familiarity was made explicit using amended R-K instructions. However, unlike the R-K instructions, participants were persuaded to use the two scales independently. For example, they were provided with examples that were meant to represent all four combinations of familiarity (Y-N) and recollection (Y-N), and were persuaded that all four combinations were possible (see Appendix B for instructions). Two different versions of Experiment 2 were run, differing in how the prime displays were controlled. Experiment 2A Method Participants. Participants were 28 undergraduate psychology students who participated in return for course credit. Participants were tested individually. Design and Materials. The design and materials were the same as in Experiment 1 except that, to test the generality of the effect of prime identification on recognition, the prime display durations were decreased slightly to 1-29 ms in the short condition and to 31-49 ms in the long condition. Procedure. All aspects of the procedure were the same as in Experiment 1, except for the following differences: 1. The recognition rating: After viewing the target item in the clear, participants were instructed to write an integer between 1 and 4 (where 1 = definitely no, 2 = probably no, 3 = probably yes, 4 = definitely yes) in each of the two spaces provided next to each identification attempt. The first value was meant to represent the degree to which the target item was familiar, and the second, Illusory Recollection 15 the degree to which that the target item was recollected. No overt yes/no recognition response was required. The detailed instructions used to prepare participants for the recognition ratings may be found in Appendix B. 2. Clarification of the instructions: Steps were taken in an attempt to ensure that participants fully understood the distinction between familiarity and recollection. First, a printout of the instructions for the test phase of the scales condition was administered to all participants; they were asked to read the instructions silently while the experimenter read them aloud. Second, after being instructed about the familiarity-recollection distinction, the experimenter asked participants to explain the distinction in their own words. When participants had displayed a satisfactory level of understanding of these concepts, the experimenter explained the four-point rating scales and read aloud examples of the different types of ratings the participants would be completing. These examples were meant to highlight the fact that the two scales could be used completely independently. Following these examples, participants were administered a four-item questionnaire to ensure that the relevant distinctions were clear to them. The questionnaire contained four scenarios, each meant to exemplify (1) both recollection and familiarity, (2) neither recollection nor familiarity (3) familiarity without recollection and (4) recollection without familiarity. The participants’ task was to read each scenario and decide which of these four combinations applied. Participants were able to proceed with the actual experiment only if they achieved a perfect score on the questionnaire. Results and Discussion Recognition Ratings. The recognition data in each experimental condition were first analysed by examining participants’ mean ratings on the recollection and familiarity scales (shown in the top panel of Table 4). A 2 (scale: recollection-familiarity) x 2 (prior presentation: old-new) x 2 (duration: short-long) within-subjects ANOVA revealed main effects of prior presentation, F(1, 27) = 92.27, MSE = .295, and duration, F(1, 27) = 15.45, MSE = .386. Old items were rated higher (2.70) than new items (2.01). More important, items displayed for long duration were rated higher (2.52) than items presented for short duration (2.19). The effect of duration was significant whether it was tested as a main effect with the data from each scale pooled, as in the previous analysis, or as orthogonal contrasts (i.e., simple effects) for each scale independently using the scale by duration interaction Illusory Recollection 16 error term [MSE = .051; familiarity: F(1, 27) = 51.41; recollection: F(1, 27) = 66.68]. The analysis that pooled the data from both scales indicated that there was a significant prior presentation by scale interaction, F(1, 27) = 9.61, MSE = .107. Orthogonal contrasts using this interaction error term revealed that new item ratings were significantly different between the scales, F(1, 27) = 18.80 (familiarity = 2.14 versus recollection = 1.87), whereas old item ratings did not differ, F < 1 (familiarity = 2.70 versus recollection = 2.70). The fact that duration did not interact with either prior presentation, F < 1, or scale, F < 1, indicates that it had a comparable effect on both old and new items and on both rated recollection and familiarity. It is possible that the previous effects on mean ratings occurred because of variations of unsure and guess responses, but not definitely yes or no responses. That is, our manipulations may only have affected participants’ use of scale values 2 and 3, leaving high confidence judgments (1 and 4) unaffected. To investigate this possibility, the frequency distributions across the four scale values for the recollection and familiarity scales were determined. Initial examination of these distributions indicated that scale values 2 and 3 were actually the least sensitive to our manipulations, and all seem to hover around .18 .20. Accordingly, we eliminated these data from the following analysis, and focused on only the proportions of “definitely yes” (4) and “definitely no” (1) responses. There was no necessary arithmetic dependence between the proportions associated with 1 and 4 (unlike the R-K procedure, where R + K = old), so “value” was included in a 2X2X2X2 within-subjects ANOVA as a factor with 2 levels (1-4), along with scale (recollection-familiarity), prior presentation (old-new) and duration (short-long). Because the experimental manipulations (duration and prior presentation) were expected to affect participants’ tendency to make positive responses, they were predicted to have opposite effects on the likelihood of choosing scale value 1 versus 4. Consequently, the effects of prior presentation and duration should be manifested as two-way interactions involving value. The 2X2X2X2 ANOVA revealed a main effect of value, F(1, 27) = 5.74, MSE = .174; participants rated more items as 1 (.35) than 4 (.26). However, as expected, value interacted with both duration, F(1, 27) = 17.66, MSE = .068, and prior presentation, F(1, 27) = 93.93, MSE = .058. Orthogonal contrasts using the interaction error term indicated that the duration by value interaction Illusory Recollection 17 occurred because items following long duration primes were assigned 4 significantly more often (.31) than items following short duration primes (.21), F(1, 27) = 8.69, but this effect was significantly reversed for 1 ratings, F(1, 27) = 8.98 (.30 and .40 for long and short duration trials, respectively). Similar orthogonal contrasts using the interaction error term revealed that the prior presentation by value interaction occurred because old items were assigned 4 significantly more often (.37) than new items (.15), F(1, 27) = 45.29, but this effect was significantly reversed for 1 ratings, F(1, 27) = 48.67 (.24 and .47 for old and new items, respectively). However, the value by prior presentation two-way interaction was qualified by a significant three-way interaction including scale, F(1, 27) = 11.38, MSE = .019. That is, the opposite effect of prior presentation on the 1 and 4 scale values (two-way) was greater with the recollection scale than the familiarity scale (three-way). Generally speaking, the pattern of ratings associated with old items remained fairly constant across the recollection and familiarity scales. However, the new item ratings differed substantially. In particular, new items were more likely to be correctly rejected with the recollection scale than the familiarity scale. The only other significant effect from the ANOVA was a two-way interaction involving prior presentation and duration, F(1, 27) = 9.41, MSE = .006. Old items in the long duration condition had a higher response proportion (.31) than old items in the short duration condition (.29), but the trend was reversed for new items (long: .30; short: .32). This interaction was not anticipated, and will not be discussed further. In short, the analysis limited to confident “yes” and “no” responses revealed the same pattern of results as the analysis on mean ratings, eliminating the possibility that the experimental manipulations were merely affecting unsure and guess responses. Because of the current emphasis on false recollection, new item recollection ratings were analysed separately. In three analyses [mean ratings, the proportion of 1 (definitely no) responses and the proportion of 4 (definitely yes) responses], the effect of duration was significant, F(1, 27) = 16.21, MSE = .103, F(1, 27) = 19.98, MSE = .015, and F(1, 27) = 7.28, MSE = .013, respectively. New items following primes displayed for long duration were assigned higher mean ratings (2.04) and were more likely to receive a response