Self-report captures 27 distinct categories of bridged by continuous gradients

Alan S. Cowena,1 and Dacher Keltnera

aBerkeley Social Interaction Laboratory, Department of , University of California, Berkeley, CA 94720

Edited by Joseph E. LeDoux, New York University, New York, NY, and approved August 7, 2017 (received for review February 9, 2017) are centered in subjective experiences that people rep- tional experiences of similar valence and , such as resent, in part, with hundreds, if not thousands, of semantic and , or and (1, 14, 19, 22–24). Varying combi- terms. Claims about the distribution of reported emotional states nations of such dimensions have been the focus of hundreds of and the boundaries between emotion categories—that is, the studies linking reported emotional experience to behavior, phys- geometric organization of the semantic space of emotion—have iology, and brain activity (25–36). sparked intense debate. Here we introduce a conceptual frame- A second approach to emotional experience details how spe- work to analyze reported emotional states elicited by 2,185 short cific emotion categories, such as , fear, and , describe videos, examining the richest array of reported emotional expe- discrete clusters of states within a presupposed semantic space. riences studied to date and the extent to which reported expe- More precisely, basic emotion theories posit that a limited num- riences of emotion are structured by discrete and dimensional ber of clusters, ranging in theoretical accounts from 6 to 15, geometries. Across self-report methods, we find that the videos describe the distribution of all emotional states (16, 37, 38). A reliably elicit 27 distinct varieties of reported emotional expe- cluster, or emotion family, may go by a prototypical label, such rience. Further analyses revealed that categorical labels such as “anger,” and contain closely related states such as irritation, as better capture reports of subjective experience , and (39) that occur in similar situations (14). than commonly measured affective dimensions (e.g., valence As with affective dimensions, such emotion families, discretely and arousal). Although reported emotional experiences are rep- partitioned into categories, have been the focus of hundreds of resented within a semantic space best captured by categorical empirical studies (16, 25, 27–29, 32, 35, 40–49). Clearly, claims labels, the boundaries between categories of emotion are fuzzy that specific affective dimensions and emotion categories cap- rather than discrete. By analyzing the distribution of reported ture how people report on their emotional experience—and, by emotional states we uncover gradients of emotion—from implication, other emotion-related processes—have shaped the to fear to horror to , to aesthetic appreciation to study of emotion. awe, and others—that correspond to smooth variation in affective Despite the pervasive influence of these theoretical approaches, dimensions such as valence and dominance. Reported emotional empirical progress in understanding how reported emotional states occupy a complex, high-dimensional categorical space. In experiences are organized within a semantic space has been mod- addition, our library of videos and an interactive map of the emo- est. Statistical approaches to testing these theoretical claims have tional states they elicit (https://s3-us-west-1.amazonaws.com/ been unable to openly explore how reported emotional expe- emogifs/map.html) are made available to advance the science of riences are organized within a more general topological space emotion. that could simultaneously involve both distinct clusters and gra- dients of relatedness in response to varied situations. As a result, emotional experience | semantic space | dimensions | discrete emotion Significance entral to the science of emotion is the principle that emo- Ctions are centered in subjective experiences that people rep- Claims about how reported emotional experiences are geo- resent with language (1–10). People represent their transient metrically organized within a semantic space have shaped the experiences within a semantic space that includes hundreds, if study of emotion. Using statistical methods to analyze reports not thousands, of semantic terms that refer to a rich variety of of emotional states elicited by 2,185 emotionally evocative emotional states (11–13) most readily characterized by the types short videos with richly varying situational content, we uncov- of situations in which they occur (14, 15). Given that experi- ered 27 varieties of reported emotional experience. Reported ence is often considered the sine qua non of emotion (1–10), the experience is better captured by categories such as “amuse- understanding of the semantic space of reported emotional expe- ment” than by ratings of widely measured affective dimen- riences is crucial to progress in characterizing emotion-related sions such as valence and arousal. Although categories are cognition, signaling, and physiology (16), as well individual dif- found to organize dimensional appraisals in a coherent and ferences in emotion (17). powerful fashion, many categories are linked by smooth gra- One line of theorizing has documented the underlying dimen- dients, contrary to discrete theories. Our results comprise an sions of the semantic space of reported emotional experience, approximation of a geometric structure of reported emotional focusing on the core affective states that make certain expe- experience. riences feel emotional (18, 19). Efforts to identify a finite set of axes central to reported experiences of emotion have Author contributions: A.S.C. and D.K. designed research; A.S.C. and D.K. performed most consistently yielded two affective dimensions—valence and research; A.S.C. contributed new reagents/analytic tools; A.S.C. analyzed data; and A.S.C. arousal—that are posited to lie at the core of all affective expe- and D.K. wrote the paper. riences, from more diffuse moods to specific emotions. These The authors declare no conflict of . dimensions are thought to describe raw, disconnected This article is a PNAS Direct Submission. as opposed to emotions felt toward specific objects or situations Data deposition: An interactive map related to this study is available at https://s3-us- (14, 18, 20, 21). To account for the occurrence of specific emo- west-1.amazonaws.com/emogifs/map.html. tions, a related line of inquiry has documented how other, more 1To whom correspondence should be addressed. Email: [email protected]. context-directed affective dimensions such as dominance, cer- This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. tainty, agency, effort, and differentiate reports of emo- 1073/pnas.1702247114/-/DCSupplemental.

