|USFDC Home | USF Electronic Theses and Dissertations||| RSS|
This item is only available as the following downloads:
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 002000893
007 cr mnu|||uuuuu
008 090423s2008 flu s 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0002510
Balderston, Catherine C.
Recognition memory for emotional words :
b an event related potential study
h [electronic resource] /
by Catherine C. Balderston.
[Tampa, Fla] :
University of South Florida,
Title from PDF of title page.
Document formatted into pages; contains 79 pages.
Thesis (M.A.)--University of South Florida, 2008.
Includes bibliographical references.
Text (Electronic thesis) in PDF format.
ABSTRACT: Evidence suggests that emotion affects memory often yielding enhanced recall and recognition of stimuli with emotional content. The nature of the relationship between emotion and memory for words has been particularly difficult to parse in part because of the stimulus characteristics. For example, emotional words tend to engender greater levels of physiological and psychological arousal, which have also been shown to enhance memory. Inter item relatedness has also been suggested as playing a part in the observed effects (i.e., emotional words belong to a closed semantic category compared to neutral words and are therefore easier to remember). While the enhancement of memory for emotional material has been demonstrated across a variety of stimuli and experimental conditions, the neural underpinnings of these effects remain unclear. The Old/New effect is an event related potential finding where electrophysiologic waveforms elicited by previously presented stimuli (i.e., old) are more positive going than those elicited by stimuli that were not previously presented (i.e., new). A few prior studies have investigated Old/New effects for emotional words, mostly comparing negative to neutral words and failing to equate their stimuli for the crucial confounding effects of arousal and inter item relatedness. The present study employed event related potentials to investigate recognition memory for words of positive, negative, and neutral valences in a sample of thirty healthy college undergraduates. It was predicted that positive and negative words would yield greater participant accuracy, response bias and Old/New effects in comparison to neutral words. The observed results yielded some variability in support for all of the hypotheses and predictions that were made a priori. Possible explanations for these results are discussed and directions for future research recommended.
Mode of access: World Wide Web.
System requirements: World Wide Web browser and PDF reader.
Advisor: Cynthia R. Cimino, Ph.D.
t USF Electronic Theses and Dissertations.
Recognition Memory for Emotional Word s: An Event Related Potential Study by Catherine C. Balderston A thesis submitted in partial fulfillment of the requirements for the degree of Master of Arts Department of Psychology College of Arts and Sciences University of South Florida Major Professor: Cynthia R. Cimino, Ph. D. Geoffery Potts, Ph. D. Douglas Rohrer, Ph. D. Date of Approval: March 27, 2008 Keywords: Old/New effects, Electrophys iology, ERPs, Neuropsychology, Cognition Copyright 2008, Catherine C. Balderston
i Table of Contents List of Tables iii List of Figures iv Abstract v Introduction 1 Emotion and Memory 2 ERPs and Emotion 6 ERPs Associated with Episodic Memory 7 Old/New Effects 7 ERPs, Emotion, and Recognition Memory 15 Purpose 20 Hypotheses and Predictions 22 Behavioral 22 Hypothesis 1 22 Hypothesis 2 22 Hypothesis 3 22 ERP 23 Hypothesis 4 23 Hypothesis 5 23 Method 24 Participants 24 Apparatus 24 Materials 25 Design and Procedure 26 Results 31 Data Diagnostics 31 Behavioral Analyses 31 Hypothesis 1 31 Hit Rate/False Alarm Rate 31 Sensitivity 33 Hypothesis 2 33 Response Bias 34 Reaction Times 34
ii Hypothesis 3 38 ERP Analyses 38 Hypothesis 4 38 Traditional Old/New Comparison 39 Objective Old/New Comparison 43 Subjective Old/New Comparison 45 Hypothesis 5 53 Discussion 54 Limitations and Directions for Future Research 61 References 63 Appendices 75 Appendix A: Word Lists 76 Appendix B: Signal Detection Form ulas 79
iii List of Tables Table 1 Signal detection theory response categories 5 Table 2 Mean (SD) ERP Category tria l counts 30 Table 3 Mean (SD) Hit and False Alarm Rate s across valences (N=30) 32 Table 4 Old/New classifications according to response type 39 Table 5 Significant Old/New effect s observed according to spatial/temporal region of inte rest (left-most column), and comparison type (top row) 58
iv List of Figures Figure 1 Electrodes used in spatial/temporal analyses. 29 Figure 2 Mean reaction times (in ms) for each re sponse type response type. 35 Figure 3 Mean reaction times for highly confidant responses for each response type. 37 Figure 4 Bilateral frontal Old/New effect (n.b., vertical bars denote analysis epoch). 40 Figure 5 Interaction between Valence, Old/New status, and Hemisphere (n.b., vertical bars denote an alysis epoch). 42 Figure 6 Grand average waveforms from the parietal region (n.b., vertical bars denote analysis epoch). 44 Figure 7 Grand average waveforms of th e bilateral frontal Old/New effect (n.b., vertical bars denote analysis epoch). 46 Figure 8 Grand average waveforms of mean amplitudes associated with each response type within the early frontal (hemispheres were averaged together) spatial temporal region of interest (n.b., vertical bars denote analysis epoch). 48 Figure 9 Grand average waveforms from parietal regions associated with each response type (n.b., vertical bars denote analysis epoch). 50 Figure 10 Grand average waveforms from the late fron tal regions associated with each response type (n.b., vertical bars denote analysis epoch). 52
v Recognition Memory for Emotional Word s: An Event Related Potential Study Catherine C. Balderston ABSTRACT Evidence suggests that emotion affects me mory often yielding enhanced recall and recognition of stimuli with emotional cont ent. The nature of the relationship between emotion and memory for words has been particul arly difficult to pars e in part because of the stimulus characteristics. For example, emotional words tend to engender greater levels of physiological and psychological arousal, which have also been shown to enhance memory. Inter item relatedness has also been suggested as playing a part in the observed effects (i.e., emotional words belong to a closed semantic category compared to neutral words and are therefore easier to reme mber). While the enhancement of memory for emotional material has been demonstrated across a variety of stimuli and experimental conditions, the neural underpinni ngs of these effects remain unclear. The Old/New effect is an event related potential finding where electrophysiologic waveforms elicited by previously presented stimuli (i.e., old) are more positive going than those elicited by stimuli that were not previously presented (i.e., new). A few prior studies have investigated Old/New effects for emo tional words, mostly comparing negative to neutral words and failing to equate their s timuli for the crucial confounding effects of
vi arousal and inter item relatedness. The presen t study employed event related potentials to investigate recognition memory for words of pos itive, negative, and neutral valences in a sample of thirty healthy college undergradua tes. It was predicted that positive and negative words would yield greater particip ant accuracy, respons e bias and Old/New effects in comparison to neutra l words. The observed results yielded some variability in support for all of the hypothese s and predictions that were made a priori. Possible explanations for these results are discu ssed and directions for future research recommended.
1 Introduction The general overarching purpose of the pr esent study was to examine the pattern and timing of electrophysiological indices of recognition memory for emotional words. One consistent event related potential (ERP) e ffect associated with episodic memory is the Old/New effect. The Old/New effect refers to the tendency for previously presented (i.e., old) words to elicit more positive goi ng ERP waveforms than words that were not previously presented (i.e., new words). Previous research investigating Old/New effects for emotional words has predominantly focused on negative and neutral words only. The current study extended previous research with the inclusion of positive words to determine if prior Old/New effects generalized to this other class of emotional stimuli. Because the decision to respond old or new can be due to accurate memory, response bias, or some combination of the two, the current study examined Old/Ne w effects from multiple perspectives other than the conventional manner where only correct ly identified trials are used to compare waveforms associated with old and new trials Specifically Old/New effects were also examined from a Subjective perspective (w here the Old/New effect consists of differences in ERPs elicited from trials in which the subject endorsed items as being previously presented, regardless of accuracy ) and an Objective perspective (where ERPs to previously presented items are compared to those not previously presented, regardless
2 of accuracy). Previous research has high lighted the importance of such additional comparisons, as they are differentially se nsitive to individual response bias. The remainder of the chapter that follows critically reviews contemporary literature regarding emotion, memory, ERPs and their relevant permut ations. The review concludes with a summary and critique of ex isting literature, followed by a more detailed discussion of the specific purposes of th e current study, hypotheses and predictions suggested by the review and examined in this thesis. Emotion & Memory Studies pertaining to the inte raction of emotion and memo ry can be separated into two distinct categories: Studies where memory performance is assessed in participants with a pre-existing or induced mood state, a nd studies where the stim uli that are to be remembered by the subject vary with respect to emotional valence (positive, negative, and/or neutral). For the purpose of the cu rrent study, only the latt er will be reviewed. It has been well establishe d that memory for emotional material is better than memory for neutral information (Bradle y, Greenwald, Petry, & Lang, 1992; Cahill & McGaugh, 1995; Heuer & Reisberg, 1990; Phel ps, LaBar, & Spencer, 1997). The constructs of arousal and va lence are most commonly imp licated as factors underlying the observed phenomena of enhanced memory for emotional stimuli. These two dimensions have been confirmed via factor an alyses to account for the majority of the variance observed in verbal judgments of emotional stimuli (Russell, 1980; Lang, Bradley & Cuthbert, 1990). Arousal can be defined as a dimension of emotion that varies from calm to excitement (LaBar & Cabeza, 2006, p. 54). More generally however, in studies of
3 emotion and memory, arousal refers to the amount of stimulation (psychological and/or physiological) engendered in the subject when presented with a particular stimulus (in this case, emotional words). Evidence suggests that arousal has a variety of effects on the various aspects of memory processes (e.g., enco ding, consolidation, storage, retrieval.). Specifically, using healthy controls and patients with amygdala lesions, LaBar and colleagues (LaBar, Gatenby, Gore, LeDoux, & Ph elps, 1998) revealed that memory for high vs. low arousal words was better in the control group after longer delays suggesting that arousal may enhance memory by modulatin g the consolidation process. Soetens and others (Soetens, Casaer, D'Hooge, & Hueting, 1995) used a series of experiments to investigate the effects of arousal (induced by oral or intramuscular administration of amphetamine) on memory. Using recall and recognition tasks over various time delay intervals the authors demonstrated that admi nistration of amphetamine either before or after encoding positively modulated recall and recognition of the subjects memory for the verbal stimuli. The authors concluded th at the increased arousal (via amphetamine administration) improved memory by acting on consolidation proce sses as evidenced by greater enhancement of memory with the passa ge of time. Furthermore, Cahill and colleagues (Cahill, Prins, Weber, & McGaugh, 1994) demonstrated that the administration of a -adrenergic receptor blockade inhibited the enhancement of emotionally arousing words presum ably by preventing activation of -adrenergic receptors within the amygdala and modulati ng (arresting) the consolidation process by influencing hippocampal function. Clearly, arousal plays an important modulat ory role in memory processes. This knowledge is particularly salient to the st udy of memory for emotional stimuli because
4 emotional stimuli tend to possess higher levels of arousal. If accurate conclusions are to be drawn about the effects of affective vale nce on memory, arousal must be controlled for across valence conditions. Unfortunately, this is by far the exception rather than the rule in the existing literature. Decades of literature generally support the idea that emotionally valenced (positive or negative) stimuli are recalled and recognized more accurately and more quickly than neutral items. These enhancing effects of valence on memory have been observed across sensory modalities and for a wide variety of stimuli including pictures (Bradley et al., 1992), sounds (Bradley & Lang, 2000), videos (Cahill et al., 1996), events (Rubin & Schulkind, 1997), and words (Vanderp loeg, Brown, & Marsh, 1987). As noted above, the exact nature and magnitude of vale nce effects on memory are difficult to glean from existing literature due to confounds posed by the effects of arousal and lack of controlling for its influence. There are, however, a modest number of studies that have successfully isolated the eff ects of valence on memory. Ochsner (2000), for example, through a series of studies using the remember/know procedure (see Tulving, 1985 for details), determined that negative stimuli (pic tures) were recognized significantly more than neutral pictures. Positive items were recognized more often than neutral pictures, but not sufficiently to reach statistical significance. In a recognition memory task, the partic ipant is presented with a number of stimuli (e.g., words) to remember (i.e., a study or target list). Afte r the participant views the study list (and often following a specified de lay period), they are presented with the so-called test list composed of the previous ly viewed targets interspersed with novel items (i.e., foils). Test list items are gene rally presented individually and the participant
5 is prompted to decide whether or not the item was old (i.e., a target previously presented as part of the study list) or ne w (i.e., a foil that was not previ ously presented as part of the study list). Responses can be classified into one of four types according to the Table 1 set fourth by signal detection theory. Table 1. Signal detection theory response categories Was this item studied? Yes No Old item Hit Miss Item Novelty New item False Alarm Correct Rejection In addition to arousal and valence drivi ng the enhancement of memory, semantic cohesion has also been implicated as a possibl e contributing factor. Semantic cohesion refers to the tendency for emotional stimuli (particularly words) to belong to the same semantic category and therefore possess high levels of inter-item association. This theory emerged relatively recently in the literat ure (Maratos, Allan, & Rugg, 2000), has been empirically tested by just two further studies yielding conflicting results. In a small sample (n=13) of healthy undergraduat es, McNeely, Dywan & Segalowitz (2004) investigated recognition memory for negativ e and neutral words equated for semantic cohesion. Results indicated that despite hi gh semantic cohesion be tween both groups of words, the emotional words were still recogni zed more often. The authors also examined response bias to address findings often repor ted in prior studies of recognition memory for emotional stimuli where a tendency exists for false alarms rates to be larger for emotional items (a phenomenon also attributed to semantic cohesion according to the
