Queens Affective Neuroscience Laboratory

Research

Mood and Processing

The general failure to find reliable instances in which mood leads to mood-congruent priming has suggested to some that emotion may not influence automatic processes at all (Bower & Meyer, 1989; Calvo & Castillo, 1997; Forgas, 2001; Matthews & Wells, 1999). Indeed, Matthews and Wells (1999) conclude that the primary influences of emotion on cognition involve controlled processing. Such conclusions, however, should be tentative because the evidence comes mainly from mood-congruent memory studies. Recent reviews of that literature (e.g., Wyer, Clore, & Isbell, 1999) indicate that, despite the popularity of the idea, affect itself may not activate similarly-valenced material in memory. Hence, it is premature to conclude that affect does not influence implicit, automatic processes.

Indeed, I have discovered that in fact mood does influence the accessibility of implicit associations as measured by priming effects. Individuals completed a comparative priming task while in either a happy or sad mood state. We found that across three priming tasks (evaluative, categorical, and lexical decision) positive mood groups showed typical priming effects, but negative mood groups failed to show evidence of priming ( Storbeck & Clore, submitted ). Consistent with the affect-as-information approach, these priming results suggest that affect can serve to guide processing, resulting in either the use of accessible cognitions (positive moods) or the inhibition of such accessible cognitions (negative moods). Other research has found that priming can be prevented by item-specific processing elicited by attentional demands or contextual manipulations (De Houwer, & Randell, 2002; Harris, Pashler, & Coburn, 2004; Stolz, & Besner, 1999). Our research shows that negative moods can also disrupt priming through the elicitation of such item-specific processing.

Sadness Influences Encoding, Reducing False Memory

Can moods govern how we encode and remember events? For example, are happy individuals more likely to recall fictitious events than sad individuals? I discovered that sad moods reduce the occurrence of false memories in the Deese-Roediger-McDermott (DRM) false memory task (Storbeck & Clore, 2005). We presented word lists that were highly associated to a single, non-presented word (the critical lure). Evidence suggests that critical lures come to mind when individuals relationally encode the lists (Brainerd & Reyna, 2003; Roediger et al., 2001).By contrast, item-specific processing is thought to impair relational encoding, thereby reducing false memories (Arndt & Reder, 2003). The affect-as-information theory suggests that affect serves as feedback during tasks and thus can govern processing effects (Clore et al., 2001; Clore & Storbeck, in press). Specifically, positive affective cues promote relational, gist processing, whereas negative affective cues promote item-specific processing. To test these ideas, positive and negative moods were induced before individuals completed the false memory task. We hypothesized and found that negative affect reduced the recall of critical lures, suggesting that sad moods promoted item-specific processing.

The next obvious question was when does affect influence the activation of the critical lures? Does sadness prevent the lures from coming to mind during encoding or does it prevent people from responding with the lure at retrieval? We ran two experiments to address these questions. First, we induced positive and negative mood states, and then participants completed the false memory task. However, this time they were asked to recall not only the presented words, but also any other words that came to mind. Such instructions are effective for determining whether critical lures were activated during encoding (Hege & Dodson, 2004). We discovered that such lures were less likely to come to mind for sad individuals, suggesting that they were not activated at encoding for these individuals. Second, we manipulated the timing of the mood induction (either before or after list learning) and used a recognition task instead of a recall task. The results were clear. Moods influenced the recognition of critical lures when induced before learning, but not when induced after learning. Moreover, using a signal-detection framework designed for assessing biased responding during retrieval, we found that response biases were not responsible for the effects. We concluded that sad moods promote item-specific encoding, which reduces false memories. These results lead us to expect that positive affect may be functional in tasks where relational encoding is required, as in learning or to generate new ideas, but that negative affect may be functional in tasks requiring item-specific encoding, as in visual search or proofreading.

The studies just described showed affective influences on encoding in a recall task. I have further examined whether affect can influence recognition judgments in similar false memory designs ( Storbeck, Dissertation ). In these studies, I replicate previous findings, demonstrating reduced false memories for sad individuals during a recognition task. In addition, with a recognition task we were able to control for response biases by using a signal-detection framework (Banks, 1970; Miller & Wilford, 1999). Thus, when I controled for such response biases, the false memory effect persisted still, suggesting that mood appears to influence encoding, but not retrieval processes.

Sadness Enhances Spatial Memory

The false memory findings suggest that negative moods promote item-specific encoding. If that is the case, then sad moods might be expected to enhance encoding of other types of item-specific information, such as source memory. Therefore, I introduced spatial location as one example of source information. Thus, words in the lists appeared in particular locations on the screen. I discovered that, in fact, sad individuals showed better spatial memory. In addition, happy individuals again recalled more critical lures.