E7900–E7909 | PNAS | Published online September 5, 2017 www.pnas.org/cgi/doi/10.1073/pnas.1702247114 Downloaded by guest on October 2, 2021 Downloaded by guest on October 2, 2021 oe n Keltner and Cowen pc orsod oadsic ait frpre emotional reported of variety distinct semantic a dimen- this to these of dimension corresponds along emo- each space report all speaking, can of Conceptually people distribution sions. the that space—and experiences indepen- the tional of in number directions dimensionality—the dent space its a by Such space. characterized semantic is a experi- within points emotional to self-reported of correspond that ences assumption readout broad direct the from a not measures is self-report is self-report that emotion experience. capture; of processes partially experience complex only subjective many The by gen- class. culture, shaped of social use and language der, on and appraisals influences awareness and emotional automatic (66), of expression granularity and the mem- in sensations, differences moods, (65)], physical [e.g., words inchoate in ories, phenomena subjective certain captur- in ing difficulties (64), basic reference including linguistic ambiguities, of to indeterminacies meaning subject the is that self-report note emotional to of important is behav- it expressive processes Nonetheless, as (60–63). such psychological ior emotion-related, other be with to thought correlate are and that 9) inter- person’s (1, experi- a state subjective about information nal of carry Self-reports measure self-reports 9). that accessible (1, ence given most “”), the or currently anger are (e.g., terms emotion of 59)? how 37, (14, in experienced transitions be smooth to or reported are jumps emotions discrete and to widely amusement Do correspond as effort? of awe such and terms categories valence emotion in between as or boundaries such of awe, dimensions, terms and affective in amusement measured as understood such do better categories, emotion experience Is situations? fol- of emotional distinct the varieties across reported experiencing to distinct report answers reliably many people provide How we questions. studied, widest lowing clips—ever the examin- video to By evoca- tive response situations—powerfully significant (1–10). in psychologically of reported experience array experiences the of emotional self-report, measure ing emotional the used by characterizing widely captured to most experience approach of empirical varieties and framework tual emotional reported for account (58). experiences to needed how are estimating dimensions generate in many not probabilities posterior do or that intervals, methods confidence approaches heuristic multivariate on simply past relied not have Further, do valence). that awe positive be of reflect may judgments that reliable items individual (e.g., on not cor- independent agree but people for awe) which and to accounted degree valence have the between (e.g., studies emotional items Such between recalled relations 57). or anal- 53, contemporaneous factor (24, upon as of experiences founded such self-reports are techniques to dis- claims multivariate ysis to many apply correspond Also, that 56). to studies (16, discovered behaviors emotional too categories expanding of tinct is emotion the variety including studies of rich distinct, past the array deem result, to people in a priori, that captured As a experiences states (55). generalize, films to emotional narrow Levenson of and array Gross Sys- the - the Picture used and Affective widely (54), International of the 10 tem true as capture is such same Schedule, stimuli, The Affect eliciting (51–53). Negative emotions and part 12 measures, Positive to large in self-report the 24), used as ref. Widely such encompassing see have reasons. (although often that methodological (46) scope, states categories for in 12 the to limited Further, 6 been 50). only have 37, 20, studied con- of 3, been areas (2, controversial claims of being space trasting issues’ semantic these larger cat- despite a emotion within emotion, between arrangement boundaries their the and both egories about known is little edu fteecnieain,i u hoiigw begin we theorizing our in considerations, these of Heedful self-reports on focusing by issues conceptual these address We concep- based mathematically a introduces investigation This P values, obntos r rqetyue omaueself-reported measure to used frequently 14 along varying are in placement rated which, its participants dimensions, combinations, of affective of terms of in group scales measured viewed final widely they A videos categories. 12 of of each set correspond- this states 34 to emotional the reported ing finer 30 feel the in of them of interrogate structure made to the each it us detail allowing which rated interest, to of participants degree categories the emotion of of group terms in expe- second videos subjective A of (75). reports arise rience unconstrained allowing spontaneously thus relatively emotion videos, of people’s 30 categories in which of ascertain each to to us response interpretations response emotional free their offered sam- of group random One a videos. to of response in pling experience emotional kinds of each three reports of by of one obtain provided elicited participants to states specifically, More Turk emotional video. of Mechanical judgments Amazon (9–17) on repeated participants to library soldiers stunts, others. risky many surgeries, and home, calls, returning close dance, and food, accidents delicious sports, handshakes, awkward celebri- films, hated nostalgic and ties, respected acts, sexual and pratfalls, feces physical warfare, vomit, beauty and explosions natural ani- disasters, architecture, natural endearing and wonders, and art snakes, elephants, and and spiders whales mals, death, and pro- and posals, weddings babies, and sig- births psychologically including s of situations, 5 range nificant wide about exceptionally lasted an average portrayed on and videos The origins). theoretical their (see (74) horror and as anxiety, such fear, states inter- between and daily differences (73); nuanced in calmness of and occur conceptualizations awkwardness, reliably , to as found such and states actions, , of findings , of 72); as observations (71, such Darwin’s , states (68–70); awe, as emotional excitement such and emotions positive , of love, studies recent summary 16); a ref. (for theorists see derived prominent “mushroom were of categories and taxonomies emotion emotion relief) from 34 (targeting The awe). call” cat- (targeting “close emotion cloud” 34 as targeting such phrases aggrega- egories, contextual content and with engines websites videos search tion The querying situations. by short emotional gathered of 2,185 were range studied: widest a ever the depicting clips stimuli gathered video evocative first emotionally We of stimuli. array visual dynamic experience by emotional elicited reported to of dataset space diverse semantic large, the a interrogate and methods inference statistical scale emo- reported of of dimensions typical the (67). experience been of tional have studies analytic than larger, factor experiences require past of may samples experience diverse emotional seman- more linear reported a of applying deriving space by accurately tic derived However, of be experiences. judgments self-report fol- emotional can to it space techniques space, reduction semantic semantic dimensionality a a of lin- that dimensions are lows the experiences of emotional this combinations reported Because ear all love. that and seman- entails other “awe” analysis to including weights perhaps suitable dimensions, applying tic by correspond obtained all could points Alterna- elation to and dimension. excitement, this joy, along terms the “elation” points tively, and various “excitement” at as the positioned such to also terms corresponding other directly with “joy,” dimension semantic term space semantic semantic the a a of example, have For combination could space. linear the define a that by dimensions seman- determined the space, within point tic single experience a emotional reported to terms reported and mathematically, term corresponds, comprise emotion every that words, individual other terms both In experiences. of for collections account the com- that and be ways can experience in reported bined of varieties These experience. epeetdteeoinlyeoaievdo rmthis from videos evocative emotionally the presented We large- modern use to out set we theorizing, this by Guided PNAS | ulse nieSpebr5 2017 5, September online Published al S1 Table o ito ttsand states of list a for | E7901

