Does Emotion Help or Hinder Immediate Memory-Arousal Versus

Does Emotion Help or Hinder Immediate Memory? Arousal Versus Priority-Binding Mechanisms Christopher B. Hadley and Donald G. MacKay University of California, Los Angeles People recall taboo words better than neutral words in many experimental contexts. The present rapid serial visual presentation (RSVP) experiments demonstrated this taboo-superiority effect for immediate recall of mixed lists containing taboo and neutral words matched for familiarity, length, and category coherence. Under binding theory (MacKay et al., 2004), taboo superiority reflects an interference effect: Because the emotional reaction system prioritizes binding mechanisms for linking the source of an emotion to its context, taboo words capture the mechanisms for encoding list context in mixed lists, impairing the encoding of adjacent neutral words when RSVP rates are sufficiently rapid. However, for pure or unmixed lists, binding theory predicted no better recall of taboo-only than of neutral-only lists at fast or slow rates. Present results supported this prediction, suggesting that taboo superiority in immediate recall reflects context-specific binding processes, rather than context-free arousal effects, or emotion-linked differences in rehearsal, processing time, output interference, time-based decay, or guessing biases. Keywords: emotion, short-term memory, arousal, binding How does emotion linked with taboo words impact immediate memory? Recent results indicate two related effects of emotion on immediate recall of rapidly presented rapid serial visual presenta- tion (RSVP) lists containing a mixture of taboo and neutral words (MacKay et al., 2004; MacKay, Hadley, & Schwartz, in press). One is taboo-superiority: better recall of taboo than of neutral words matched for length and familiarity. The other is poorer recall of neutral words at least one word after a taboo word (the word-after effect) and up to two words before a taboo word (the word-before effect; for related “retrograde amnesia” effects in- volving other distinctive or emotionally salient stimuli, see Loftus & Burns, 1982, and Tulving, 1969). The goal of the present study was to further our understanding of immediate memory and emo- tion by testing alternative explanations of these effects. The two main alternatives were arousal theory (see, e.g., Cahill & van Stegeren, 2003; Hamann, Ely, Grafton, & Kilts, 1999; Kensinger, & Corkin, 2004; Le Doux, 1996, pp. 206–208; Maratos, Allan, & Rugg, 2000; and Phelps et al., 1998) and the priority-binding assumption of binding theory (see, e.g., MacKay et al., 2004; MacKay et al., in press; see also MacKay, Burke, & Stewart, 1998; and MacKay, Stewart & Burke, 1998). Arousal Theory and the Taboo-superiority Effect Arousal theory explains taboo superiority as a facilitation effect. Under the arousal hypothesis tested here, low-level sensory inputs directly engage an emotional reaction system (say, the basolateral amygdala) that triggers release of neurotransmitters or endogenous stress hormones (such as epinephrine and cortisol) that facilitate memory consolidation for emotional events that are suprathreshold and not overly traumatic or repression-prone. These amygdala- mediated encoding processes enable consolidation of emotional events in a brain region such as the hippocampus so that the brain can achieve memory strength that is directly proportional to mem- ory importance and arousal without interference from other ongo- ing events or stimulus factors (see, e.g., Cahill & McGaugh, 1998; Cahill & van Stegeren, 2003). Like other emotion-linked stimuli associated with enhanced event recall, taboo words are not repression-prone when presented at suprathreshold rates (see MacKay et al., 2004), and taboo (but not neutral) words induce enhanced skin conductance, an uncon- scious index of sympathetic nervous system activity and emotional arousal (see, e.g., LaBar & Phelps, 1998; and Harris, Aycicegi, & Gleason, 2003). The arousal hypothesis therefore applies to the present tasks: immediate recall of taboo and neutral words pre- sented at varied rates in pure (taboo-only and neutral-only) lists (Experiment 1) and in mixed taboo–neutral lists (Experiment 2). The arousal hypothesis predicted taboo superiority, independent of stimulus factors such as presentation rate and pure- versus mixed- list type in Experiments 1 and 2. Binding Theory, Priority-binding, and the Taboo-superiority Effect Like arousal theory, binding theory applies across a wide range of emotion-linked contexts. However, we will reserve general or task-independent binding theory claims for the General Discus- sion. Here we focus on applying binding theory to immediate list recall (MacKay & Burke, 1990) and the taboo-superiority effect (MacKay et al., 2004). Like other distributed memory theories (e.g., Burgess & Hitch, Christopher B. Hadley and Donald G. MacKay, Psychology Depart- ment, University of California, Los