An allied line of work has been done by Jeremy Gray and colleagues who find that happy moods enhance verbal working memory and that sad moods enhance spatial working memory (Gray, 2001; Gray, Braver, & Raichle, 2002). Using fMRI techniques, they find greater BOLD response in left dorsal lateral prefrontal cortex when people are approach motivated and engaged in verbal working memory tasks. They find greater BOLD response in corresponding right prefrontal areas when individuals are avoidance motivated and engaged in spatial processing. I am following up this research that uses cerebral blood flow measures by examining electrical activation (EEG) in the prefrontal cortex (Davidson, 1998). With Jim Coan, a clinical neuropsychologist, I am examining performance as a function of mood on n-back tasks designed to assess verbal and spatial working memory. At issue is whether the relationship between affective states and performance on verbal and spatial tasks is indeed mediated by activity in left and right prefrontal cortical areas.

Affect and Cognition: Is one more basic?

One broad theoretical question arises: Are affect and cognition independent systems or are they integrated? Several prominent emotion theories suggest that the affective system may be independent from cognition (Bargh, 1997; Zajonc, 2001). However, I and several colleagues (Storbeck, Robinson & McCourtt, 2006, GRP ) have reviewed the social psychological and neuroscience evidence, which led us to the conclusion that affect and cognitive processes are highly integrated. With this assumption, we have proposed that specific emotions modulate selective cognitive processes. Moreover, we suggest that affect can regulate cognitive processes from initial stages of perception to the final stages of memory consolidation. For example, we believe that both dimensions of affect (valence and arousal) modulate cognitive processes (Clore & Storbeck, in press). Specifically, we propose that valence governs processing style (e.g., relational vs. item-specific), whereas arousal governs attention and consolidation. In addition, both valence and arousal can interact to influence affective evaluations and behavior (see Robinson, Storbeck, Meier, & Kirkeby, 2004; Robinson, 1998). Thus, I propose that affect is integrated with cognitive processes, and as a result, that affect routinely modulates cognition, and cognition constantly modulates affect.

Affective Priming and Mere Exposure are highly visible and intriguing phenomena that have captured people's imagination. From a strong emphasis on such phenomena, an “affective primacy” view has arisen, which says that affect dominates the early processing of stimuli. Although it is true that affect can be elicited quickly, current data and neurological models suggest that affect does not in fact arise before other cognitive and semantic processes occur ( Storbeck, Robinson & McCourtt, 2006, GRP ). It appears that a truly useful view of emotion and its effects will ultimately require that it be integrated with a larger understanding of the cognitive functions of the brain.

For instance, traditional studies of semantic priming suggest that memory is organized by descriptive features (e.g., McClelland, 2003; Neely, 1991). On the other hand, recent studies of affective priming by social psychologists imply that memory is organized by affective features (e.g., Ferguson & Bargh, 2003). One way to resolve this conflict is to examine empirically whether individuals are faster to access descriptive or affective information from memory. We studied this question in four priming studies, some employing words and some pictures. Consistently, the dominance of descriptive over affective priming was observed, and we concluded that implicit descriptive relations were more likely to be accessed than affective relations. We found affective priming only when affective task demands were made salient and when the stimuli used eliminated the possibility of descriptive categorization ( Storbeck & Robinson, 2004, PSPB ). Such evidence suggests that what appears to be affective priming is actually a special case of semantic priming. This conclusion is supported by other studies that also suggest that affective priming depends on semantic associations in memory (e.g., De Houwer & Randell, 2004; Kemp-Wheeler & Hill, 1992; Erdley & D'Agostino, 1988).

The studies described above concerned the relative accessibility of descriptive and affective attributes. To complement these studies, we then asked which type of information is accessed faster. In answering this question, we employed a forced-choice response window paradigm. This paradigm forces participants to make rapid descriptive and affective categorizations and examines accuracy as the dependent measure. We found that within the shortest response window, participants were more accurate at making descriptive than affective classifications ( Storbeck, Robinson, Ram, & Meier, in prep ). This discovery, that semantic associations are accessed faster than affective associations, again implies that semantic memory is organized around descriptive features, which can include, but are not dominated by, affective features. Using subliminal priming techniques, Greenwald and colleagues (2004) have also produced evidence that supports such conclusions.

In addition to behavioral studies, I have conducted a systematic review that included relevant work in cognitive neuroscience. This evidence also leads to the conclusion that semantic rather than affective information is primary, and that affective reactions to lexical and pictorial stimuli depend on prior semantic analysis. The review ( Storbeck, Robinson & McCourtt, 2006, GRP) presented evidence that stimuli are first categorized and identified at an unconscious level on the basis of their semantic features. This information can then serve as a basis for affect elicitation. In addition, the review provides substantial evidence that both affective priming and mere exposure effects actually rely on cognitive processing.

Can mood influence the accessibility of attitudes?

Our previous studies found that mood influences the accessibility of general semantic associations ( Storbeck & Clore, submitted ). In addition, we are also currently investigating whether mood can influence the accessibility of implicit attitudes. The implicit association test (IAT) was designed to measure implicit attitudes by assessing their underlying automatic evaluations (Greenwald, McGhee, & Schwartz, 1998). With Brian Nosek, I am conducting a web-based study to investigate whether daily mood can influence the accessibility of implicit attitudes, as measured by the IAT. We expect that individuals who report a higher level of positive affect before taking the IAT on the web will have greater accessibility for their implicit attitudes. If so, we should see a larger IAT effect size for these individuals when compared to individuals in less positive moods ( Storbeck & Nosek – Project Implicit ). Such findings would be consistent with the affect-as-information approach, which suggests that by virtue of increased relational processing, individuals in positive moods should have greater access to target-relevant attitudes than those in negative moods.