PSYCHOLOGICAL AND PNAS PLUS COGNITIVE SCIENCES emotional experience (24–36, 76, 77). All told, these procedures to produce overly conservative estimates of dimensionality; see yielded a total of 324,066 individual judgments (27,660 multiple- Fig. S2.) choice categorical judgments, 19,710 free-response judgments, To address concerns that forced choice methods may inflate and 276,696 nine-point dimensional judgments; see Materials and the apparent specificity of emotion self-reports (75), we also Methods and Tables S1 and S2 for more information on the rat- assessed how many dimensions of variance were reliably shared ings gathered). between the emotion category ratings and the free response labels participants used in reporting on their experience in Results viewing the videos (see Fig. S1C for representation of fre- Emotion Elicitation. Critical to our conceptual endeavor is a pre- quency of use of free response terms). In other words, we liminary question: Did distinct videos elicit reports of distinct determined how many distinct varieties of emotion captured emotional experiences? To test whether the videos reliably by the categorical ratings (e.g., fear vs. horror) were also reli- elicited reports of distinct emotional experiences, we assessed ably associated with distinct terms in the free response task how many videos elicited significant concordance in judgment (e.g., “suspense” vs. “shock”). We did so using CCA, which rates of each of the 34 emotion categories. By concordance, we finds linear combinations within each of two sets of variables mean multiple raters judging a given video as eliciting the same that maximally correlate with each other. In this analysis, we category of emotion among the 34 choices. We found that 75% found 27 significant linearly independent patterns of shared of the videos elicited significant concordance for at least one variance between the categorical and free response reports of category of emotion across raters [false discovery rate (FDR) emotional experience (P < 0.01), meaning people’s multiple- <0.05], with concordance averaging 54% (chance level being choice and free-response interpretations identified 27 of the 27%, obtained from simulated raters choosing randomly with same distinct varieties of emotional experience. The near con- the same base rates of category judgment observed in our data). vergence in the number of significant linearly independent pat- Importantly, all 34 emotion categories were found to be reported terns across two methods and datasets—SH-CCA within the cat- at significantly above-chance rates in response to at least one egorical judgment ratings and CCA between the categorical and video (Fig. S1A). These results show that all 34 categories of free response judgment ratings—serves as convergent validity for emotion are meaningful in that they are reliably reported as fit- up to 27 semantically distinct varieties of reported emotional ting descriptions for the experience of emotion. However, these experience. findings also leave open the possibility that some categories are How do the 27 distinct semantic dimensions we have doc- synonyms, or, more generally, that not all are linearly indepen- umented correspond to reported emotional experiences? To dent. The latter could also be the case, for example, if one cat- extract the meaning of the 27 dimensions within the category egory, such as joy, was equivalent to a conceptual grouping of judgments, we first used PCA to extract the 27 dimensions others, such as adoration and triumph. This concern over linear explaining the most variance, then applied factor rotation to dependence can be resolved by deriving principal components their loadings on the 34 categories, as shown in Fig. 1. Factor from the ratings—dimensions that are linearly uncorrelated but rotation yields a set of semantic dimensions that span the same have continuous loadings for each category (78). Determining space as the principal components but are more easily inter- how many of these distinct dimensions were reliably rated by dif- pretable in that they will each tend to load on a small number ferent observers would reveal the number of distinct emotional of categories. After factor rotation, many of the semantic dimen- experiences that can be reported using the 34 categories that sions have loadings on single categories, such as awe. In fact, guided this investigation. there were only seven emotion categories not mapped to distinct dimensions: pride and triumph, which coloaded on experiences Evidence for 27 Distinct Varieties of Reported Emotional Experience. of admiration; and , which coloaded on To examine how many semantically distinct categories struc- experiences of anger; sympathy, which coloaded on experiences tured participants’ reports of emotional experience, we devised of both empathic and ; and and envy, which had a method called split-half canonical correlations analysis (SH- only negligible loadings on any semantic dimensions. Essentially, CCA). SH-CCA is a generalization of split-half reliability anal- these findings show that approximately 27 categories of emo- ysis, in which the averages obtained from half of the ratings tion had distinct meaning in describing the reported experiences of each video clip for a single item (e.g., awe) are correlated elicited by the 2,185 videos, given that each semantic dimen- with the averages obtained from the other half of the ratings, sion loaded maximally on a distinct category. Where different across stimuli. In SH-CCA, the averages obtained from half categories coload on the same semantic dimension they were of the ratings of all items simultaneously are compared using used in an approximately linearly dependent manner, perhaps CCA to the averages obtained from other half of the ratings, as synonyms; where categories do not have strong loadings on yielding an estimate of the number of independent, reliable any semantic dimensions (e.g., envy) they were used insufficiently dimensions of variance in category judgments across raters (see or not consistently enough to contribute much reliable variance. Supporting Information and Fig. S2 for details of the method However, those loading on separate semantic dimensions—27 and its validation in 2,312 simulated studies). In other words, in total—were reliably separable in meaning with respect to SH-CCA accounts for shared variance (correlations) between the emotional states elicited by the videos. (In Fig. S3 we also items, such as awe and “aesthetic appreciation,” without dis- repeat this analysis with 24, 25, and 26 dimensions to understand carding the reliable variance in ratings of each individual item, how the dimensions may have differed under stricter criteria for such as the extent to which awe may differentially be evoked significance.) The 27 dimensions we derive from emotion self- by some stimuli (e.g., explosions) while aesthetic appreciation report in response to short videos demonstrate a semantic space may differentially be evoked by others (e.g., pastoral scenes of of emotions far richer in distinct varieties of reported experience nature). Using SH-CCA we found that between 24 (P < 0.05) than anticipated by emotion theories to date (for a summary see and 26 (P < 0.1) statistically significant semantic dimensions ref. 16). Not only do we find evidence for traditionally understud- of reported emotional experience (i.e., 24–26 linear combina- ied varieties of positive emotion, such as excitement (68–70), but tions of the categories) were required to explain the reliabil- also for differences between nuanced states relevant to more spe- ity of participants’ reports of emotional experience in response cific theoretical claims, such as the distinctions between roman- to the 2,185 videos. So far, this would suggest that the cate- tic love and sexual desire (79), interest and (80), hor- gorical ratings capture at least 24–26 semantically distinct vari- ror and fear, and aesthetic appreciation or beauty and feelings eties of reported emotional experience. (In fact, SH-CCA tends of awe (81).

