Can false memory for critical lures occur without conscious awareness of list words?

Consciousness and Cognition xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Consciousness and Cognition journal homepage: www.elsevier.com/locate/concog

Can false memory for critical lures occur without conscious awareness of list words? ⁎

Daniel D. Sadler , Sharon M. Sodmont, Lucas A. Keefer1 Indiana University of Pennsylvania, Department of Psychology, Uhler Hall, 1020 Oakland Avenue, Indiana, PA 15705, United States

AR TI CLE I NF O

AB S T R A CT

Keywords: False memory Semantic associates Critical lures Subliminal processing Conscious awareness

We examined whether the DRM false memory effect can occur when list words are presented below the perceptual identification threshold. In four experiments, subjects showed robust veridical memory for studied words and false memory for critical lures when masked list words were presented at exposure durations of 43 ms per word. Shortening the exposure duration to 29 ms virtually eliminated veridical recognition of studied words and completely eliminated false recognition of critical lures. Subjective visibility ratings in Experiments 3a and 3b support the assumption that words presented at 29 ms were subliminal for most participants, but were occasionally experienced with partial awareness by participants with higher perceptual awareness. Our results indicate that a false memory effect does not occur in the absence of conscious awareness of list words, but it does occur when word stimuli are presented at an intermediate level of visibility.

1. Introduction The study of false memory has grown exponentially since the introduction of the Deese-Roediger-McDermott (DRM) paradigm. The procedure originated with Deese (1959) as a technique for investigating the effects of associative context on intrusions into free recall for wordlists. The technique was later modified by Roediger and McDermott (1995) and has since become the most widely used procedure for studying associative memory errors (Gallo, 2006). The basic method entails testing recall or recognition after presenting a series of wordlists, each consisting of words that are associates of a nonpresented theme word (called the “critical lure”). Roediger and McDermott (1995) found that participants falsely recalled and falsely recognized the critical lures at very high levels. For example, participants often falsely recalled or recognized the critical lure, sleep, after studying a list of 15 associates, including bed, rest, awake, tired, dream, etc. Among several theoretical explanations for the DRM false memory effect, an activation-monitoring account seems to have the broadest acceptance (Roediger & McDermott, 2000). By this account, list words activate their conceptual representations in a semantic memory network and this activation automatically spreads along pathways to related concepts (Collins & Loftus, 1975), including the critical lure (Roediger, Balota, & Watson, 2001). Activation from multiple list items converges on the conceptual representation for the critical lure and summates. The level of activation of the critical lure is a function of the sum of its associative strengths with list words (Roediger, Watson, McDermott, & Gallo, 2001). The greater the activation of the critical lure the more likely it will be falsely recalled or recognized. The activated critical lure may enter conscious awareness during list presentation, and subsequently be falsely recalled when participants misattribute the source of their memory of having thought of the word to its

⁎

Corresponding author. E-mail addresses: [email protected] (D.D. Sadler), [email protected] (L.A. Keefer). 1 Present address: University of Southern Mississippi, Department of Psychology, 118 College Drive #5025, Hattiesburg, MS 39406, United States. https://doi.org/10.1016/j.concog.2017.10.018 Received 7 March 2017; Received in revised form 9 October 2017; Accepted 27 October 2017 1053-8100/ © 2017 Elsevier Inc. All rights reserved.

Please cite this article as: Sadler, D.D., Consciousness and Cognition (2017), https://doi.org/10.1016/j.concog.2017.10.018

Consciousness and Cognition xxx (xxxx) xxx–xxx

D.D. Sadler et al.

having been presented (Johnson, Hashtroudi, & Lindsay, 1993; Roediger & McDermott, 1995). Similarly, false recognition may also result from source-monitoring errors as the familiarity of a critical lure on a recognition test is mistakenly attributed to the word having been presented (Roediger & McDermott, 1995, 2000; Underwood, 1965). An additional tenet of the activation-monitoring theory is that, although activated during study, a critical lure will not necessarily become conscious, but may still be falsely recognized due to its familiarity from residual activation (Roediger & McDermott, 2000; Roediger et al., 2001). The theory also allows for unconscious activation of critical lures in the absence of conscious processing of list items (Roediger et al., 2001), and draws support from studies that have demonstrated unconscious semantic priming using a subliminal priming paradigm (e.g., Balota, 1983; Marcel, 1983). As discussed in the review that follows, the question of whether false memory can arise from unconscious activation of critical lures has been a subject of considerable debate. Attempts to demonstrate false memory from unconscious activation of critical lures have been unable to produce unambiguous evidence. The major limitation of these earlier studies is that they have attempted to demonstrate false memory from unconscious activation of list words by trying to minimize, but not eliminate, conscious awareness of list words during study. As long as list words are being consciously processed, the possibility that critical lures are also consciously processed cannot be excluded. The rationale for the present study is that the subliminal priming paradigm offers the best hope of demonstrating that associative memory errors can arise from unconscious processing of critical lures. Seamon, Luo, and Gallo (1998) examined whether false recognition memory in DRM studies can arise from unconscious activation of the critical lure by creating study conditions that minimized conscious processing of list items. They manipulated conscious awareness of list items by varying two factors: exposure duration of items (2 s, 257 ms, or 20 ms) during list presentation and concurrent memory load (either no load or retention of a single sequence of seven digits throughout the presentation of all study lists). Corrected recognition rates were calculated for studied words by subtracting the false alarm rates for nonstudied list words from the hit rates for studied list words and for critical lures by subtracting the false recognition rates for unrelated critical lures from the false recognition rates for related critical lures. Their results showed a decrease in correct recognition for studied words, but not false recognition of critical lures, as exposure duration decreased from 2 s to 20 ms. At 20 ms, the false recognition rate for critical lures was significantly higher than the correct recognition rate for studied words. While claiming that this finding is consistent with the hypothesis that unconscious activation of critical lures at study can produce a false memory effect, they noted that memory for studied words had been reduced, but not eliminated. They claimed to provide further evidence for the unconscious activation hypothesis from a comparison of subjects classified as good- and poor-memory subjects on the basis of a median split on studied word recognition performance in the 20 ms memory load condition. Whereas good-memory subjects had similar corrected recognition rates for studied words (.22) and critical lures (.27), poor-memory subjects had a significantly higher false recognition rate for critical lures (.20) than their near zero rate for studied words (.03). Seamon et al. (1998) concluded that their results support the view that false recognition of critical lures can be caused by unconscious activation of these items at study. The main support for their conclusion was the finding that poor-memory subjects in the 20 ms memory load condition demonstrated false memory for critical lures while showing virtually no correct recognition of studied words. As an explanation for the obvious question of why participants would forget the studied words that produced the unconscious activation of critical lures, they proposed that, whereas activation of briefly presented studied words may be short lived, activation of a critical lure may be stronger and more persistent due to its repeated activation as an unconscious associative response to the semantically related list words. At test, with their activations having decayed, studied words are not recognized, while the still unconsciously activated critical lures lead to false recognition. The conclusion by Seamon et al. (1998) that false memory for critical lures can arise from unconscious associative responses to list words is based on the assumption that the lack of explicit memory for list words in the poor memory subjects implies that critical lures were not consciously activated during list presentation. However, the evidence for a lack of explicit memory for list words in the poor-memory subjects is questionable due to the problematic use of a median split on correct recognition for list words to define good and poor memory (Zeelenberg, Plomp, & Raaijmakers, 2003). It is likely that this technique capitalizes on chance. When a list of 15 words is presented at the very rapid rate of 20 ms per word, during the 300 ms presentation of the entire list, subjects may be able to consciously identify only a few words. Given that the subset of identified words will vary across subjects, those who, by luck, happen to fairly consistently identify one or more of the three words selected for the recognition test (the first, third, and tenth items in Seamon et al., 1998) can be expected to have higher correct recognition rates at test than those who have worse luck by identifying the same number of words in lists, but words that were less often included among the three test words selected from each list. Thus the lack of evidence for explicit memory in poor-memory subjects may simply be an artifact of the median-split technique and a recognition test that included only 20 percent of list words, and their false memory for critical lures may be due to conscious activation of those words arising from conscious activation of several list words that were not later tested for recognition. The goal of presenting word lists at the rate of 20 ms per item with a 0 ms interstimulus interval (ISI) would seem to be to test whether unconscious semantic priming of critical lures might lead to false memory in the absence of conscious awareness of list words. However, Gallo and Seamon (2004) have stated that the purpose of the 20 ms condition in the Seamon et al. (1998) study was not to eliminate perception of items, but rather to minimize conscious processing of items by making it extremely difficult, and thereby minimize conscious generation of critical lures. This goal is implicitly based on the assumption that there is some minimal level of conscious processing of list words that is sufficient to activate the critical lure, but will not bring it into conscious awareness. Zeelenberg et al. (2003) have questioned the validity of this assumption, arguing that as long as there is any conscious awareness of list items, the possibility that critical lures were consciously generated cannot be ruled out. Their study was designed to determine whether unconscious semantic priming of critical lures would occur when conscious awareness of list words is eliminated. Like Seamon et al. (1998), they used rapid presentation of list items: a 20 ms exposure duration with a 0 ISI. Stimuli were presented using 2

Consciousness and Cognition xxx (xxxx) xxx–xxx

D.D. Sadler et al.

hardware and software that allowed precise millisecond timing accuracy, allowing them to avoid the apparent hardware and software limitations in the Seamon et al. (1998) experiments which likely permitted stimulus displays to exceed the intended 20 ms duration, working against the goal of minimizing conscious processing of the list items. For their slow (2000 ms) exposure duration, the results showed the typical high correct recognition rate for studied words and high false recognition rate for critical lures. However, for the 20 ms condition, recognition rates for studied words and critical lures were quite low, and not significantly different from false alarms to unrelated control stimuli. This finding demonstrates that the false memory effect does not occur when word stimuli are presented below the perceptual identification threshold, and it does not support the view that the effect can arise from unconscious processes, as proposed by Seamon et al. (1998). In response to the finding by Zeelenberg et al. (2003), Gallo and Seamon (2004) argued that, when list items are presented too rapidly to identify, they may not receive the minimal level of processing required to create a false memory either by conscious or unconscious activation of the critical lure. They designed a new task with the goal of identifying specific critical lures that were not processed consciously during list presentation, and then testing for false memory for these items. They used a 20 ms exposure duration for list words, with each word embedded between 80 ms forward and backward masks. To identify critical lures that were not consciously activated during study, they used immediate recall after list presentation, and assumed that any consciously activated critical lure would be included in the recall along with words that were perceived during the list presentation. Following the presentations of all lists, a surprise recognition test was given consisting of pairs of a critical lure related to a studied list and a critical lure related to a nonstudied list. Participants were told to choose which item in each pair was from a studied list, and to guess if necessary. Gallo and Seamon’s results showed that the critical lures from studied lists were chosen at a rate significantly above chance when all pairs were included (.58) and also when only pairs whose critical lure was not recalled during study were included (.57). They concluded the false recognition of the critical lure can occur without it having been consciously activated during study. Unfortunately, the conclusion by Gallo and Seamon (2004) hinges on the validity of the assumption that consciously activated critical lures would have been falsely recalled immediately after list presentation, an assumption that implies the complete failure of source monitoring by which participants would reject these items as nonstudied items (Raaijmakers & Zeelenberg, 2004). Furthermore, although the purpose of the rapid presentation rate was to minimize conscious processing of list words, and thereby minimize conscious activation of critical lures, the use of immediate recall following each list provided the opportunity for extended, elaborative processing of those list words that were perceptually identified. It would not be surprising if, during recall, some critical lures were consciously activated and participants recognized them as nonpresented associates of list words, and therefore did not include them in their recall. A study by Bredart (2000) provides evidence that participants often consciously think of critical lures during study without including them in their recall. After standard recall following each of eight lists, participants were presented with recalled items and asked to write down words that had come to mind during the list presentation, but which they had not written down at the time because they thought the word was not a presented item. For each list, Bredart found that, among participants not recalling a critical lure during the standard first test, a substantial majority recalled the critical lure in the second test. This finding clearly challenges Gallo and Seamon’s (2004) assumption that all consciously activated critical lures will be produced as false recall, allowing them to be excluded from subsequent data analyses to obtain pure measures of false memory for unconsciously activated critical lures. The support for the unconscious activation hypothesis in Gallo and Seamon (2004) also relies on an assumption that studied critical lures in the test pairs were not selected based on their familiarity arising from association with remembered list words (e.g., choosing sleep because you remember bed) (Gallo & Seamon, 2004; Raaijmakers & Zeelenberg, 2004). An alternative explanation is that some, if not all, of a participant’s above chance selection of critical lures from studied rather than nonstudied lists might be based on conscious activation of the critical lures during study or familiarity with recollected test items. In fact, the other major finding by Gallo and Seamon supports this alternative explanation. They found that selection of the critical lure from a studied list occurred only when at least one studied word was recalled from the list. With conscious processing of one or more list words, any false recognition for nonpresented critical lures might be due to conscious activation of the critical lures during study or to their familiarity through association with recollected list words. To provide a stronger test of the hypothesis that false memory for critical lures can arise from unconscious processes, Cotel, Gallo, and Seamon (2008) used essentially the same study condition as Gallo and Seamon (2004), consisting of rapid (40 ms), masked presentation of list words followed by immediate list recall, but revised the test condition such that related critical lures (i.e., related to studied lists) and unrelated critical lures (i.e., related to nonstudied lists) were presented individually using rapid (20 ms), masked presentation. For each test item, participants were asked to identify the word (or respond “none” if they could not), and then rate their belief that the test word was a studied item (regardless of whether they had identified it) on a scale from 1 (definitely not studied) to 10 (definitely studied). Next, for test words not identified, participants were asked again to identify it, guessing if necessary. The perceptual identification test was either immediate (following recall for each list) or delayed (following the presentation of all lists). The strategy of Cotel et al. appears to have been the following. Rapid presentation of list words will result in minimal processing of these items, with the effect of minimizing conscious activation of nonpresented critical lures. Immediate free recall of list words will allow identification and removal of all critical lures that were consciously activated during study, based on the assumption that any consciously activated critical lure will automatically be falsely recalled. Rapid presentation of test items (after excluding critical lures falsely recalled) will avoid the problems of previous studies in which test-based associations between list words and nonpresented critical lures might have influenced test performance for critical lures. The Cotel et al. (2008) experiment produced two important results. First, correct perceptual identification was significantly higher for related than for unrelated critical lures for the immediate test (.58 versus .49) and for the delayed test (.54 versus .47). Cotel et al. attributed the facilitation of perceptual identification of related critical lures to implicit priming from the unconscious activation of 3