We are also examing examines how emotion may facilitate the learning of attitudes towards novel groups. This project (Storbeck, Ranganath, & Nosek) investigates whether emotional arousal can enhance the learning about the characteristics of novel groups, and whether the affect biases behavioral impressions of two fictitious groups. We expect that emotional arousal will facilitate the acquisition of attitudes towards novel groups as well as their preservation over time.

Arousal and Valence

Other conclusions from my research that challenge conventional thinking about affect flow from a discovery that, rather than functioning independently as is usually assumed, valence and arousal interact to influence evaluation. Models of affect typically include both an arousal and a valence component (e.g., Larsen & Diener, 1991). Emotional theorists generally find that self-reports of valence and arousal show two independent dimensions (e.g., Lang, 1995; Russell & Feldman-Barrett, 1999). Recent papers have suggested a simple model focused only on valence and suggest that positive items elicit approach and negative items elicit avoidance tendencies (e.g., Cacciopo et al., 1993; Neumann, Forster, & Strack, 2003). Some other models have focused only on arousal and suggest that intense, arousing stimuli trigger avoidance, whereas mild, familiar stimuli trigger approach tendencies (Zajonc, 2001). However, we proposed, and have found, that evaluations and approach-avoidance tendencies depend on both the arousal and the valence components.

In a series of studies, we discovered that arousal and valence interact to influence evaluations and approach-avoidance tendencies ( Robinson, Storbeck, Meier, & Kirkeby, 2004, PSPB ). Both words and pictures were viewed in a design that varied high vs. low arousal and positive vs. negative affect. We found a robust and interesting interaction involving speed of response. The quickest evaluations were for negative, high arousal stimuli and positive, low arousal stimuli. Two experiments found that the quickest decisions for both approach and avoidance occurred when items were negative and high in arousal or positive and low in arousal. This particular finding is interesting, because most approach-avoidance models propose that individuals are especially fast when approaching positive or avoiding negative objects (e.g., Neumann et al., 2003). However, varying arousal in addition to valence allowed us to see that this simple dichotomous model does not hold up. These findings show that, despite popular assumptions, arousal and valence do not, in fact, function independently, but interact in a systematic way to influence evaluations and emotion-related behavior ( Robinson, Storbeck, et al., 2004, PSPB).

Arousal and mood

Studies from Neurobiology have found that arousal from an emotional experience can enhance memory consolidation for that experience (e.g., McGaugh et al., 2001). When arousal has been examined in the mood and memory literature, results so far have not proven particularly illuminating (Bower & Forgas, 2000; Ellis & Moore, 1999). We find this to be particularly surprising given the robust and reliable evidence from neurobiology that suggests arousal should be associated with enhanced memory. One issue regarding previous studies involving mood and memory is that most studies have failed to investigate long-term memory.

I am investigating whether a happy individual will remember different information than a sad individual, without the intention of doing so. I have already shown that happy and sad individuals remember different information during false memory and spatial memory tasks, and now I want to investigate what is remembered several days later. Affective states will be induced to govern encoding processes, and emotional arousal will be induced to govern the consolidation of what is encoded into long-term memory. Thus, we will be able to examine how affect influences encoding and how affect modulates consolidation of biased perceptions into long-term memory. This basic process, in which emotional arousal leads to indelible memories is important for many reasons, and is also believed to be causal in PTSD, in which memories of traumatic events remain distressingly vivid.

Arousal and perception

In addition to these influences on memory, related processes appear to be involved in the perception of physical space. Fifty years ago, work on the New Look in perception (Bruner, 1957) hypothesized that perception was influenced not only by objects and the conditions of their viewing, but also by the emotions and motivations of the viewer. The phenomena proved unreliable at the time, but more solid data and a more elaborated understanding of the visual system are renewing this idea. With recent Phd. Jeanine Stefanucci, I am examining whether early stages of cognition, such as visual perception, can be biased by emotions. Initial research (e.g., Stefanucci & Proffitt, 2006) showed that that anxiety about heights can lead to overestimations of perceived heights. To extend this line of research, we are currently examining whether general emotional arousal can also lead to such height estimation biases. Subsequently, we plan to assess whether cognitive strategies for emotion regulation will be effective in reducing these emotion-induced biases in estimation. Finally, this project dovetails with the project discussed above to investigate how these affective influences on height perception enter into long-term memory for specific heights. One question is whether overestimations of height get stored in long-term memory in a way that then influences beliefs and attitudes about heights in general? Just as the long term memory project taps processes basic to PTSD, this anxiety and perceptual overestimation project appears to tap some of the processes involved in phobias and anxiety-related disorders.