E7902 | www.pnas.org/cgi/doi/10.1073/pnas.1702247114 Cowen and Keltner Downloaded by guest on October 2, 2021 Downloaded by guest on October 2, 2021 n na2 pc nwihec ie ssrone yother by surrounded is Keltner video and Cowen each which in space 2D a result- in experience, emotional ing of varieties distinct semantically 27 by separated distances. are accu- approximate, points more are data longer, points distinct data more between while distances preserved rately local nonlinear the two that uncovered—onto such axes, have reported we of experience dimensions emotional 27 data—the high-dimensional dimen- projects anal- semantic 2A previous Fig. 27 the In in the ysis. documented along experience emotional distributed of to are sions responses visu- videos emotional visual- data 2,185 how to modern the interrogate use difficult can to is we techniques it cloud, alization While dimensions point (82). 27-dimensional affective a arousal along ize and distributed valence as emotional evenly such that number more suggests a are possibility approximate states Another to clusters. distribution theorists distinct the emotion of of Discrete shape together? the blended predict experience or emotional endur- of combined an varieties to be distinct germane can is How inquiry question: of ing along line states a emotional of Such examin- distribution dimensions. reported the experi- these also requires emotional of report?—but of people do dimensions varieties emotion ence distinct semantic what of the is, experience—that space only not semantic Meaning. ing of the Gradients Continuous standing Cat- by “Discrete” Bridged Experience: Are Emotional egories Reported of Distribution The aesthetic as nuanced awe). theories and as beauty emotion states of most feelings emotional (i.e., in emo- differentiating appreciation proposed of reliably been complexity date, has a to what demonstrate beyond videos emotion structure emo- short from tion the to derive we to response dimensions respect in The videos. self-report with the meaning by elicited in states tional separable reliably separate were on used loading frequently dimensions Categories more reduction. other dimensionality with during dependent categories or linearly reliably roughly judged as not taken either disappointment, were were triumph) Categories (contempt, and sympathy, categories. dimensions relief, any few guilt, envy, on relatively loadings rotation, on maximal varimax load without that using dimensions dimensions have finds interpretable (PCA), which more analysis into components rotated principal categor- experi- been using the emotional within extracted variance corre- reported judgments, of components each of ical principal variance, 27 variety first of the distinct Here, dimensions semantically ence. separable a 27 to to sponding judgments up categorical captured that revealed reliably analyses Statistical responses. egorical 1. Fig. nFg 2A Fig. In atraayi odnso 7dmnin fvrac ihntecat- the within variance of dimensions 27 on loadings analysis Factor eapytSEt a h ,8 iesaogall along videos 2,185 the map to t-SNE apply we eueamto aldtSE(3.Ti method This (83). t-SNE called method a use we Under- rvd ul neatv eso fFg 2A we Fig. implications, of version conceptual interactive To their fully awe. a and to provide appreciation findings experiences, aesthetic these to reported calmness illustrate of from emotion that chains of as categories, such related categories distinct other of conceptually two majority or forming one the with dimen- that boundaries semantic fuzzy infer emotional share 27 can along of that videos we distribution view of sions, the thousands opposite by on the elicited Based states support (82). indepen- to dimensions entirely by fail dent defined but are nature experiences “cut (20), emotional reported categories joints” emotion find- its dis- that These and at videos. same awe with the as converge of any such ings Further by space, elicited times). seldom t-SNE were eight were the gust, (just fear horror within videos locations and were distant same anxiety as 2B yet the total), Fig. of times), of in inspection elicited 55 few times were R; by (75 Q) and elicited videos and (Q (F same horror fear the and and of anxiety many 2B, categories. distinct by Fig. of in pairs shown particular to As specific highly are aries discrete. distinct than many rather between fuzzy 2A. boundaries are categories Fig. videos, the emotion by that of revealed reveal set distribution results overlapping the These an of by shape the elicited corroborating were fear, as such and oth- gradients, these anxiety among to along awkwardness, dimensions appreciation to semantic aesthetic Adjacent amusement ers. to to to calmness adoration anxiety and disgust, from awe, to the gradients horror example, smooth to For along fear distributed experience. are emotional videos particular reported between of transitions dis- varieties smooth semantically emotional forming other of categories, with categories tinct love, gradients many romantic smooth time, share and same experience desire the sexual At exam- (desire), . for and craving experience, of 2A emotional those Fig. of of ple, clusters Inspection certain nei- uniform. reveal is simply does nor that clustered experiences simply emotional com- ther reported a of reveals analysis distribution This color. plex in transitions between smooth corre- to gradients experience spond reported smooth of varieties Thus, distinct on semantically positively. dimensions these of semantic loaded interpolation the they weighted of a each which to using correspond- corresponding colored colors letters are the video the each Specifically, to elicited. ing it emotional colored that reported is of video experience varieties each specific this which the in within to map, according videos chromatic uniquely the a by use elicited we To space emotion 2D of experiences. varieties emotional 27 reported the plot similar evoked that videos ttswr ucino h fetv ieso judgments, could dimension judgments affective dimension emotional affective the of the of judgments whether examined function categorical we a the were whether states test (25– To emotional dimensions measuring affective 35). studies of 77, many combination 24, to some using led (18, experience has par- dimensions claim by This affective formed 84). these space 82, a of within