Consciousness and Cognition xxx (xxxx) xxx–xxx

D.D. Sadler et al.

these items during study. Unfortunately, this conclusion is based on the assumption that, by excluding critical lures that were falsely recalled following list presentation from the perceptual identification test, the test will measure facilitation solely for words that were unconsciously activated during study. However, as argued above, this technique is unlikely to identify all consciously activated critical lures during study, and those that remain in the set of items used in the analysis of perceptual identification data may be responsible for the small facilitation effect for related critical lures. Their second major result was that, in the immediate test condition, the mean rating on the likely-to-have-been studied scale was higher for related critical lures (5.19) than for unrelated critical lures (3.88). Cotel et al. noted that this finding of a false-recognition-without-identification effect replicated a similar finding by Cleary and Greene (2004), whose experiment provided the recognition-without-identification procedure used in their study. The two studies differed in that Cleary and Greene did not use rapid presentation of list items. Cotel et al. concluded that their finding replicated Cleary and Greene’s finding of false-recognition-without-identification, but also demonstrated that this effect can occur for critical lures that were unconsciously activated during study. However, Cleary and Greene discussed an important limitation of their results, noting that the basis for false-recognition-without-identification in their participants might not be word meaning, but instead could be orthographic information encoded in a memory trace formed as the critical lure is consciously activated during study. Thus, false recognition-without-identification may occur through similarity between orthographic information encoded during study and orthographic information detected during the perceptual identification test. As the foregoing review suggests, the studies by Seamon et al. (1998), Gallo and Seamon (2004), and Cotel et al. (2008) have provided intriguing, but ambiguous, evidence that false memory in DRM studies can occur through unconscious activation of critical lures during study. The logic of using rapid list presentation was based on the questionable assumption that minimizing conscious processing of list words during study would create conditions that virtually eliminate conscious activation of critical lures, while still allowing them to be activated unconsciously. However, as long any list words are consciously processed, the possibility of conscious activation of the critical lure remains. The use of immediate recall to identify and eliminate from analyses those critical lures that are consciously activated (and therefore falsely recalled) ironically creates rehearsal conditions that are likely to increase the probability that the critical lure will be consciously activated during recall, but not falsely recalled due to effective source monitoring decisions. The use of the perceptual identification paradigm to eliminate conscious processing of critical lures at test eliminates test-based associative guessing using lexical information, but introduces another alternative explanation, that of recognition ratings based not on lexical information, but instead on matching orthographic information between study and test. If unambiguous evidence for unconscious processing of critical lures cannot be found under conditions that allow conscious processing of list words, perhaps it can be found under conditions in which conscious awareness of list words is prevented by presenting items just below the perceptual threshold. Although the evidence is not without controversy (Holender, 1986), unconscious semantic priming has been demonstrated in studies that have used a subliminal priming paradigm (e.g., Balota, 1983; Marcel, 1983). Much of the recent debate concerning whether subliminal processing can occur at a semantic level has revolved around interpretations of congruity effects obtained in studies that have manipulated prime-target congruity in a categorization task. Findings by some researchers (e.g., Dehaene et al., 1998; Naccache & Dehaene, 2001) have led them to claim that unconscious processing of subliminal stimuli can include not only semantic access, but manipulation of semantic content as well. For example, Dehaene et al. (1998) used a semantic comparison task in which participants categorized the value of a target number as smaller or larger than 5. On each trial, a subliminal prime (also a number smaller or larger than 5) preceded the target. The results showed slower response times on incongruent trials when the prime and target were from different categories than on congruent trials when they were from the same category. Dehaene et al. concluded that participants were unconsciously applying the task instructions requiring semantic categorization of the visible targets to the unseen primes, resulting in response competition on incongruent trials. Additional support for the conclusion that primes were processed at a semantic level came from the finding that the congruity effect was cross-notational, occurring for both same format prime-target pairs (e.g., 6–1) and different format pairs (e.g., SIX-1). Brain imaging data from both event-related potentials and functional magnetic resonance imaging also supported their conclusion of unconscious semantic processing of primes by showing that target-induced activation in the motor cortex was preceded by primeinduced activation on the same response side on congruent trials, but on the opposite response side on incongruent trials, resulting in response competition. From the combined findings of their behavioral and brain activity measures, Dehaene et al. concluded that subliminal primes are processed through the same stream of perceptual, conceptual, and motor response levels as the conscious target. The conclusions of Dehaene et al. (1998) were challenged by subsequent research supporting nonsemantic interpretations of subliminal congruity effects, including automaticity (Damian, 2001; Logan, 1988) and action-trigger (Kunde, Kiesel, & Hoffmann, 2003) accounts. However, using a number comparison task and an improved design in a replication of Dehaene et al. (1998), Naccache and Dehaene (2001) obtained results that are inconsistent with the automaticity hypothesis, and the action trigger hypothesis has been challenged by Van Opstal, Reynvoet, and Verguts (2005a) and subsequently debated in Kunde, Kiesel, and Hoffmann (2005) and Van Opstal, Reynvoet, and Verguts (2005b). In their review of studies of subliminal congruity effects, Kouider and Dehaene (2007) mention a study by Dell’Acqua and Grainger (1999) that seems to provide unambiguous support for a semantic interpretation of the subliminal congruity effect. Using pictures (line drawings) as subliminal primes, Dell’Acqua and Grainger found a congruity effect when participants categorized target words (names of the picture stimuli) as natural things or artifacts. As discussed by Kouider and Dehaene, this effect is not adequately explained by automaticity because the prime and target stimuli were different (i.e., in pictorial versus verbal format), preventing the development of automatized stimulus-response mappings for primes that also appear as targets. They also noted that it seems implausible that participants, influenced by task instructions and context, could have generated action triggers for the large set of 84 concepts used as stimuli, particularly given that they had no knowledge of the stimuli prior to their presentation during experimental trials, other than they would be shown instances of the broad categories of natural 4

Consciousness and Cognition xxx (xxxx) xxx–xxx

D.D. Sadler et al.

things and artifacts. Assuming that subliminal stimuli can be processed semantically, there is the additional question of whether their effects would persist long enough to influence performance on the delayed episodic recognition test used in the present study. Unconscious semantic priming effects are generally thought to be short-lived (Dehaene, Changeux, Naccache, Sackur, & Sergent, 2006; Greenwald, Draine, & Abrams, 1996), producing activation in lexical representations that influences processing of targets presented at short SOAs, but decaying rapidly without leaving an episodic memory trace (Balota, 1983; Forster & Davis, 1984; Raaijmakers & Neville, 2015). Findings by Meade, Watson, Balota, and Roediger (2007) suggest that false recognition of critical lures in a DRM study arises from reactivation during retrieval of episodic associative structures encoded during list presentation rather than from the activation of the critical lure during list presentation persisting into the test phase. In their experiments, following each study list participants either performed a lexical decision task (LDT) or a speeded recognition task. Presentation of the test list, identical for the two tasks, began 1 s after the last study item, with the critical lure appearing at position 1, 3, 6, or 11 in the list. For the LDT, Meade et al. found a priming effect for critical lures only when they appeared as the first test item, but they obtained a false memory effect at all test positions. Meade et al. proposed an updated activation/monitoring account of their findings that more precisely specifies the role of activation during the encoding and retrieval stages of a DRM task. During the study phase, the semantic associates on a DRM list activate their conceptual representations and this activation automatically spreads to related concepts, including the critical lure. The output from this processing is a well-integrated associative network, potentially including the critical lure, which is retained in episodic memory. However, the activation rapidly decays once attention is withdrawn from the network, becoming too weak to produce a priming effect on the critical lure beyond a short delay following the last list item. Although activation induced by processing the study list has dissipated before the test phase begins, each critical lure on the recognition test functions as a probe that reactivates the relevant associative structure encoded during the study phase, recreating the heightened activation that accumulated during list presentation, and providing information on which a recognition judgment is based, with false recognition depending on the outcome of the source monitoring process. Meade et al. note that reactivation of these episodic associative structures requires that the participant be in episodic retrieval mode (Tulving, 1983), a state that is unlikely for the LDT because it is not essential to performing the task. If the DRM false memory effect is attributable to reactivation of episodic associative networks (Meade et al., 2007), but episodic memory traces are not formed for subliminal stimuli (Raaijmakers & Neville, 2015), then there would be no possibility in the present study of obtaining a subliminal false memory effect. However, research by Reber and Henke (2011) challenges the notion that consciousness is essential for the formation of episodic memories. They examined encoding of pairs of unrelated words (e.g., rainscrew, coffee-tango) presented subliminally. The subliminal encoding trial for each word pair consisted of a series of 12 presentations of the stimuli, with each presentation embedded between pattern masks. To encourage sustained attention for the duration of each 6 s encoding trial, the participant performed an attention task consisting of responding whenever a fixation cross, which appeared once per second, was replaced by a vertical or horizontal bar. The encoding phase was followed by a 5 min rest period before beginning the retrieval phase, which consisted of making semantic fit judgments for supraliminal stimuli consisting of experimental word pairs that were “analogs” composed of semantic neighbors for the words in an encoding word pair (e.g., snow-nail was the analog for rain-screw) and control word pairs that were “broken analogs” of the encoding word pairs (e.g., hail-waltz). The researchers also manipulated awareness of task structure by using a suprathreshold version of the task in which encoding word pairs (not used in the subliminal task) were presented supraliminally for 3.5 s and participants were asked to judge the semantic fit of the two words using a loose response criterion (e.g., for the word pair cow-grill, although an association is not immediately obvious, cow and grill can be viewed as associated based on the semantic relation between beef and barbecue). Participants performed the suprathreshold version of the task either before or after the subliminal version. The results for the suprathreshold version showed more “yes” semantic fit judgments for analogs than for broken analogs. The same effect occurred for the subliminal version, but only for the condition in which participants performed the suprathreshold version before the subliminal version. The dependence of the effect on awareness of task structure was also evident in the results for their second experiment: all 36 participants performed the suprathreshold version before the subliminal version, but only the 15 who demonstrated insight into the task structure (i.e., the conceptual connection between encoding pairs and their retrieval pair analogs) showed a higher rate of semantic fit judgements for analogs than for broken analogs. In both experiments, an objective measure of visibility showed chance discrimination of subliminal stimuli. Reber and Henke’s results demonstrate unconscious formation of episodic memories consisting of semantic associations for subliminal word pairs (rain-screw) and the unconscious influence of these associations (via unconscious reactivation at retrieval) on semantic distance computations for analog word pairs (snow-nail). This effect depended crucially on the participant’s understanding of the conceptual analogy between encoding word pairs and their retrieval analogs; however, once acquired, this task-set appears to have unconsciously guided encoding and retrieval processes for the subliminal stimuli. The goal of the present study was to determine whether the false memory effect can occur through unconscious activation of critical lures. The rationale for Experiments 1 and 2 assumes the existence of unconscious semantic processing of subliminal stimuli. Within- and between-subjects designs were used to test for a DRM false memory effect following subliminal and supraliminal presentation of lists of semantic associates. Both experiments produced results indicating strong veridical and false memory when wordlists were presented just above the perceptual threshold, but veridical memory virtually disappeared and false memory vanished when wordlists were presented below the perceptual threshold. Experiments 3a and 3b were conducted to obtain visibility data to test the assumption that the subliminal wordlist presentations in Experiments 1 and 2 were truly subliminal. These experiments replicated the within-subjects design of Experiment 1 with the addition of subjective visibility ratings integrated into the encoding task during wordlist presentation.

5

Consciousness and Cognition xxx (xxxx) xxx–xxx

D.D. Sadler et al.