of “definitely yes” (.15), than new items following primes presented for short duration (1.70 and .07, respectively). However, new items were less likely to receive a response of “definitely no” in the long (.44) than short (.58) condition (see Table 4). Illusory Recollection 18 As in Experiment 1, discrimination (A’) and bias (B”D) were analysed. The results are shown in the middle panel of Table 2. The previous analyses suggest that the recollection scale should have a higher discrimination index because it had lower new item ratings, and about equal old item ratings, compared to the familiarity scale. In contrast, prime display duration should have an effect on the bias measure (i.e., long duration more liberal bias than short) because it affected both old and new item ratings about equally. To conduct the analyses, responses of “3” or “4” on the 4-point rating scale were scored as “yes” responses so that hit and false alarm rates could be computed for each participant. A 2 (scale: recollection-familiarity) x 2 (duration: short-long) within-subjects ANOVA on A’ revealed a significant effect of scale type, F(1, 27) = 9.80, MSE = .016. As expected, old-new discrimination using the recollection scale was greater (.73) than with the familiarity scale (.66). To determine whether discrimination was different between the scales because of variations in the hit rate, false alarm rate, or both, the hit and false alarm rates were entered into a 2 (scale: recollectionfamiliarity) X 2 (response type: hit-false alarm) X 2 (duration: short-long) within-subjects ANOVA. The ANOVA revealed main effects of response type, F(1, 27) = 85.11, MSE = .041, and duration, F(1, 27) = 12.04, MSE = .062, but more important, a response type by scale interaction, F(1, 27) = 6.79, MSE = .017. Orthogonal contrasts using the interaction error term indicated that discrimination differed between the scales because the recollection scale had significantly fewer false alarms (.28) than the familiarity scale (.38), F(1, 27) = 15.89, whereas the number of hits was constant, F < 1 (recollection = .57; familiarity = .58). Also as expected, a 2 (scale: recollection-familiarity) x 2 (duration: short-long) within-subjects ANOVA on B”D revealed a significant effect of duration, F(1, 27) = 12.64, MSE = .273. Response bias was more liberal in the long duration condition (.01) than in the short duration condition (.36). This effect of duration on response bias was comparable to the duration effect found for R response bias in Experiment 1. The final analysis on the recognition ratings involved deriving R and K responses from the independent scale data. The proportion of R judgments was estimated as the proportion of items that were assigned “yes” on the recollection scale, regardless of the associated familiarity value. In contrast, the proportion of K judgments was estimated as the proportion of items assigned “yes” on the familiarity scale, but “no” on the recollection scale. Four 2 (recollection: Y-N) X 2 (familiarity: Y-N) Illusory Recollection 19 contingency tables are presented in the top panel of Table 5, one corresponding to each experimental condition in the experiment (old-new X short-long), along with the derived R and K responses. As in Experiment 1, the proportions of derived K and R responses were analysed with separate 2 (duration: short-long) X 2 (prior presentation: old-new) within-subject ANOVAs. The analysis on derived R responses yielded results that were similar to those on actual R responses in Experiment 1: significant main effects of prior presentation, F(1, 27) = 97.99, MSE = .026, and duration, F(1, 27) = 12.91, MSE = .035, were found. Old items were more likely to be remembered (.57) than new items (.27), and items preceded by long duration primes (.48) were more likely to be remembered than items following short duration primes (.36). Thus, although the absolute scores differed somewhat in that both the R hit rate and the R false alarm rate were higher in Experiment 2A relative to Experiment 1 (see Tables 1 and 5), actual R judgments and derived R judgments appear to be sensitive to similar variables. These similarities occurred despite methodological differences between Experiments 1 and 2A, which included different recognition memory test formats, and slightly different display durations. In contrast, the analogous ANOVA on the derived K responses revealed a very different pattern of results than the actual K responses in Experiment 1. In addition to a counterintuitive marginal effect of prior presentation, F(1, 27) = 3.28, MSE = .022, p < .09, where more new items (.22) than old items (.17) were rated as familiar-but-not-recollected, a significant prior presentation by duration interaction was obtained, F(1, 27) = 9.00, MSE = .005. Orthogonal contrasts using the interaction error term indicated that derived K ratings were significantly higher for new items (.24) than old items (.15) in the long duration condition, F(1, 27) = 24.33, but no such difference existed for targets in the short duration condition, F < 1 (old = .19; new = .20). Based on the results from Experiment 1, no interaction of any nature was expected on K judgments, but particularly not one involving higher familiarity for new items than old items! Although this result was unexpected, others have also found that manipulations that were expected to increase knowing (familiarity), actually decrease it, when the R-K procedure is used. For example, Yonelinas and Jacoby (1995) found that matching the size of stimuli (perceptual Illusory Recollection 20 reinstatement of context) between study and a recognition test decreased knowing relative to a condition where sizes were mismatched (see also Rajaram & Coslett, 1992). This result was counterintuitive in that reinstating the perceptual qualities of the stimuli between study and test should enhance the fluency of processing at test, resulting in more K responses. However, they attributed the counterintuitive pattern of results to the nature of the standard R-K task; because participants cannot make both an R and a K judgment to the same item, and because the presence of recollection always leads to an R judgment, R effectively limits K. Indeed, they found if K was divided by one-minus-R (such that R and K were independent), the adjusted K proportion increased in response to size matching. Because the independent scales method eliminates the either-or nature of responding, it also alleviates the problem of R judgments limiting K judgments indeed, this is one reason we designed the method in the first place. However, the problem is only alleviated if the rating on the familiarity scale is taken as a measure of the underlying familiarity process. If, instead, the estimate of the underlying process is a rating of familiarity-in-the-absence-of-recollection, then the R-K problem identified by Yonelinas and Jacoby (1995) is also a problem with the independent scales method: the recollection rate limits the proportion of available responses that can be rated familiar-but-notrecollected. Following Yonelinas and Jacoby (1995), we divided the derived K proportion by one minus the derived R proportion to gain an unbiased (by R) estimate of familiarity. The results are shown in the bottom row of the top panel of Table 5. Indeed, the adjustment to K rendered a pattern of familiarity that was much more reasonable; a 2 (prior presentation: old-new) X 2 (duration: short-long) within-subjects ANOVA on the independent estimate of familiarity revealed main effects of prior presentation, F(1, 27) = 4.33, MSE = .035, and duration, F(1, 27) = 4.73, MSE = .028. Familiarity was higher for old items (.39) than new items (.32), and higher for items following long duration primes (.39) than short duration primes (.32). Thus, the odd results obtained with derived K in the previous analysis, which suggested higher familiarity for new items than old items, were likely due to recollection limiting the number of familiar-but-not-recollected judgments. The interaction from the ANOVA on the adjusted K proportions was also significant, F(1, 27) = 4.31, MSE = .014. Orthogonal Illusory Recollection 21 contrasts using the interaction error term indicated that duration had a significant effect on new items, F(1, 27) = 13.56 (short = .26 versus long = .37), but not on old items, F < 1 (short = .38 versus long = .40). The effect of duration specifically on new items (which also resulted in an overall main effect of duration in the ANOVA) was the only point of departure between the two experiments, although there was a trend for new items to receive more K responses in Experiment 1 also. Identification performance. The mean likelihoods of identifying primes in Experiment 2A, as a function of prior presentation and duration, are shown in Table 3. A 2 (prior presentation: old-new) X 2 (duration: short-long) within-subjects ANOVA revealed main effects of prior presentation, F(1, 27) = 52.87, MSE = .003, and duration, F(1, 27) = 689.74, MSE = .021. Old items were more likely to be identified (.49) than new items (.41) (perceptual priming; Jacoby & Dallas, 1981), and primes presented for long duration were more likely to be identified (.81) than primes presented for short duration (.09). Additionally, the