combinations position ticular by emotion of largely representations explained categorical are theo- that Some suggested valence. have and rists conditions), of safety, (improvement obstruction, commitment, judg- upswing identity, dimension fairness, certainty, effort, affective dominance, attention, 14 control, the arousal, by carried ments—approach, information experience the emotional Emotional to reported Dimensions. of Reported structure categorical Affective the in pared Proposed Variance than More Experience Explain Labels Categorical to opposed as gradients clusters. continuous the discrete occupy expe- with emotional states over reported most within of hovered riences, that space is qualitatively semantic confirms 27-dimensional map map the the this of in Inspection position cursor. its when dis- is played video each which in 1.amazonaws.com/emogifs/map.html) nmr n-rie nlss efidta hs uz bound- fuzzy these that find we analyses, fine-grained more In PNAS eel hteoinctgre apdto mapped categories emotion that reveals | ulse nieSpebr5 2017 5, September online Published ( https://s3-us-west- enx com- next We | E7903

PSYCHOLOGICAL AND PNAS PLUS COGNITIVE SCIENCES Fig. 2. The structure of reported emotional experience: Smooth gradients among 27 semantically distinct categorical judgment dimensions. (A) A chromatic map of average emotional responses to 2,185 videos within a 27-dimensional categorical space of reported emotional experience. t-distributed stochastic neighbor embedding (t-SNE), a data visualization method that accurately preserves local distances between data points while separating more distinct data points by longer, more approximate, distances, was applied to the loadings of the 2,185 videos on the 27 categorical judgment dimensions, generating load- ings of each video on two axes. The individual videos are plotted along these axes as letters that correspond to their highest loading categorical judgment dimension (with ties broken alphabetically) and are colored using a weighted interpolation of the unique colors corresponding to each of the categorical judgment dimensions on which they loaded positively. The resulting map reveals gradients among distinct varieties of reported emotional experiences, such as the gradients from anxiety to fear to horror to disgust (also see the interactive map at https://s3-us-west-1.amazonaws.com/emogifs/map.html). (B) Number of significant coloadings of each video on each categorical judgment dimension. The significance of individual loadings of each video on each categorical judgment dimension was determined via simulation of a null distribution (Supporting Information). We then counted the number of instances in which videos loaded significantly (FDR <0.05) on pairs of two categorical judgment dimensions. These results validate the emotion gradients observed in A. For example, anxiety and fear (F and Q) were elicited by many of the same videos (75 times in total), as were fear and horror (Q and R; 55 times), yet anxiety and horror were seldom elicited by the same videos (just eight times). (C) Top free response terms associated with each categorical judgment dimension. The free response judgments were regressed onto the categorical judgment dimensions, across videos. For 22/27 dimensions, the highest loading category is among the three (out of 600) top-weighted free response terms, strongly validating the categorical ratings as measures of subjective experience.

E7904 | www.pnas.org/cgi/doi/10.1073/pnas.1702247114 Cowen and Keltner Downloaded by guest on October 2, 2021 Downloaded by guest on October 2, 2021 oe n Keltner and Cowen h n htrsle ntegets vrg xlie ainefreach for variance explained average i.e., k, greatest optimal the also the See choosing in from prediction. resulted results the that show one and we 50 the neighbors, to k-nearest For 1 video. from per k ratings tested nine used we variance. analysis, explainable this estimated For the leave-one- by experiences using divided calculated emotional then was and reported variance out-cross-validation (Explained explicating cate- videos. in short the by value that elicited most suggests the This have (Middle). gories neighbors k-nearest (Left squares with least regression ordinary judg- with dimension regression linear affective both using the (P versa in vice variance than ments more significantly versa. explain vice sions and judgments, dimension 3. Fig. (P reverse the than better substantially judg- ments dimension affective the explained dimensions judgment affective ical categor- the the That in Information). judgment variance (Supporting explainable judgments categorical dimension the of the 78% explained in tently consis- dimensions variance judgment categorical the explainable we whereas most dimensions, 3) the at explained (Fig. of judgments regression dimension 61% nonlinear affective the and tech- that linear found regression both cross-validated com- With using We niques. person’s possibilities dimensions? the these affective explain well-validated or pared better fear, these fear of of as reports reports experience the person’s an a of explain submis- labeling better judgments arousal, does on, displeasure, dimension so elicits and affective video siveness, each do which or to is, dimensions, degree That judgment categorical versa. the vice of each explain better itntrpre mtoa xeine eetdsot or of smooth light reflected In between experiences 39). gradients emotional whether 38, reported tested (22, we distinct emotions claims, pat- the other conceptual from these of of fashion, instances discrete responses in all different, terned across but similar category behav- are given a that subjective, physiological) associated (e.g., and are responses ioral, emotions patterned basic emotion prototypical that basic with predict support which changes? (85), would dimensions theories emotion affective categories associated of neighboring in where jumps between category sharp points of reported at presence dimensions The frequently affective most in the jumps the do sharp or instance, gradients, there For category as are across judgments? smoothly vary categorical vary dimensions dimensions the affective affective of the function do by then, a explained How, largely categories. were dimensions the affective in the Differences of judgments Smooth to Experience. Emotional Correspond