2. Experiment 1 We examined the effects of word exposure duration on veridical and false recognition within three word display conditions: no mask displays, mask displays, and no mask-no delay displays. Because veridical and false recognition at long exposure durations have been well-demonstrated in prior research, we used shorter durations of 257 ms, 43 ms, and 29 ms. In the 257 ms condition we expected to find robust veridical recognition of studied words and false recognition of critical lures, replicating the finding of Seamon et al. (1998). The main question of interest was whether the shorter exposure durations with masked list words would provide evidence for false recognition through unconscious priming of critical lures. 2.1. Method 2.1.1. Participants The participants were 108 students attending Indiana University of Pennsylvania who received credit towards a general psychology course requirement for their participation. 2.1.2. Materials We used the 36 15-item wordlists provided in Stadler, Roediger, and McDermott (1999). In their study, Stadler et al. ranked wordlists by probability of false recognition of the critical lure. From these wordlists, three sets of 24 lists were created in the following way. We formed eight blocks of three consecutively ranked lists, and then randomly assigned lists within each block such that each list appeared in two of the three sets of lists. Within each set of 24 wordlists, we formed three subsets of 8 lists, equating the subsets on average probability of false recognition. The 15 words in each list were ordered from most to least semantically related to the critical lure. We obtained three additional wordlists to use as practice lists from Watson, Balota, and Roediger (2003). The recognition test had the same structure as tests used in prior research (Roediger & McDermott, 1995; Seamon et al., 1998). There were 144 items in the recognition test: 72 old and 72 new. For each participant, the new items included 36 words from nonstudied list words and 36 critical lures (24 related to studied lists, 12 to nonstudied lists). All old and new list items came from the 1st, 8th, and 10th positions of the lists. For each word, subjects had to determine if the word was old or new (never seen). Remember/ Know judgments were made for each word judged as old, with Remember used for a vivid memory of the word’s presentation and Know used when the word was believed to have been presented, but there was no recollection of the presentation. Each test item had 3 response options: 1 = new word, 2 = old remember, and 3 = old know. The items were randomly ordered. The recognition test following the practice lists also included the list items from the 1st, 8th, and 10th positions of each studied list and from each of three nonstudied lists, but did not include the critical lures because we wanted to avoid the possibility that participants might speculate on the purpose of these words and the effect such speculation might have on how they would study the experimental lists. 2.1.3. Design and procedure Thirty-six participants were run in each of three different word display conditions: no mask, mask, and no mask-no delay. Participants were selected from a randomized list of students in a general psychology subject pool. They were contacted by phone and scheduled for individual sessions based on their availability. Fig. 1 presents the event sequence for a single wordlist presentation for each word display condition. Word exposure duration was a within-subjects variable with three levels: 29 ms, 43 ms, and 257 ms. For the no-mask and mask conditions, the display interval in which each list word appeared was 1500 ms, replicating the word presentation rate of 1.5 s for the spoken wordlists used in Roediger and McDermott (1995). We used a constant stimulus onset asynchrony (SOA), the sum of exposure duration and interstimulus interval (ISI), to provide equivalent time for automatic activation of list words and spread of activation to critical lures at all levels of exposure duration. Our rationale for holding SOA constant is supported by prior research presented below in the discussion of results for the no mask condition. It should be noted that, by holding SOA constant, we allowed interstimulus interval (ISI) to vary inversely with word exposure duration in the no mask and mask conditions, which precludes any definitive conclusions on the relative effects of these two variables on recognition performance. However, our goal was not to determine the separate contributions of duration and ISI to veridical and false memory, but rather to determine the effects of varying the visibility of studied items on recognition by presenting them above and below the perceptual threshold under conditions that allowed equivalent time for automatic spread of activation. The purpose of including the no mask condition was to provide data under conditions of high visibility for comparison to the mask condition in which visibility was varied. The visibility manipulation was achieved by masking word stimuli while varying exposure duration. The expected effect of this manipulation was that less visible stimuli would have lower veridical recognition. Given that this visibility effect is more likely a function of the decrease in word exposure duration than the corresponding increase in ISI, it seems appropriate to simplify our terminology in reporting and discussing results by referring to “effects of exposure duration” rather than using the more complex term, “effects of exposure duration/ISI.” For the no-mask condition, a list word appeared at the beginning of the display interval, followed by a blank screen at the end of its exposure duration for the remainder of the interval. In the mask condition, the list word appeared between forward (200 ms) and backward (257 ms) masks, with the interval beginning with the forward mask and a blank screen following the backward mask for the remainder of the display interval. The 200 ms forward mask for each successive word in the list is the final 200 ms of the 1500 ms SOA for consecutive words. Both masks were an alternating sequence of eight pound signs and eight ampersands (#&#&#&#&#&#&#&#&). In the no mask-no delay condition, the word-to-word ISI was 0 ms. Within each word display condition, each participant was presented with the 24 lists in one of the three wordlist sets, each comprising 6

Consciousness and Cognition xxx (xxxx) xxx–xxx

D.D. Sadler et al.

No Mask 5000 ms

No Mask-No Delay

Mask

End of list. Wait for instructions.

5000 ms

End of list. Wait for instructions.

5000 ms

End of list. Wait for instructions.

Repeat until end of wordlist 1471, 1457, or 1243 ms

1014, 1000, or 786 ms

29, 43, or 257 ms

29, 43, or 257 ms bed

1000 ms 500 ms

#&#&#&

257 ms

rest

1471, 1457, or 1243 ms 29, 43, or 257 ms

29, 43, or 257 ms

200 ms

bed #&#&#&

500 ms

29, 43, or 257 ms

awake

29, 43, or 257 ms 29, 43, or 257 ms

rest bed

1000 ms

1000 ms

******

tired

500 ms

******

******

Fig. 1. Sequence of events for a single wordlist presentation in Experiment 1. Type of word display (no mask, mask, no mask-no delay) was a between-subjects variable and word exposure duration was a within-subjects variable. Eight wordlists were presented at each exposure duration. In the mask and no mask conditions, each word appeared within a 1500 ms stimulus presentation interval. The blank screen duration following each word presentation equals 1500 minus the word duration in the no mask condition and 1500 minus the sum of the word and mask durations in the mask condition. In Experiment 2, word stimuli were masked and exposure duration was a four-level between-subjects variable with the 4th level consisting of a 14 ms exposure duration and a 1029 ms blank screen for the remainder of the stimulus presentation interval. In a fifth condition, participants received the no mask-no delay display at 29 ms. Twelve wordlists were presented to each participant. Experiment 3a and 3b were replications of the Experiment 1 mask condition with two changes: (1) in Experiment 3b, a 57 ms exposure duration replaced the 257 ms duration (with a 986 ms blank screen for the remainder of the stimulus presentation interval), and (2) in both experiments, participants entered a subjective visibility rating after each word stimulus.

three wordlist subsets presented in one of two subset orders: Subset 1, Subset 2, Subset 3 or Subset 2, Subset 3, Subset 1, but not Subset 3, Subset 1, Subset 2. Complete counterbalancing of the order of the three word exposure durations was used for each wordlist subset order, with one of the six permutations of exposure duration used for each participant. Thus, one participant was run in each of the 36 combinations of three wordlist sets, two wordlist subset orders, and six word exposure duration orders, with each wordlist subset being used an equal number of times at each word exposure duration. All eight wordlists in each wordlist subset were presented at the duration assigned to that subset before presenting wordlists in the subset for the next duration. Table 1 presents a summary of the design. As part of the task instructions, participants were told that it would be difficult to see the words, were asked to pay close attention, and were informed that their memory would be tested. The experimenter waited 5 s between lists before telling participants to continue. Following the practice lists, participants were given a free recall test. The recall test was intended to increase attention and motivation during the study of the experimental lists above the level participants might make if they thought their memory would be Table 1 Word exposure duration (ms) for each combination of wordlist subset order, wordlist subset, and exposure duration order. Wordlist subset order and wordlist subset (WS) Subset order 1

Subset order 2

Exposure duration order

WS1

WS2

WS3

WS2

WS3

WS1

1 2 3 4 5 6

29 29 43 43 257 257

43 257 257 29 29 43

257 43 29 257 43 29

29 29 43 43 257 257

43 257 257 29 29 43

257 43 29 257 43 29

Note. This design was used for each of the three sets of 24 wordlists used in each word display condition (no-mask, mask, and no mask-no delay). For each wordlist set, a single participant was assigned to each combination of wordlist subset order and exposure duration order. All eight wordlists in each wordlist subset were presented at the duration assigned to that subset before presenting wordlists in the subset for the next duration.

7

Consciousness and Cognition xxx (xxxx) xxx–xxx

D.D. Sadler et al.

assessed only by a recognition test. Following the recall test, instructions for the recognition test were presented, including the explanation of Remember versus Know judgments for items classified as old, and a request to not make wild guesses. The experimenter scored the practice recognition test, and provided feedback to the participant on new items. If the number of correct new responses was seven or more, the participant was told that s/he did as well as we were expecting and to continue with the same approach to judging whether the items were new or old. If the number correct was 6 or less, the participant was told that s/he had scored a little lower than we were expecting and was asked to be careful not to guess on the experimental lists. 2.1.4. Apparatus The program to run the experiment was written in Turbo Pascal for DOS (Version 7.0) and run on Dell Dimension T500 computer systems in real-mode MS-DOS using Dell Ultra Scan P991 monitors with the vertical refresh rates set to 70 Hz. Source code for a millisecond timer was obtained from Hamm (2001). 2.2. Results and discussion The mean old response rates are presented in Table 2 along with the mean proportion of responses classified as old-remember or old-know. Corrected recognition rates (Pr) (Feenan & Snodgrass, 1990) are presented in Table 3. For studied words, the values of Pr Table 2 Experiment 1: Mean recognition rates. Word exposure duration Word display and item type No mask List words Studied Non-studied Critical lures Related Unrelated Mask List words Studied Non-studied Critical lures Related Unrelated No Mask-no delay List Words Studied Non-studied Critical lures Related Unrelated

29 ms (R/K)

43 ms (R/K)

257 ms (R/K)

.65 (.40/.25) .13 (.03/.10)

.59 (.38/.21) .16 (.04/.12)

.63 (.43/.20) .13 (.04/.09)

.67 (.38/.29) .19 (.03/.15)

.61 (.39/.22) .18 (.03/.15)

.58 (.31/.27) .19 (.02/.17)

.22 (.08/.14) .17 (.06/.11)

.61 (.32/.29) .15 (.05/.10)

.65 (.38/.27) .16 (.05/.11)

.26 (.08/.17) .18 (.08/.10)

.65 (.33/.32) .22 (.08/.14)

.66 (.37/.29) .14 (.03/.10)

.30 (.14/.16) .22 (.10/.12)

.34 (.14/.20) .25 (.08/.16)

.59 (.36/.22) .27 (.12/.15)

.33 (.15/.19) .24 (.09/.15)

.42 (.20/.22) .28 (.12/.17)

.74 (.48/.26) .35 (.17/.19)

Note. R = Old remember judgment and K = Old know judgment.

Table 3 Experiment 1: Single-sample t tests for corrected recognition rates (Pr). Word exposure duration 29 ms Word display and item type No mask Studied words Critical lures Mask Studied words Critical lures No mask-no delay Studied words Critical lures

43 ms

257 ms

M

SE

t

M

SE

t

M

SE

t

.52 .48

.031 .047

16.70*** 10.17***

.43 .43

.036 .047

12.11*** 9.16***

.49 .39

.036 .043

13.65*** 9.15***

.05 .08

.024 .048

2.12* 1.60a

.46 .43

.041 .044

11.27*** 9.89***

.49 .52

.032 .043

15.31*** 12.09***

.07 .09

.022 .037

3.17** 2.45**

.09 .13

.033 .040

2.82** 3.26**

.32 .38

.030 .052

10.67*** 7.29***

Note. df = 35 for all t tests (one-tailed). * p < .05. ** p < .01. *** p < .001. a p = .060.

8

Consciousness and Cognition xxx (xxxx) xxx–xxx

D.D. Sadler et al.

were calculated by subtracting the recognition rate for nonstudied words from the rate for studied words and for critical lures by subtracting the recognition rate for unrelated critical lures from the rate for related critical lures. The values of Pr provide measures of veridical and false recognition after correcting for guessing at each word exposure duration, and they were tested for significance above a rate of zero representing no memory and tested for differences between the levels of word exposure duration. Partial omega squared (ωp2) was used to estimate the size of an effect of duration on recognition rates (Keppel & Wickens, 2004; Lakens, 2013). The standardized mean difference for paired observations (Cohen’s dz) was used as a measure of effect size for pairwise comparisons (Lakens, 2013; Rosenthal, 1991). 2.2.1. No mask condition For the no-mask word display condition, the test for an effect of word exposure duration on corrected recognition scores was nonsignificant for studied words, F(2, 70) = 2.58, MSE = .026, p = .083, and nonsignificant for critical lures, F(2, 70) = 1.34, MSE = .051, p > .25. The results of one-tailed single-sample t tests to determine whether the Pr for each item type at each word exposure duration was significantly above zero are presented in Table 3. The results show that for each item type, veridical memory for studied words and false memory for critical lures were present at all word exposure durations. These results showing equally robust veridical memory for studied words and false memory for critical lures at all levels of exposure duration provide baseline rates by which to judge the effects of masking or rapid presentation on veridical and false memory. Our findings are somewhat at odds with those of McDermott and Watson (2001) who found that veridical and false recall increased with a change in exposure duration from 33 ms to 250 ms and those from an unpublished study by Roediger, Robinson, and Balota (as cited in McDermott & Watson, 2001) who also found increasing veridical and false memory across several presentation durations varying from 33 ms to 333 ms2. Similarly, Arndt and Hirshman (1998) found increases in veridical and false recognition with increases in word presentation duration. McDermott and Watson attributed their increase in veridical recall to the extraction of more semantic information during longer presentation durations and the increase in false recall to the spread of activation from list words to the critical lures, with longer durations leading to a greater accumulation of semantic activation over time. Although they used an interstimulus interval (ISI) of 33 ms, giving SOAs of 67 ms for the 33 ms duration and 283 for the 250 ms duration, because the ISI was constant, the increase in veridical and false recall appears to be due to the presentation duration. However, prior research suggests that SOA, rather than presentation duration, is the important determinant of the time course of semantic activation. For example, in their second experiment, Smith and Kimball (2012) factorially manipulated presentation duration (33, 50, or 67 ms) and ISI (0 or 33 ms). Both veridical and false recall increased significantly as SOA (duration + ISI) increased. More importantly for the present discussion, there was no significant difference in either veridical or false recall for the two conditions that had equivalent SOAs of 67 ms (one having a 67 ms presentation duration combined with a 0 ms ISI and the other having 33 ms presentation duration combined with a 33 ms ISI). One implication of this result is that greater activation of semantic information occurs for longer SOAs, whether or not the stimulus remains on throughout the entire SOA. Smith and Kimball’s result parallels an earlier finding by Posner and Snyder (1975). Using five values of prime to target SOA varying from 10 to 510 ms, they found no difference in the time course of facilitation as a function of whether the prime exposure duration lasted the full SOA (e.g., 510 ms prime duration, 0 ms ISI) or it lasted for only the first 10 ms of the SOA (e.g., 10 ms prime duration, 500 ms ISI). It appears that longer SOAs allow for greater extraction of semantic information and accumulation of semantic activation, with perhaps no advantage of longer exposure duration beyond the brief duration needed for initial activation of list words. In the present study, we held our SOA constant at 1500 ms to allow for equivalent activation time at all three levels of exposure duration, and our finding for the no-mask word display condition is what we predicted, equivalent veridical and false recognition for all levels of exposure duration. 2.2.2. Mask condition For the mask condition, there was a significant effect of word exposure duration on corrected recognition rates for both studied words and critical lures, F(2, 70) = 72.16, MSE = .030, p < .001, ωp2 = .569, and F(2, 70) = 29.66, MSE = .068, p < .001, ωp2 = .347, respectively. For studied words, the small difference between the 257 ms and 43 ms exposure durations was nonsignificant, t(35) = 0.72, SE = .040, p > .25, dz = 0.12, and both durations showed significantly higher veridical recognition than did the 29 ms duration, t(35) = 11.15, SE = .039, p < .001, dz = 1.86 and t(35) = 9.59, SE = .042, p < .001, dz = 1.60, respectively. As shown in Table 3, the mean Pr values for studied words were significantly greater than zero at all exposure durations, including the small rate of .05 for the 29 ms duration. For critical lures, the corrected false recognition rates were significantly higher at 257 and 43 ms than at 29 ms, t(35) = 6.64, SE = .067, p < .001, dz = 1.11 and t(35) = 5.96, SE = .059, p < .001, dz = 0.99, respectively, but did not differ from each other, t(35) = 1.66, SE = .057, p > .10, dz = 0.28. As reported in Table 3, the false recognition rate was significantly greater than zero at 257 ms and 43 ms, but nonsignificant at 29 ms. In contrast to the robust veridical memory found at all exposure durations in the no mask condition, the results for masked presentation of list words showed robust veridical memory at 257 ms and 43 ms, with a dramatic falloff at 29 ms. Corresponding to the strong veridical memory at the 257 ms and 43 ms durations were strong false memory effects at these durations. Remarkably, the small 14 ms decrease in exposure duration from 43 ms to 29 ms appears to have dropped the presentation of masked stimuli below the perceptual threshold of the average participant, resulting in the virtual disappearance of veridical memory and the absence of false 2 Following Smith and Kimball (2012), we are using the actual presentation durations calculated from the monitor refresh rate (60 Hz) provided by McDermott and Watson (2001), rather than the nominal durations they used to label their presentation conditions. For example, 33 ms (or more precisely, 33.33 ms) is the actual duration for the 20 ms nominal duration.