analysis revealed a significant interaction, F(1, 27) = 9.34, MSE = .006; the size of the perceptual priming effect was larger for primes presented for long duration (.12) than for primes presented for short duration (.03). This result is likely due to a floor effect, as identification likelihoods in the short duration condition were less than 10% (see Table 3). Experiment 2B As explained earlier, there was some (uniform) variability of the prime display durations within the short and long conditions of Experiments 1 and 2A. Also, because the short and long durations conditions were only separated by an average of 30 ms, there was virtually no difference between the longest duration in the short condition and the shortest duration in the long duration condition. The lack of a marked division between the two duration categories within each experiment may have increased the likelihood that participants misattributed prime identification performance to the prior exposure of the target items. That is, the uniform distributions may have rendered display duration a less salient attributional source of identification performance, resulting in more misattributions to prior presentation. To test for this possibility, in this next experiment, the uniformity of the short and long prime display duration distributions was eliminated by synchronising the displays with the vertical raster-scan of the computer monitors. Experiment 2B was intended as a replication of the independent scales procedure of Illusory Recollection 22 Experiment 2A, but with fixed display durations. Sixty Hz video monitors were used in this experiment, so to be synchronised with the raster-scan of the monitors, prime display durations had to be some multiple of 1/60 sec. Accordingly, prime display durations of 1/60 and 3/60 sec (16.67 and 50 ms) were used as the short and long display durations, respectively. Method Participants. Twenty undergraduate psychology students participated in exchange for course credit. Participants were tested individually and in groups of 2-4 at individual work stations. Five participants were assigned at random to each of four counterbalancing conditions produced by rotating the items through the two levels of prior presentation and the two levels of prime display duration. Apparatus. To achieve synchronisation with the raster-scan, Apple //GS computers and monitors were used. These monitors have vertical refresh rates of 60 Hz. All items were displayed in lower-case using the default 40x24, white-on-black text mode. Design and Materials. The design and materials were similar to those of Experiment 2A. The principal difference was that the prime displays were synchronised to the 60 Hz raster-scan of the video monitors by detecting the vertical blanking interval (or VBLINT) of the video monitors, and updating the video memory and, hence, the display only during this interval.2 One further difference was that participants were prompted for their recognition judgments by the presentation of two response fields on the computer monitor, one labeled “familiarity” and one labeled “recollection”, into which they were to type their ratings. Only scale ratings of 1-4 were accepted, and participants had to type a response into both fields before they were allowed to proceed to the next trial. Procedure. The procedure was the same as in Experiment 2A, except that participants in this experiment read the training list silently rather than aloud because some participants were tested in small groups. Results and Discussion Recognition ratings. As in Experiment 2A, the recognition data in each experimental condition were first analysed by examining participants' mean ratings (shown in the bottom panel of Table 4) on the recollection and familiarity scales. Old items received significantly higher ratings Illusory Recollection 23 (2.73) than did new items (1.98), F(1, 19) = 48.50, MSE = .465, and items displayed for long duration received higher ratings (2.54) than did items presented for short duration (2.18), F(1, 19) = 20.24, MSE = .259. The effect of duration was also significant when effects of duration were tested as orthogonal contrasts (i.e., simple effects) for each scale independently using the scale by duration interaction error term [MSE = .028; familiarity: F(1, 19) = 92.60; recollection: F(1, 19) = 96.59]. As in Experiment 2A, there was a prior presentation by scale interaction, F(1, 19) = 6.69, MSE = .043. Orthogonal contrasts using the interaction error term revealed that ratings to new items were significantly different between the scales, F(1, 19) = 28.57 (familiarity = 2.11 versus recollection = 1.86), whereas ratings to old items did not differ significantly, F(1, 19) = 2.84 (familiarity = 2.77 versus recollection = 2.69). As In Experiment 2A, duration did not interact significantly with either prior presentation, F < 1, or scale, F < 1, indicating that duration had comparable effects on both old and new items, and on both rated recollection and familiarity. The relative frequencies of the two extreme scale values (1 = definitely no and 4 = definitely yes) for the recollection and familiarity scales were analysed as in Experiment 2A. The main effect of scale value was marginally significant, F(1, 19) = 3.63, MSE = .180, p = .07. Participants rated marginally more items as "definitely no" (.34) than "definitely yes" (.25). However, as in Experiment 2A, scale value interacted with both duration, F(1, 19) = 17.90, MSE = .053, and prior presentation, F(1, 19) = 44.94, MSE = .096. An orthogonal contrast using the interaction error term indicated that the duration by value interaction occurred because items following long duration primes were assigned "definitely yes" ratings significantly more often (.30) than items following short duration primes (.20), F(1, 19) = 7.40, but this effect was significantly reversed for "definitely no" ratings, F(1, 19) = 10.65 (.28 and .40 for long and short duration primes, respectively). A similar orthogonal contrast revealed that the prior presentation by scale value interaction occurred because old items were assigned ratings of "definitely yes" significantly more often (.38) than new items (.11), F(1, 19) = 28.89, but this effect was significantly reversed for ratings of "definitely no", F(1, 19) = 16.86 (.24 and .44 for old and new items, respectively). However, again as in Experiment 2A, this cross-over interaction of scale value and prior presentation was significantly larger for the recollection scale than the familiarity scale, F(1, 19) = 7.41, MSE = .009. Unlike Experiment 2A, however, there was no Illusory Recollection 24 significant interaction between prior presentation and duration, F(1, 19) = 2.02, MSE = .007, p > .17. As in Experiment 2A, new item recollection ratings were analysed separately to assess the degree of false recollection. The effect of duration was significant for all three analyses: mean ratings, F(1, 19) = 19.25, MSE = .074, the proportion of "definitely no" responses, F(1, 19) = 11.18, MSE = .013, and the proportion of "definitely yes" responses, F(1, 19) = 8.05, MSE = .009. New items following primes displayed for long duration were assigned higher mean ratings (2.05) and were more likely to receive a response of "definitely yes" (.14), than new items following primes presented for short duration (1.67 and .05, respectively). However, new items were less likely to receive a response of "definitely no" in the long (.44) than in the short (.55) condition. The recognition ratings were again analysed in terms of the signal detection indices of discrimination (A') and bias (B”D) as described for Experiment 2A. The results are shown in the bottom panel of Table 2. As in Experiment 2A, old-new discrimination (A') using the recollection scale was significantly greater (.73) than with the familiarity scale (.71), F(1, 19) = 6.73, MSE = .003. Similarly, response bias (B"D) was more liberal in the long duration condition (-.02) than in the short duration condition (.44), F(1, 19) = 19.95, MSE = .215. To determine whether discrimination was different between the scales because of variations in the hit rate, false alarm rate, or both, the hit and false alarm rates were analysed as a function of duration (short-long), and scale (familiarityrecollection). As in Experiment 2A, there were significant main effects of prior exposure (hit vs. false alarm), F(1, 19) = 50.87, MSE = .065, and duration, F(1, 19) = 20.87, MSE = .040, and, more important, a prior exposure by scale interaction, F(1, 19) = 4.50, MSE = .007. Orthogonal contrasts using the interaction error term showed that, as in Experiment 2A, discrimination differed between the scales primarily because the recollection scale had a significantly lower false alarm rate (.26) than did the familiarity scale (.35), F(1, 19) = 25.79, whereas the hit rates associated with each scale were only marginally different, F(1, 19) = 4.31, p < .06 (recollection = .57 versus familiarity = .61). The derived R and K rates were computed and analysed as in Experiment 2A. The results are shown in the bottom panel of Table 5. For derived R rates, as in Experiment 2A, significantly more old items were remembered (.57) than new items (.26), F(1, 19) = 51.72, MSE = .038, and items