6 Maratos group). In this part icular study, emotional foils elicited higher false alarm rates than the highly related (animals) neutra l comparison condition suggesting that the observed memory enhancement (as well as the response bias) is not due to semantic cohesion of emotional words. Conversely, in a study published in the same timeframe, Talmi & Moscovitch (2004) used three sets of words (emotionally related, neutral related, and neutral unrelated) in a se ries of experiments to test Maratos semantic cohesion hypothesis. Overall, the 60 participants recall ed significantly more of the related (both negative and neutral) than the unrelated words but the emotional words provided no additional recall enhancement than the rela ted unemotional words. Clearly, further replications of these experiments are needed before the role of semantic cohesion in memory for emotional words can be determined. ERPs and Emotion While emotion has been studied for centuries it was only relatively recently that scientists were able to measure the electro cortical activity associated with emotional processing. Because of their high temporal resolution (milliseconds), and stimulus locked nature, ERPs provide information abou t the timing and scalp distribution of neural activity that cannot be obtained through other methods. As such, emotional processing of pictures (Carretie, Iglesias & Garcia, 1997; Carretie, Igle sias, Garcia, & Ballesteros, 1997; Schupp et al., 2000) and words (Naumann et al., 1992) has generally been observed to evoke P300 components greater in amplitude when compared to nonemotional pictures and words. With the exception of perhaps clinical research comparing healthy control subjects and groups with neurological or psyc hiatric pathology ERPs are used most often
7 to investigate the neural underpinnings of em otion and another construct, most commonly memory. ERPs Associated with Episodic Memory There are two main ERP phenomena associat ed with episodic memory. The first, known as the subsequent memory effect or the difference due to memory (dm) effect, refers to differences in ERPs elicited at en coding by words that are la ter recalled at test. Specifically, ERPs recorded during the st udy phase and elicited by items that are subsequently remembered are more positive than those to items that are not remembered at test (see Johnson, 1995 for a review). Subs equent memory effects have been noted across a variety of stimuli (e.g., words, faces et c.) and in varying test formats (e.g., recall, recognition etc.) (e.g., Benson & Kutas, 1993; Fabiani & Donchin, 1995). Studies employing the subsequent memory paradigm th at use emotional stimuli consistently find the electrophysiological brain activity elicite d by subsequently remembered emotional items occurs significantly faster than activity associated with neutral stimuli (e.g., Dolcos & Cabeza, 2002). Because the paradigm u tilizes waveforms elicited by words that are later successfully retrieved and not those that are forgotten, the subsequent memory effect is thought to index activity associated with successful encoding of stimuli. The second ERP effect related to episodic memory refers to differences in ERPs elicited by items during the recognition phase of a memory test. These so called Old/New effects are the focus of this thesis and will be discussed in detail in the following sections. Old/New Effects The Old/New ERP effect is a robus t finding in recognition memory tasks where previously presented items (i.e., old or so called target items) elicit more positive
8 going (i.e., larger) waveforms than items that were not previously presented (i.e., new or foil items). Traditionally, the Old/New wavefo rms are composed solely of trials where the subject correctly responded to the stimuli; that is, hits and correct rejections. The remaining responses (misses and false alarms) are not used. Several spatiotemporal patterns have been identified within ERP studies of recognition memory and linked to various aspects of the recognition memory process (e.g., encoding, retrieval, etc.). The first component generally observed occurs at approximately 300-500 ms post stimulus, and is distributed bilaterally over frontal brain regions. The dual process model of recognition memory posits that recognition decisions may be based on both the actua l recollection of the item (i.e., the memory trace), and some degree of feeling of familiarity it may engender in the subject (e.g. Yonelinas, 1994). This so called early frontal Old/New effect has been described by Rugg and others (e.g., Donaldson & Rugg, 1998; M. D. Rugg, 1995; Tendolkar & Rugg, 1998; Wilding, Doyle, & Rugg, 1995) as represen ting the familiarity subcomponent of recognition within this dual process model. Support for th is idea is also provided by Curran (2000). Specifically, he had participan ts study lists of singul ar and plural words and then discriminate previously presented words from new words and related lures that were presented in the opposite plurality to that of the study words. Hypothesizing that the lures were comparable to the studied item s in terms of familiarity the finding that the early, frontally distributed Ol d/New effect differentiated the new words from the old words and the related lures was interpreted as evidence that this early Old/New effect represents familiarity. Additionally, this early frontal effect has been associated more so with know judgments when remember/know experiments are employed (Gardiner,
9 Java, & Richardson-Klavehn, 1996). The s econd Old/New effect occurs around 400-800 ms post-stimulus, is positive in polarity, and has a left parietal scalp distribution. This left parietal Old/ New effect is thought to represent the putative neural correlate of recollection for previously presented items. Support for this view has been provided by Rugg (1987). Specifically, the item enc oding phase of a recognition task was manipulated by instructing subjects to st udy items in a shallow manner, (determining whether or not the first and last letters of each word fell in alphabetical order) or more deeply (incorporating the word into a senten ce) and recorded subsequent ERPs during the test phase in an attempt to dissociate neural generators associated with recognition and familiarity. Rugg reasoned that words from the shallow study phase recognized at test were representative of familiarity whereas the deeply encoded words would reflect the recollection process more purely. Results revealed that only the deeply studied items (not shallow) elicited a left parietal Old/New ef fect leading the authors to interpret this particular ERP component/effect as an index of recollection. This left parietal Old/New effect has also been associated with remember judgments in remember/know experiments (Duzel, Yonelinas, Mangun, Heinze, & Tulving, 1997) and has been observed in association with the recollection of specific contextual information such as source and temporal time tags (i.e., the pariet al Old/New effect is observed when such information about an item can be recalled a nd conversely is not t ypically observed when such information cannot be recalled). Recal ling specific details about studied items is thought to provide evidence for strength of the underlying memory trace (recollection). The third and final observed Old/New effect occurs between 500-700 ms post stimulus, is also positive in polarity and has a right front al focus. This so called right frontal
10 Old/New effect is thought to reflect post retrieval/recollection processing (Rugg & Allan, 2000). Much of the empirical support for this interpretation is garnered from studies of false recognition. False recognition or fa lse memory refers to a phenomenon where individuals tend to endorse nontarget items that are highly associated with study list items (i.e., targets, a.k.a., lures) as being previously presented wh en they are in fact novel. Such tasks are thought to rely heav ily on post-retrieval monitoring because the memory trace is weak and/or compromise d by the highly associated lures. Curran, Schacter, Johnson and Spinks (2001) compared subjects with high false recognition rates to those with low false recognition rates from a false memory experiment and found that only the ERPs from subjects that were better at discriminating between studied items and lures showed right frontal differences between new items and the studied items or lures. Old/New effects were initially thought of as being comprised of the N400, the amplitude of which was attenuated by old items, and a subsequent late positive component, the amplitude of which was enhanced. The N 400 is a well-studied ERP component thought to index semantic aspects of sentence proces sing. In ERP studies of recognition memory where Old/New effects are extrapolated the firs t frontal Old/New effect is often referred to as the frontal N400 (FN400) as it is simila r (both in polarity and latency) to the wellknown N400 component. Importantly though, th e frontal scalp dist ribution observed in the Old/New effect differs from the N400 seen in language studies as that particular N400 has a centro-parietal focus. With the advent of dual-process theories of recognition memory came a need to integrate ERP findings. Prior theoretical explanations of recognition memory performance generally focused on single proc ess models that posited a single strength-
11 like memory trace or signal e.g. Shiffrin and Steyvers (1 997) retrieving effectively from memory; REM Dual process models are conceptualized as having two sub processes: recollection and familiarity (Jacoby & Kelley, 1992; Mandler, 1981) that interact and better account for previously unexpl ained or incongruent scientific findings. Recollection generally refers to the ability to locate and retrieve the specific memory trace associated with a particular item dur ing a recognition test. Familiarity is often conceptualized as a feeling of association or prior expe rience with a stimulus; an automatic process that gives rise to a se nse of pastness (Yonelinas & Jacoby, 1996). Recollection is accompanied by explicit details or evidence (e.g., contextual & temporal tags) from the encoding phase of the test whereas familiarity is not. ERP studies provided a unique contributi on to the understanding of dual-process theories because of their high temporal reso lution in addition to their aid in making inferences about potential ne uroanatomical loci of active neuronal populations. Despite the unique contributions of this methodology, di ssociation of recollection and familiarity was not evident at the outset. Rather, as is the case in most scientific research, elucidation was a lengthy process that even tually led to a better understanding of the subcomponents of recognition memory (according to dual-process models). Smith and Halgren (1989) believed the N 400 was elicited by the integration of the semantic attributes for the evoked item and th e current cognitive context. The result, in their view, was the formation of an episodic memory trace. Subsequent repetition of the item in the same context then reactivated the trace, preventing the formation of a standard (i.e., that typically seen in studies of language comprehens ion) N400. Instead, the late positive component was enhanced. Support for these views as well as lesion data to
12 bolster the dual-process model came from Smith and Halgrens report that left temporal lobectomy patients did not generate reliabl e old new effects to verbal information whereas right-sided lobectomy patients did. Furthermore, the authors noted that only a mild decrement in behavioral performance wa s observed in the left lesioned patients suggesting they had poor recollection but intact abilities to make familiarity judgments (thus yielding only mild accuracy impairments). In a subsequent attempt to tease apart r ecollection from familiarity Potter et al. (1992) used the anticholiner gic drug scopolamine to impair explicit memory and spare implicit memory. Prior research demonstrated that administration of such a receptor antagonist renders subjects impaired on free recall tasks (explicit memory; Kopelman & Corn, 1988) yet spares the ability to comp lete word-stem tasks accurately (implicit memory; Nissen, Knopman, & Schacter, 1987). Re sults indicated that, contrary to the findings of Smith and Halgren, their results s howed that impairments of recollection (i.e., explicit memory) do not yield reduced Old/New effects (early or late). Rugg and Nagy (1989) also demonstrated a lack of associ ation between memory retrieval and early Old/New effects using normal subjects. Specifically, the two used a continuous recognition task where study words were repeat ed after either 6 or 19 intervening words followed by a 45-minute delay period and administration of a traditional recognition memory task using words from the con tinuous recognition test. The continuous recognition task yielded Ol d/New effects beginning around 250ms post stimulus (thus encompassing both the P300 and the N400). The traditional recognition task also elicited Old/New effects the amplitudes of which were smaller and the latency longer (starting around 550ms post stimulus). Because of the increased latency (des pite continuing to
13 perform quite accurately) Rugg and Nagy postulated that the N400 component was not necessary to perform the task. Furthermore, they argued that early Old/New effects were related to the time delay (i.e., recollection, or the memory trace was weak or absent after 45 minutes so judgments were made according to familiarity levels) between the two tasks rather than discrimination itself. Ta ken together it has been suggested (e.g., Rugg, Brovedani, & Doyle, 1992; Smith, 1993) that early Old/New effects do not seem to contribute to recognition memory processes ove r intervals of more than a few minutes. Unlike early effects, late Old/New eff ects persist over long delays between study and test phases. For this reason, it is genera lly assumed that late Old/New effects reflect the neural underpinnings of recognition memory. In 1990, Rugg had participants detect non-words mixed amongst a list of actual words, some of which were repeated at ei ther high or low levels to investigate electrophysiological responses to words of varying frequencies (an effect observed behaviorally that words of lower frequency show enhanced recollection compared to high frequency words). ERP results indicated th at the repetition of low frequency words tended to enhance the amplitude of the late positive wave but repetition of high frequency words did not. In light of his findings, Rugg hypothesized that the observed frequency effect could be indexing familiarity and raised the question of whether or not recollection and familiarity could be teased apart within late Old/New effects. Follow-up experiments were conducted (Rugg & Doyle, 1994; Rugg et al., 1992) to inves tigate the possible dissociation of Old/New effect s using this indirect task. Using a delay period of 20 minutes Old/New effects were larger for lo w than for high frequency words suggesting