Angeles. The authors gratefully acknowledge support from the Samuel A. MacKay Memorial Research Fund and thank Ronit Menashe, Melissa Murren, and Sam Soleimany for running participants, coding data, and providing general assistance. Correspondence concerning this article should be addressed to Christo- pher B. Hadley, Psychology Department, UCLA, Los Angeles, CA 90095- 1563. Email: chadley@psych.ucla.edu Journal of Experimental Psychology: Copyright 2006 by the American Psychological Association Learning, Memory, and Cognition 2006, Vol. 32, No. 1, 79–88 0278-7393/06/$12.00 DOI: 10.1037/0278-7393.32.1.79 79 1999; Howard & Kahana, 2002), binding theory assumes that list recall depends on the formation of new bindings or associative links between each word in a list and its episodic context, that is, an internal representation of where or when the word occurred, for example, in a particular list or experiment. To recall that a particular word occurred in a particular list under binding theory, people activate the word via its link to the episodic context. However, words are complex stimuli with at least three aspects (phonology, orthography, and semantics), and binding theory specifies precisely what aspect of a word becomes bound to its episodic context during list learning: the lexical node that represents the meaning of the word in the cortex. Binding theory also specifies the process whereby a lexical node becomes bound to its episodic context: An activated lexical node primes or readies for activation a binding node (located, say, in the hippocampus1) that specializes in binding two or more general classes of cortical nodes. The general classes of cortical nodes with links to a particular binding node are known as the domain of the binding node, and different binding nodes have different domains. The present study focuses especially on a binding node with two classes of cortical nodes in its domain: nodes representing episodic context and lexical nodes representing taboo words. Under binding theory, activating the binding node with that domain determines whether a taboo word will be linked to its episodic context and recalled following list presentation. For pure or unmixed lists containing either taboo or neutral words, the processing sequence is as follows: Lexical nodes rep- resenting word meaning are activated in the cortex and call up or prime their connected binding node word by word in sequence; each binding node in turn forms a new connection between what- ever nodes are currently activated in its domain, here, the classes of cortical nodes representing word meaning and episodic context. However, because binding normally proceeds sequentially one link at a time, and connection formation is a relatively time-consuming process, the time to form a new connection between a lexical node and its episodic context can greatly exceed stimulus duration for rapidly presented stimuli. This means that a word in a long or rapidly presented list will be forgotten or irretrievable if its lexical node is no longer activated when the binding node for forming its link to episodic context is applied. Because of emotion-linked priority binding, the processing se- quence differs somewhat for mixed lists containing both taboo and neutral words. When children first learn a taboo word, a strong link is formed between the lexical node representing the meaning of the taboo word and the system that generates emotional reactions (say, the amygdala). This assumption seems plausible because emo- tional reactions to taboo words are based on word meaning rather than acoustics, phonology, or orthography; it is word meaning rather than acoustics, phonology, or orthography that makes taboo words taboo and emotionally arousing. By way of illustration, a word such as ask is nonarousing and neutral in emotional tone despite extensive overlap in acoustics, phonology, and orthogra- phy with the taboo word ass, and many examples of such low-level overlap between taboo and neutral words could be cited. In short, word meaning triggers emotional reactions in binding theory, unlike in arousal theory, in which sensory- rather than semantic- level inputs can directly engage the amygdala. However, neutral words lack links to the amygdala for triggering strong emotional reactions under binding theory. Activating a lexical node therefore engages the binding node system for taboo and neutral words under binding theory, but only taboo words strongly engage the emotional reaction system. This emotional reaction system responds immediately and, via direct amygdala-to-hippocampus links, gains access to the binding nodes that link stimuli to their episodic contexts. The resulting interac- tions between the amygdala and hippocampus serve to delay the activation of currently primed binding nodes for (less important) neutral stimuli, which only become bound to their episodic context after binding for (more important) emotion-linked stimuli is com- plete. These