Illusory Recollection 25 preceded by long duration primes (.49) were significantly more likely to be remembered than items following short duration primes (.34), F(1, 19) = 19.98, MSE = .021. For derived K rates, as in Experiment 2A, there was a counterintuitive marginal effect of prior presentation, F(1, 19) = 3.73, MSE = .012, p < .07, in which more new items (.18) than old items (.13) were rated as familiar-but-notrecollected. However, as in Experiment 2A, dividing the K rates by one-minus-R (Yonelinas and Jacoby, 1995) rendered a more reasonable pattern of results; the bottom row of the bottom panel of Table 5 shows that the derived independent K rate was significantly higher for old items (.32) than new items (.24), F(1, 19) = 5.98, MSE = .021, and significantly higher for items following long duration primes (.31) than short duration primes (.24), F(1, 19) = 5.20, MSE = .016. Unlike Experiment 2A, there was no significant interaction of prior presentation and prime duration (F < 1) on derived independent K estimates: the effect of duration did not differ significantly between old and new items, as in Experiment 1. Identification performance. The mean likelihoods of identifying primes in Experiment 2B, as a function of prior presentation and duration, are shown in Table 3. A 2 (prior presentation: old-new) x 2 (duration: short-long) within-subjects ANOVA revealed main effects of prior presentation, F(1, 19) = 24.18, MSE = .003, and duration, F(1, 19) = 226.51, MSE = .046. Old items were more likely to be identified (.46) than new items (.40) (perceptual priming; Jacoby & Dallas, 1981), and primes presented for long duration were more likely to be identified (.80) than primes presented for short duration (.07). As in Experiment 2A, the analysis revealed a significant interaction, F(1, 19) = 4.49, MSE = .003; the size of the perceptual priming effect was larger for primes presented for long duration (.08) than for primes presented for short duration (.03). As we suggested for Experiment 2A, this interaction is most likely due to a floor effect as identifications in the short duration condition were less than 10% (see Table 3). In summary, the pattern of results from Experiment 2B replicates that of Experiment 2A almost exactly. The only mentionable exception was the lack of a significant interaction of prior exposure and duration on derived independent K estimates in Experiment 2B -a result that confirmed a similar lack of interaction in Experiment 1. Illusory Recollection 26 Overall Summary of Experiments 2A and 2B. (1) Relative to shorter prime duration, longer prime display duration significantly increased mean recognition ratings, the likelihood of confident acceptance (i.e., choosing scale value 4), and the likelihood of responding “yes” more generally, whereas it decreased response bias and the likelihood of an item being assigned 1 (confident rejection). These effects of duration were comparable between old and new items, and between the recollection and familiarity scales. (2) When examining recollection and new items separately, longer prime duration (relative to shorter prime duration) significantly increased mean recollection ratings, recollection false alarms, and the likelihood of an item being confidently accepted (4) on the recollection scale, but it decreased the likelihood of an item being confidently rejected (1) on the recollection scale. That is, the duration manipulation affected the degree of rated false recollection. (3) The recollection scale had better old-new discrimination than the familiarity scale, but this effect was attributable specifically to lower new item ratings (false alarm rate), not higher old item ratings (hit rate). (Indeed, the recollection hit rate was marginally lower than the familiarity hit rate in Experiment 2B.) This pattern of results was not limited to unsure and guess responses, but rather was mostly apparent in “definitely yes” (4) and “definitely no” (1) responses. (4) R judgments derived from the recollection and familiarity scales showed similar sensitivity to the experimental manipulations as actual R judgments in Experiment 1. (5) K judgments derived from the scales were not sensitive in the same way to the same experimental manipulations as actual K judgments in Experiment 1 and showed a counterintuitive pattern of results (e.g., K new > K old). However, if the K rate was divided by one-minus-R to render it independent of the R rate (Yonelinas & Jacoby, 1995), the pattern of results was more reasonable. This suggests that the R rate was limiting the uncorrected K rate. (6) Old items were identified better than new items (perceptual priming; Jacoby & Dallas, 1981). These results were obtained in this experiment despite the added effort to delineate carefully the distinction between recollection and familiarity in our instructions, the added test that was used to ensure that participants understood the distinction, and the evidence that suggested that Illusory Recollection 27 participants were using the scales differently (i.e., use of the scales was not haphazard as evinced by better old-new discrimination for the recollection scale than the familiarity scale). General Discussion The current research has replicated the memory illusion reported by Higham and Vokey (2000), but has extended that earlier work by exploring the attributions associated with the illusion. In short, increasing the duration, and hence identification, of a briefly presented stimulus immediately prior to judging the same stimulus presented in the clear for recognition, enhanced both reported recollection (Experiments 1, 2A, and 2B) and familiarity (Experiments 2A and 2B) of that stimulus relative to misidentifying it. This result was obtained both with the standard R-K methodology, and with our new, independent scales methodology. The identification effect on subjective familiarity and recollection was not limited to hits, but occurred also for false alarms. Probably because dual-process theory arose largely as a theory of target recognition (Brainerd et al., 2001), provisions are simply not in place in the theory to account for recollection false alarms. Although this gap in dual-process theory has been alluded to in the literature (e.g., see Xu & Bellezza, 2001), and some theorists have even amended their dual-process models in light of the problem (e.g., Brainerd et al., 2001), we think its seriousness for dual-process models has been underemphasized in recognition memory research. Accordingly, we now consider two alternative revisions to the assumptions underling dual-process theory that directly address the false recollection problem. Revisions to Dual Process Theory Solution 1: Change the relationship between recollection (the process) and items. Instead of viewing recollection as something that can only support hits, dual-process theory might be altered to have recollection occur for both hits and false alarms. This approach is similar to Higham’s (1997) account of specific similarity effects in artificial grammar learning. He found that test items that were specifically similar to single training items were more likely to be endorsed as “grammatical”, and he argued that this effect was recollection-based because it was eliminated when attention was divided at test. Following this example, then, independent evidence for the existence of recollection is derived from the effect of manipulations (e.g., dividing attention, lengthening the retention interval, and so on) that, a priori, are expected to affect recollection. Evidence for recollection is not based on Illusory Recollection 28 the type of item (old versus new) associated with the response. However, it is not clear how straightforward it would be for dual-process theory to incorporate such a modification to its assumptions given that recollection has always been conceived as something that reflects a veridical encounter with an item. Certainly, rather drastic amendments would have to be made to process dissociation equations (Jacoby, 1991), and formal dual-process models (e.g., SAC, Reder et al., 2000) to implement such a change. Solution 2: Change the relationship between recollection (the process) and recollection judgments. A second, perhaps preferable, solution would involve changing the assumed identity between the recollection process and recollection judgments (the RRI assumption) by allowing the familiarity process to affect such judgments. Over ten years ago, Jacoby (1991) argued that memory researchers were making a “process purity” assumption whereby it was assumed that performance on certain tasks was a pure measure of the underlying process. In our view, a similar, equally questionable “process purity” assumption is being made in R-K research today, only the assumption is that a particular response (R judgment) is a pure measure of the underlying process (recollection). Changing the RRI assumption by allowing R judgments to be influenced by familiarity has the advantage of providing an explanation for research results in domains such as memory and suggestibility in which participants come to “relive” false memories (e.g., Higham, 1998; Roediger et al., 1996). Presumably, because there are no veridical experiences that could have given rise to these “memories,” the phenomenology of false recollection must derive from familiarity (experienced as recollection). Indeed, a number of false R effects seem to be best explained by variations in familiarity. For example, the fact that new high frequency words receive higher R ratings than new low frequency words, whereas the opposite is true of old words (Reder et al., 2000; Joordens & Hockley, 2000), suggest that familiarity is “leaking” into R judgments of new items. To allow R judgments to measure both recollection (from old items) and familiarity experienced as recollection (from both old and new items) is a step that fuzzy trace theorists (e.g., Brainerd et al., 2001) have taken to expand their model. They suggested that excellent retrieval of a particularly strong gist representation (i.e., high familiarity) produces phantom recollection, such that high new item familiarity is reflected in R judgments. To explain our current results, these theorists could argue Illusory Recollection 29 that identifying a new word at test, before making a recognition response, caused a gist representation of the word to be created. Retrieval of this gist representation should be excellent during the recognition component of the test task, given the temporal proximity of the identification and recognition tasks, and the fact that the recognition stimulus always matches the newly formed gist representation. Thus, phantom recollection (i.e., high familiarity for new items that is experienced as recollection, leading to a false R judgment; Brainerd et al., 2001) is more likely for identified new words than misidentified ones. Although FTT with phantom recollection can explain some aspects of our current results, it cannot explain data from some of our other research using this paradigm. For example, in two experiments (Higham and Vokey, 2000, Experiment 2 and the 16.67/250 msec condition of Experiment 3), we made the manipulation of prime duration extreme; the manipulated difference in prime duration was over 200 msec, rather than 30 msec as used in our current research. Presumably, FTT predicts that gist representations of identified new words with the extreme manipulation would be even more accessible during recognition compared to misidentified words because, generally, they were presented for longer as primes (i.e., over 200 msec). Thus, FTT with phantom recollection predicts that the prime identification illusion that we demonstrated in our current research should be more pronounced with the more extreme manipulation of prime duration. However, we found the opposite; under these extreme circumstances, identification of the prime reduced the likelihood of an “old” response compared to misidentification, a result which is directly counter to the prediction of FTT. Recollection versus Familiarity Both Experiments 2A and 2B showed that the false alarm rate differed between the familiarity and recollection scales, but the hit rate did not. The false alarm portion of this difference would be expected from a dual-process perspective: For example, it could be argued that the absence of retrieval of veridical study information rendered recollection ratings near floor, whereas fluent processing (intraitem integration) of some new items renders somewhat higher familiarity ratings. Although the recollection-familiarity false alarm difference is interpretable with dual-process theory, the similarity in the hit rates, particularly in Experiment 2A, is more awkward to explain. DualIllusory Recollection 30 process theory holds that there are two bases supporting hits, with the recollection basis presumably measured by the recollection scale and the familiarity basis measured by the familiarity scale. The central question is: If the recollection and familiarity scales are measuring different components of memory, then why were the old item ratings virtually indistinguishable between the scales? One possibility is that the similarity was mere coincidence. That is, although the memory substrate measured by the two scales was different, participants chose the same scale values for a given item, such that the pattern of responding between the scales remained invariant. The temporal proximity of the recollection and familiarity ratings may have contributed to the use of similar scale values. However, this argument leaves unexplained why new item ratings differed between the scales. Additionally, the credibility of such an argument is stretched to the limit if one also wants to dismiss as coincidence the fact that prime display duration had almost exactly the same size effect on familiarity and recollection ratings to old items, and this similarity was maintained across two different experiments (see Table 4). A more plausible possibility is that the scales were actually measuring the same thing -that is, participants simply did not distinguish between recollection and familiarity in the presence of studied items. This conclusion is clearly at odds with dual-process theory and with an essential assumption underlying the R-K methodology, but nonetheless, it is the most parsimonious explanation of the current data. Our finding that recollection correct rejections form the basis of the distinction between the scales indicates a clear advantage of using the independent scales methodology over the more standard R-K methodology. The independent scale methodology can identify and separate cases of “probably” (3) and “definitely” (4) not recollected, regardless of the level of familiarity. This possibility is denied with standard R-K methodology. The R-K method was designed only to flag cases where the presence of veridical study information leads to a positive response, not cases leading to confident rejection. The nature of the R-K design means that both “definitely not” and “probably not” recollection judgments are assigned ambiguously to the default categories of “new” and “known”, making it impossible to pinpoint confident recollection correct rejections. Elsewhere, we have argued that rejection mechanisms form an important basis of responding in recognition and classification tasks (Higham & Brooks, 1997; Higham, Vokey, & Pritchard, 2000; see also, Johns & Mewhort, 2002; Illusory Recollection 31 Mewhort & Johns, 2000; Whittlesea, in press; Wright & Burton, 1995). By using the independent scales methodology, the importance of such mechanisms has been highlighted yet again, suggesting that they form a central basis of the distinction between judgments of recollection and familiarity, at least in the current experimental context. R-K Judgements and Donaldson’s (1996) Two-Criteria Signal Detection Model Donaldson (1996; see also Hirshman & Master, 1997; Inoue & Bellezza, 1998; Xu & Bellezza, 2001) proposed that most R-K data could be accounted for with a two-criteria signal detection (SD) model with a single dimension of familiarity. Specifically, he suggested that R judgments are made when a test item possesses familiarity above a response criterion that is more conservative than the recognition (old-new) criterion. K judgments, on the other hand, are made when a test item possesses familiarity above the recognition criterion, but below the more conservative R criterion and “new” judgments are made to test items with familiarity below both criteria. This SD model of R and K judgments has the clear advantage over dual-process theory in its ability to account for false R data. Such judgments in SD models are based on the fact that the tail of new item distribution often falls above the R criterion. Nonetheless, such a model cannot easily explain some of our data derived from the independent scales technique. For example, familiarity and recollection ratings in a two criteria SD model would most plausibly be represented by two different criteria placed on the same underlying memory dimension, much like R judgments are thought to be represented by a more conservative criterion than K judgments in the original model. Importantly, however, the SD model predicts that discrimination should not differ between the scales because each are assumed to be measuring the same underlying strength dimension. However, the analyses on the SD parameters in Experiments 2A and 2B indicated that this prediction was not supported: the recollection scale yielded better discrimination than the familiarity scale. An Attibutional Account The reliable effect of prime duration (identification) on R judgments in Experiment 1, and on recollection ratings in Experiments 2A and 2B may be surprising to some readers. Prime duration is likely to have affected the perceptual fluency of the recognition target, and Rajaram (1993) has Illusory Recollection 32 shown that enhancing the perceptual fluency of recognition targets by preceding them with briefly presented matching versus mismatching primes resulted in more subjective familiarity (K), but not more subjective recollection (R). Furthermore, our prime duration manipulation was unlikely to have affected the conceptual or elaborative encoding of the study items, the traditional underpinning of conscious recollection processes from the dual-process perspective (e.g., Jacoby & Dallas, 1981; Mandler, 1979, 1980), nor was it likely to have affected the distinctiveness or salience of