14 that the previously observed word frequenc y effect was driven by the influence of familiarity. Paller and Kutas (1992) used a novel appro ach to investigate Old/New effects. Specifically, subjects were requ ired to process a list of words both orthographically and, in a separate list, imaginal processing. They subsequently presented an additional list of words composed of items from the orthographi c and imaginal lists along with new, never presented words. ERPs elicited by correctly identified items diffe red according to how they were encoded (studied) in that words fr om the imagery task elicited more positive waveforms then those from the orthographi c processing study phase. The authors reasoned that words from the imaginal task were judged by subjects as having been experienced more recently then the words from the orthographic task and interpreted the findings as evidence for a neural sign ature for conscious recollection. In an attempt to dissociate recollection and familiarity, Gardiner and Java (1991) employed a paradigm in which subjects endor sed descriptors to re flect their recognition judgments. Specifically, after identifying a test item as being old, participants reported whether their decision was based on an explic it memory of its initial presentation (a remember judgment) or on the basis of a fam iliarity feeling in absence of an explicit memory of the initial presentation (a know judgment). Results indicated Old/New effects were larger for the words judged as being remembered than for the know judgments. Smith (1993) late r replicated these findings and concluded that Old/New effects were generated by neural processes engaged during recollection (not familiarity) in recognition memory tasks. Importantly, in all of the Old/New studies employing remember/know judgments, words judged ol d on the basis of familiarity (know
15 judgments) elicited reduced, but not absent Old/New effects when compared to remember judgment words and with the same accompanying scalp distribution (i.e., midline foci). In sum, Old/New effects refer to the te ndency for ERPs elicited by previously presented words to be more positive going th an those elicited by words that were not previously presented. These robust effects are conceptualized as consisting of an increased positivity superimposed on the we ll-studied N400 component and are often divided into two categories: early and late Old/New e ffects. Early effects are hypothesized to be associated with the familiarity component of recognition memory whereas the late counterpart is thought to index recollection. ERPs, Emotion, and Recognition Memory Numerous ERP studies exist within the domains of emotion and memory separately. Relatively few, however, have examined the ERPs resulting from recognition memory for emotional stimuli. Johansson, Mecklinger, & Treese (2004) re cently examined Old/New effects in recognition memory for faces differing in affect (i.e., positive, negative, and neutral). Results indicated that the negative faces we re remembered more frequently than the neutral and positive faces as evidenced by a greater parietal Old/New effect associated with the negative faces. Furthermore, frontally distributed ERPs elicited by correctly rejected new faces were modulated in a posit ive direction by the ne gative valence in the correctly rejected new faces (i.e., negative faces elicited more positive going waveforms than positive and neutral faces). Therefore, the authors concluded that emotional stimuli appear to relax the criterion set by prefrontal cortical area s resulting in a bias towards
16 such stimuli as evidenced by better recognition and greater parietal Old/New effects. These findings were a replication and extens ion of Ochsners (2000) experiment where participants recognized negative photos more accurately than their positive and neutral counterparts. Several investigators have examined the effects of emotional context on recognition memory performance. By testing recognition for emotionally neutral objects that were associated with positively or ne gatively valenced background context during the tasks study phase, Smith, Dolan, and Rugg (2004) were able to probe ERP correlates of retrieval for emotional (and nonemotional) material. Results did not support the previous fMRI findings of Maratos and Rugg ( 2001) in that the emotional context of the stimuli failed to affect the amplitude of the left parietal Old/New effect. Although words have not traditionally been used to study the interplay between emotion and memory because th eir elicited responses are le ss intense than that of pictures, they possess many desirable qualities as stimuli. Most importantly, a substantial literature regarding the factor s thought to affect memory for emotional stimuli (e.g., frequency, recallability, concrete ness) has given rise to norm ative data (e.g., arousal and valence ratings) for stimuli allowing for greater experimental contro l of these important variables. ERP studies of recognition memo ry for emotional words have been conducted most recently and are therefore fewer in number. Using a small (n = 16) sample of healthy young adults Maratos et al. (2000) examined Old/New effects of emotionally ne gative and emotionally neutral words in a recognition memory paradigm. Behavioral data revealed a false alarm rate nearly twice as large for negative words when compared to their neutral c ounterparts. That is, subjects
17 were more likely to incorrectly identify ne w negative words as having previously been presented on the study word list. In additi on, electrophysiological responses elicited by correct rejections of negative items were la rger in magnitude and identical in scalp topography to those elicited by the neutral words. Because the authors did not find evidence for qualitatively distinct neural systems as a function of valence they concluded that the results bolstered th eir semantic cohesion hypothesis. That is, emotional words influence memory because they belong to the same (or similar) semantic category. Windmann and Kutas (2001) later attempted to replicate the findings of Maratos, Allan, and Rugg (2001) and extend the study to investigate subjective Old/New effects. Traditionally, the Old/New waveforms are comp osed solely of trials where the subject correctly responds to stimuli; that is, hits and correct rejections. The remaining responses (misses and false alarms) are not used. Subjective Old/New effects refer to ERP waveforms elicited by words that the particip ant judges to be old (i.e., hits or false alarms) or new (i.e., correct rejections or mi sses) regardless of whether or not the items are actually old or new. Unlike Mara tos findings, Windma nn and Kutas found no quantitative Old/New differences as a result of valence within the Traditional/Correct comparison (i.e., using correct tr ials only). Analysis of the Subjective Old/New effects, however, revealed that only neutral (not negative) items elicited a large Old/New difference over prefrontal scalp regions during the early epoch (300-500ms). Behaviorally, however, participan ts classified negative words faster and more frequently as being old regardless of whether or not the items were actually old. The authors interpreted the findings as being congruent with the idea of the prefr ontal cortex relaxing the criterion for negative stimuli so that they are better remembered and are less
18 frequently overlooked. The aut hors also argued that this f unction plays an adaptive role to protect against potentially threatening situations. Dietrich et al. (2001) employed a co ntinuous recognition memory test to investigate the effects of negative, neutral, and positive words on Old/New effects. Subjects were presented with an ongoing (as opposed to th e typical recognition memory format where subjects are presented with a list of words followed by a delay period and then the recognition por tion of the test) list of words and asked to make decisions about whether each word had been seen before (i .e., old) or not (i.e., new). The greatest Old/New effects were elicited by negative words, followed by positive and then neutral items. This was the first study to utilize positive words as stimuli in their experiment. However, two significant limitations of this study are that 1) vale nce ratings of words were obtained from a very small sample of subjects (n = 12) and 2) words were not balanced for levels of arousal across the th ree different valence conditions. Emotional stimuli are associated with higher levels of arousal and have been shown to affect recall and recognition (see Christianson, 1992) for a review. Furthermore, negative words tend to possess higher levels of arousal than th eir positive or neutral counterparts (Hamann, 2001). In a subsequent study, Windmann, Urbach, & Kutas, (2002) divided their sample (n = 30) into two groups based on individua l response styles (i.e., bias; high or low predisposition to endorse items as having b een previously presented) and examined Old/New effects for negative and neutra l words according to the following classifications: Traditional/Correct Old/New effects where only the correct trials are compared, Subjective Old/New effects where co mparisons are made between trials that
19 the subject believes to be old are those the subject believes to be new (regardless of whether or not the items are actually old or new), and Objective Old/New effects where all of the trials are used to compare wave forms elicited by old vs. new items (regardless of accuracy). Recognition accuracy and ERP pa tterns elicited by items that actually were old versus new (Objective Old/New effect) we re comparable between the two groups. It was not until items were analyzed according to the subjects perspectives as to which items were old and new (Subjective Old/New effect) that differences emerged. These differences were maximal over prefrontal sights and occurred between 300-500 ms poststimulus leading the authors to conclude that response bias influences early memory retrieval processes. The Traditional/Correct classificat ion (i.e., hits and correct rejections) yielded differences intermediate to the Objective and Subjective Old/New effects. Most recently, Inaba, Nomura, and Ohir a (2005) used positive, negative, and neutral words to examine Old/New effects in 20 students. They found greater Old/New effects for negative targets, followed by positive, and then neutral targets. These differences were maximal at midline and left cen tro-parietal sites. As was the case for all but perhaps one (Maratos study is unclear) of the above ERP studies of recognition memory for emotional words, items were not balanced for arousal; only valence. Therefore, interpretation of results is difficult, at best, due to the tendency for arousal to influence memory performance and to be higher for emotional words as compared to neutral words. Furthermore, for all bu t the two studies from the Windmann group, Old/New effects were extrapolated from ERPs elicited by correct trials only. This is the traditional method of measuring Old/New e ffects. However, as the Windmann group
20 revealed, Old/New effects composed of correct trials only can be confounded by response bias in that correct trials could be due to accurate memory, response bias, or a combination of memory and bias. Therefor e, accurate conclusions cannot be made without accounting for this individual subject bias. Additionally, by examining the full range of responses (hits, misses, false alarms, and correct rejections) and the accompanying Old/New effects one is able to obtain a more comprehensive view of the Old/New effects and how they interact with varying levels of response bias. Purpose The current study had two specific purpos es. The first, overarching purpose was to investigate the pattern and timing of electrophysiol ogical indices of Old/New recognition memory effects for negative, neutra l, as well as positive words. Secondly, this study examined behavioral and electrophys iological indices of subject response bias in Old/New recognition memory. This investigation attempted to replicate previous findings of increased positivity elicited by negative words previously presented in a recognition memory paradigm (i.e., the Old/New effect) and extend those findings to include positively valenced words. Of the few existing studies in the literature on the influence of valence on Old/New effects, only two have used positive words in addi tion to negative and neutral ones (both investigations have yielded findings indi cating greatest Old/New effects for negative words, followed by positive words, and then neutral). However, neither of these studies controlled for the confounding effects of arousal on memory. Investigating positive words will help elucidate the mechanisms by which these phenomena operate.
21 Windmann, Urbach, and Kutas (2002) ha ve demonstrated the importance of examining Subjective Old/New effects wh ere the Old/New status of words are determined solely by the subjects response of old versus new as opposed to the standard comparison where only the ERPs elicited by correct classification of responses. Traditional analyses of Old/ New effects are confounded by response bias due to the fact that recognition memory judgments can resu lt from accurate memory, bias, or some combination of the two. The Windmann st udy used three separate types of trial comparisons to probe varying degrees of res ponse bias within Old/Ne w effects that will also be used in the present study. They are: Traditional/Correct Old/New Effects. Th is classification utilizes ERP waveforms elicited only by trials where the participant correctly re sponded targets (i.e., hits) or foils (i.e., correct rejections). Objective Old/New Effects. This classi fication utilizes ERP waveforms elicited by targets (i.e., hits and misses) and foils (i .e., correct rejections and false alarms). This includes all the particip ants responses (correct and incorrect) regardless of whether or not they were correctly identif ied by the participant as old or new (i.e., hits, correct rejections, false alarms, and misses) Subjective Old/New Effects. This classi fication utilizes ERPs elicited by words that the participant judges to be old (i.e., hits or false alarms) or new (i.e., correct rejections or misses) regardless of whet her or not the items are actually old or new. Using groups with comparable levels of recognition memory accuracy and ERP patterns to Objective Old/New items, but either with or without a predilection (i.e., high
22 bias and low bias, respectively) to re spond old, the Windmann group found significant differences between the high and low bias groups when the Subjective Old/New classification was used. Furthermore, they found the Traditional/Correct Old/New comparison to have peak amplitude differe nces intermediate to the Objective and Subjective comparisons. These findings suggest that response bias influences the magnitude of the observed Old/New effects and that valid interpretation of Traditional/Correct Old/New effects cannot be made without accounting for this subject response bias. Hypotheses and Predictions Behavioral Hypothesis 1: Based on prior research, it is hypothesized that positive and negative words will elicit more hits and false alar ms than neutral items. Hypothesis 2: The bias to respond old will be greater for negative words and positive words (compared to neutral items) and reaction times will be shorter (i.e., participant responses will be faster) for correct responses and responses to negative and positive words compared to neutral words. Hypothesis 3: If emotional words are truly remembered more frequently due to their high semantic cohesion then controlling for seman tic cohesion across posit ive, negative, and neutral words should eliminate any enhancement of the emotional words as compared to the neutral items. If, however, the bias fo r negative words seen in previous research
23 exists as a function of adaptation to potenti ally threatening stimuli, then positive words should yield behavioral data equivalent to that of the neutral items. ERP Hypothesis 4: With respect to the ERP data, it is hypot hesized that all three word classes (i.e., negative, neutral, and positive) will elicit Old/New effects. The largest effects are predicted to be within the Traditional/Correct classifica tion, followed by the Subjective and the Objective classifications (respectively). Hypothesis 5: Based on Windmann and Kutas assertion th at bias for negative words serves an adaptive function, whereby the cognitive sy stem is prompted to assign greater significance and a higher priority to the processing of potentially threatening stimuli is correct it is predicted that Old/New effects will be greatest for negative items compared to positive and neutral because the positive and neutral items lack threatening connotations.