9

Consciousness and Cognition xxx (xxxx) xxx–xxx

D.D. Sadler et al.

memory. Although veridical memory was not fully eliminated at 29 ms, the sharp drop to a 5% corrected recognition rate is consistent with the goal of subliminal presentation of words lists, and subjective visibility data supporting this conclusion is presented in Experiments 3a and 3b. Assuming that the sharp increase in veridical recognition from the 29 to 43 ms duration is attributable to a corresponding increase in participant awareness at the longer duration, the exact nature of this awareness at 43 ms is unclear. However, the visibility ratings obtained in Experiments 3a and 3b also suggest that, at 43 ms duration, the masked stimuli were experienced most often at intermediate levels of awareness (Windey, Vermeiren, Atas, & Cleeremans, 2014). There are at least two possible sources of the small amount of veridical recognition at 29 ms. Given that perceptual thresholds vary across individuals (Cheesman & Merikle, 1984) it would not be surprising if some participants with below average thresholds experienced some degree of conscious awareness for some stimuli, resulting in the small amount of veridical recognition observed at 29 ms. The ideas of partial awareness (Kouider & Dupoux, 2004) and individual differences in perceptual thresholds are discussed in greater depth in the background for Experiments 3a and 3b. An alternative explanation is that the small amount of veridical recognition at the 29 ms duration might be the result of subliminal processing that left memory traces for studied words that are activated during the recognition test. This possibility is consistent with the previously mentioned findings by Reber and Henke (2011) suggesting that episodic memories can form through subliminal processing. However, a different conclusion was reached by Raaijmakers and Neville (2015), who found a long-term priming effect for subliminal word stimuli on an implicit (perceptual identification) memory test, but no effect on an explicit (forced-choice recognition) memory test. They speculated that subliminal stimuli update long-term memory by adding contextual features to pre-exiting lexical-semantic memory traces, providing the basis for repetition priming effects, but not by creating new episodic memory traces that could be retrieved while performing an explicit memory test. Regardless of the nature of the processing that underlies the small amount of veridical memory at 29 ms, one straightforward conclusion can be made: to the extent that subliminal semantic processing of list words did in fact occur at this duration, the absence of false recognition of critical lures does not support a view that unconscious activation of these words can spread and converge on the critical lure, leading to a well-integrated associative structure that is stored in episodic memory and then reactivated by the critical lure when it appears as a test item (Meade et al., 2007). 2.2.3. No mask-no delay condition For the no mask-no delay condition, both studied words and critical lures showed significant effects of word exposure duration on corrected recognition scores, F(2, 70) = 27.34, MSE = .025, p < .001, ωp2 = .328 and F(2, 70) = 11.41, MSE = .078, p < .001, ωp2 = .162, respectively. For studied words, mean veridical recognition was significantly higher for the 257 ms word exposure duration than for both the 43 ms and 29 ms durations, t(35) = 6.06, SE = .037, p < .001, dz = 1.01 and t(35) = 6.83, SE = .036, p < .001, dz = 1.14, respectively, which did not differ significantly from each other, t(35) = 0.58, SE = .038, p > .25, dz = 0.10. As shown Table 3, significant Pr values indicate that veridical recognition occurred at all three word exposure durations. The results of pairwise comparisons for critical lures paralleled those for studied words: mean false recognition was significantly greater at the 257 ms word exposure duration than at durations of 43 ms and 29 ms, t(35) = 3.42, SE = .073, p < .01, dz = 0.74 and t(35) = 4.45, SE = .065, p < .001, dz = 0.57, respectively, which did not differ significantly from each other, t(35) = 0.71, SE = .059, p > .25, dz = 0.12. As reported in Table 3, significant Pr values show a false memory effect at all three exposure durations. For the no mask-no delay condition, our corrected veridical and false recognition rates of .32 and .38, respectively, for the 257 ms word exposure duration are comparable to those of Seamon et al. (1998) for their 250 ms duration (.44 and .43), with both studies showing strong veridical and false memory at this duration. However, our veridical and false recognition rates for the 29 ms duration differ substantially from the same rates for the Seamon et al. 20 ms duration (which, as noted by Zeelenberg et al. (2003), was in all likelihood longer, possibly 33 ms, assuming their monitor had a 60 Hz refresh rate). Whereas our corrected veridical and false recognition rates of .07 and .09, respectively, were both significantly lower at 29 ms than at 257 ms, those reported by Seamon et al. were much larger, .20 and .41, and only their veridical recognition rate was significantly lower at 20 ms than at 250 ms. Seamon et al. interpreted the findings for their 20 ms duration as evidence demonstrating that false memory can occur through unconscious activation of the critical lure, which they claimed can be inferred under presentation conditions that permit only minimal conscious processing of list words. Alternatives to the conclusions of Seamon et al. (1998) were discussed in the Introduction. Adding to that discussion, we believe our results for the no mask-no delay condition demonstrate the limitations of combining brief stimulus exposure durations and rapid presentation to create conditions for examining whether unconscious activation of critical lures can occur during list presentation. At the 43 ms exposure duration, our results for the no mask-no delay condition show sharp declines from the robust veridical and false recognition found at this duration for the no mask and mask conditions. Because the exposure duration is the same, and because the mask does not prevent conscious awareness of list words at this duration, the large difference between the mask and no mask-no delay conditions is necessarily related to their different SOAs. For the mask condition, the total SOA (including the forward and backward masks) is 1500 ms, which allows ample time for semantic access of list words and the spread of their activation to the critical lure. However, for the no mask-no delay condition, the back-to-back presentation of displays with a 0 ISI effectively results in list words being embedded in forward and backward masks consisting of other list words. The effect of list words masking list words on semantic access and automatic spread of activation is uncertain, but it seems plausible that such conditions might disrupt these processes in ways that differ from the effect of a nonword mask composed of nonletter symbols (# and &). The rapid presentation of unmasked stimuli used in Seamon et al. (1998) is also problematic because successive words are often much shorter than their preceding words (e.g., the word ill follows physician in the doctor list from Stadler et al., 1999) and the last list word is not masked

10

Consciousness and Cognition xxx (xxxx) xxx–xxx

D.D. Sadler et al.

(e.g., cure). It is perhaps impossible to know the combined effects of presenting very brief displays of words that double as both stimuli and masks at 0 ISI on the conscious and unconscious processes involved in the activation of list words and the spread of that activation. 3. Experiment 2 Zeelenberg et al. (2003) considered the possibility that participants in the within-subjects design of their first experiment might have used a high criterion for old responses, and thereby not responded old to critical lures that had a lower degree of familiarity that was insufficient to exceed a high criterion. They conducted a second experiment using a between-subjects design, reasoning that participants who experience a single short duration, rather than a mix of slower and faster durations, might adopt a lower old response criterion, allowing for a greater frequency of false recognition of critical lures. Experiment 2 examined this possibility and, in addition, used an additional word exposure duration of 14 ms. 3.1. Method 3.1.1. Participants The participants were 111 students attending Indiana University of Pennsylvania who received credit towards a general psychology course requirement for their participation. 3.1.2. Materials Eighteen wordlists from Stadler et al. (1999) were used to create three sets of 12 wordlists that were equated on average probability of false recognition for critical lures. The 15 words in each list were ordered from most to least semantically related to the critical lure. There were 72 items in the recognition test, 36 old and 36 new. The new items included 18 nonstudied list words and 18 critical lures (12 related to studied lists, 6 to nonstudied lists). All list items came from the 1st, 8th, and 10th positions of the lists. For each word, subjects had to determine if the word was new (never seen) or old. Remember/Know judgments were made for each word judged as old, with Remember used for a vivid memory of the word’s presentation and Know used when the word was believed to have been presented, but there was no recollection of the presentation. 3.1.3. Design and procedure Word exposure duration was a between-subjects variable with four levels: 14 ms (n = 23), 29 ms (n = 22), 43 ms (n = 23), and 257 ms (n = 22). The forward and backward masks and display interval were the same as those in the mask condition of Experiment 1. We also included a no mask-no delay condition (n = 21) with a 29 ms word exposure duration and 0 ms ISI as a replication of the 20 ms exposure duration in Seamon et al. (1998). The experimenter waited 5 s between lists before telling the participant to continue. The instructions for the recognition test were the same as those used in Experiment 1. Unlike the first experiment, the practice lists were followed by a recognition test, but not a recall test. 3.2. Results and discussion The mean proportion old responses appear in Table 4 and the mean Pr values are presented in Table 5. For the masked conditions, the effect of exposure duration on veridical recognition of studied words was significant, F(3, 86) = 34.05, MSE = .033, p < .001. Pairwise comparisons using Bonferroni t tests (α = .0167, one-tailed) showed a nonsignificant increase in veridical recognition from 14 to 29 ms, t(43) = 0.72, SE = .033, p > .25, d = 0.22, but significant increases from 29 to 43 ms and 43 to 257 ms, t(43) = 4.59, SE = .059, p < .001, d = 1.37 and t(43) = 2.43, SE = .068, p < .01, d = 0.73, respectively. The effect of exposure duration on false recognition of critical lures was significant, F(3, 86) = 32.74, MSE = .064, p < .001. Table 4 Experiment 2: Mean recognition rates. Condition and Word exposure duration Mask

No Mask-No Delay

Item type

14 ms (R/K)

29 ms (R/K)

43 ms (R/K)

257 ms (R/K)

29 ms (R/K)

List words Studied Non-studied

.21 (.08/.13) .18 (.05/.14)

.32 (.12/.20) .26 (.07/.19)

.55 (.24/.31) .23 (.12/.10)

.71 (.44/.27) .22 (.10/.12)

.34 (.16/.18) .29 (.14/.15)

Critical lures Related Unrelated

.22 (.07/.15) .20 (.08/.12)

.33 (.13/.20) .37(.13/.24)

.67 (.29/.38) .17 (.07/.10)

.74 (.44/.31) .21 (.10/.12)

.41 (.14/.26) .31 (.11/.20)

Note. R = Old remember judgment and K = Old know judgment.

11

Consciousness and Cognition xxx (xxxx) xxx–xxx

D.D. Sadler et al.

Table 5 Experiment 2: Single-sample t tests for corrected recognition rates (Pr). Studied words Word exposure duration 14 ms – Mask 29 ms – Mask 43 ms – Mask 257 ms – Mask 29 ms – No mask-No delay

M .03 .05 .33 .49 .05

Critical lures SE

t *

.015 .031 .052 .045 .033

2.05 1.78* 6.32*** 11.08*** 1.51a

M

SE

t

.02 −.04 .50 .53 .10

.033 .056 .057 .064 .050

0.64 −0.79 8.79*** 8.30*** 1.93*

Note. t tests are one-tailed. df = 22 for 14 ms and 43 ms mask; df = 21 for 29 ms and 257 ms mask; df = 20 for 29 ms no mask-no delay. * p < .05. *** p < .001. a p = .073.