the encoded items (Rajaram, 1998). Indeed, the fact that we observed prime duration effects on false R data renders all explanations based on differential study encoding of old words logically unfeasible. We believe the discrepancy between our results and those of Rajaram (1993), and the different pattern of results obtained with mild versus extreme manipulations of prime duration (cf. Higham & Vokey, 2000), may be best understood in terms of the attributional processes involved. We have argued that participants faced with 100% matching prime/target pairs may have been using an identification heuristic, whereby prime identification performance was used as a basis of recognition (i.e., identified = “old”; misidentified = “new”, with more items identified in the long duration condition than in the short duration condition). In contrast, participants faced with a mixture of matching and mismatching prime/target pairs, as in Rajaram’s research, may have relied more on the perceptual fluency of the target processing (i.e., fluent = “old”; nonfluent = “new”, with more targets fluently processed in the matching condition than in the mismatching condition). If, in fact, Rajaram’s participants were using the fluency heuristic, whereas our participants were relying on the identification heuristic, then there is no reason to expect that either responding, or the phenomenology associated with responding, would be the same. By this attributional account, particular judgments (e.g., K, R, “familiar” or “recollected”) are not directly tied to retrieval of particular traces or the activation of specific processes or systems. Rather, we believe that judgments of this sort are made in much the same way that Johnson and colleagues have suggested that people make source attributions (e.g., Johnson et al., 1993; although see below for points of departure between our and Johnson and colleagues’ viewpoints). During a memory test, participants have available to them myriad kinds of information. This information includes not just retrieved study details, but retrieved pre-experimental experiences, aspects of Illusory Recollection 33 current processing, information regarding recent performance, and other available details. The nature of the phenomenological judgment, in our view, will depend on which components of the information complex participants “assess” to complete the task at hand and how these components are interpreted. The interpretation, in turn, may depend on lay theories about the general workings of memory and on more personal knowledge regarding the working of participants’ own memory. Which information is salient and how it is interpreted will depend on the particular task and on the test context. For example, if the task involves reality monitoring, then highly salient aspects of this information complex will include memory for perceptual details (perhaps leading to a external source attribution) and memory for cognitive operations performed on the material (perhaps leading to a internal source attribution). On the other hand, if the task involves judgments of “familiar” or “recollected”, then any source-specifying details in the information complex will be made salient and likely lead to a recollection judgment. Similarly, our current research suggests that the lack of such details will support a “definitely not recollected” judgment. Our account of recollection and familiarity judgments is similar to the functionalist account forwarded by Gruppuso, Lindsay and Kelley (1997; see also Bodner & Lindsay, 2002 and Whittlesea, in press). Like us, they argued that recollection and familiarity are not necessarily tied to the operation or activation of particular processes or systems, but what counts as recollection or familiarity depends on the situation and context. In particular: Recollection allows one to exert control based on retrieval of memory for particular aspects or attributes of a prior event, and familiarity refers to retrieval of information that falls short of the specificity required by the task (Gruppuso et al., 1997, p. 273). We agree wholeheartedly with the notion that participants’ definitions of recollection and familiarity are not fixed and are contextually dependent. However, in our view, Gruppuso et al.’s (1997) functionalist account, as well as the source monitoring framework more generally, is still limited in that it, like dual process models, assumes that recollection is based primarily on retrieval of (sourcespecifying) veridical study details. We concur that the presence of veridical study details in the information complex available at test will certainly be important for supporting recollection judgments, as long as participants recognize them as such and incorporate them into their judgments (i.e., Illusory Recollection 34 monitor them correctly). However, more than this is needed. The functionalist account does not explain, for example, participants’ translation of successful test-time identification performance into attributions of recollection when that success has occurred in the context of hard-to-identify primes. With this attribution account in mind, consider again the opposite (to the current findings) pattern of results obtained when the prime duration manipulation was made more extreme (Higham & Vokey, 2000, Experiments 2 & 3) that posed problems for FTT. We argued that with an extreme manipulation of prime duration, participants no longer relied on the identification heuristic to make recognition judgments. This was so because identification performance was no longer predictive of prior presentation; for example, items were identified correctly at the very long duration (over 200 ms) regardless of whether they were old or new. Instead, participants adopted the fluency heuristic and made attributions regarding target processing fluency to various sources. When primes were identified, a ready source of target processing fluency was made available (prime identification itself), and so misattributions of high target processing fluency to prior presentation were less likely than if the prime was misidentified. Consequently, the effect of prime duration, seen in the current experiments, reverses with more extreme manipulations of prime duration. The identification heuristic is probably just one strategy of many that participants use in a recognition memory experiment. Recognition memory research has been primarily focused on better understanding participants’ use of fluency, not on discovering the various heuristics and strategies participants might use across different experimental paradigms (although see Whittlesea, in press; Whittlesea & Leboe, 2000, 2002 for exceptions). Nonetheless, participants’ use of the identification heuristic in the current setting, at the very least, highlights the insufficiency of veridical retrieval, source-specifying or not, as the sole basis of recollection judgments. In summary, FTT with phantom recollection, two criteria SD models, and particularly dual process theory with the RRI assumption all have difficulty accounting for the results we have obtained in this and associated research. In contrast, an attributional account, which argues that participants assess the evidence available to them at the time of the recognition response (e.g., prime identification success; target processing fluency), assess it for its predictive value, and make (mis)attributions accordingly, is able to explain the extant data generated by this paradigm. Illusory Recollection35 ReferencesBodner, G.E., & Lindsay, D.S. (2002). Remembering and knowing in context. Manuscriptsubmitted for publication.Brainerd, C.J., Reyna, V.F., & Brandse, E. (1995). Are children’s false memories morepersistent than their true memories? Psychological Science, 6, 359-364.Brainerd, C.J., Reyna, V.F., & Mojardin, A.H. (1999). Conjoint recognition. PsychologicalReview, 106, 160-179.Brainerd,C.J., Wright, R., Reyna, V.F., & Mojardin, A.H. (2001). Conjoint recognition andphantom recollection. Journal of Experimental Psychology: Learning, Memory and Cognition, 27,307-327.Dalla Barba, G. (1993). Confabulation: Knowledge and recollective experience. CognitiveNeuropsychology, 10, 1-20.Deese, J. (1959). On the prediction of occurrence of particular verbal intrusions in immediaterecall. Journal of Experimental Psychology, 83, 266-273.Dewhurst, S. (2001). Category repetition and false recognition: Effects of instance frequencyand category size. Journal of Memory and Language, 44. 153-167.Dewhurst, S., & Conway, M. (1994). Pictures, images and recollective experience. Journal ofExperimental Psychology: Learning, Memory, and Cognition, 20, 1088-1098.Donaldson, W. (1992). Measuring recognition memory. Journal of Experimental Psychology:General, 121, 275-277.Donaldson, W. (1996). The role of decision processes in remembering and knowing. Memory& Cognition, 24, 523-533.Gardiner, J.M. (1988). Functional aspects of recollective experience. Memory & Cognition,16(4), 309-313.Gardiner, J.M., Gregg, V.H., Mashru, R., & Thaman, M. (2001). Impact of encoding depth onawareness of perceptual effects in recognition memory. Memory & Cognition, 29, 433-440.Gardiner, J.M., & Parkin, A.J. (1990). Attention and recollective experience in recognitionmemory. Memory & Cognition, 18, 579-583. Illusory Recollection36 Gardiner, J.M., Richardson-Klavehn, A., & Ramponi, C. (1997). On reporting recollectiveexperiences and “direct access to memory systems.” Psychological Science, 8, 391-394.Glanzer, M., & Bowles, N. (1976). Analysis of the word frequency effect in recognitionmemory. Journal of Experimental Psychology: Human Learning and Memory, 2. 