emotion-linked adjustments in binding order can only occur for stimuli such as taboo and neutral words that engage different binding nodes: Repeated application of the same type of binding node as in pure taboo-only lists is not subject to prioriti- zation. However, emotion-linked delays in the activation of neutral binding nodes neither speed up the activation of emotion-linked binding nodes nor increase the “binding resources” (e.g., time, energy, or rate of processing) available to emotion-linked stimuli. Nor do emotion-linked delays in the activation of neutral binding nodes reduce the binding resources available to neutral stimuli when their binding nodes are applied. The priority-binding as- sumption therefore contrasts with other hypotheses in which an emotion-linked stimulus prematurely terminates ongoing encoding of a preceding neutral stimulus (Loftus & Burns, 1982; MacKay et al., 2004) or in which the time or energy available for binding emotion-linked stimuli trades off with the time or energy for binding neutral stimuli (see Meinhardt & Pekrun, 2003). The priority-binding assumption readily explains the two al- ready observed types of interference in mixed lists containing taboo and neutral words: word-before and word-after effects, that is, poorer recall of neutral words before and after taboo words in mixed lists presented at 100 ms/word (MacKay et al., in press) or 170 ms/word (MacKay et al., 2004). The word-before effect occurs because priority binding for a taboo word delays activation of the binding node for linking the immediately prior neutral word to its episodic context, and because of the rapid RSVP rate, this neutral word is no longer activated when its binding node is applied. As a result, the link to episodic context necessary for retrieving this “word-before” as part of that particular list cannot be formed at fast presentation rates (200 ms/word or less). Binding priority for a taboo word likewise prevents episodic encoding of the immedi- ately following neutral word with even greater likelihood in rap- idly presented lists: By the time that binding processes for the taboo word have been completed and the lower-priority episodic binding node for the neutral “word-before” has been applied, activation of the neutral “word-after” has decayed with high prob- ability, so that when its episodic binding node is finally applied, it cannot form the episodic link necessary for retrieving the word- after as part of the list. To summarize, the arousal hypothesis explains taboo superiority as being due to encoding facilitation for taboo words independent of the occurrence of neutral words, whereas the priority-binding assumption explains taboo superiority as being due to interference with the encoding of neutral words before and after taboo words in rapidly presented mixed lists. By extension, binding theory pre- dicted that word-before and word-after effects will cause taboo 1 We assigned different types of nodes to neuroanatomical loci here to allow same-level comparison with arousal theory. However, these hypothetical neu- roanatomical assignments are not central to how binding theory functions: Different neuroanatomical assignments would yield the same predictions. 80 HADLEY AND MACKAY superiority for the rapidly presented lists of randomly intermixed taboo and neutral words in Experiment 2: Taboo words will be more retrievable than neutral words in these lists because priority- binding enables taboo words to become linked to their episodic context with higher probability than neighboring neutral words. Experiment 1 tested the “taboo-equality” prediction derived from the priority-binding assumption: There will be no better recall of taboo than of neutral words in pure (unmixed) lists at fast or slow presentation rates because emotion-linked delays in bind- ing only occur for stimuli that engage different types of binding nodes, regardless of presentation rate. As a result, the emotion- linked delays in the binding of neutral words that occur in mixed taboo–neutral lists will not occur in taboo-only lists such as ass dyke piss that involve repeated application of the same type of binding node. Experiment 1 also tested predictions derived from two subsid- iary hypotheses for explaining taboo superiority in mixed lists: the rehearsal hypothesis and the processing time hypothesis. The re- hearsal hypothesis applies especially to presentation rates that are slow enough to allow rehearsal, for example, 1,000 ms/word or more. Because people are more likely to rehearse emotional than neutral events (see, e.g., MacKay et al., 2004, for a recent review), taboo superiority may reflect greater rehearsal of taboo than of neutral words presented at 1,000 ms/word. This rehearsal hypoth- esis predicted greater taboo superiority at the 1,000 ms/word than at the 200ms/word rate in Experiment 1 because 200 ms/word is too rapid to allow rehearsal (see, e.g., Murdock, 1974, p. 168). The processing-time