Pairwise comparisons showed a nonsignificant difference between false recognition rates for 14 and 29 ms, t(43) = −1.01, SE = .064, p = .158, d = 0.30, a significant increase from 29 to 43 ms, t(43) = 6.82, SE = .080, p < .001, d = 2.04, and a nonsignificant increase from 43 to 257 ms, t(43) = 0.32, SE = .086, p > .25, d = 0.10. The results for the 257 ms word exposure duration are typical of DRM studies: a high correct recognition rate for studied words and an equally high false recognition rate for critical lures. The significant drop in correct recognition from the 257 ms to the 43 ms exposure duration suggests a reduced capacity to consciously identify list words. Nevertheless, the false memory effect for the 43 ms duration was robust, comparable to that of the 257 ms duration. The near zero recognition rates for studied words for the 29 ms and 14 ms exposure durations indicate that conscious awareness of list items was virtually eliminated in these two conditions. Two possible explanations for the source of the small amount of veridical recognition at 29 and 14 ms were previously discussed in relation to the similarly small effect for 29 ms in Experiment 1. Although both veridical recognition rates were significantly above zero, whatever the level of awareness of list items, it must have been insufficient to prime critical lures, as the false recognition rates for both the 29 ms and 14 ms durations did not differ from zero.

4. Experiments 3a and 3b Given the very low (5%) veridical recognition rates for studied items, one might speculate that subliminal semantic processing occurred for some word stimuli, contributing to a sense of familiarity for those items that subsequently appeared on the recognition test. However, we believe that a more plausible explanation for the occasional correct recognition of 29 ms masked stimuli is that some form of partial awareness occurred for some list items, including those later recognized, perhaps due to imperfect masking. Kouider and Dupoux (2004) have proposed a distinction between partial and global states of awareness in the processing of complex visual stimuli such as words that are represented at multiple, hierarchically organized levels (e.g., at the feature, letter, and word levels). According to this view, masked short duration presentation of word stimuli may prevent processing at a higher, wholeword level, thus preventing the global awareness that arises only when processing occurs at all levels, but allow partial awareness through identification of letters or fragments. In their first experiment, Kouider and Dupoux used a Stroop priming task in which participants decided whether a target consisting of a string of ampersands was presented in red or green. Targets followed masked primes that were either French color words (e.g., VERT for green) or their pseudo color word counterparts (e.g., VRET). Participants were also told of the four color words used as primes, but not their pseudoword variants, and tasked with trying to read the prime. Priming effects were obtained for both color words and pseudowords at 29 ms, but for color words only at 43 ms. These results suggested that conscious awareness of the more difficult to perceive 29 ms primes (real color words and pseudo color words alike) was limited to partial information (i.e., letters and fragments), and that this information was used by participants to reconstruct the real color word. The reconstructed word is then processed semantically, producing the priming effect that gives the illusory appearance of subliminal semantic priming. For the more visible 43 ms primes, conscious awareness was global, and the priming effect occurred for real, but not pseudo, color words because the latter were not incorrectly recognized as words. In their second experiment, Kouider and Dupoux increased the strength of the mask, with result that the priming effect disappeared for 29 ms primes, but occurred for both real and pseudo color words for 43 ms primes. Using the same strong masking in a third experiment, participants were not informed of the presence of color word primes, as they were in the first two experiments. Without top-down expectations, the priming effects observed in the second experiment disappeared. Conceivably, for the 29 ms masked word stimuli in Experiment 1 of the present study, participants might have occasionally used partial information to reconstruct word stimuli which were then processed semantically and later correctly identified on the recognition test. However, our participants would not have had the strong top-down expectations that participants in the Kouider and Dupoux (2004) experiments had from having been told that the primes were color words and from having to guess the prime identity. Nevertheless, they were tasked with remembering lists of words for a subsequent memory test, and it seems possible that some participants in the 29 ms masked condition might have used a strategy of attempting to complete word fragments whenever partial information was available. If a single list word were generated from a fragment, a participant would then have contextual information that would potentially facilitate word reconstruction for any subsequent item within the list perceived as a fragment. For example, for the word ill from the doctor list, sufficient information might be available for a participant to perceive it as the fragment, i_l. If the 12

Consciousness and Cognition xxx (xxxx) xxx–xxx

D.D. Sadler et al.

participant then generates ill as a possible completion, and the next word, patient, is perceived as the fragment, p_t__n_, the contextual constraint from ill might facilitate the reconstruction of the fragment as patient. The extent to which participants engaged in word reconstruction is, of course, unknown, but the fact that the veridical recognition rate for studied words was very low at 5% suggests that this strategic processing was used minimally, if at all, possibly due to infrequent experience of partial awareness at 29 ms. In addition to word reconstruction, participants might have retained word fragments and responded old when these fragments were recognized in test items. Although the very low veridical recognition rate for studied items in the 29 ms masked condition suggests that participants were rarely consciously aware of word stimuli, a determination of the extent to which processing in the 29 ms masked conditions in Experiments 1 and 2 was subliminal would be greatly facilitated by subjective reports of the clarity of participants’ visual experience during list presentation. The Perceptual Awareness Scale (PAS) (Overgaard, Rote, Mouridsen, & Ramsoy, 2006) is one measure of subjective visibility that has been used successfully to compare graded and dichotomous views of conscious awareness. In a study by Windey, Gevers, and Cleeremans (2013), participants viewed masked colored number stimuli presented at durations ranging from 10 to 80 ms. In separate conditions, they performed a low-level task involving identifying whether the color was red or blue and a highlevel task involving judging whether the number was smaller or larger in value than the number 5. At the end of each trial, participants used the PAS to rate their subjective visual experience. Separate psychophysical functions relating accuracy and subjective visibility to stimulus duration showed greater nonlinearity for numerical judgments than for color identification. In addition, although stimuli were identical in both tasks, compared to the color identification task, stimuli in the numerical judgment task were perceived as less visible pre-threshold, but more visible post-threshold. These results suggest that level of processing modulates visual awareness, determining whether performance accuracy and subjective visual experience are graded or dichotomous (see also Windey et al., 2014). When attention is focused on low-level, non-semantic information (e.g., a single feature such as color), awareness and performance accuracy are graded. When attending to high-level, semantic information (e.g., as in numerical judgments), awareness and accuracy are less graded, undergoing a sharper transition at the perceptual threshold. In Experiment 1 of the present study, the veridical recognition rates in the masked stimulus condition were .05, .46, and .49 for the 29, 43, and 257 ms exposure durations, respectively. This pattern of results is more consistent with a dichotomous visual experience in which there was an abrupt transition from virtually no awareness pre-threshold to full awareness post-threshold. In Experiment 2, however, although there was a substantial increase in veridical recognition rate corresponding to the change from 29 to 43 ms duration, the significant 50% increase in veridical recognition rate that occurred with the change in duration from 43 to 257 ms suggests graded visual awareness across the three levels of duration. Although the purpose of the present study is not to address the question of whether visual awareness is graded or dichotomous, reports of subjective visual experience would help clarify whether the masked stimuli in the 29 ms conditions of Experiments 1 and 2 were truly subliminal, as the low veridical recognition rates for studied items suggests, and provide insights into how the nature of awareness changes as word exposure duration increases. Experiments 3a and 3b were designed to measure subjective visual awareness as it occurs in the word-by-word processing of masked list words presented at brief durations. As in Experiment 1, participants studied word lists at three exposure durations under instructions that they would later be tested on their memory for the words. In addition, they rated the clarity of their visual experience immediately after each word was presented. 4.1. Method 4.1.1. Participants The participants were 72 students (36 in each experiment) attending Indiana University of Pennsylvania who received credit towards a general psychology course requirement for their participation. Students who chose to participate in this study selected their session dates and times using an online experiment management system. 4.1.2. Materials The wordlists and recognition test were the same as those used in Experiment 1. 4.1.3. Design and procedure Both experiments used masked word displays. The designs of both experiments replicated the mask presentation condition in Experiment 1, with one exception: in Experiment 3b, a 57 ms word exposure duration replaced the 257 ms duration. The practice and experimental wordlist presentation and test procedures were identical to those of Experiment 1. In addition to being asked to remember the words in each list, participants were asked to enter a visibility rating after each word using the Perceptual Awareness Scale (Overgaard et al., 2006; Windey et al., 2013). The following representation of the PAS was taped to the top of the keyboard.

1 No experience

What was the visibility of the word? 2 3 Brief Almost glimpse clear experience

4 Clear experience

The experiments used the same definitions for the four response categories with the exception of the second category, Brief Glimpse. The response categories were defined as follow: 13

Consciousness and Cognition xxx (xxxx) xxx–xxx

D.D. Sadler et al.

1. No experience. When you have no experience of the word at all. 2. Brief glimpse. Experiment 3a. When you experience only some letters or parts of letters. Experiment 3b. A feeling that something was present. 3. Almost clear experience. When you feel that you almost had a clear experience of the word. 4. Clear experience. When you have a clear experience of the word. The numbers 1–4 were written on round color-coding labels that were placed on the c, v, n, and m keys, respectively, on QWERTY keyboards. The left middle and index fingers were used to press the 1 and 2 keys, and the right index and middle fingers were used to press the 3 and 4 keys. Participants were asked to enter their response after the stimulus disappeared, but before the next stimulus was presented. They were also asked to use the scale taped to the keyboard as a reminder of the response options, but to remain focused on the center of the screen during the presentation of a wordlist. 4.2. Results and discussion Three participants from Experiment 3a, and two from Experiment 3b, with false alarm rates for nonstudied list items (combining both list words and critical lures from nonstudied lists) above .75 were replaced. After removing these outliers, the mean false alarm rates for Experiments 3a and 3b were .206 (SD = .181) and .222 (SD = .200), respectively. A third participant in Experiment 3b with a high proportion of missing PAS ratings (.225) was also replaced. The mean proportion of missing ratings for Experiments 3a and 3b were .010 (SD = .014) and .015 (SD = .022), respectively. 4.2.1. Recognition rates The mean old response rates are provided in Table 6 and the mean Pr values are presented in Table 7. In Experiment 3a, the effect of word exposure duration on corrected recognition rates was significant for both studied items and critical lures, F(2, 70) = 52.59, MSE = .034, p < .001, ωp2 = .489 and F(2, 70) = 31.31, MSE = .072, p < .001, ωp2 = .360, respectively. For studied words, both the 257 ms and 43 ms exposure durations resulted in significantly higher veridical recognition than did the 29 ms duration, t (35) = 10.99, SE = .039, p < .001, dz = 1.83 and t(35) = 7.25, SE = .046, p < .001, dz = 1.21, respectively. The veridical recognition rate at 257 ms was also higher than the rate at 43 ms, a difference that narrowly missed significance, t(35) = 1.99, SE = .046, p = .054, dz = 0.33. The results of single-sample t tests (see Table 7) show that the Pr values for studied words were significantly greater than zero for the 257 ms and 43 ms durations, but not for the 29 ms duration. For critical lures, the corrected false recognition rates were significantly higher at 257 ms and 43 ms than at 29 ms, t(35) = 7.29, SE = .058, p < .001, dz = 1.21 and t(35) = 6.76, SE = .065, p < .001, dz = 1.13, respectively, but did not differ from each other, t(35) = 0.26, SE = .066, p > .25, dz = 0.04. As shown in Table 7, the Pr values for critical lures were significantly greater than zero at the two longer durations, but not at the 29 ms duration. Exposure duration also had a significant effect on Experiment 3b corrected recognition rates for both studied words and critical lures, F(2, 70) = 44.27, MSE = .033, p < .001, ωp2 = .445 and F(2, 70) = 38.72, MSE = .068, p < .001, ωp2 = .411, respectively. For studied words, veridical recognition was significantly higher at 57 ms and 43 ms than at 29 ms, t(35) = 8.51, SE = .047, p < .001 dz = 1.42 and t(35) = 5.85, SE = .045, p < .001, dz = 0.98, respectively, and higher at 57 ms than at 43 ms, t(35) = 3.68, SE = .037, p < .001, dz = 0.61. As shown in Table 7, the Pr values for studied words were significantly greater than zero at all exposure durations. For critical lures, false recognition was significantly greater at 57 ms and 43 ms than at 29 ms, t(35) = 7.97, Table 6 Experiments 3a and 3b: Mean recognition rates. Word exposure duration Item type Experiment 3a List words Studied Non-studied Critical Lures Related Unrelated Experiment 3b List Words Studied Non-studied Critical Lures Related Unrelated

29 ms (R/K)

43 ms (R/K)

257/57 ms (R/K)

.21 (.08/.13) .20 (.09/.10)

.49 (.22/.27) .15 (.07/.08)

.63 (.35/.28) .19 (.06/.13)

.24 (.11/.12) .25 (.12/.12)

.61 (.27/.34) .19 (.08/.10)

.67 (.40/.27) .26 (.10/.16)

.23 (.07/.16) .18 (.06/.12)

.50 (.29/.21) .19 (.05/.14)

.64 (.39/.25) .19 (.09/.11)

.26 (.12/.15) .28 (.08/.21)

.59 (.33/.26) .22 (.06/.16)

.75 (.47/.28) .26 (.07/.19)

Note. R = Old remember judgment and K = Old know judgment. The longest exposure durations were 257 and 57 ms for Experiments 3a and 3b, respectively.

14

Consciousness and Cognition xxx (xxxx) xxx–xxx

D.D. Sadler et al.

Table 7 Experiments 3a and 3b: Single-sample t tests for corrected recognition rates (Pr). Word exposure duration 29 ms

43 ms

257/57 ms

Item type

M

SE

t

M

SE

t

M

SE

t

Experiment 3a Studied words Critical lures

.01 −.01

.026 .046

0.35 −0.30

.34 .43

.043 .061

7.94*** 7.04***

.44 .41

.032 .039

13.70*** 10.48***

Experiment 3b Studied words Critical lures

.05 −.02

.025 .052

1.84* −0.40

.31 .37

.040 .048

7.74*** 7.68***

.44 .50

.034 .046

13.14*** 10.72***

Note. df = 35 for all t tests (one-tailed). The longest exposure durations were 257 and 57 ms for Experiments 3a and 3b, respectively. * p < .05. *** p < .001.