24 Method Participants Forty-five participants were initially recruited from the on line participant pool of undergraduate psychology students at the University of South Florida. Fifteen participants were subsequen tly excluded from the study for the following reasons: nonnative English speakers (2 females), stimulus presentation software crash (1 male), elevated depression scores (>10 on the Beck Depression Inventory; BDI) (3 females), braided hair/extensions rendering proper el ectrode placement impossible (2 females), extremely low trial counts in multiple categories (e.g., 2 negative misses) (1 male), and too many (>15%) bad channels (due to motion ar tifact, eye blinks etc.) for the EEG data to be averaged (5 females). Also, one partic ipant refused to consent, as she did not wish to get her hair wet. The final sample consis ted of thirty participants (15 male, 15 female) with a mean age of 20.3(SD = 3.3) years. Of these remaining participants all were right handed, had normal or corrected to normal vision, reported a history free of major neurological and psychiatric symptoms (inc luding any head injury with a loss of consciousness > 10 minutes), and were native English speakers, above the age of 18. Apparatus All stimuli (i.e., words) that composed the recognition memory test were presented using a DELL Genuine Intel x86 Fa mily 6 model 8 computer and a 21-inch Sony Multiscan 220GS monitor. The computer software E-prime (version 3.0; Schneider, Eschman, & Zuccolotto, 2002) was used to present all recognition memory
25 stimuli, collect responses and valence ratings. Participants responses were recorded via a standard keyboard. Collection of ERP da ta was carried out through the use of the Electrical Geodesics Incorporated Sy stem200 (EGI, Eugene, OR). Brain electrophysiology was recorded with a 128-channel Electrical Geode sics Incorporated sensor net in conjunction with NETSTATION 4.2 acquisition software powered by a Macintosh G4 computer. Electroencphalographic data were sampled at 250Hz. Materials In addition to the aforementioned ERP e quipment, materials included two selfreport questionnaires and two published bodies of words containing extensive normative data on many variables. Due to the nature of the current study (p erception/judgment of emotional stimuli) depressive symptoms were assessed using the Beck Depression Inventory (2nd edition; BDI-II; Beck & Steer, 1987) scale to avoi d any confounds from participants with affective disturbances. The BDI-II has demonstr ated excellent reliability/validity, is brief to administer, and is often used as a screen ing tool in research settings (Beck, Steer, & Garbin, 1988). The Edinburgh Handedness Inve ntory (Oldfield, 1971) was used to assess type and degree of laterality. This 22 item self-report measure is preferred because it assesses not only hand preference, but foot and eye preferences as well. These additional preference types have been shown to reflect a more accurate representation of degree and type of laterality (Williams, 1991) Five hundred and twenty-eight total (i.e., 176 positive, 176 negative and 176 neutral) words were obtained from the A ffective Norms for English Words (ANEW; Bradley & Lang, 1999). This large body of word s has been rated in terms of pleasure,
26 arousal, and dominance. A random number gene rator was used to assign a number to the words in each valence list (numbers 1 though 176) The first half (numbers 1 through 88) of each valence list became part of the pool of targets and the second half (numbers 89 through 176) of each valence list became part of the pool of foils. Target words did not differ significantly according to arousal F (2,261) = .196, p=.822 or frequency F (2,252) = 2.69, p=.070 Additionally, the University Colorados Latent Semantic Analysis (LSA; Landauer, Foltz, & Laham, 1998) program was used to equate for effects due to high semantic cohesion among target words. LSA is a theory and method for extracting and representing the contextual -usage meaning of words by statistical calculations applied to a large body of text. The analysis is based on the hyperspace analogue to language (HAL) model of memo ry and works by encoding a co-occurrence learning algorithm of the context in which words occur. Through the use of the LSA program target words did not differ si gnificantly across the three valences F (2,261) = 1.02, p=.363 Word lists can be found in Appendix A. Design and Procedure After procedures were explained, and an opportunity for participants to ask questions was provided, written informed c onsent was obtained. The first portion of the study included administration of the pape r and pencil measures (i.e., demographic information, BDI-II, and Edinburgh Handedness Inventory). During this time the experimenter prepared an electrolyte solution composed of 1 liter distilled H2O, 1.5 teaspoons of NaCl, and .75 teaspoons of baby shampoo and submersed the appropriately sized net for absorption of said solution. Upon completion of all questionnaires, the participants head was measured, their ve rtex was marked and the128 channel net
27 described above was fitted on the participants head and adjusted as needed for proper fit and to insure channel im pedances were below 50k Once the participants were seated in the experiment room alone in front of th e monitor the preprogram med instructions lead them through the remainder of the experiment The participants were instructed to memorize the list of words for a subsequent me mory test and to try to remain still. Words were randomly presented (one at a time) in the middle of the screen for 300 milliseconds (ms) with an interstimulus in terval of 2200 ms. After presentation of the two-hundred and sixty-four target/study words, participants were moved to a different room and asked to engage in a five minute distracter task consisting of multiplication problems. Immediately following the delay pe riod subjects were m oved back into the experiment room, impedances were rechecked (electrodes rehydrated as needed) and the study commenced with the recognition memory test Participants used either their right or left hand (counterbalanced across subjects ) to indicate whether each word (presented individually) was old or new according to a c onfidence rating scale. Test list words were also presented in random order. Instruc tions to the participant were as follows: If you are: Highly confidant the word was studied, press 1 Less confidant the word was studied, press 2 Less confidant the word was NOT studied, press 3 Highly confidant the word was NOT studied, press 4 Two hundred and sixty-four foil words ( 88 positive, 88 negative, 88 neutral) in addition to the 264 target words from the study phase comprised the recognition portion
28 of the test. All words were displayed for a period of 400ms with each subsequent word appearing 1600ms after the pa rticipant gave a response. Upon completion of the study, ERPs were digitally filtered with a 20hz low-pass filter and segmented into 1600ms epochs around the test list words (200ms before/1400ms after the presentation of each test word). Epochs were screened for noncephalic artifact and marked as bad (i.e., excluded from further analysis) if they contained more than 10 bad channels. Indivi dual participant files with fewer than 15% good trials per category were excluded from further analysis. ERPs were baseline corrected by 200ms and transformed using an average reference montage. Individual participant ERPs were averaged together to create grand average waveforms. Mean amplitude values were used as the depende nt variables in the following time windows: 300-640 ms post stimulus over frontal leads (electrodes 12, 18, 19, 23 and 24 composed the left frontal region, and electrodes 3, 4, 5, 10, and 124 composed the right frontal region) to capture the early bilateral frontal Old/New e ffect, 400-1000 ms post stimulus over parietal leads (electrodes 31, 36, 37, 41, 42, 47, 51, 52, 53, 54, 59, 60, and 61 composed the left parietal region, and electrodes 78, 79, 85, 86, 87, 91, 92, 93, 97, 98, 103, and 104 composed the right parietal region) to capture the left parietal Old/New effect, and 400-1400ms over frontal lead s (electrodes 12, 18, 19, 20, 22, 23, 24, 26, 27, 28, 29, 33, 34, and 35 composed the left frontal region, and electrodes 2, 3, 4, 5, 9, 10, 110, 111, 116, 117, 118, 122, and 123 composed the right frontal region) to capture the late right frontal Old/New effect. Spatial/te mporal regions of inte rest are graphically depicted in Figure 1. All of the aforementioned mean am plitudes reflect the average voltage value of channels within each montag e. See Table 2 for descriptive statistics for
each of the trial categories (i.e., number of hit, miss, false alarm, correct rejection trials retained for analysis). Grand average wave form amplitudes were topographically plotted and visually inspected to determine th e regions of intere st listed above. Figure 1. Electrodes used in spatial/temporal analyses. 29
30 Table 2. Mean (SD) ER P category trial counts Positive Hit: 48.17(11.12) False Alarm: 36.66(15.01) Miss: 30.53(10.02) Correct Rejection: 42.63(15.84) Negative Hit: 52.57(9.45) False Alarm: 33.27(11.29) Miss: 26.17(8.79) Correct Rejection: 45.10(13.10) Neutral Hit: 46.30(10.73) False Alarm: 28.43(14.15) Miss: 33.43(9.85) Correct Rejection: 51.30(15.13)
31 Results Data diagnostics Before hypothesis testing all data were screened to determine whether or not the necessary assumptions were met to conduct va lid statistical analyses. Histograms and boxplots were visually examined to confirm th at the assumption of normality was met. Several significant outliers existed within the behavioral data so an alyses that included these variables were run twice (once with and once without th e addition of the participant with the extreme score). Exclusion of outlier s did not yield a differe nt pattern of results so analyses reported below included the entire sample. Skewness and kurtosis values were also examined for each of the variable s and determined to be within normal limits. Examination of sphericity tests revealed that all of the variables met this assumption. Behavioral Analyses Hypothesis 1: Positive and negative words will elicit more hits and false alarms than neutral items. Hit Rate/False Alarm Rate Participant responses of either a (i.e., Highly confidant the word was studied) or a (i.e., Less confidant the word was studied) were coded as an old endorsement. Conversely, a response of either a (i.e., Less confidant the word was NOT studied) or a (i.e., Highly confid ant the word was NOT studied) was coded as a new endorsement. As such, a hit can be thought of as an old response to an old
word (i.e., target). A false alarm is an old response to a new word (i.e., foil). A correct rejection is a new response to a new word and a miss is a new response to an old word. Each participants Hit Rate (probability of old items that are correctly classified as old) and False Alarm Rate (i.e., the probabi lity of new items that were incorrectly classified as old) were calc ulated and served as the depe ndent variables in a one-way MANOVA; Valence served as the independent variable. Means and standard deviations for each Hit Rate and False Alarm Rate from the three experimental conditions (i.e., positive, negative, and neutral) are listed in Table 3. Results of the overall MANOVA indicated the Hit Rate and False Alarm Rate differed significantly from each other as a function of Valence = .314; F (4,26) = 14.17, p < .001 Table 3. Mean (SD) Hit and False Alarm Rates across valences (N=30) Hit Rate False Alarm Rate Positive .60(.12) .45(.19) Negative .67(.10) .43(.14) Neutral .58(.12) .36(.17) Univariate ANOVAs for Hit Rate and Fals e Alarm Rate both re vealed significant main effects of Valence F (2,58) = 14.96, p < .001 and F (2,58) = 12.68, p < .001 respectively For Hit Rate data, Bonferroni adjust ed pairwise comparisons revealed that negative words elicited a significantly higher hit rate (M=.67, SD = .10) then positive 32
33 words (M = 60, SD = .12, p<.01) and neutral wo rds (M = .58, SD = .12). The Hit Rate for positive words was not significantly differe nt than the Hit Rate for neutral words (p=.18). Analysis of the False Alarm Rate using B onferroni adjusted pairwise comparisons revealed negative and positive words elicited significantly higher False Alarm Rates (M =.43, SD = .14 and .45, SD = .19) than neutral words (M = .36, SD = .17) words (p<.01 and p<.001 respectively. The False Alarm Rate for negative words did not differ significantly from the False Alarm Rate for positive words (p=.86). Sensitivity An ancillary analysis was conducted using the sensitivity index d as this index takes into account both hits and false alarms. The formula used to derive d can be found in Appendix A. When d was entered as the dependent va riable and Valence was kept as the independent variable in a one-way repe ated-measures ANOVA, the previously seen main effect of Valence emerged again F (2,58) = 15.03, p < .001 Bonferroni adjusted paired comparisons revealed accuracy was si gnificantly greater for negative (M = .65, SD = .36) and neutral words (M = -.61, SD = .34) compared to positive words (M = .41, SD = .37). The difference between neutral and ne gative words was not significantly different (p=1.0). Hypothesis 2: The bias to respond old will be greater for negative words and positive words (compared to neutral items) and reaction times will be shorter (i.e., participant responses will be faster) in re sponse to negative and positive words compared to neutral words.
34 Response Bias Results of the repeated-measures ANOVA where the three level (i.e., positive, negative, and neutral) independent variable of Valence was entered and the bias index C served as the dependent variable re vealed a main effect of Valence F(2,58) = 10.67, p <.001 Bonferroni adjusted paired comparis ons revealed that participants adopted a significantly more liberal res ponse bias for negative (M = -.13, SD = .30) and positive (M = -.04, SD = .35) compared to neutral words (M = .10, SD = .39; p<.001 and p = .011 respectively). The difference between negative and positive words was not significantly different (p=.32). The formula used to derive C can be found in Appendix A. Reaction Times A two-way repeated measures ANOVA was employed to test differences in median reaction times of the four possible respon se types (i.e., hit, miss, false alarm, and correct rejection). For this analysis Valence (with the three levels of positive, negative, and neutral) as well as Response Type were entered as independent variables and Median Reaction Time as the dependent variable. Re sults indicated a trend towards a significant main effect of Valence F(2,58) = 2.66, p = .078 and a main effect of Response Type F(3,87) = 9.74, p < .001 Bonferroni adjusted pair wi se comparisons revealed reaction times were significantly faster (i.e., smalle r) when participants responded correctly to targets (i.e., hits; p<.001) compared to all ot her response types. Median reaction times for each of the four response types (i.e., hit, miss, false alarm, and correct rejection) can be seen in Figure 2. Although a Valence tre nd was observed it was not in the expected direction as the longest reaction time was observed for negative words.