hypothesis applies especially to fast rather than to slow presentation rates. Under the processing-time hypoth- esis, taboo superiority arises because emotion enables faster acti- vation (rather than priority-binding) of emotion-linked words, so that taboo words enjoy surplus encoding time relative to neutral words in pure lists. This longer encoding time (say, tens of milli- seconds) is of little consequence at slow presentation rates but can spell the difference between successful versus unsuccessful encod- ing at rates such as 200 ms/word. The processing-time hypothesis therefore predicted relatively better recall of taboo than of neutral words at fast rather than slow presentation rates. Experiment 1 Experiment 1 compared immediate recall of taboo-only versus neutral-only lists at fast (200 ms/word) versus slow (1,000 ms/ word) RSVP rates. We matched the taboo and neutral words for length and familiarity as in MacKay et al. (2004, in press), and we selected the neutral words from a set of foodstuff- and cooking- related terms, a restricted semantic category with high category coherence resembling taboo words. We also attempted to equate our taboo and neutral words on factors such as set size, imagery, connotative strength, syntax, and semantic similarity. However, we make no a priori claim that we successfully equated these memory-related factors for the taboo and neutral words and par- ticipants in Experiment 1, especially because in the case of taboo words, some of these factors vary as a function of participant gender and sexual orientation (see Jay, 2000, p. 168) and cannot be matched a priori. Instead we will present results that speak post hoc to the success of our stimulus matching procedures. The arousal hypothesis predicted taboo superiority at both fast and slow presentation rates in Experiment 1. Binding theory pre- dicted taboo equality; that is, no better memory for taboo-only than for neutral-only lists at the slow or fast presentation rates. The rehearsal hypothesis predicted taboo superiority at the slow rate but not at the fast rate. The processing-time hypothesis predicted relatively greater taboo superiority at the fast than at the slow rate. Method Participants Participants were 16 University of California, Los Angeles, undergrad- uates (5 men and 11 women, aged 20–25 years, M � 21.5, SD � 1.56), who spoke fluent English, reported normal or corrected-to-normal vision, and received partial course credit or payment of $5 for participating. Participants were informed that they might see offensive or taboo words and could have served in another ongoing experiment without taboo words, but none chose that option. Materials and Design The within-subject design was 2 (word type: taboo vs. neutral)� 2 (rate: 200 ms/word vs. 1,000 ms/word). The Appendix shows the materials: 68 single-syllable taboo and neutral words matched in pairs for initial conso- nant and length in letters. The taboo words (N � 34) were socially proscribed insults, sexual references, and profanities, and the neutral words (N � 34) were foodstuff- or cooking-related words. A group of 27 undergraduates resembling participants in Experiments 1 and 2 rated all 68 words for familiarity and “tabooness” on scales of 1–5 (see MacKay et al., 2004; in press), with mean results shown in the Appendix. Familiarity ratings did not differ for taboo (M� 2.96, SD� .901) versus neutral (M� 2.97, SD� .903) words, t(66)� .047, p� .963, but obscenity ratings were reliably greater for taboo (M � 3.16, SD � .793) than for neutral (M � 1.08, SD � .171) words, t(66) � 15.012, p � .001. We created experimental lists containing 7, 8, 9, or 10 words2 by randomly sampling without replacement from the taboo or neutral word sets. We then modified the lists to ensure that adjacent words never formed familiar phrases that could aid recall. Each participant saw and recalled 32 lists: 16 lists (two of each list type and length) at each rate, with order of presentation rates and list types counterbalanced across participants. To equate proactive interference across list type and presentation rate, we arranged for each word to appear only once in each eight-list block, for a total of four repetitions across the eight four-list blocks, with order of list type and list length randomized within each block to discourage the adoption of length-based strategies in recall. Procedure We presented the lists in a 36-point Chicago font using standard RSVP procedures with words centered on the screens of Macintosh G3 computers (Apple Computer, Cupertino, CA) running PsyScope (Cohen, MacWhin- ney, Flatt, & Provost, 1993). Verbal and computer-presented instructions called for item recall, in order if possible. To ensure comprehension of the instructions, we showed participants four practice lists containing neutral words from the categories of animals, minerals, or clothing presented at the same rate as the first block of experimental lists. To initiate each trial, participants p