SE = .065, p < .001, dz = 1.33 and t(35) = 6.70, SE = .059, p < .001, dz = 1.12, respectively, and greater at 57 ms than at 43 ms, t (35) = 2.06, SE = .061, p < .05, dz = 0.34. The Pr values for critical lures were significantly above zero for the two longer durations, but did not differ from zero for the 29 ms duration. The results for the two longer exposure durations in Experiments 3a and 3b are consistent with findings for the mask condition in Experiment 1 in showing that exposure durations for masked stimuli that are sufficiently long to produce robust veridical recognition also produce robust false recognition. However, the results are less clear for the 29 ms condition, with Experiment 3b showing exactly the same small (5%) significant veridical recognition rate found to be significant in Experiment 1, but Experiment 3a showing a nonsignificant 1% rate. Importantly, all four experiments have consistently shown that a masked word exposure duration that was so brief (29 ms) that it produced little, if any, veridical recognition, also failed to produce any false recognition. An additional finding of interest is that, whereas the veridical recognition rates in the masked condition in Experiment 1 did not differ between the 257 and 43 ms durations, this rate was significantly lower for the 43 ms duration than for the 57 ms duration in Experiment 3b, and lower (but narrowly missing significance) than the 257 ms duration in Experiment 3a. Veridical recognition was also lower for the 43 ms than the 257 ms duration in Experiment 2. Collectively, these results suggest graded awareness across the three exposure durations. 4.2.2. Subjective visibility ratings The frequency distributions for PAS ratings (see Fig. 2) show that participants’ conscious awareness of word stimuli increased as exposure duration increased. In both experiments, stimuli were seldom visible, at any level of awareness, at 29 ms. At 43 ms, participants most often experienced stimuli with intermediate levels of awareness, but seldom full awareness. In Experiment 3a, virtually all stimuli were experienced with full awareness at 257 ms, and, in Experiment 3b, stimuli were rarely invisible, and most often perceived with full awareness or close to full awareness, at 57 ms. Table 8 presents the mean PAS ratings by word exposure duration and experiment. In Experiment 3a, mean subjective visibility increased significantly with each increase in exposure duration, t (35) = 9.69, SE = 0.085, p < .001, dz = 1.62 for the increase in duration from 29 ms to 43 ms and t(35) = 22.41, SE = 0.085, p < .001, dz = 3.73 for the increase from 43 ms to 257 ms. Similarly, subjective visibility in Experiment 3b increased significantly with increases in exposure duration, t(35) = 10.84, SE = 0.094, p < .001, dz = 1.81 for the increase in duration from 29 ms to 43 ms

Fig. 2. Experiments 3a and 3b: (A) and (B) are frequency distributions for PAS ratings by word exposure duration.

15

Consciousness and Cognition xxx (xxxx) xxx–xxx

D.D. Sadler et al.

Table 8 Experiments 3a and 3b: Mean PAS ratings. Experiment 3a

Experiment 3b

Word exposure duration

M

SE

M

SE

29 ms 43 ms 257/57 ms

1.17 2.00 3.90

0.053 0.090 0.039

1.25 2.27 2.99

0.042 0.110 0.123

Note. The longest durations were 257 and 57 ms in Experiments 3a and 3b, respectively.

and t(35) = 8.07, SE = 0.089, p < .001, dz = 1.35 for the increase from 43 ms to 57 ms. The PAS ratings suggest a possible source for the small amount of veridical recognition observed at the 29 ms exposure duration. The very high mean frequency of No Experience responses (87% and 78% for Experiments 3a and 3b, respectively) indicate that participants seldom experienced conscious awareness of word stimuli at any level. To the extent that they experienced some degree of awareness, they almost never reported an experience of full awareness or one approaching full awareness. Instead, in both experiments, the response categories Clear Experience and Almost Clear Experience together comprised only 2% of responses. The 11% PAS response frequency for the Brief Glimpse category in Experiment 3a indicates that at least some participants experienced partial awareness consisting of letters or parts of letters for some stimuli presented for 29 ms, and it seems likely that much of the 20% response frequency for this category in Experiment 3b, using a less specific definition of Brief Glimpse, is also due to awareness of letters or features. Collectively, the subjective visibility distributions for both experiments suggest that the word stimuli at 29 ms usually fell below the perceptual threshold, but when they did not, they were often experienced with some degree of partial awareness of letters or features. This partial awareness suggests an alternative to a subliminal semantic processing explanation for the small, but significant 5% veridical recognition rate observed in Experiment 3b. As discussed previously, Kouider and Dupoux (2004) have proposed that when a few letters are perceived, a word stimulus might be reconstructed, and then processed semantically. It is also possible that word fragments from list items might be encoded and then later recognized in test items (Cleary & Greene, 2004). Although subliminal semantic processing cannot be eliminated as a possible source of the veridical recognition, the partial awareness hypothesis provides a more plausible explanation of the results, and it is supported by the following reanalysis of the data that divides participants into lower and higher perceptual threshold groups based on their subjective visibility ratings. 4.2.3. Analysis of recognition rates by participant awareness at 29 ms The visibility ratings suggest that word stimuli presented at the 29 ms duration were usually subliminal. However, these results are ambiguous regarding the question of whether the very small, but significant veridical recognition rates at this duration in Experiments 1, 2, and 3b were due to subliminal semantic processing occurring for some stimuli or, instead, to some stimuli being processed with some degree of conscious awareness. Some insight into this question can be gained by examining recognition rates as a function of level of perceptual awareness for word stimuli presented with a 29 ms exposure duration. Although individual detection thresholds (Cheesman & Merikle, 1984) were not determined, we examined the extent to which the amount of veridical recognition that occurred at 29 ms differed for participants with higher versus lower frequencies of PAS ratings above the No Experience category. For 29 ms exposure duration in each experiment, we created a dichotomous PAS measure by combining the response categories of Brief Glimpse, Almost Clear Experience, and Clear Experience into a single awareness category representing any level of awareness above the level of No Experience. For each participant, we calculated an awareness score representing the percentage of PAS ratings for the 120 studied items (15 items in each of 8 lists) at the 29 ms duration that fell into the condensed awareness category. In both experiments, we used the 75 th percentile of the distribution of the participant awareness scores to divide participants into two groups: a higher perceptual awareness group (HPA, n = 9) and a lower perceptual awareness group (LPA, n = 27). The 75 th percentile was chosen based on the goal of having a low awareness group sufficiently large to be considered a large majority of participants while having a high awareness group that would not be too small for reliable results. In Experiment 3a, the mean awareness scores were 42.1% (SD = 26.2) and 3.9% (SD = 4.5) for HPA and LPA groups, respectively. In Experiment 3b, mean awareness scores for participants in the HPA and LPA groups averaged 55.8% (SD = 22.6) and 10.8% (SD = 10.8), respectively. Table 9 presents corrected veridical and false recognition rates and one-sample t tests. Figs. 3 and 4 present frequency distributions for the PAS ratings. As shown, the veridical recognition rates for the 29 ms duration in both experiments were not significantly above zero for the LPA groups, but these rates were significantly above zero for the HPA groups. False recognition rates were nonsignificant for both awareness groups in both experiments. Both awareness groups showed substantial veridical and false recognition at 43 ms and the longer durations, 257 ms and 57 ms. These results suggest that the small amount of veridical recognition found at 29 ms in three of four experiments is attributable not to subliminal semantic processing of word stimuli, but rather to some degree of awareness for some stimuli experienced by participants in the higher awareness group. Whereas the perceptual threshold appears to have been close to 29 ms for participants in the LPA groups, making virtually all stimuli invisible, it appears to have been somewhat lower for participants in the HPA groups, allowing some stimuli to be perceived at some level of awareness. Although this awareness appears to be source of the familiarity that underlies the small amount of veridical recognition observed for the 29 ms duration, the underlying stimulus processing appears to have been insufficient for activating the semantically-related critical lures, at least in a way that would lead to episodic memory traces that later would be the basis for false recognition. 16

Consciousness and Cognition xxx (xxxx) xxx–xxx

D.D. Sadler et al.

Table 9 Experiments 3a and 3b: Single-sample t tests for corrected recognition rates (Pr) by awareness level. Word exposure duration 29 ms Awareness level and item type Experiment 3a LPA Studied words Critical lures HPA Studied words Critical lures Experiment 3b LPA Studied words Critical lures HPA Studied words Critical lures

43 ms

257/57 ms

M

SE

t

M

SE

t

M

SE

t

−.04 −.05

.022 .055

−1.65 −0.85

.33 .42

.048 .065

6.72*** 6.47***

.46 .42

.034 .041

13.51*** 10.36***

.14 .08

.067 .078

.40 .44

.097 .152

4.09** 2.93**

.36 .38

.073 .102

4.91*** 3.67**

.02 −.01

.028 .065

0.71 −0.21

.28 .37

.047 .056

5.94*** 6.57***

.44 .49

.041 .055

10.77*** 8.81***

.12 −.04

.047 .078

2.68* −0.54

.39 .38

.070 .100

5.64*** 3.75**

.45 .53

.060 .088

7.52*** 6.01***

2.14* 1.07

Note. t tests are one-tailed. df = 26 for lower perceptual awareness (LPA) group; df = 8 for higher perceptual awareness (HPA) group. The longest exposure durations were 257 and 57 ms for Experiments 3a and 3b, respectively. * p < .05. ** p < .01. *** p < .001.

Fig. 3. Experiment 3a: (A) and (B) are frequency distributions for PAS ratings by word exposure duration and participant level of perceptual awareness at the 29 ms duration: low perceptual awareness (LPA, n = 27) and high perceptual awareness (HPA, n = 9).

The goal of the present study was to determine whether false recognition for critical lures occurs when list words are presented subliminally, preventing participants from consciously recognizing and thinking about them. A more detailed examination of the subjective visibility data is useful in understanding the extent to which conscious awareness of list words at the level of whole-word recognition was eliminated in the 29 ms condition. The PAS measures degrees of conscious awareness (Overgaard et al., 2006), with ratings of 2 to 4 indicating increasing levels of awareness above the complete absence of awareness indicated by a rating of 1 (No Experience). For the analysis that follows, we created two visibility categories by collapsing the PAS 4-point scale into a category representing no whole-word recognition (ratings 1 and 2) and a category representing whole-word recognition (ratings 3 and 4). In dichotomizing the PAS in this way, we are assuming that whenever participants experienced partial awareness (e.g., word fragments), they would have rated the experience as Brief Glimpse if it did not result in an experience at the whole-word level (e.g., through word reconstruction from a word fragment, Kouider & Dupoux, 2004). We are also assuming that in rating an experience as Almost Clear Experience participants were indicating that they experienced a stimulus at the whole-word level (including a word coming to mind through reconstruction from perception of a word fragment), but were not completely certain about its identity. Although these assumptions are speculative, they are consistent with a similar no experience-experience dichotomizing of the PAS used by Ramsoy and Overgaard (2004) to show a parallel between the PAS and the traditional not perceived-perceived dichotomy used in subliminal perception studies. At the 29 ms duration, PAS ratings indicating no experience of stimuli at the whole-word level were assigned to virtually all 120 17

Consciousness and Cognition xxx (xxxx) xxx–xxx

D.D. Sadler et al.

Fig. 4. Experiment 3b: (A) and (B) are frequency distributions for PAS ratings by word exposure duration and participant level of perceptual awareness at the 29 ms duration: low perceptual awareness (LPA, n = 27) and high perceptual awareness (HPA, n = 9).

list items by the average participant, with mean combined percentages for ratings 1 and 2 of 97.6% in Experiment 3a and 97.8% in Experiment 3b. For the 75% of participants classified as having lower perceptual awareness at 29 ms, the mean percentages of ratings indicating no whole-word awareness of list items at 29 ms were 99.8% and 98.1% for Experiments 3a and 3b, respectively. In each experiment, corresponding to this nearly complete absence of whole-word awareness for list items for LPA participants was the complete absence of veridical recognition of studied items and false recognition of critical lures on the memory test. For the 25% of participants classified as having higher perceptual awareness at 29 ms, the mean percentages of ratings indicating no whole-word awareness of list items at 29 ms were also very high (91.1% and 96.8% for Experiments 3a and 3b, respectively). However, in contrast to the LPA group, the HPA group had small, but significant veridical recognition rates in both experiments, although their false recognition rates were nonsignificant. One possible source for this veridical recognition is the small percentage of list items experienced at the whole-word level: 8.9% (an average of 1.3 items per 15-item list) in Experiment 3a and 3.2% (an average of 0.5 items per 15-item list) in Experiment 3b. However, given such a low probability of experiencing a list item at the whole-word level, there is even a much lower probability that the item recognized at study would be one of the three items from that list included as a test item. Another possible source of the veridical recognition in the HPA group is orthographic information retained at study and recognized at test. Given the high rates of Brief Glimpse ratings for the HPA participants (33.1% and 52.6% in Experiments 3a and 3b, respectively), it would not be surprising if some word fragments perceived at study were retained an later recognized at test. 5. General discussion The results of four experiments produced no evidence for unconscious activation of critical lures without conscious awareness of list words. Instead, we found that false memory for critical lures vanishes when veridical recognition of studied words approaches zero. In within- and between-subjects designs, after viewing masked presentations of list words participants demonstrated robust veridical recognition for studied words and false recognition for critical lures at word exposure durations of 257, 57 and 43 ms. However, reducing exposure duration to 29 ms resulted in very low veridical recognition for studied words, and consistently nonsignificant false recognition for critical lures. Experiments 3a and 3b provided evidence that the low level of veridical recognition for masked stimuli at 29 ms is attributable to a subset of participants who have higher perceptual awareness and experienced some degree of intermediate awareness for some stimuli, but at a level that was insufficient to activate critical lures. For the 75% of participants classified as having lower perceptual awareness, veridical and false memory were completely absent at the 29 ms duration. As discussed previously, whatever the nature of the processing occurred for the subliminal stimuli, it apparently did not lead to the formation and encoding of associative structures during study that are then reactivated at test and underlie veridical recognition and false recognition of critical lures (Meade et al., 2007). The purpose in using a subliminal perception paradigm in the present study was to examine whether false memory for critical lures would occur under study conditions in which conscious awareness of list items has been eliminated. Had we obtained false memory in our 29 ms masked conditions, a strong case for false memory through unconscious activation of critical lures could have been made, not only under conditions of subliminal list presentation, but for conscious list presentation as well. False memory in the absence of conscious awareness of list words would have strongly implied unconscious semantic activation of critical lures, and supported an activation monitoring explanation (Roediger et al., 2001) by which spreading activation from multiple subliminally processed list words converges on the critical lure and summates, leading to its false recognition. Previous studies that appeared to demonstrate unconscious activation of critical lures (e.g., Cotel et al., 2008; Gallo & Seamon, 2004; Seamon et al., 1998) did not eliminate the alternative explanation that merely minimizing conscious processing of list items might not be sufficient to prevent conscious processing of critical lures. The finding by Zeelenberg et al. (2003) demonstrated that when rapid, brief presentations of list 18