Reported Gradients Category Smooth videos. evocative to response in peo- experiencing emotions affective report the reliably 14 to ple respect the with what more, of and most capture, elicited dimensions capture experience they emotional videos—that short of dimensions by reports affective people’s the explicating cat- than the value in that semantic suggests more regression) have nonlinear egories and linear both test, aineepandb h aeoia uget nteaffective the in judgments categorical the by explained Variance uprigInformation. Supporting < 10 − 6 otta et.Teefidnshl when hold findings These test). bootstrap , h bv nlssidct that indicate analyses above The h aeoia ugetdimen- judgment categorical The < 10 − n nonlinear and ) 6 bootstrap , i pc frpre mtoa xeine o aydistinct many seman- How the experience: about emotional claims reported are of emotion space of tic science the to Central in Discussion increase relative a with individual. associated an gradient to is the commitment sadness example, gradi- For to the 2A. disgust and of expe- from 1 emotional some Figs. reported in interpret of represented varieties further riences 27 to the along us observed allow ents maps (C dimensions The affective E). of combinations linear particular to (B judgments category of binations t.Hne h xeto iiaiyi oo between color in between similarity as (red, of well channel extent the color 4 Hence, Each Fig. ate. in map. blue) t-SNE and the green, onto colors as jected (B egorical In map. t-SNE our 4 onto Fig. loadings dimension affective and gorical attach- to related closely are (90). experiences that processes emotional adoration ment of and commit- reports sadness of as in Judgments such variation 89). for 88, account (76, anger ment of reports in 6 in and variation 4 variates of dominance, judgments and example, approach likewise For experience. dimensions approach emotional affective reported and drive dominance social of More feelings states as negative pain). such most empathic the and from highly joy) (horror of and experiences emo- (calmness differentiates reported states positive of and studies (18), corre- analytic experience variate factor dimension tional canonical all initial nearly same first in the the valence, uncovered instance, to exclusively analy- For almost similar 24). sponds (for ref. experience see emotional sis of categories to relate iesos hsaayi ile 3sgicn aoia variates canonical (P significant 13 yielded analysis This judgment dimensions. categorical of extracted combinations linear CCA corre- with that maximally Here, lated judgments 4). dimension affective (Fig. of video combinations dimen- linear between each judgment CCA with categorical applying associated 27 by and sions so dimensions affective did 14 We the 84). of (24, dimensions emotion judgment categorical elicited between with relationship covary the dimensions affec- tive examine how ascertained To we categories, 87). and dimensions 86, experi- affective the current 24, coalescence of 22, the labels (18, reflect categorical ences and that Experi- dimensions factors affective Emotional by of Reported defined are of experiences Meaning the ences. of Factors Unifying shift the not category. along emotion do smoothly each with vary ratings associated they gradients these rather, pat- ratings, categories; specific across dimension a abruptly with affective associated of is given tern category a of each instances While all prototypi- across category. that similar are theories responses emotion patterned basic cal from notion into the calls categories question and between dimen- gradients affective arousal across the smoothly That as vary S4). sions (Fig. such individual dimensions an to affective commitment smooth in the a variations reflected that categories to smooth found particular emotions we between different analyses, documented two boundaries these of In each extent. elicited significant that dimensions videos we affective the so, across to do ratings To categorical the dimensions. from affective regressed associated in changes abrupt ohactgrcladafciedmninloading. dimension with affective com- experiences, and central emotional categorical or reported a factor, both of unifying meaning a the as of canonical of ponent, Each thought S5 ). be might (Fig. variate experiences emotional of reported judgments categorical their and dimensions affective of self-reports ovsaieec aoia ait,w rjc ohiscate- its both project we variate, canonical each visualize To ls npcinof inspection Close < .1,sae iesoso ainebtenparticipants’ between variance of dimensions shared 0.01), odnso h rtsxcnnclvrae—ohcat- variates—both canonical six first the of loadings B–E, eta li coseointere sta emotional that is theories emotion across claim central A and n fetv ieso (C dimension affective and D) D and PNAS E i.S5 Fig. niaeteetn owihlna com- linear which to extent the indicate | B–E ulse nieSpebr5 2017 5, September online Published eel o fetv dimensions affective how reveals orsod oacnnclvari- canonical a to corresponds and r iia nmeaning in similar are D) con for account S5, Fig. and —r pro- E)—are B and | E7905 C and as

PSYCHOLOGICAL AND PNAS PLUS COGNITIVE SCIENCES Fig. 4. Canonical correlations analysis between the categorical judgment dimensions and the affective dimensions. (A) The first 13 canonical correlations between the categorical judgment dimensions and the affective dimensions were found to be significant (P < 0.01). We assigned labels to each canonical variate by interpreting its coefficients on the affective dimensions (see Fig. S5 for the coefficients). (B and C). Categorical variates (B) and dimensional variates (C) for the first three canonical correlations, projected as red, green, and blue color channels onto the t-SNE map from Fig. 2A. Color legends are given in the titles for each map. Similarity in colors between E and F illustrates the degree of shared information between the categorical and dimensional judgments for these three dimensions of emotional experience. Labels on each map reflect the combination of loadings on the three dimensions that give rise to each color. (D and E) Like B and C, but for the fourth through sixth canonical correlations. Altogether, B–E illustrate that the categorical gradients correspond to smooth differences in affective dimension (see also Fig. S4 for analysis of gradient smoothness).