21-31.Grier, J.B. (1971). Nonparametric indexes for sensitivity and bias: Computing formulas.Psychological Bulletin, 75, 424-429.Gruppuso, V, Lindsay, D.S., & Kelley, C.M. (1997). The process-dissociation procedure andsimilarity: Defining and estimating recollection and familiarity in recognition memory. Journal ofExperimental Psychology: Learning, Memory and Cognition, 23 , 259-278.Hicks, J.L., & Marsh, R.L. (1999). Remember-know judgments can depend on how memory istested. Psychonomic Bulletin and Review, 6, 117-122.Higham, P.A. (1997). Dissociations of grammaticality and specific similarity effects in artificialgrammar learning. Journal of Experimental Psychology: Learning, Memory and Cognition. 23. 1029-1045.Higham, P.A. (1998). Believing details known to have been suggested. British Journal ofPsychology. 89, 265-283.Higham, P.A., & Vokey, J.R. (2000). Judgement heuristics and recognition memory: Primeidentification and target-processing fluency. Memory & Cognition, 28, 574-584.Higham, P.A., Vokey, J.R., & Pritchard, J.L. (2000). Beyond dissociation logic: Evidence forcontrolled and automatic influences in artificial grammar learning. Journal of ExperimentalPsychology: General, 129. 457-470.Higham, P.A., & Brooks, L.R. (1997). Learning the experimenter’s design: Tacit sensitivity tothe structure of memory lists. Quarterly Journal of Experimental Psychology: Human ExperimentalPsychology, 50. 199-215.Hirshman, E., & Henzler, A. (1998). The role of decision processes in conscious recollection.Psychological Science, 9, 61-65.Hirshman, E., & Master, S. (1997). Modelling the conscious correlates of recognition memory:Reflections of the remember-know paradigm. Memory & Cognition, 25, 345-351. Illusory Recollection37 Holmes, J.B., Waters, H.S., & Rajaram, S. (1998). The phenomenology of false memory:Episodic content and confidence. Journal of Experimental Psychology: Learning, Memory, &Cognition, 24, 1026-1040Inoue, C., & Bellezza, F.S. (1998). The detection model of recognition using know andremember judgments. Memory & Cognition, 26, 299-308.Jacoby, L.L. (1991). A process dissociation framework: Separating automatic fromintentional uses of memory. Journal of Memory and Language, 30, 513-541.Jacoby, L.L. (1998). Invariance in automatic influences of memory: Toward a user’s guide forthe process-dissociation procedure. Journal of Experimental Psychology: Learning, Memory, andCognition, 24, 3-26.Jacoby, L.L., & Dallas M. (1981). On the relationship between autobiographical memory andperceptual learning. Journal of Experimental Psychology: General, 110, 306-340.Jacoby, L.L., Jones, T.C., & Dolan, P.O. (1998). Two effects of repetition: Support for a dual-process model of know judgments and exclusion errors. Psychonomic Bulletin and Review, 5, 705-709.Jacoby, L.L., & Whitehouse, K. (1989). An illusion of memory: False recognition influencedby unconscious perception. Journal of Experimental Psychology: General, 118, 126-135.Java, R.I. (1994). States of awareness following word stem completion. European Journal ofCognitive Psychology, 6, 77-9.Johns, E.E., & Mewhort, D.J.K. (2002). What information underlies correct rejections in short-term recognition memory? Memory & Cognition, 30, 46-59.Johnson, M.K., Hashtroudi, S., & Lindsay, D.S. (1993). Source monitoring. PsychologicalBulletin. 114, 3-28.Joordens, S., & Hockley, W.E. (2000). Recollection and familiarity through the looking glass:When old does not mirror new. Journal of Experimental Psychology: Learning, Memory andCognition, 26, 1534-1555.Lane, S.M., & Zaragoza, M.S. (1995). The recollective experience of cross-modality errors,Memory & Cognition, 23, 607-610. Illusory Recollection38 Mandler, G. (1979). Organization and repetition: Organizational principles with specialreference to rote learning. In L.G. Nilsson (Ed.), Perspectives on memory research: Essays in honorof Uppsala University's 500th Anniversary (pp. 293-327). Hillsdale, NJ: Erlbaum.Mandler, G. (1980). Recognizing: The judgment of previous occurrence. PsychologicalReview, 87, 252-271.Mewhort, D.J.K., & Johns, E.E. (2000). The extralist-feature effect: Evidence against itemmatching in short-term recognition memory. Journal of Experimental Psychology: General, 129, 262-284.Norman, K.A., & Schacter, D.L. (1997). False recognition in younger and older adults:Exploring the characteristics of illusory memories. Memory & Cognition, 25, 838-848.Payne, D.G., Elie, C.J., Blackwell, J.M., & Neuschatz, J.S. (1996). Memory illusions:Recalling, recognizing, and recollecting events that never occurred. Journal of Memory andLanguage, 35, 261-285.Rajaram, S. (1993). Remembering and knowing: Two means of access to the personal past.Memory & Cognition, 21 , 89-102.Rajaram, S. (1996). Perceptual effects on remembering: Recollective processes in picturerecognition memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 22 ,365-377.Rajaram, S. (1998). The effects of conceptual salience and perceptual distinctiveness onconscious recollection. Psychonomic Bulletin and Review, 5, 71-78.Rajaram,S., & Coslett, H.B. (1992) Further dissociations between “remember” and “know”judgments in recognition memory. Presented at the 33 annual meeting of the Psychonomic Society.Read, J.D. (1996). From a passing thought to a false memory in 2 minutes: Confusing realand illusory events. Psychonomic Bulletin and Review, 3, 105-111.Reder, L.M., Nhouyvanisvong, A., Schunn, C.D., Ayers, M.S., Angstadt, P., & Hiraki, K.(2000). A mechanistic account of the mirror effect for word frequency: A computational model ofremember-know judgments in a continuous recognition paradigm. Journal of ExperimentalPsychology: Learning, Memory and Cognition, 26, 294-320. Illusory Recollection39 Reyna, V.F., & Brainerd, C.J. (1995). Fuzzy-trace theory: An interim synthesis. Learning andIndividual Differences, 7, 1-75.Roediger, H. L., Jacoby, J. D., & McDermott, K. B. (1996). Misinformation effects in recall:Creating false memories through repeated retrieval. Journal of Memory and Language, 35, 300-318.Roediger, H.L., & McDermott, K.B. (1995). Creating false memories: Remembering words notpresented in lists. Journal of Experimental Psychology: Learning, Memory, & Cognition, 21, 803-814.Schacter, D.L., Verfaellie, M., & Anes, M.D. (1997). Illusory memories in amnesic patients:Conceptual and perceptual false recognition. Neuropsychology, 11, 331-342.Strack, F., & Förster, J. (1995). Reporting recollective experience: Direct access to memorysystems? Psychological Science, 6, 352-358.Tulving, E. (1985). Memory and consciousness. Canadian Psychology, 26, 1-12.Wagner, A.D., & Gabrieli, J.D.E., & Verfaellie, M. (1997). Dissociations between familiarityprocesses in explicit recognition and implicit perceptual memory. Journal of Experimental Psychology:Learning, Memory, and Cognition, 23, 305-323.Watkins, M.J., & Gibson, J.M. (1988). On the relation between perceptual priming andrecognition memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 14 ,477-483.Whittlesea, B.W.A. (in press). Two routes to remembering (and another to remembering not).Journal of Experimental Psychology: General.Whittlesea, B.W.A., & Leboe, J.P. (2000). The heuristic basis of remembering andclassification: Fluency, generation, and resemblance. Journal of Experimental Psychology: General,129, 84-106.Whittlesea, B.W.A., & Leboe, J.P (2002). The inferential basis of familiarity and recall:Evidence for a common underlying process. Journal of Memory and Language, 46. 804-829.Wright, R.L., & Burton, A.M. (1995). Implicit learning of an invariant: Just say no. QuarterlyJournal of Experimental Psychology: Human Experimental Psychology. 48A. 783-796.Yonelinas, A.P., & Jacoby, L.L. (1995). The relation between remembering and knowing asbases of recognition: Effects of size congruency. Journal of Memory and Language, 34, 622-643. Illusory Recollection40 Yonelinas, A.P. (2002). The nature of recollection and familiarity: A review of 30 years ofresearch. Journal of Memory and Language, 46, 441-517.Xu, M., & Bellezza, F.S. (2001). A comparison of the multimemory and detection theories ofknow and remember recognition judgements. Journal of Experimental Psychology: Learning, Memory,and Cognition, 27, 1197-1210. Illusory Recollection41 Appendix AInstructions for the R-K condition of Experiment 1During this phase, words will be flashed one at a time on the computer monitor. Once a wordis flashed, your first job is first to try to identify it and write it on paper in front of you. After that, pressthe space bar and the word that was flashed will appear in the clear. At this point, regardless ofwhether you identified the word correctly or not, rate whether you think the word occurred in thetraining phase or not. Write either 'O' (for 'old'; it did occur in training) or 'N' (for 'new'; it did not occurin training) in the space provided next to the word.Remember judgmentsIf you decide that an item is 'old' (i.e., it occurred in the training list), then you will need tomake a further judgment as well. If your recognition of the word is accompanied by a consciousrecollection of its prior occurrence in the study list, then put down 'R' (for 'remember'). 