Figure 2. Mean reaction times (in ms) for each response type. Since participant responses were made acco rding to various levels of confidence (as opposed to yes/no only) ancillary analyses were conducted to dete rmine what role, if any, level of confidence contributed to reac tion times. Chi square tests were employed first to determine if the frequency of each po ssible rating response differed within each of the four response types (hit, miss, false alarm, correct rejection). Within the hit response type there were significantly more (i .e., Highly confidant the word was studied) responses 2 (1, N = 30) = 393.32, p < .001, indicating hits were responded to more frequently with a highly conf idant rating as opposed to th e less confidant rating (i.e., ). Within the miss response type there were significantly more responses (i.e., Less confidant the word was NOT studied) 2 (1, N = 30) = 57.84, p < .001, indicating 35
36 misses were responded to more frequently with a less confidant rating as opposed to the highly confidant rating (i.e., Highly confid ant the word was NOT studied). Within the false alarm response type there were significan tly more responses (i.e., Less confidant the word was studied) 2 (1, N = 30) = 21.54, p < .001, indi cating false alarms were responded to more frequently with a less confidant rating as opposed to the highly confidant rating (i.e., Highly confidant th e word was studied). Within the correct rejection rating there were significantly more responses (i.e., Less confidant the word was NOT studied) 2 (1, N = 30) = 33.50, p < .001, indica ting correct rejections were responded to more frequently with a less confidant rating as opposed to the highly confidant rating (i.e., Highly confid ant the word was NOT studied). Given the significant differences among th e frequency of confidence (i.e., highly confidant vs. less confidant) ratings for each response type (i.e., hit, miss, false alarm, correct rejection) reaction time data was re-a nalyzed using a two-way repeated measures ANOVA where Confidence Level (h ighly confidant, less confidant) and Response Type (hit, miss, false alarm, correct rejection) were entered as independent variables and reaction time stood as the depe ndent variable. Results i ndicated a main effect of Response Type F(3,87) = 3.45, p = .02 as well as a main effect of Confidence Level F(1,29) = 12.29, p = .002 A significant interaction between Response Type and Confidence Level was not observed F(3,87) = 1.99, p = .121 Bonferroni adjusted post hoc tests indicated that highly confidant responses (M = 1174.69, SE = 72.74) were made significantly faster than less c onfidant responses (M = 1567.45, SE = 119.80). In order to reveal possible reaction tim e differences between the four response types, less confidant rati ngs were discarded and highl y confidant trials were
reanalyzed. In a one-way repeated measur es ANOVA where Response Type (hit, miss, false alarm, correct rejection) served as th e independent variable and Reaction Time as the dependent variable results indicated a main effect of Response Type F(3,87) = 13.29, p < .001 Bonferroni adjusted post hoc tests revealed hit respons es (M = 1006.62, SE = 56.20) were made significantly faster than false alarms (M = 1097.88, SE = 67.80; p = .025), misses (M = 1295.43, SE = 98.69; p < .01), and correct rejections (M = 1298.82, SE = 92.16; p < .001). False alarms were made significantly faster than correct rejections (p= .022) and there was a trend towards false alarms being made signi ficantly faster than miss responses (p = .058). Reaction times asso ciated with correct rejections and misses did not differ from each other. These r eaction times can be seen in Figure 3. 0 200 400 600 800 1000 1200 1400 Hit MissFalse AlarmCorrect RejectionMs Figure 3. Mean reaction times for highly c onfidant responses for each response type. 37
38 Hypothesis 3: If emotional words are truly remembered more frequently due to their high semantic cohesion th en controlling for semantic cohesion across positive, negative, and neutral words should eliminate any accuracy enhancement of the emotional words as compared to the neutral items. If, however, the accuracy enhancement of negative words seen in previous research exists as a function of adaptation to potentially threatening stimuli, then positi ve words should yield behavioral data equivalent to that of the neutral items. As noted above, participan ts did exhibit accuracy di fferences as a function of valence (i.e., accuracy, as indexed by d was significantly greater for negative and neutral words compared to positive items). The current study did control for possible effects of semantic cohesion. Thus, contro lling for semantic cohesion across stimuli did not eliminate the increased accuracy for negative words in this sample. ERP Analyses Hypothesis 4: With respect to the ERP dat a, it is hypothesized that all three word classes (i.e., negative, neutral, and positive) will elicit Old/New effects. The largest effects are predicted to be within the Traditional/Correct classification, followed by the Subjective and the Objective clas sifications (respectively). For analysis of ERP data, repeated measures ANOVAs were conducted using Valence (negative, neutral, positive), Old/New status (old, new), and Hemisphere (left, right) as IVs, and mean amplitudes (average d across channels within each region of interest) from each of the spatial/temporal regions of interest as DVs. Refer to Table 4 for Old/New classification type information.
39 Table 4. Old/New classifications according to response type OLD NEW Traditional Hits Correct Rejections (CRs) Subjective Hits & False Alarms (FAs) CRs & Misses Objective Hits & Misses CRs & FAs Traditional Old/New Comparison A repeated measures ANOVA was conducte d where Valence (negative, neutral, positive), Old/New status (old, new), and Hemis phere (left, right) served as IVs and the mean amplitude (averaged across leads) fr om frontal leads recorded from 300-640ms served as the DV. Results revealed a main effect of Old/New status F(1,29) = 6.35, p = .02 indicating old waveforms (M = .68, SE = .192) were significantly more positive in mean amplitude than new waveforms (M = .19, SE = .71). The effects of Valence F(2,58) = .663, p = .52 and Hemisphere F(1,29) = .067, p = .80 as well as all interactions were not statis tically significant indicating that the Old/New effect was bilaterally distributed, but did not vary as a function of Valence (see Figure 4).
-3 -2 -1 0 1 2 3 4 -2000200400600800100012001400 Old NewV Ms Figure 4. Bilateral frontal Old/New effect (n.b., vertical bars denote analysis epoch). When a repeated measures ANOVA was conducted where Valence (negative, neutral, positive), Old/New stat us (old, new), and Hemisphere (left, right) served as IVs and the mean amplitude (averaged across leads) from parietal leads recorded from 4001000ms served as the DV the main effect of Old/New status F(1,29) = 7.77, p < .01 emerged again, where old waveforms (M = .28, SE = .33) were significantly more positive in amplitude than old waveforms (M = .009, SE = .36). This analysis also revealed a very strong trend towards a main effect of Hemisphere F(1,29) = 3.99, p = .055 although not in the expected direction as the mean amplitude from the right hemisphere (M = .49, SE = .42) was larger than that of the left hemisphere (M = -.20, SE = .34). Although a main effect of Valence was not observed F(2,58) = 2.39, p = .10 40
41 there was a significant three-way interacti on between Valence, Old/New status, and Hemisphere F(2,58) = 3.82, p = .03 where the Old/New effect was greatest in response to positive words in the left hemisphere (see Figure 5).
Figure 5. Interaction between Valence, Old/New status, and Hemisphere (n.b., vertical bars denote analysis epoch). When mean amplitudes (averaged across l eads) from frontal leads recorded from 400-1400ms served as the DV and Valence (nega tive, neutral, positive), Old/New status 42
43 (old, new), and Hemisphere (left, right) se rved as IVs in a repeated measures ANOVA a main effect of Hemisphere was observed F(1,29) = 9.90, p < .01 indicating mean amplitudes were greater in the right hemisphe re (M = 1.67, SE = .75) compared to the left hemisphere (M = -1.21, SE = 1.13). Although significant main effects of Valence F(2,58) = 1.10, p = .34 and Old/New status F(1,29) = 1.90, p = .18 were not observed in this analysis, there was a trend toward s a significant two-way interaction between Old/New status and Hemisphere F(1,29) = 3.64, p = .07 where the Old/New effect was greater in the left hemisphere. Objective Old/New Comparison A repeated measures ANOVA was conducte d where Valence (negative, neutral, positive), Old/New status (old, new), and Hemis phere (left, right) served as IVs and the mean amplitude (averaged across leads) fr om frontal leads recorded from 300-640ms served as the DV. There were no significant main effects of Valence F(2,58) = .845, p = .44 Old/New status F(1,29) = 1.43, p = .24 or Hemisphere F(1,29) = .302, p = .59 When the mean amplitudes (averaged across leads) from parietal leads recorded from 400-1000ms served as the DV and Valen ce (negative, neutral, positive), Old/New status (old, new), and Hemisphere (left, ri ght) served as IVs in a repeated measures ANOVA a significant main effect of Old/New status emerged F(1,29) = 7.48, p = .01 indicating old waveforms (M = .13, SE = .33) were significantly more positive than the new waveforms (M = -.04, SE = .34). Grand av erage waveforms illustrating this parietal Old/New effect can be seen in Figure 6. The effects of Valence F(2,58) = 2.04, p = .14
and Hemisphere F(1,29) = 2.49, p = .13 on mean amplitudes were not statistically significant. Significant interactions did not result from this analysis. -1.5 -1 -0.5 0 0.5 1-2000200400600800100012001400 Old NewV Ms Figure 6. Grand average waveforms from the pa rietal region (n.b., ve rtical bars denote analysis epoch). When mean amplitudes (averaged across l eads) from frontal leads recorded from 400-1400ms served as the DV and Valence (nega tive, neutral, positive), Old/New status (old, new), and Hemisphere (left, right) se rved as IVs in a repeated measures ANOVA a significant main effect of Old/New status was again observed F(1,29) = 5.81, p = .023 as well as a significant main eff ect of Hemisphere on amplitude F(1,29) = 9.25, p < .01 Inspection of amplitude means revealed that th e old waveforms in this analysis were less positive (M = .176, SE = .85) than the new waveforms (M = .490, SE = .84) and 44
45 amplitudes recorded from the right hemisphere leads were significantly more positive (M = 1.7, SE = .73) than those from the correspond ing region of the left hemisphere (M = =1.04, SE = 1.14). This analysis did not yi eld a significant main effect of Valence F(2,58) = 2.10, p =.132 or any signifi cant interactions. Subjective Old/New Comparison A repeated measures ANOVA was conducte d where Valence (negative, neutral, positive), Old/New status (old, new), and Hemis phere (left, right) served as IVs and the mean amplitude (averaged across leads) fr om frontal leads recorded from 300-640ms served as the DV. A significant main effect of Old/New status was observed F(1,29) = 10.98, p < .01 where old waveforms were more positive in amplitude (M = .77, SE = .64) than new waveforms (M = .14, SE = .67). Grand average waveforms (collapsed across valence and hemisphere) of this Old/ New effect can be seen in Figure 7. Significant main effects of Valence F(2,58) = .845, p = .44 and Hemisphere F(1,29) = .303, p = .59 as well as significant interactions were not observed in this analysis.