varieties of emotion do people report experiencing? What are variety of states than considered earlier in the field (37). By no the boundaries between distinct varieties of reported emotional means do we mean to claim that this is the definitive taxonomy experience? To what extent is reported emotional experience of emotional states, for which studies of other types of stimuli, rooted in the categorical labeling of the state, vs. judgments of other approaches to self-report, other modalities of emotional proposed dimensions of affect such as valence and arousal? What response, and other cultures will need to be incorporated. Nev- are the unifying factors that relate proposed affective dimensions ertheless, the present investigation reveals the rich varieties of to emotion categorization? reliably reported emotional experience that may shape human The findings from the present investigation emerge from behavior. a mathematical framework positing that reports of emotional Our next finding concerns how reported emotional experi- experience can be characterized as points within a semantic ences are distributed in relation to one another, another matter space, distributed along semantic dimensions corresponding to central to theoretical debate regarding the structure of emo- distinct varieties of reported experience. To interrogate the tion. Past theorists have suggested that the distribution of emo- semantic space of reported emotional experience, we had partic- tional states is shaped in one of two ways: either that emotional ipants report on their emotional responses, using three different states occupy a limited number of distinct clusters or emotion methods, to 2,185 emotionally evocative short videos, heedful of families (16, 37–39) or that they are more evenly distributed the ambiguities and indeterminacies of self-report. across more independent dimensions (82). Our approach, which Our first finding concerns the richness of the semantic space interrogates the distribution of elicited emotional states within of reported emotional experience. Using statistical methods to a dimensional space using an open-ended statistical framework, determine the dimensionality of reported categories of experi- can identify both discrete clusters and continuous gradients. ence, we obtain evidence for up to 27 varieties of experience While our findings suggest that there may be constraints on from the categories of emotion reliably reported in response to which varieties of emotional experiences can be reported simul- over 2,000 emotionally evocative short videos. Our finding that taneously in response to a single stimulus, most categories of 27 distinct varieties of reported experience are reliably associ- emotion share continuous gradients with at least one other cat- ated with distinct situations converges with recent developments egory. These correspond to smooth gradients in affective mean- in the emotion signaling literature suggesting that upwards of ing, as one can see in Fig. 2A, where we observe gradients linking 20 states have distinct nonverbal signals (56). The space of dis- experiences of admiration, awe, and aesthetic appreciation; anx- tinct reported emotional experiences in English involves a richer iety, fear, horror, and disgust; and a number of other emotion

E7906 | www.pnas.org/cgi/doi/10.1073/pnas.1702247114 Cowen and Keltner Downloaded by guest on October 2, 2021 Downloaded by guest on October 2, 2021 oe n Keltner and Cowen ftebanrsosbefrhge-re onto 1) consis- (15), cognition higher-order parts for in conscious- represented responsible are brain emotional states the emotional of of that theory predicts also higher-order evoca- ness The the to stimuli. reactions emotional partici- their tive label structures to symbolic upon relied the pants in emo- reliabil- commonalities reported videos— the from their short emerges in theorizing, situations—viewing particular participants this to across responses within representa- tional observed Cast we symbolic that it states. are ity these terms (15), of emotion signif- Brown tions states and psychologically introspective and of situations, are representations icant LeDoux schematic experiences by by emotional conscious- defined that forward emotional posited For of put is generated. theoretical theory recently are higher-order recent experiences ness the such inform in how example, also explicate to may efforts experiences emotional of patterns distinct by represented activity. are distinct neural that of emotion number of the determine varieties estimation to here reliability introduced multidimensional techniques and the elicitation with may combined emotion be fruitfully and could New approaches awe) categories. modeling discrete fMRI and than rather appreciation gradients doc- continuous have aesthetic reflect we (e.g., states positive here many the umented been might as have such than activation far, states so of neural investigated array of broader much patterns a among distinct distinguish intrigu- the that these raise possibility of investigation present light ing the In of music). results and the (film findings, by stimuli elicited of states modalities emotional on separate tested above- when and even with trained activation, were brain discriminated classifiers of patterns be , on could based accuracy anger, surprise chance and amusement, emotions. sadness, distributed discrete of fear, that among experiences documenting distinguish brain, activity Specifically, brain human of the concomi- patterns of from experiences fMRI emotional tant decoded (95) LaBar Kragel notably, and Most experience. emotional of representation ral than rather fuzzy are the nature. that in categories discrete find emotion we theories, between (16, cate- discrete theories boundaries to emotion of contrary discrete terms although, by emotional forward 37–39), in put reported conceptualized often that more precisely gories suggest more findings is present experience The (24–36, of dimensions 77). measurements affective 76, specific on other focused and have arousal, phys- valence, effects cognitive, neural behavioral, and emotion-related hun- iological, descrip- of Nevertheless, studies (94). exhaustive of experience an dreds emotional seldom offer reported theories of dimensions appraisal tion specific and that that constructivist a recognize propose current to in important most dimensions is It the versa affective fashion. vice organize powerful and than labels coherent judgments Categorical dimension 3). affective (Fig. the experiences in and their appraisal explained variance placing powerfully affective more by judgments of categorical or scales the motivation, categories dimensional 34 different of 14 choos- along list by state a emotional from their ing judge to When asked terms. were semantic participants in experiences emotional experiences their ceptualize emotional mixed for admira- and from 91), (e.g., ref. 93). people next (92, see how the awe; for to to experience account tion one may of from document gradients transition war- smooth we questions the meaning (16, example, intriguing affective For theories raise research. also dimensional further findings and ranting These discrete dis- 82). in uniform 37–39, more at or clustered hinted distribu- the complex tributions than more states emotional far of a tion suggest findings These