Consciousness and Cognition xxx (xxxx) xxx–xxx

D.D. Sadler et al.

items fall below the perceptual threshold, neither correct recognition of studied items nor false recognition of critical lures occurs. However, Zeelenberg et al. presented unmasked stimuli at presentation durations of 20 ms with a 0 ms ISI, perhaps not allowing enough time for the spread of semantic activation from list item to critical lure. The present study has demonstrated that, when conscious awareness of list items is prevented by using 29 ms masked presentation, but with a long ISI to allow for activation to spread from list words and converge on the critical lure, there is no evidence in false recognition rates that the critical lures are activated. There are three possible explanations for our failure to find unconscious semantic activation for critical lures. Perhaps the most parsimonious explanation is that activation of critical lures requires conscious awareness of list words, which by design was virtually nonexistent in our subliminal presentation condition. As discussed above, PAS ratings in Experiment 3a and 3b indicate that masked list words presented at 29 ms duration were unidentifiable for the LPA group and rarely identifiable (at most averaging 1.3 and 0.5 items per list in Experiments 3a and 3b, respectively) for the HPA group. By this account, whether subliminal semantic processing of list words occurred at 29 ms is irrelevant; there was no possibility of an unconscious DRM false memory effect because critical lures are not activated without conscious recognition of list words. A similar conclusion was reached by Gallo and Seamon (2004) who found that false recognition of a critical lure required conscious perception/recall of at least one list item. They also concluded that conscious activation of the critical lure during study is not necessary for it to be falsely recognized, but for reasons discussed above, this conclusion was based on faulty assumptions (Raaijmakers & Zeelenberg, 2004). The present study does not include a task that demonstrates subliminal processing of stimuli in the 29 ms masked condition. Consequently, another possible explanation for the absence of an unconscious false recognition effect is that subliminal semantic processing of list items did not occur. Given the evidence from prior research (see discussion above) supporting the existence of unconscious semantic processing, one possible reason why it might not occur in the context of a particular study is that lower-level visual processing was prevented by overly strong masking (Kouider & Dehaene, 2007). However, the distribution of PAS ratings for our 43 ms masked stimuli indicate substantial visual processing occurred at multiple levels at this duration, and it seems unlikely that the same masking used at 29 ms would weaken the visual signal so much that it prevents lower-level visual processing. Unconscious priming effects have been found in previous studies using masked prime durations at or near 29 ms. For example Dehaene et al. (2001) obtained an unconscious repetition priming effect for word stimuli using 29 ms masked prime words in a semantic classification task (i.e., judging whether the target was natural or manmade). Van Opstal et al. (2005a) demonstrated unconscious semantic categorization with their finding of a congruity effect using a numerical judgement task with 33 ms masked primes. Assuming that subliminal semantic processing did occur for 29 ms masked stimuli, a third explanation for the absence of subliminal false recognition is that this processing differs from conscious semantic processing such that it does not lead to storage of new episodic traces (Raaijmakers & Neville, 2015). One possibility is that conscious awareness requires recurrent neural processing, but subliminal processing is limited to feedforward activation (Lamme, 2006; Lamme & Roelfsema, 2000). Lamme (2006) has proposed that recurrent processing, but not feedforward activation, activates the synaptic plasticity processes between pre- and post-synaptic neurons that are the basis of learning and memory. As discussed previously, the DRM false memory effect may depend on encoding episodic associative structures during list presentation that are later reactivated at test. If episodic traces are not encoded for subliminal list presentation, then false recognition would not occur. However, the study by Reber and Henke (2011) discussed above demonstrated that semantic associations between pairs of unrelated words can be encoded during subliminal presentation, and later influence judgments of semantic fit. Given the importance of the Reber and Henke findings to the underlying rationale for the present study, we discuss below several possible explanations for the different outcomes. Although the present study and the Reber and Henke (2011) study differ considerably in the type of stimuli used and nature of the encoding and retrieval tasks, they share the common goal of examining whether unconscious semantic processing will occur for subliminal stimuli and whether such processing will result in the formation of episodic associations that later influence retrieval behavior. It is therefore surprising, in light of the finding by Reber and Henke that episodic associations between pairs of unrelated words can form without conscious processing of those words, that the present study found no evidence for the formation of episodic associative structures for lists of semantic associates when presented subliminally at an exposure duration of 29 ms. There are, however, three major methodological differences that might have contributed to the different outcomes for the two studies. First, whereas a subliminal encoding trial in the Reber and Henke (2011) experiments consisted of 12 repetitions of a word pair, in the present study each subliminal trial consisted of a single presentation of each of 15 list items. The repeated masking paradigm used by Reber and Henke has previously been shown to increase prime strength without increasing prime awareness. For example, Wentura and Frings (2005) found a subliminal priming effect using a lexical decision task for 10 alternating presentations of mask and prime, but not for a single presentation, with no difference between conditions in perceptual awareness of the primes. However, Atas, Vermeiren, and Cleeremans (2013) noted that the absence of a priming effect in the single prime condition might have been due to Wentura and Frings using a much longer prime-target delay in the single prime versus repeated prime condition, potentially allowing substantial decay of prime activation before target presentation. In their study, Atas et al. used a number comparison task similar to Dehaene et al. (1998), but varied the number of masked prime repetitions, with a constant delay between the final prime repetition and target. Their major finding was that both visual priming and prime awareness were absent for one presentation of the masked prime, but increased significantly with increasing repetitions of the prime, a finding predicted by several higher-order theories of consciousness (see Atas et al., 2013, for a discussion). Perhaps future research will resolve conflicting findings on the relationship between number of masked prime repetitions and perceptual awareness. For the current discussion, if we assume the validity of the Reber and Henke (2011) results indicating chance discrimination of their subliminal stimuli after 12 repetitions, it is conceivable that using repeated masked presentations of DRM lists might similarly strengthen activation and encoding of subliminal list items and critical lures while still preventing conscious 19

Consciousness and Cognition xxx (xxxx) xxx–xxx

D.D. Sadler et al.

awareness of the items. Second, Reber and Henke (2011) found their subliminal processing effect to be dependent on the participant developing an understanding of the task structure by first performing a supraliminal version of the task before the subliminal version. Their task instructions were very specific in encouraging participants to form associations between the unrelated words in each word pair. Although our encoding instructions were less specific than those of Reber and Henke (our participants were merely told that they would be tested on their memory for the list items), we believe that the three practice trials (one list at each exposure duration) in our experiments were sufficient for our participants to acquire a good understanding of the task structure. The first practice trial always presented list items at the longest of the three durations, and it seems very likely that nearly all participants noticed that the list was composed of semantic associates and engaged in associative processing. The practice recall and recognition tests emphasized the importance of long-term retention of the list items. Nevertheless, it is possible that semantic processing of subliminal stimuli would be more likely to occur, or enhanced when it does occur, if participants were explicitly encouraged to engage in associative memory strategies during list presentation. Assuming that our participants did have a good understanding of the standard DRM task structure, the absence of veridical and false recognition for the subliminal list presentations indicates that this task set did not unconsciously influence encoding processes for these lists. Third, Reber and Henke (2011) used an implicit memory task (semantic fit judgements) as a retrieval task. Implicit tasks might provide more sensitive measures of subliminal processing effects than the explicit recognition memory test used in the present study. Explicit memory tasks such as recall and recognition entail conscious retrieval of prior experience, whereas implicit tasks such as lexical decision, word stem or fragment completion, and perceptual identification can be used to reveal memory indirectly, without conscious retrieval, by showing the influence of prior experience on test performance (Roediger, 1990; Schacter, 1987). In a classic demonstration of the dissociation between explicit and implicit memory tests, Jacoby (1983) varied the encoding context for target words (e.g., cold) such that they received different amounts of perceptual and conceptual processing. When the encoding task required conceptual processing (an antonym generation task, e.g., hot - ????), performance on a recognition test was much higher than when the task required greater reliance on perceptual processing (read aloud the target word that replaces the X’s, e.g., XXXX - cold). Recognition accuracy was at an intermediate level when the encoding task encouraged a more balanced mix of perceptual and conceptual processing (read aloud the antonym of the word it replaces, e.g., hot - cold. The opposite pattern of results was obtained for a perceptual identification test, with greater priming (higher identification probabilities) for the no context condition than the generate condition. Evidence from studies demonstrating dissociations between implicit and explicit memory tasks suggest that they access different forms of retention (Roediger, 1990). Implicit memory measures have also been used to show retention of prior experience in the absence of awareness. For example, Jacoby and Witherspoon (1982) had participants (five amnesic Korsakoff patients and five university students) listen to questions that biased low frequency interpretations of homophones (e.g., “Name a musical instrument that employs a reed.”). Performance on a subsequent spelling test demonstrated memory for the prior presentation of the homophones; both participant groups spelled homophones using the low frequency interpretation (e.g., reed rather than read) with a much higher probability for old homophones that had appeared in biasing contexts than for new homophones that were not previously presented. However, on a subsequent memory test requiring awareness, the Korsakoff patients performed poorly, achieving recognition accuracy only a third that of the high performing university students. Jacoby and Witherspoon noted that, when performing the spelling task, the Korsakoff patients seemed to be unaware that they were remembering. Evidence for memory without awareness was also found for the university students as analyses showed that recognition memory and bias in spelling were independent in both groups. Various theoretical accounts have been proposed for dissociations between implicit and explicit memory tasks (see Reder, Park, & Kieffaber, 2009, for an overview of several major theories). One of these accounts, the transfer appropriate processing (TAP) framework (e.g., Blaxton, 1989; Morris, Bransford, & Franks, 1977; Roediger, 1990), has particular relevance to the present study. Essentially, TAP posits that memory performance is optimal when there is overlap between the processing operations of encoding and retrieval. An implicit memory task that does not depend on conscious retrieval of prior stimuli may be better suited for measuring unconscious semantic activation of subliminal stimuli and the spread of that activation to critical lures. In four experiments using an explicit memory test, we were unable to obtain evidence for unconscious activation of critical lures following subliminal presentation of word lists. Although these null findings seem to favor a conclusion that participants did not encode associative representations for list items and critical lures, the possibility remains that subliminal list items are processed semantically and spread activation to critical lures. If it does occur, this unconscious activation of critical lures may be detectable through an implicit task such as lexical decision that is sensitive to semantic activation (Neely, 1991), but does not require conscious retrieval (Meade et al., 2007; Tse & Neely, 2005). Prior studies (McKone, 2004; Zeelenberg & Pecher, 2002) that have used the LDT following supraliminal presentations of multiple DRM lists failed to find long-term semantic priming of critical lures, and therefore it seems unlikely that the LDT would be successful in showing unconscious semantic activation of critical lures from subliminal list items after a long delay such as was used in the present study. However, several studies have found semantic priming effects for critical lures following presentation of single lists (e.g., Hancock, Hicks, Marsh, & Ritschel, 2003; Meade et al., 2007; Tse & Neely, 2005, 2007). If a similar finding were obtained for subliminal DRM lists, it would support a conclusion that critical lures can be unconsciously activated. Acknowledgements We thank Susan Zimny for helpful discussions, and Kayla Bell and John Wunderlich for their assistance in data collection and data entry. 20

Consciousness and Cognition xxx (xxxx) xxx–xxx

D.D. Sadler et al.