46 -3 -2 -1 0 1 2 3 4 -2000200400600800100012001400 Old NewVMs Figure 7. Grand average waveforms of the bila teral frontal Old/New effect (n.b., vertical bars denote analysis epoch). When the mean amplitudes (averaged across leads) from parietal leads recorded from 400-1000ms served as the DV and Valen ce (negative, neutral, positive), Old/New status (old, new), and Hemisphere (left, ri ght) served as IVs in a repeated measures ANOVA neither Valence F(2,58) = 2.04, p = .14 Old/New status F(1,29) = 1.27, p = .27 nor Hemisphere F(1,29) = 2.50, p = .13 yielded significant main effects. When mean amplitudes (averaged across l eads) from frontal leads recorded from 400-1400ms served as the DV and Valence (nega tive, neutral, positive), Old/New status (old, new), and Hemisphere (left, right) se rved as IVs in a repeated measures ANOVA a significant main effect of Ol d/New status was observed F(1,29) = 9.54, p < .01 where
47 amplitudes of old waveforms were significan tly more positive (M = .64, SE = .82) than those of new waveforms (M = .02, SE = .88). Additionally, a significant main effect of Hemisphere was observed F(1,29) = 9.25, p < .01 where amplitudes of waveforms from the right hemisphere (M = 1.70, SE = .73) were more positive than those of waveforms recorded from the left hemisphere (M = -1.04, SE = 1.14). A significant interaction between Old/New status and Hemisphere also emerged F(1,29) = 4.90, p = .04 reflecting a larger Old/New eff ect in the left hemisphere. To test the prediction that Old/New effects would be largest within the Traditional/Correct classification, followed by the Subjective and the Objective classifications ERP waveforms associated with each of the four response types (i.e., hit, miss, false alarm, correct rejection) were an alyzed separately. This was done to avoid violations of the independence assumptions of MANOVA that would come with an analysis directly comparing Old/New effects within the three classification types (i.e., Traditional, Objective, Subjective). Within the early frontal spatial tempor al region of interest a two-way ANOVA was employed where Response Type (Hit, Mi ss, False Alarm, Correct Rejection), and Hemisphere (Left, Right) served as the independent variables and mean amplitudes (averaged across frontal leads recorded from 300-640ms) served as the dependent variable. Results indicated a main effect of Response Type F(3,87) = 4.91, p = .003 only. Bonferroni adjusted post hoc tests rev ealed that mean amplitudes elicited by False Alarms (M = .86, SE = .66) were significantly more positive than misses (M = .08, SE = .64; p = .03). A trend towards mean amplitude s elicited by Hits (M = .68, SE = .65) to be significantly greater than Misses was also observed (p = .052). Mean amplitudes
associated with Correct Rejections (M = .19, SE = .71) did not differ from the other three response types. This effect can be seen in Figure 8. -3 -2 -1 0 1 2 3 4 -2000200400600800100012001400 Hit Miss False Alarm Correct RejectionVMs Figure 8. Grand average waveforms of m ean amplitudes associated with each response type within the early frontal (hem ispheres were averaged together) spatial temporal region of interest (n.b., ver tical bars denote analysis epoch). Since false alarms were significantly greater (i.e., more positive) than misses in the above analysis and greater than the difference between hits and correct rejections the conclusion can be made that within this spatia l temporal region of interest (early frontal), subjective Old/New effects were larger than traditional, and objective (respectively). 48 Within the parietal spatial temporal region of interest a two-way ANOVA was employed where Response Type (hit, miss, false alarm, correct rejection), and
49 Hemisphere (left, right) served as the independent variables and mean amplitudes (averaged across parietal leads recorded from 400-1000ms) served as the dependent variable. Results indicated a main effect of Response Type F(3,87) = 5.33, p = .002 and a significant inter action between Response Type and Hemisphere F(3,87) = 3.26, p = .025 Bonferroni adjusted post hoc tests re vealed that mean amplitudes associated with hit (M = .28, SE = .33) responses were si gnificantly greater th an those associated with false alarms (M = -.97, SE = .34; p = .007) and misses (M = -.24, SE = .34; p = .03). There was also a trend towards significantly gr eater mean amplitudes associated with hits when compared to correct rejections (M = .01, SE = .36; p = .056). Misses, false alarms and correct rejections did not differ from each other. The significan t interaction between Response Type and Hemisphere was driven by less positive amplitudes associated with false alarms in the left hemisphere (M = -.23, SE = .36) when compared to their right hemisphere counterpart (M = .04, SE = .42). These results are depi cted graphically in Figure 9.
-1.5 -1 -0.5 0 0.5 1 1.5 -2000200400600800100012001400Left Parietal Hit Miss False Alarm Correct RejectionVMs -1.5 -1 -0.5 0 0.5 1 1.5 -2000200400600800100012001400Right Parietal Hit Miss False Alarm Correct RejectionVMs Figure 9. Grand average waveforms from parietal regions associated with each response type (n.b., vertical bars denote analysis epoch). 50
51 Since the difference between hits and false alarms is greater than the difference between hits and correct rejec tions, the conclusion can be made that within the parietal spatial temporal region of interest, the obj ective Old/New effect is larger than the traditional Old/New effect and the subjective Old/New effect (respectively). Within the late frontal spatial temporal region of interest a two-way ANOVA was employed where Response Type (hit, miss, false alarm, correct rejection), and Hemisphere (left, right) served as the independent variables and mean amplitudes (averaged across frontal leads recorded from 400-1400ms) served as the dependent variable. Results indicated a main effect of Response Type F(3,87) = 6.69, p < .001 a main effect of Hemisphere F(1,29) = 11.20, p = .002 and a significant interaction between Response Type and Hemisphere F(3,87) = 8.18, p < .001 Bonferroni adjusted post hoc comparisons revealed mean amplitude s associated with false alarms (M = .90, SE = .83) were significantly greater than t hose associated with misses (M = -.03, SE = .89; p = .006), and correct rejections (M = .08, SE = .88; p = .02). There was no difference between mean amplitudes associated with hits (M = .38, SE = .82) and those of false alarms, correct rejections, or misses. The main effect of Hemisphere resulted from greater mean amplitudes in the right hemisphere (M = 1.14, SE = .77) when compared to the left hemisphere (M = -.47, SE = .97). The significant interaction between Response Type and Hemisphere resulted from more positive amplitudes resulting from false alarms within the left hemisphere (M = 2.04, SE = 3.83) when compared to amplitudes associated with fals e alarms measured in the right hemisphere (M = -.23, SE = 6.20). These results ar e depicted graphically in figure 10.
-4 -2 0 2 4 -2000200400600800100012001400Left Frontal Hit Miss False Alarm Correct RejectionVMs -4 -2 0 2 4 -2000200400600800100012001400Right Frontal Hit Miss False Alarm Correct RejectionVMs Figure 10. Grand average waveforms from the la te frontal regions associated with each response type (n.b., vertical ba rs denote analysis epoch). 52
53 Since the traditional comparison of Old/New effects (i.e., hits vs. correct rejections) was not significan t we can disregard the influence of hits and correct rejections in the objective and subjective compar isons. As such, the absolute value of the difference between old and new in each of these comparisons (i.e. objective and subjective) is identical. Therefore, the conclu sion can be made that within this particular spatial temporal region of interest (i.e., late frontal) the objective and subjective Old/New effects are larger in absolute magnitude than those observed in the Traditional comparison, but that they do not differ from each other. Hypothesis 5: Based on Windmann and Kuta s assertion that bias for negative words serves an adaptive function, whereby the cognitive system is prompted to assign greater significance and a higher priority to the processing of potentially threatening stimuli is correct then it is predicted that Old/New effects will be greatest for negative items compared to positive and neutral b ecause the positive a nd neutral items lack threatening connotations. Although main effects of Valence were not observed, there was a significant three-way interaction between Valen ce, Old/New status, and Hemisphere F(2,58) = 3.82, p = .03 where the Old/New effect was greatest in response to positive words in the left hemisphere.
54 Discussion The purpose of the present study was to investigate the pattern and timing of electrophysiological indices of Old/New r ecognition memory effects for negative, neutral, as well as positive words. S econdly, this study examined behavioral and electrophysiological indices of subject respon se bias in Old/New recognition memory. Previous research in this area employed (almost exclusively) negativ e and neutral stimuli only. Furthermore, the majority of studies neglected to control for the potentially confounding characteristics of the stimuli such as arousal and inter-item relatedness (i.e., semantic cohesion). The current study extended the previous literatur e by including a list of positively valenced words in addition to the standard negative and neutral word lists. All word lists were carefully balanced (i.e., equated) for valence, frequency, and semantic cohesion. The behavioral data prediction that emotional words (i.e., positive and negative) would elicit more hits than ne utral words was partially suppo rted in that the Hit Rate associated with negative words was significan tly greater compared to the Hit Rate for positive and neutral stimuli. This finding is c onsistent with others in the literature (e.g., Kensinger & Corkin, 2003) where only negati ve and neutral words were compared. However, when positive words are also used as stimuli in addition to negative and neutral words findings generally support enhancement of recognition memory for both negative
55 and positive words when compared to neutral stimuli (e.g., Ku chinke, Jacobs, Vo, Conrad, Grubich, & Herrmann, 2006). The frequenc y statistics associated with the list of positive stimuli employed in the current study were, on average, higher than those from the negative and neutral lists. Although this frequency difference did not achieve statistical significance, there was a trend (p = .07) towards the positive word list having a higher frequency estimate than the negative a nd neutral lists. Th is trend could account for the finding that negative words yielded a higher Hit Rate than positive words since less frequent items tend to be remembered better (Shiffrin & Steyvers, 1997). False Alarm Rate analyses revealed that both positive and negative words elicited a significantly higher False Alarm Rate compared to neutral words. Along with an elevated Hit Rate for negative words being co mmonly reported in the literature, there is an equally prevalent finding of a significantl y elevated False Alarm Rate for negative and positive words (e.g., Maratos et al., 2000; Mc Neely et al., 2004; Windmann & Kutas, 2001; Vo, Jacobs, Kuchinke, Hofmann, Conra d, Schacht, & Hutzler), although there are far fewer studies that have used positive stim uli. These results highlight the importance of examining positive words in addition to negative and neutral as not all types of emotional words produce similar results. Obviously, one cannot rely on Hit Rate alone as a measure of accuracy as it yields an incomplete picture of performance. Su ch a method could, theoretically, yield equal Hit Rate values for the participant who res ponds Old to every item and the participant who only responds Old to th e actual old items. One must combine both the Hit Rate and the False Alarm Rate in order to obtain a clear picture of partic ipant performance. This combination is best reflected in an accu racy/sensitivity index that takes into account
56 both correct and incorrect responses to old item s. For these reasons, an ancillary analysis using the sensitivity index d was employed. This analysis re vealed a somewhat different pattern of results. Specificall y, participants made more accurate judgments in response to negative and neutral words compared to positiv e stimuli. The previous note regarding increased frequency ratings within the positive word list applies to the results of this analysis as well. That is, the trend towa rds positive words having significantly higher frequency ratings compared to the negative a nd neutral word lists could account for the decreased accuracy for positive words. These results also underscore the importance of including positive stimuli, as they may yield important differences across categories of emotional words that would not be apparent had only negative stimuli been used and. Moreover, the use of a sensitiv ity/accuracy index provides uni que information that is not captured by Hit Rate and False Alarm Rate analyses alone. With respect to tests of the first hypothes is that Hit Rate and False Alarm Rate would be greater for emotional words, there was some variability in support across the three dependent measures (i.e., Hit Rate, False Alarm Rate, d ). Main effects of valence were found across all three dependent measur es but not always consistently greater performance for both positive and negative words as Hit Rate was greater for negative words only and sensitivity/accuracy was gr eater for negative and neutral words. The second prediction that both classes of emotional words (i.e., positive and negative) would elicit greater levels of response bias was fully supported. These effects are congruent with published studies that gene rally find an increased bias for emotional words (positive and negative) compared to neutral (e.g., Maratos, Allan, & Rugg, 2000; Windmann & Kutas, 2001; Vo et al., 2008).
57 With respect to reaction time, only a significant main effect of response type was observed where hit responses (M = 1095.17, SE = 66.21) were made faster than misses (M = 1308.44, SE = 86.14), false alarms ( 1253.04, SE = 85.910, and correct rejections (M = 1273.39, SE = 78.59) is in keeping with si milar behavioral lite rature (e.g., Bentin & McCarthy, 1994; Windman & Chimielewski, 2007). However, the prediction that reaction times would be fastest in respons e to emotional (i.e., positive and negative) words was not supported. A trend was observed for this analysis (p=.078) but it was not in the correct direction (i.e., reaction times to negative words were slowest). As noted above, the prediction that valen ce effects in accuracy would exist after controlling for the confounding effects of arousal and semantic cohesion was supported. To our knowledge, the current stud y was the first to control for both of these confounding variables so as to make legitimate inferences about the interplay between them. With respect to the ERP data, the hypot hesis that all thr ee classes of words (negative, neutral, and positive) would elic it Old/New effects was partially supported. When significant Old/New effects were observed they occurred in all valences. However, significant Old/New effects were not observed in every sp atial/temporal region of interest (i.e., early frontal, parietal, late frontal) across all three Old/New comparison types (i.e., traditional, objective, su bjective). See Table 5 for details.