'Remember' isthe ability to become consciously aware again of some aspect or aspects of what happened or whatwas experienced at the time the word was presented (e.g., aspects of the physical appearance ofthe word, or of something that happened in the room, or of what you were thinking and doing at thetime). In other words, the 'remembered' word should bring back to mind a particular association,image, or something more personal from the time of study, or something about its appearance orposition (i.e., what came before or after the word).Know judgments'Know' responses should be made when you recognise that the word was in the list of wordsthat you read aloud, but you cannot consciously recollect anything about its actual occurrence. Inother words, put down 'K' (for 'know') when you believe you recognise the word, but the word fails toevoke any specific conscious recollection from the study list.To further clarify the difference between these two judgments, (i.e., 'R' versus 'K'), here are afew examples. If someone asks for your name, you would typically respond in the 'know' sense,without becoming consciously" aware of anything about a particular event or experience; however,when asked about the last movie you saw, you would typically respond in the 'remember' sense, thatis, becoming consciously aware again of some aspects of the experience. If you have any questions Illusory Recollection42 regarding these judgments, please ask the experimenter. Illusory Recollection43 Appendix BInstructions for the independent scales condition of Experiments 2A and 2BIn the test phase, you will be presented with more words, one at a time on the computermonitor. However, at first, each word will be displayed very briefly and then covered up. Your first jobis to try to identify the word and to write it down on the paper provided. Even though you may nothave been able to see the word, guess what it is. It is important that you write down something onevery trial. To help you keep track of what trial you are on, the trial number will be displayed on themonitor. Make sure that the trial number you are at on the test sheet is the same as the numberdisplayed on the monitor.After you have attempted to identify the word and have written it down on the paper, pressthe space bar to find out if you were right. Once the word is presented in the clear, then two ratingsare required. One is a recollection rating and the other is a familiarity rating.Recollection RatingIf the word is accompanied by a conscious memory of its prior occurrence in the study list,then you are recollecting it. "Recollection" is the ability to become consciously aware again of someaspect or aspects of what happened or what was experienced at the time the word was presented(e.g. aspects of the physical appearance of the word, or of something that happened in the room, orof what you were thinking or doing at the time). In other words, the "recollected" word should bringback to mind a particular association, image, or something more personal from the time of study, orsomething about its appearance or position (i.e. what came before or after that word). One half ofthe items in the test list were presented in the study list and one half were not.Familiarity RatingSometimes you may know a word occurred in the study list because it provides a feeling offamiliarity. This feeling can be thought to occur independently of recollection. A word might seemfamiliar whether or not you recollect anything from the time you studied it. Likewise, recollection canoccur with or without a feeling of familiarity. As stated above, one half of the items in the test listwere presented in the study list and one half were not.To clarify the difference between these two ratings, consider the following examples. Illusory Recollection44 High recollection-high familiarityIf a word evokes a feeling of familiarity, and you can recollect something about the word’soccurrence in the study list, then you should rate both recollection and familiarity as high.High-recollection-low-familiarityFor any particular test item, you might recollect your encounter with the test word, but it doesnot seem familiar. For example, you might remember coughing when this word was presented earlier,but you have no feeling of familiarity associated with this memory. If this is the case, recollectionshould be rated high but familiarity should be rated low.Low recollection-high familiarityFor any particular test item, you might have a strong feeling of familiarity associated with it,but not recollect anything about your encounter with it in the study phase. If this is the case,recollection should be rated low but familiarity should be rated high.Low recollection-low familiarityItems that evoke no feeling of familiarity or recollection should be rated low on bothrecollection and familiarity.Make the familiarity and recollection ratings for each word by inserting a number from 1 to 4 inthe space provided where:1 = definitely no2 = probably no3 = probably yes4 = definitely yesThis rating scale is shown at the top of the first page of your answer sheets.If you have any questions regarding these judgments, please ask the experimenter. Illusory Recollection45 Author NotesPhilip A. Higham, Department of Psychology, University of Southampton and John R. Vokey,Department of Psychology and Neuroscience, University of Lethbridge. Preparation of this articlewas supported by British Academy and Natural Sciences and Research Council of Canada researchgrants. Portions of this research were presented at the NATO Advanced Study Institute on Memoryand Trauma in Port de Bourgenay, France, June 15-25, 1996. We thank David Gallo and ananonymous reviewer for helpful comments on an earlier version of this manuscript. We also thankBob Cochrane, Michelle Arnold and David Brook for research assistance. Correspondenceconcerning this article should be addressed to Philip A. Higham, Department of Psychology,University of Southampton, Highfield, Southampton, ENGLAND, SO17 1BJ. E-mail:[email protected]. Illusory Recollection46 Footnotes1. We also included experiments in which the long and short prime display durations wereseparated by over 230 ms which yielded very different results from those described in the text(Higham & Vokey, 2000; Experiment 2 and the 16.67/250 ms condition of Experiment 3). However,the methodology of the current experiments was more similar to that used in Experiment 1, so whenwe discuss our previous research on memory illusions induced by prime identification, we are referringto that experiment.2. We also adjusted the K proportions in Experiment 1 by dividing them by one minus R andreanalysed the data. Unlike Experiment 2A, the results of the analyses on the adjusted Kproportions were the same as those for the raw K proportions.3. The VBLINT is the period during which the raster-gun of the video monitor returns from thebottom-right to the top-left of the display without affecting the display, after which it again re-paintsthe display with the contents of video memory. Altering the contents of video memory only during thisinterval synchronises the display of the contents of the video memory with the raster-scan of themonitor.4. Relatively little discussion or debate has centered on the relationship between R judgmentsand recollection; most agree that R judgments directly reflect the “all-or-none” process of recollection.As Jacoby (1998) has stated, “it should be noted that the correspondence between the probabilityof ‘remember’ and R ... remains regardless of the assumption made about the relation betweenconsciously controlled and automatic processes” (p. 18 [emphasis added]). Illusory Recollection47 Table 1Mean Likelihoods of Recognition, R, and K Judgements in Experiment 1 as a Function of PriorPresentation and Prime Duration. Experimental Condition Memory MeasureShort-Old Long-Old Short-New Long-New Recognition.71.73.26.32R.40.41.06.09K.32.31.20.23 Note. R and K means may not sum to the mean for recognition because of rounding error. Illusory Recollection48 Table 2Old/New Discrimination (A’) and Response Bias (B”D) in Experiments 1, 2A and 2B as a Function ofResponse Type and Duration Discrimination (A’)Response Bias (B”D)Memory MeasureShortLongShortLong Experiment 1Recognition.81.79.00-.11K.61.59.75.70R.79.78.89.82 Experiment 2AFamiliarity.67.65.25-.07Recollection.73.74.48.10 Experiment 2BFamiliarity.71.71.29-.15Recollection.76.72.59.10 Illusory Recollection49 Table 3Mean Prime Identification Likelihoods in Experiments 1, 2A and 2B as a Function of PriorPresentation and Prime Duration. Experimental Condition ExperimentShort-Old Long-Old Short-New Long-New Experiment 1.38.90.29.82Experiment 2A.11.87.08.75Experiment 2B.08.84.06.75 Illusory Recollection50 Table 4Mean Familiarity and Recollection Ratings in Experiment 2A and 2B as a Function of PriorPresentation and Prime Duration. Experimental Condition Memory MeasureShort-Old Long-Old Short-New Long-New Experiment 2AFamiliarity2.562.841.982.30Recollection2.532.881.702.04