References Arndt, J., & Hirshman, E. (1998). True and false recognition in MINERVA2: Explanations from a global matching perspective. Journal of Memory and Language, 39, 371–391. http://dx.doi.org/10.1006/jmla.1998.2581. Atas, A., Vermeiren, A., & Cleeremans, A. (2013). Repeating a strongly masked stimulus increases priming and awareness. Consciousness and Cognition, 22, 1422–1430. http://dx.doi.org/10.1016/j.concog.2013.09.011. Balota, D. A. (1983). Automatic semantic activation and episodic memory encoding. Journal of Verbal Learning and Verbal Behavior, 22, 88–104. http://dx.doi.org/10. 1016/S0022-5371(83)80008-5. Blaxton, T. A. (1989). Investigating dissociations among memory measures: Support for a transfer appropriate processing framework. Journal of Experimental Psychology: Learning, Memory, and Cognition, 15, 657–668. Bredart, S. (2000). When false memories do not occur: Not thinking of the lure or remembering that it was not heard? Memory, 8, 123–128. http://dx.doi.org/10.1080/ 096582100387669. Cheesman, J., & Merikle, P. M. (1984). Priming with and without awareness. Perception & Psychophysics, 36, 387–395. Cleary, A. M., & Greene, R. L. (2004). True and false memory in the absence of perceptual identification. Memory, 12, 231–236. http://dx.doi.org/10.1080/ 09658210244000577. Collins, A. M., & Loftus, E. F. (1975). A spreading-activation theory of semantic processing. Psychological Review, 82, 407–428. Cotel, S. C., Gallo, D. A., & Seamon, J. G. (2008). Evidence that nonconscious processes are sufficient to produce false memories. Consciousness and Cognition, 17, 210–218. http://dx.doi.org/10.1016/j.concog.2007.01.009. Damian, M. F. (2001). Congruity effects evoked by subliminally presented primes: Automaticity rather than semantic processing. Journal of Experimental Psychology: Human Perception and Performance, 27, 154–165. http://dx.doi.org/10.1037//0O96-1523.27.1.154. Deese, J. (1959). On the prediction of occurrence of particular verbal intrusions in immediate recall. Journal of Experimental Psychology, 58, 17–22. http://dx.doi.org/ 10.1037/h0046671. Dehaene, S., Changeux, J. P., Naccache, L., Sackur, J., & Sergent, C. (2006). Conscious, preconscious, and subliminal processing: A testable taxonomy. TRENDS in Cognitive Sciences, 10, 204–211. http://dx.doi.org/10.1016/j.tics.2006.03.007. Dehaene, S., Naccache, L., Cohen, L., Le Bihan, D., Mangin, J.-F., Poline, J.-B., et al. (2001). Cerebral mechanisms of word masking and unconscious repetition priming. Nature Neuroscience, 4, 752–758. Dehaene, S., Naccache, L., Le Clec’H, G., Koechlin, E., Mueller, M., Dehaene-Lambertz, G., et al. (1998). Imaging unconscious semantic priming. Nature, 39, 597–600. http://dx.doi.org/10.1038/26967. Dell’Acqua, R., & Grainger, J. (1999). Unconscious semantic priming for pictures. Cognition, 73, B1–B15. http://dx.doi.org/10.1016/S0010-0277(99)00049-9. Feenan, K., & Snodgrass, J. G. (1990). The effect of context on discrimination and bias in recognition memory for pictures and words. Memory & Cognition, 18, 515–527. Forster, K. I., & Davis, C. (1984). Repetition priming and frequency attenuation in lexical access. Journal of Experimental Psychology: Learning, Memory, and Cognition, 10, 680–698. http://dx.doi.org/10.1037/0278-7393.10.4.680. Gallo, D. A. (2006). Associative illusions of memory: False memory research in DRM and related tasks. New York: Psychology Press. Gallo, D. A., & Seamon, J. G. (2004). Are nonconscious processes sufficient to produce false memories? Consciousness and Cognition, 13, 158–164. http://dx.doi.org/10. 1016/j.concog.2003.09.001. Greenwald, A. G., Draine, S. C., & Abrams, R. L. (1996). Three cognitive markers of unconscious semantic activation. Science, 273, 1699–1702. http://dx.doi.org/10. 1126/science.273.5282.1699. Hamm, J. P. (2001). Object-oriented millisecond timers for the PC. Behavior Research Methods, Instruments, & Computers, 33, 532–539. Hancock, T. W., Hicks, J. L., Marsh, R. L., & Ritschel, L. (2003). Measuring the activation level of critical lures in the Deese-Roediger-McDermott paradigm. American Journal of Psychology, 116, 1–14. Holender, D. (1986). Semantic activation with and without conscious identification in dichotic listening, parafoveal vision, and visual masking: A survey and appraisal. Behavioral and Brain Sciences, 9, 1–23. http://dx.doi.org/10.1017/S0140525X00021269. Jacoby, L. L. (1983). Remembering the data: Analyzing interactive processes in reading. Journal of Verbal Learning and Verbal Behavior, 22, 405–508. Jacoby, L. L., & Witherspoon, D. (1982). Remembering without awareness. Canadian Journal of Psychology, 36, 300–324. Johnson, M. K., Hashtroudi, S., & Lindsay, D. S. (1993). Source monitoring. Psychological Bulletin, 114, 3–28. http://dx.doi.org/10.1037/0033-2909.114.1.3. Keppel, G., & Wickens, T. D. (2004). Design and analysis: A researcher’s handbook. Upper Saddle River, NJ: Pearson Prentice Hall. Kouider, S., & Dehaene, S. (2007). Levels of processing during non-conscious perception: A critical review of visual-masking. Philosophical Transactions of the Royal Society B, 362, 857–875. http://dx.doi.org/10.1098/rstb.2007.2093. Kouider, S., & Dupoux, E. (2004). Partial awareness creates the “illusion” of subliminal semantic priming. Psychological Science, 15, 75–81. http://dx.doi.org/10.1111/ j.0963-7214.2004.01502001.x. Kunde, W., Kiesel, A., & Hoffmann, J. (2003). Conscious control over the content of unconscious cognition. Cognition, 88, 223–242. http://dx.doi.org/10.1016/S00100277(03)00023-4. Kunde, W., Kiesel, A., & Hoffmann, J. (2005). On the masking and disclosure of unconscious elaborate processing: A reply to Van Opstal, Reynvoet, and Verguts (2005). Cognition, 97, 99–105. http://dx.doi.org/10.1016/j.cognition.2005.03.005. Lakens, D. (2013). Calculating and reporting effect sizes to facilitate cumulative science: A practical primer for t-tests and ANOVAs. Frontiers in Psychology, 4, 1–12 doi:03389/fpsyg.2013.00863. Lamme, V. A. F. (2006). Towards a true neural stance on consciousness. Trends in Cognitive Sciences, 10, 494–501. http://dx.doi.org/10.1016/j.tics.2006.09.001. Lamme, V. A. F., & Roelfsema, P. R. (2000). The distinct modes of vision offered by feedforward and recurrent processing. Trends in Neurosciences, 23, 571–579. Logan, G. D. (1988). Toward an instance theory of automatization. Psychological Review, 95, 492–527. Marcel, A. J. (1983). Conscious and unconscious perception: Experiments on visual masking and word recognition. Cognitive Psychology, 15, 197–237. http://dx.doi. org/10.1016/0010-0285(83)90009-9. McDermott, K. B., & Watson, J. M. (2001). The rise and fall of false recall: The impact of presentation duration. Journal of Memory and Language, 45, 160–176. http:// dx.doi.org/10.1006/jmla.2000.2771. McKone, E. (2004). Distinguishing true from false memories via lexical decision as a perceptual implicit test. Australian Journal of Psychology, 56, 42–49. Meade, M. L., Watson, J. M., Balota, D. A., & Roediger, H. L., III (2007). The roles of spreading activation and retrieval mode in producing false recognition in the DRM paradigm. Journal of Memory and Language, 56, 305–320. http://dx.doi.org/10.1016/j.jml.2006.07.007. Morris, C. D., Bransford, J. D., & Franks, J. J. (1977). Levels of processing versus transfer appropriate processing. Journal of Verbal Learning and Verbal Behavior, 16, 519–533. Naccache, L., & Dehaene, S. (2001). Unconscious semantic priming extends to novel unseen stimuli. Cognition, 80, 215–229. http://dx.doi.org/10.1016/S00100277(00)00139-6. Neely, J. H. (1991). Semantic priming effects in visual word recognition: A selective review of current findings and theories. In D. Besner, & G. W. Humphreys (Eds.). Basic processes in reading: Visual word recognition (pp. 264–336). Hillsdale, NJ: Erlbaum. Overgaard, M., Rote, J., Mouridsen, K., & Ramsoy, T. Z. (2006). Is conscious perception gradual or dichotomous? A comparison of report methodologies during a visual task. Consciousness and Cognition, 15, 700–708. http://dx.doi.org/10.1016/j.concog.2006.04.002. Posner, M. I., & Snyder, C. R. R. (1975). Facilitation and inhibition in the processing of signals. In P. M. A. Rabbitt, & S. Dornic (Eds.). Attention and performance V. New York: Academic Press. Raaijmakers, J. G. W., & Neville, D. A. (2015). Is conscious processing required for long-term memory? In J. G. W. Raaijmakers, A. H. Criss, R. L. Goldstone, R. M. Nosofsky, & M. Steyvers (Eds.). Cognitive modeling in perception and memory: A festschrift for Richard M. Shiffrin (pp. 162–172). Retrieved from < https://bookshelf. vitalsource.com/#/books/9781134629329/ > .

21

Consciousness and Cognition xxx (xxxx) xxx–xxx

D.D. Sadler et al.

Raaijmakers, J. G. W., & Zeelenberg, R. (2004). Evaluating the evidence for nonconscious processes in producing false memories. Consciousness and Cognition, 13, 169–172. http://dx.doi.org/10.1016/j.concog.2003.10.002. Ramsoy, T. Z., & Overgaard, M. (2004). Introspection and subliminal perception. Phenomenology and the Cognitive Sciences, 3, 1–23. Reber, T. P., & Henke, K. (2011). Rapid formation and flexible expression of memories of subliminal word pairs. Frontiers in Psychology, 2. http://dx.doi.org/10.3389/ fpsyg.2011.00343. Reder, L. M., Park, H., & Kieffaber, P. D. (2009). Memory systems do not divide on consciousness: Reinterpreting memory in terms of activation and binding. Psychological Bulletin, 135, 23–49. Roediger, H. L., III (1990). Implicit memory: Retention without remembering. American Psychologist, 45, 1043–1056. Roediger, H. L., III, Balota, D. A., & Watson, J. M. (2001). Spreading activation and arousal of false memories. In H. L. Roediger, J. S. Nairne, & A. M. Surprenant (Eds.). The nature of remembering: Essays in honor of Robert G. Crowder (pp. 95–11). Roediger, H. L., III, & McDermott, K. B. (1995). Creating false memories: Remembering words not presented in lists. Journal of Experimental Psychology: Learning, Memory, and Cognition, 21, 803–814. http://dx.doi.org/10.1037/0278-7393.21.4.803. Roediger, H. L., III, & McDermott, K. B. (2000). Tricks of memory. Current Directions in Psychological Science, 9, 123–127. http://dx.doi.org/10.1111/1467-8721.00075. Roediger, H. L., III, Watson, J. M., McDermott, K. B., & Gallo, D. A. (2001). Factors that determine false recall: A multiple regression analysis. Psychonomic Bulletin & Review, 8, 385–407. http://dx.doi.org/10.3758/BF03196177. Rosenthal, R. (1991). Meta-analytic procedures for social research. Newbury Park, CA: Sage. Schacter, D. L. (1987). Implicit memory: History and current status. Journal of Experimental Psychology: Learning, Memory, and Cognition, 13, 501–518. Seamon, J. G., Luo, C. R., & Gallo, D. A. (1998). Creating false memories of words with or without recognition of list items: Evidence for nonconscious processes. Psychological Science, 9, 20–26. Smith, T. A., & Kimball, D. R. (2012). Revisiting the rise and fall of false recall: Presentation rate effects depend on retention interval. Memory, 20, 535–553. http://dx. doi.org/10.1080/09658211.2012.684881. Stadler, M. A., Roediger, H. L., III, & McDermott, K. B. (1999). Norms for words that create false memories. Memory & Cognition, 27, 494–500. Tse, C.-S., & Neely, J. H. (2005). Assessing activation without source monitoring in the DRM false memory paradigm. Journal of Memory and Language, 53, 532–550. Tse, C.-S., & Neely, J. H. (2007). Semantic and repetition priming effects for Deese/Roediger-McDermott (DRM) critical items and associates produced by DRM and unrelated study lists. Memory & Cognition, 35, 1047–1066. Tulving, E. (1983). Elements of episodic memory. Oxford: Clarendon Press. Underwood, B. J. (1965). False recognition produced by implicit verbal responses. Journal of Experimental Psychology, 70, 122–129. http://dx.doi.org/10.1037/ h0022014. Van Opstal, F., Reynvoet, B., & Verguts, T. (2005a). How to trigger elaborate processing? A comment on Kunde, Kiesel, and Hoffmann (2003). Cognition, 97, 89–97. http://dx.doi.org/10.1016/j.cognition.2004.12.011. Van Opstal, F., Reynvoet, B., & Verguts, T. (2005b). Unconscious semantic categorization and mask interactions: An elaborate response. Cognition, 97, 107–113. http:// dx.doi.org/10.1016/j.cognition.2005.04.005. Watson, J. M., Balota, D. A., & Roediger, H. L., III (2003). Creating false memories with hybrid lists of semantic and phonological associates: Over-additive false memories produced by converging associative networks. Journal of Memory and Language, 49, 95–118. http://dx.doi.org/10.1016/S0749-596X(03)00019-6. Wentura, D., & Frings, C. (2005). Repeated masked category primes interfere with related exemplars: New evidence for negative semantic priming. Journal of Experimental Psychology: Learning, Memory, and Cognition, 31, 108–120. http://dx.doi.org/10.1037/0278-7393.31.1.108. Windey, B., Gevers, W., & Cleeremans, A. (2013). Subjective visibility depends on level of processing. Cognition, 129, 404–409. http://dx.doi.org/10.1016/j.cognition. 2013.07.012. Windey, B., Vermeiren, A., Atas, A., & Cleeremans, A. (2014). The graded and dichotomous nature of visual awareness. Philosophical Transactions of the Royal Society B, 369, 201302882. http://dx.doi.org/10.1098/rstb.2013.0282. Zeelenberg, R., & Pecher, D. (2002). False memories and lexical decision: Even twelve primes do not cause long-term semantic priming. Acta Psychologica, 109, 269–284. http://dx.doi.org/10.1016/S0001-6918(01)00060-9. Zeelenberg, R., Plomp, G., & Raaijmakers, J. G. W. (2003). Can false memories be created through nonconscious processes? Consciousness and Cognition, 12, 403–412. http://dx.doi.org/10.1016/S1053-8100(03)00021-7.

22