58 Table 5. Significant Old/New effects observe d according to spatial/temporal region of interest (left-most column), and comparison type (top row). Traditional Objective Subjective Early frontal Old > New Old > New Parietal Old > New Old > New Late frontal New > Old Old > New As can be seen in Table 5, significant Old/ New effects in the early frontal regions were observed in the traditional and subjec tive comparisons, but not in the objective comparison. In the parietal regions, signi ficant Old/New effects were noted in the traditional and objective comparisons, but not in the subjective comparison. Within the objective Old/New comparison, miss trials are used to constr uct the Old waveform, and trials where false alarms are made are us ed to construct the New waveform. The opposite is true for the subjective comparison. In order for this dissociation to occur (i.e., significant subjective but not objective Old/New effects in the early frontal regions and significant objective but not subj ective Old/New effects in pari etal regions) the difference between false alarms and misses must be pos itive in the early frontal regions between 300ms and 640ms whereas the difference betw een false alarms and misses must be negative in parietal regions between 400ms and 1000ms. This finding is in keeping with dual-process models of recogni tion memory from an ERP perspective. Specifically, the parietal Old/New effect is thought to index recollecti on (Rugg, 1987; Rugg et al, 1998) and is larger for remember (vs. know ) judgments in remember/know experiments (Duzel, Yonelinas, Mangun, Heinze, & Tulvi ng, 1997). Logically, the neural signature
59 of the memory trace (as indexed by the pariet al Old/New effect) s hould be larger for misses than for false alarms because miss trials are elicited by words that were actually studied (i.e., targets). Conversely, false alarms are elicited by words that cannot be recollected because they were never actually studied (i.e., foils). Early frontal Old/New effects are hypothesized as representing the ne ural correlate of familiarity (Curran, 2000; Paller, Voss, & Boehm, 2007) and associated more so with know judgments when remember/know experiments are employed (G ardiner, Java, & Richardson-Klavehn, 1996). As such, familiarity-driven False alarms should be most apparent (i.e., greater/more positive) in this cortical regi on (i.e., the putative neural correlate of familiarity). Recent evidence provided by Goldman et al. (2003) suggests that the late frontal component observed in some ERP studies of recognition memory is an index of postretrieval processing. This component is thought to be most prominent when Old/New discrimination is difficult. With this idea in mind, incorrect responses (i.e., misses and false alarms) are considered more effortful (as opposed to correct responses) and more likely to engage the post-retrie val processing indexed by the la te frontal Old/New effect. This would explain why a signifi cant late frontal Old/New effect within the traditional comparison was not observed (the rendering of correct judgments was not effortful enough to engage post-retrieval processing) The addition of incorrect (and thus effortful) trials (i.e., false alarms and misse s) to Old/New comparis ons (i.e., objective and subjective) allows the late frontal Old/New eff ect to be revealed. The inverse pattern of late frontal Old/New effects seen in the current study (i.e ., Old waveforms greater than New waveforms in the subjective comparis on and new waveforms greater than old
60 waveforms in the objective comparison) is ag ain driven by potentia ls associated with false alarms and misses. Incorrect acceptance of a word that was not previously studied (i.e., false alarm) would have to be more pos itive going compared to incorrect rejections of words that were not previously studied (i.e., misses) in order to make the objective new waveform more positive than its old count erpart. The notion that false alarms are associated with more positive amplitudes has been observed in previous research (Goldmann et al., 2003 & Windmann et al., 2002) although the epochs analyzed were arguably too early to capture this late effect in one of these studies (Windmann et al., 2002) and the other (Goldmann et al., 2003) did not analyze the amplitudes of their waveforms according to the subjective comparison. Analysis of waveforms associated with ea ch response type (hit, miss, false alarm, correct rejection), along with the pattern of results observe d between the three Old/New comparison types revealed that inclusion of error trials (i.e., misses and false alarms) yields varying results across spatial temporal regions of interest. This finding is in keeping with that of Windmann et al. (2002) who found differential effects between the three comparison types although th e results dont map directly onto those of the current study as Windmann also employed a grouping factor (high response bias vs. low response bias). With respect to the lack of significan t valence effects within the ERP data previous research provides little aid in the explanation of such results. There is a paucity of research investigating Old/New effects for negative, neutral, and positive words and, as noted in the introduction, the results ar e somewhat equivocal. The most obvious difference, however, between the existing literature and the results of the current study is
61 that the current study assessed, and controll ed for, the possible confounding effects of arousal, whereas the others did not. It is also possible that the degree of emotionality within the valenced word lists differs between studies (and is smalle r in the current study) and has differential effects on the observed results. There may exist a valence threshold of sorts below which these effects go undetect ed. Perhaps the word lists used in the current study were below this threshold (i.e., the negative words were not negative enough to produce valence effects in the ERPs etc.). Limitations and Directions for Future Research As noted above in the discussion, some of the behavioral findings can be attributed to the increased level of frequenc y among the words in the positive list. This difference can be considered a limitation as th e results would certainly be clearer without having to account for this possible confound. Additionally, reaction times may have been shorter, and valence effects observed had the response style been altered. Specifically, if the number of choices the participant coul d respond with was shortened (e.g., a basic yes/no decision instead of a confidence rating) then the participants would likely be faster to respond. Explicitly instructing the participants to respond as quickly as possible (an element not employed in the current study ) also has the potential to reduce reaction times. Overall, the present study helped elucid ate the relationship between emotion and recognition memory for words. Specificall y, increased accuracy and response bias for emotional words was observed even after word lists were equated to prevent possible memory enhancement by stimulus characteristics (e.g., arousal and semantic cohesion/inter item relatedness). The need to meticulously control for arousal and
62 semantic cohesion in studies within this ar ea is important in order to make clear inferences about the effects of emotion on memo ry and to ensure that it is not actually differences in arousal level or semantic c ohesion that account for differential rates of recall across emotional and nonemotional condi tions. Objective ratings from available normative data should be used to confirm e quivalence on these dimensions rather than relying on subjective judgments and statistical analyses used to confirm equivalence across conditions. Furthermore, as noted a bove, different types or classes of emotional words may not always behave similarly given the same cognitive operation (i.e. accuracy, false alarm rate, etc.) and these differences ma y provide future insights into differential influences the emotional nature of a stimul us has on specific cognitive processes within memory. Therefore, including positive as well as negative stimuli may allow more precise understanding of differential influen ces of these stimuli on distinct memory processes. The effect(s) of emotion on memory will undoubtedly continue to intrigue scientists and the general public for decades to come. This re lationship is not only interesting but has implications for many ar eas of research beyond the field of cognitive neuroscience. Forensic (e.g., conceptualizati on and credibility of eye-witness testimony) and clinical implications (e.g., classification and treatment of panic disorder and post traumatic stress disorder) will likely become more evident as research advances and knowledge of the interplay be tween emotion and memory in the healthy adult brain increases.
63 References Beck, A.T., Steer, R.A., (1987). Manual fo r the Revised Beck Depression Inventory. Psychological Corporation, San Antonio, TX. Beck, A.T., Steer, R.A., & Garbin, M.G., (1988). Psychometric properties of the Beck Depression Inventory: twenty-five years of evaluation. Clinical Psychology Review, 8, 77-100. Bentin, S., McCarthy, G. (1994). The effects of immediate stimulus repetition on reaction time and event-related potentials in task s of different complexity. Journal of Experimental Psychology. 20, 130-149. Bradley, M. M., & Lang, P. J. (1999). Af fective norms for English words (ANEW): Stimuli, instruction manual and affective ratings. Technical report C-1. Gainesville, FL: The Center for Research in Psych ophysiology, University of Florida. Bradley, M. M., Greenwald, M. K., Petry, M. C., & Lang, P. J. (1992). Remembering pictures: Pleasure and arousal in memory. Journal of Experimental Psychology. Learning, Memory, and Cognition; Journa l of Experimental Psychology.Learning, Memory, and Cognition, 18(2), 379-390.
64 Bradley, M. M., & Lang, P. J. (2000). Af fective reactions to acoustic stimuli. Psychophysiology; Psychophysiology, 37(2), 204-215. Cahill, L., Haier, R. J., Fallon, J., Alkire, M. T., Tang, C., & Keator, D., et al. (1996). Amygdala activity at encoding correlated with long-term, free recall of emotional information. Proceedings of the National A cademy of Sciences of the United States of America; Proceedings of the National A cademy of Sciences of the United States of America, 93(15), 8016-8021. Cahill, L., & McGaugh, J. L. (1995). A nove l demonstration of enhanced memory associated with emotional arousal. Consci ousness and Cognition; Consciousness and Cognition, 4(4), 410-421. Cahill, L., Prins, B., Weber, M., & McGa ugh, J. L. (1994). Beta-adrenergic activation and memory for emotional events. Nature; Nature, 371(6499), 702-704. Carretie, L., Iglesias, J., & Garcia, T. (1997). A study on the emotional-processing of visual stimuli through event-related potentials. Brain and Cognition; Brain and Cognition, 34(2), 207-217. Carretie, L., Iglesias, J., Garcia, T., & Ballesteros, M. (1997). N300, P300 and the emotional processing of visual stimuli. Electroencephalography and Clinical Neurophysiology; Electroencephalography and Clinical Neurophysiology, 103(2), 298-303.
65 Christianson, S. (Ed.). (1992). The handbook of emotion and memory: Research and theory. London: Lawrence Erlbaum Associates. Curran, T. (2000). Brain potentials of recollection and familiarity. Memory & Cognition; Memory & Cognition, 28(6), 923-938. Curran, T., Schacter, D. L., Johnson, M. K., & Spinks, R. (2001). Brain potentials reflect behavioral differences in true and false recognition. Journal of Cognitive Neuroscience; Journal of Cognitiv e Neuroscience, 13(2), 201-216. Dietrich, D. E., Waller, C., Johannes, S., Wier inga, B. M., Emrich, H. M., & Munte, T. F. (2001). Differential effects of emotional c ontent on event-related potentials in word recognition memory. Neuropsychobiol ogy; Neuropsychobiology, 43(2), 96-101. Dolcos, F., & Cabeza, R. (2002). Event-re lated potentials of emotional memory: Encoding pleasant, unpleasant, and neut ral pictures. Cognitive, Affective & Behavioral Neurosci ence, 2(3), 252-263. Donaldson, D. I., & Rugg, M. D. (1998). Recognition memory for new associations: Electrophysiological evidence for the ro le of recollection. Neuropsychologia; Neuropsychologia, 36(5), 377-395. Duzel, E., Yonelinas, A. P., Mangun, G. R., Heinze, H. J., & Tulving, E. (1997). Eventrelated brain potential correla tes of two states of cons cious awareness in memory. Proceedings of the National Academy of Scie nces of the United States of America;
66 Proceedings of the National Academy of Scie nces of the United States of America, 94(11), 5973-5978. Gardiner, J.M., Java, R.I., Richardson-Klave hn. (1996). How level of processing really influences awareness in recognition memo ry. Canadian Journal of Experimental Psychology, 50, 114-122. Gardiner, J. M., & Java, R. I. (1991). Forget ting in recognition memo ry with and without recollective experience. Memory & Cognition, 19(6), 617-623. Gratton, G., Coles, M. G., Sirevaag, E. J., Eriksen, C. W., & Donchi n, E. (1988). Preand poststimulus activation of response ch annels: A psychophysiological analysis. Journal of Experimental Ps ychology. Human Perception and Performance; Journal of Experimental Psychology. Human Percep tion and Performance, 14(3), 331-344. Goldmann, R. E., Sullivan, A. L., Droller, D. B. J., Rugg, M. D., Curran, T., Holcomb, P. J., et al. (2003). Late frontal brain potent ials distinguish true and false recognition. NeuroReport, 14(13), 1717-1720. Hamann, S. (2001). Cognitive and neural mechanisms of emotional memory. Trends in Cognitive Sciences, 5(9), 394-400. Heuer, F., & Reisberg, D. (1990). Vivid memories of emotional events: The accuracy of remembered minutiae. Memory & Cogniti on; Memory & Cognition, 18(5), 496-506. Inaba, M., Nomura, M., & Ohira, H. (2005). Neural evidence of ef fects of emotional valence on word recognition. International Journal of Psychophysiology : Official
67 Journal of the Internationa l Organization of Psychophysiology; International Journal of Psychophysiology : Official Journal of the Internationa l Organization of Psychophysiology, 57(3), 165-173. Inaba, M., Kamishima, K., & Ohira, H. (2007). An electrophysiol ogical comparison of recollection for emotional words using an exclusion recognition paradigm. Brain Research, 1133, 100-109. Jacoby, L. L., & Kelley, C. (1992). Unconscious influences of memory: Dissociations and automaticity. In A. D. Milner, & M. D. Rugg (Eds.), The neuropsychology of consciousness (pp. 201-233). London: Academic Press. Johansson, M., Mecklinger, A., & Treese, A. C. (2004). Recognition memory for emotional and neutral faces: An event-rela ted potential study. Journal of Cognitive Neuroscience; Journal of Cognitiv e Neuroscience, 16(10), 1840-1853. Kopelman, M. D., & Corn, T. H. (1988). Cholinergic 'blockade' as a model for cholinergic depletion. A comparison of the memory deficits with those of alzheimertype dementia and the alcoholic korsakoff syndrome. Brain; a Journal of Neurology; Brain : A Journal of Neurology, 111 ( Pt 5)(Pt 5), 1079-1110. Kensinger, E.A., & Corkin, S. (2003). Memo ry enhancement for emotional words: Are emotional words more vividly remembered than neutral words? Memory & Cognition, 31(8), 1169-1180.
68 Kuchinke, L., Jacobs, A. M., Vo, M. L.-H., Conrad, M., Grubich, C., & Herrmann, M. (2006). Modulation of prefrontal cortex act ivation by emotional words in recognition memory. Brain Imaging, 17(10), 1037-1041. LaBar, K. S., & Cabeza, R. (2006). Cognitive neuroscience of emotional memory. Nature Reviews. Neuroscience; Nature Re views. Neuroscience, 7(1), 54-64. LaBar, K. S., Gatenby, J. C., Gore, J. C., Le Doux, J. E., & Phelps, E. A. (1998). Human amygdala activation during conditioned f ear acquisition and extinction: A mixedtrial fMRI study. Neuron; Neuron, 20(5), 937-945. Landauer, T. K., Foltz, P. W., & Laham, D. (1998). Introduction to Latent Semantic Analysis. Discourse Processes, 25, 259-284. Macmillian, N. A. & Creelman, C. D. (1991) Detection theory: A users guide. Cambridge, England: Cambridge University Press. Mandler, G. (1981). The recogni tion of previous encounter s. American Scientist; American Scientist, 69(2), 211-218. Maratos, E. J., Allan, K., & Rugg, M. D. (2000). Recognition memory for emotionally negative and neutral words: An ERP study. Neuropsychologia; Neuropsychologia, 38(11), 1452-1465. Maratos, E. J., & Rugg, M. D. (2001). Electroph ysiological correlates of the retrieval of emotional and non-emotional context. Journa l of Cognitive Neuroscience; Journal of Cognitive Neuroscience, 13(7), 877-891.
69 McNeely, H. E., Dywan, J., & Segalowitz, S. J. (2004). ERP indices of emotionality and semantic cohesiveness during rec ognition judgments. Psychophysiology; Psychophysiology, 41(1), 117-129. Naumann, E., Huber, C., Maier, S., Plihal, W., Wustmans, A., & Diedrich, O., et al. (1992). The scalp topography of P300 in the visual and auditory modalities: A comparison of three normalization methods and the control of statis tical type II error. Electroencephalography and Clinical Neurophysiology; El ectroencephalography and Clinical Neurophysiology, 83(4), 254-264. Nelson, D. L., McEvoy, C. L. & Schreiber, T. A. (1998). The university of south Florida word association, rhyme, and word fragment norms.http://www.usf.edu/FreeAssociation/ Nissen, M. J., Knopman, D. S., & Schacter, D. L. (1987). Neuroche mical dissociation of memory systems. Neurology; Neurology, 37(5), 789-794. Ochsner, K. N. (2000). Are affective events richly recollected or simply familiar? the experience and process of recognizing f eelings past. Journal of Experimental Psychology. General; Journal of Expe rimental Psychology. General, 129(2), 242261. Oldfield, R. C. (1971). The assessment a nd analysis of handedness: The edinburgh inventory. Neuropsychologia; Neur opsychologia, 9(1), 97-113.
70 Paller, K. A., & Kutas, M. (1992). Br ain potentials during retrieval provide neurophysiological support for the distinc tion between conscious recollection and priming. Journal of Cognitive Neuroscience, 4, 375-391. Paller, K. A., Voss, J. L., & Boehm, S. G., Validating neural correlates of familiarity. Trends in Cognitive Sciences. 11(6), 243-250. Phelps, E. A., LaBar, K. S., & Spencer, D. D. (1997). Memory for emotional words following unilateral temporal lobectomy. Br ain and Cognition; Brain and Cognition, 35(1), 85-109. Potter, D. D., Pickles, C. D., Roberts, R. C., & Rugg, M. D. (1992). The effects of scopolamine on event-related potentials in a continuous recognition memory task. Psychophysiology; Psychophysiology, 29(1), 29-37. Rubin, D. C., & Schulkind, M. D. (1997). The distribution of autobiographical memories across the lifespan. Memory & Cogniti on; Memory & Cognition, 25(6), 859-866. Rugg, M. D. (1987). Dissociation of seman tic priming, word and nonword repetition by event-related potentials. Journal of Experimental Psychology, 39, 123-148. Rugg, M.D., Allan, K. (2000). Memory retrieval: An electrophysiologi cal perspective. In M. S. Gazzaniga (Ed.), The cognitive neurosciences (2nd ed., pp. 805-816). Cambridge, MA: MIT Press.
71 Rugg, M. D., & Doyle, M. C. (1994). Event rela ted potentials and stimulus repetition in direct and indirect tests of memory. In H. Hienze, T. Munte & G. R. Mangun (Eds.), Cognitive electrophysiology (pp. 124-148). Boston: Birkhauser. Rugg, M. D. (1995). Memory and consciousness: A selective review of issues and data. Neuropsychologia; Neuropsychologia, 33(9), 1131-1141. Rugg, M. D. (1990). Event-related brain potentia ls dissociate repetiti on effects of highand low-frequency words. Memory & Cognition; Memory & Cognition, 18(4), 367379. Rugg, M. D., Brovedani, P., & Doyle, M. C. (1992). Modulation of event-related potentials (ERPs) by word repetition in a task with inconsis tent mapping between repetition and response. Electroencephal ography and Clinical Neurophysiology; Electroencephalography and Clin ical Neurophysiology, 84(6), 521-531. Rugg, M. D., Mark, R. E., Walla, P., Schloers cheidt, A. M., Birch, C. S., & Allan, K. Dissociation of the neutral correlates of implicit and explicit memory. Nature 392, 595-598. Rugg, M. D., & Nagy, M. E. (1989). Event-rela ted potentials and recognition memory for words. Electroencephalography and Clinical Neurophysiology; Electroencephalography and Clin ical Neurophysiology, 72(5), 395-406. Schneider, W., Eschman, A., & Zuccolo tto, A. (2002). E-Prime users guide Computer software Pittsburgh: Psychology Software Tools, Inc.
72 Schupp, H. T., Cuthbert, B. N., Bradley, M. M., Cacioppo, J. T., Ito, T., & Lang, P. J. (2000). Affective picture processing: The late positive potential is modulated by motivational relevance. Psychophysiol ogy; Psychophysiology, 37(2), 257-261. Shiffrin, R.M., Steyvers, M. (1997). A model for recogni tion memory:REM-retrieving effectively from memory. Psychonomic Bulletin & Review, 4, 145-166. Smith, M. E. (1993). Neurophysiological manife stations of recollective experience during recognition memory judgments. Journal of Cognitive Neuroscience, 5, 1-13. Smith, A. P., Dolan, R. J., & Rugg, M. D. ( 2004). Event-related poten tial correlates of the retrieval of emotional and nonemotional cont ext. Journal of Cognitive Neuroscience; Journal of Cognitive Neuroscience, 16(5), 760-775. Smith, M. E., & Halgren, E. (1989). Disso ciation of recognition memory components following temporal lobe lesions. Journa l of Experimental Psychology. Learning, Memory, and Cognition; Journal of E xperimental Psychology.Learning, Memory, and Cognition, 15(1), 50-60. Snodgrass, J. G., & Corwin, J. (1988). Pr agmatics of measuring recognition memory: Applications to dementia and amnesia. Journal of Experimental Psychology: General, 117(1), 34-50. Soetens, E., Casaer, S., D'Hooge, R., & Hueti ng, J. E. (1995). Effect of amphetamine on long-term retention of verbal mate rial. Psychopharmacology; Psychopharmacology, 119(2), 155-162.
73 Swets, J. A. (1996). Signal Detection Theory and ROC Analysis in Psychology and Diagnostics: Collected Papers. Mahwa h, NJ: Lawrence Erlbaum Associates. Talmi, D., & Moscovitch, M. (2004). Can sema ntic relatedness explain the enhancement of memory for emotional words? Me mory & Cognition; Memory & Cognition, 32(5), 742-751. Tendolkar, I., & Rugg, M. D. (1998). Electroph ysiological dissociation of recency and recognition memory. Neuropsychologia; Neuropsychologia, 36(6), 477-490. Tulving, E. (1985). Memory and consciousness. Canadian Psychology, 26(1), 1-12. Vanderploeg, R. D., Brown, W. S., & Marsh, J. T. (1987). Judgments of emotion in words and faces: ERP correlates. Inte rnational Journal of Psychophysiology : Official Journal of the In ternational Organiza tion of Psychophysiology; International Journal of Psychophysiology : Official Journal of the International Organization of Psychophysiology, 5(3), 193-205. Vo, M. L.-H., Jacobs, A. M., Kuchinke, L., Ofmann, M., Conrad, M., Schacht, A., & Hutzler, F. (2008). The coupling of emoti on and cognition in the eye: Introducing the pupil old/new effect. Psychophysiology, 45, 130-140. Wickens, T. D. (2002). Elementary signal detection theory. Oxford; New York: Oxford University Press.
74 Wilding, E. L., Doyle, M. C., & Rugg, M. D. (1995). Recognition memory with and without retrieval of context: An event -related potential study. Neuropsychologia; Neuropsychologia, 33(6), 743-767. Windman, S., & Chmielewski, A. (2007). Em otion-induced modulation of recognition memory decisions in a Go/NoGo task: Response bias or memory bias?. Cognition and Emotion, 1-16. Windmann, S., & Kutas, M. (2001). Electrophys iological correlates of emotion-induced recognition bias. Journal of Cognitive Neuroscience; Journal of Cognitive Neuroscience, 13(5), 577-592. Windmann, S., Urbach, T. P., & Kutas, M. ( 2002). Cognitive and neural mechanisms of decision biases in recognition memory. Ce rebral Cortex (New York, N.Y. : 1991); Cerebral Cortex (New Yo rk, N.Y.: 1991), 12(8), 808-817. Williams, S.M., (1991). Handedness invent ories: Edinburgh versus Annett. Neuropsychology, 5, 43-48. Yonelinas, A.P., (1994). Receiver-operating characteristics in recognition memory: Evidence for a dual-process model. Journa l of Experimental Psychology: Learning, Memory, & Cognition, 20, 1341-1354. Yonelinas, A. P., & Jacoby, L. L. (1996). Noncriterial recollection: Familiarity as automatic, irrelevant recollection. Consci ousness and Cognition; Consciousness and Cognition, 5(1-2), 131-141.
76 Appendix A: Word Lists Negative Targets accident alimony alone beggar blind bored broken bullet burial burn cell coward crime criminal crisis cut dead death debt deceit defeated deformed depressed depression despise detached disdainful dummy dump dustpan failure false fat fault fear feeble filth flood frustrated funeral fungus garbage grime handicap headache hell hinder hungry ignorance immature immoral impotent infection injury jail knife lonely malaria manure measles meek mold noose nuisance overcast poison prison pungent punishment resent rigid rusty sad scorching scornful severe sin slave slum spanking stupid tobacco tomb tragedy trouble ugly war waste Neutral Targets alien alley anxious army autumn bake beast blond book boxer boy cat chance chaos city cliff clock coast cold concentrate dark dawn defiant dentist derelict detail diver elevator embattled employment excuse fabric face fall field foam fur garter gymnast hammer haphazard hard hat hide hospital industry lantern legend listless market material modest mystic naked nursery obey office overwhelmed paint patient person rattle reunion revolt revolver rough runner saint salute save shadow ship skeptical skull stomach swift truck trumpet vampire vanity virgin virtue voyage whistle white wife wonder writer
77 Appendix A: (Continued) Positive Targets ace advantage affection agreement angel answer art baby bath bird bless breast bride bright bunny butterfly car carefree child chocolate circus cozy dancer dollar earth easy easygoing eat food gift glory heal honest hug impressed innocent inspire jewel lake learn leisurely lottery luscious mobility money mother movie natural nectar ocean optimism palace pasta peace pet politeness rabbit radiant rainbow relaxed resp ectful restaurant reward river sapphire satisfied silk smooth snow snuggle song spirit spouse sun sunrise sunset tender terrific thankful toy travel untroubled useful valent ine vision waterfall wedding woman Negative Foils ache addict agony allergy blackmail blister blubber coffin controlling corpse cruel crutch dagger damage despairing destruction devil dirty disappoint discomfort discouraged dreary fatigued fever foul fraud frigid germs gloom greed grief guilty hurt idiot illness impair inferior insult lice lie loneliness messy mildew mistake moody morbid mosquito mucus nasty needle obesity offend pest pity poverty rat ridicule robber rotten scapegoat scar scum shamed sick sickness slime slow snob sour spider stench stink suffocate suicide terrible thorn timid traitor trash unhappy upset urine venom victim wasp weapon weary wounds
78 Appendix A: (Continued) Neutral Foils air alert aloof ankle avenue bar black body bottle busybody cane cannon cellar church clumsy coin cook curious custom doctor doll dress event favor flag fragrance garment glass grass green hand hawk highway hit hotel icebox idol kick knot lightning limber lion manner medicine mischief month moral muddy mushroom name news noisy nonsense nurse obsession odd opinion pancakes pie pig plane priest python quality queen razor red rock scissors serious skyscraper spray startled stiff stool storm stove tank teacher tennis thought tool trunk vehicle village watch wine yellow Positive Foils adult beach beauty bed beverage blue breeze brother cake candy charm color comedy comfort crown cuddle cute decorate diamond dignified dog dove dream family fantasy father flower freedom friend game garden gentle girl god gold grateful health home honey honor house humor intellect kind kindness king kitten knowledge letter life loyal luxury magical mail melody memory mountain music nice perfume pillow pizza pleasure prestige pretty protected safe secure sky sleep soft soothe space spring star talent taste treat tune twilight vacation warmth wise wish wit world young youth
Appendix B: Signal Detection Formulas Sensitivity/accuracy: d = zFAzH Response bias: C = zFAd /2 = 0.5 (zFA+ zH), where zH is the z score in the old distribution having H proportion (i.e., Hit Rate) above it and zFA is the z score in the new distribution having FA proportion a bove it (i.e., False Alarm Rate; Snodgrass & Corwin, 1988). 79