categories. h rsn nig eadn h tutr fself-reported of structure the regarding findings present The neu- the to related findings recent with dovetail findings Our con- people how of question the to speak findings our Finally, utrsadterlnugs ie hthr ehv nystudied only have geo- we here ascertain that other given to languages, within their critical experience and cultures emotional be reported also of will structures activities, metric It daily music, interactions. as social such and elicitors, emotion other of studies emo- reported of experience. structure uncov- tional geometric have a we of date, approximation to studying an elicitors by ered and and emotions cate- of methods, array analytic emotion widest With the distributed particular space. this are between within and gories gradients situations continuous distinct dimensions debate. along reliably 27 intense least with such at sparked associated inhabit has experiences that emotional theorizing Reported is the than analytic answers in complex geometric typical nuanced, our more and yielded self-reports space have emotional semantic techniques a how within in of to situated claims conceptualization are theoretical Answers Our central emotion. most field. of the the upon science bear the questions to such foundational are ences dis- in of represented patterns. be varieties activity potentially brain other could dis- tinct many in that occur and the reliably situations 29, find out tinct we 28, leaving that 26, experience 47–49), (25, emotional emotions 44, reported basic 42, six 35, the focused or 34, have arousal and experiences valence emotional on neuro- regions reported brain to the during of respect studies activated physi- With of hundreds behavior, (14). example, for to differences physiology, experience individual reported and relate implications ology, that generative have studies findings for These with association situations. their distinct in structured reliably are concepts emotion semantic of of experience. structure sources emotional the of culture-related shape spaces such self-report for emotion how crucial in be examine variation will to it studies Nevertheless, future S6). variance (Fig. the commonalities participants of with proportion across compared small personal- experience emotional a and reported most class, in at social explained age, suggest factors performed gender, ity we as analysis such additional differences an that of How- 99). results (98, the situation given ever, a their to each in responses fashion, emotional in striking of experiencing reports in often report differ, also people individuals emotions situation; the in monalities emotional of space extent semantic experience. the the research to characterize further processes, Future to such 97). possible, examine cul- systematically (18, to to experience situ- need according emotional will cause a of should that norms or of experience tural cause perception emotional potentially (iii) the could or is, ation (18); that labeled how quality, color is affective nevertheless experience own may emotional their but in an emotional, not considered may be that right experiences con- perceptual three (ii) and itself; of cognitive experience combination emotional (i) a phenomena: distinct reflect ceptually reported may that note experience we emotional point, latter besides this self-report On of experience. determinants emotional potential other self-report, are elude ear- there may and noted experience emotional As of aspects experiences. some lier, subjective correspondence con- and of degree the self-reports the experience, on between emotional depend draw of might studies we many clusions so in As gation. have we that emotion might of findings regions varieties brain distinct Our uncovered. of such cortex). high- dozens how the in orbitofrontal about encode questions found representa- (e.g., intriguing are that regions raise (96) experiences brain LaBar emotional and level subjective Kragel of by tions findings with tent twl eipratt xedteemtosadfidnsto findings and methods these extend to important be will It experi- emotional reported of structure the about Questions how reveal results our caveats, and limitations these Granting com- on focused have we that mention to important also is It investi- present the of limitations important noting worth is It PNAS | ulse nieSpebr5 2017 5, September online Published | E7907

PSYCHOLOGICAL AND PNAS PLUS COGNITIVE SCIENCES emotional experience reported by US participants using English The 2,185 videos and their mean ratings can be requested here: https:// emotion concepts (100). The methods developed here could be goo.gl/forms/XErJw9sBeyuOyp5Q2. Please exercise discretion in viewing the fruitfully applied to studies of emotion-related peripheral physio- videos, many of which contain highly graphic violence, nudity, and/or sexual logical response, central nervous system response, and nonverbal content. Videos with highly graphic content are blurred in the chromatic map expression, once again to shift toward an understanding of how linked elsewhere in the paper (https://s3-us-west-1.amazonaws.com/emogifs/ map.html). However, an uncensored chromatic map is also available to read- emotions are how emotional states are arranged within a geo- ers of age 18+ by replacing the word “map” with the word “uncensored” in metric space. the previous URL (although please exercise careful discretion in viewing the uncensored map, which, again, contains extremely graphic content). In both Materials and Methods maps, floating over the number corresponding to each video for an extended Emotion judgments of the videos were obtained using Amazon Mechani- period will reveal the video’s unique numeric tag, which, followed by “.mp4,” cal Turk. A total of 853 English-speaking US participants took part in the also serves as its filename within our database. Note that videos within the study (403 females, mean age = 36 y). The experimental procedures were map can be clicked and dragged. approved by the Institutional Review Board at the University of California, Berkeley. All participants gave their informed consent. See Supporting Infor- ACKNOWLEDGMENTS. This research was supported in part by a grant from mation for details. the John Templeton Foundation.

1. Barrett LF, Mesquita B, Ochsner KN, Gross JJ (2007) The experience of emotion. Ann 32. Quigley KS, Lindquist KA, Barrett LF (2014) Inducing and measuring emotion and Rev Psychol 58:373–403. affect: Tips, tricks, and secrets. Handbook of Research Methods in Social and Person- 2. Izard CE (2007) Basic emotions, natural kinds, emotion schemas, and a new paradigm. ality Psychology, ed Judd CJ (Cambridge Univ Press, New York), pp 220–252. Perspect Psychol Sci 2:260–280. 33. Schuller B, et al. (2011) Avec 2011–The first international audio/visual emotion chal- 3. Izard CE (2007) Emotion feelings stem from evolution and neurobiological develop- lenge. International Conference on and Intelligent Interaction ment, not from conceptual acts: Corrections for Barrett et al., 2007. Perspect Psychol (Springer, Berlin), pp 415–424. Sci 2:404–405. 34. Wager TD, Phan KL, Liberzon I, Taylor SF (2003) Valence, gender, and lateralization of 4. Goldie P (2004) Emotion, , and knowledge of the world. Thinking About Feel- functional brain anatomy in emotion: A meta-analysis of findings from neuroimag- ing: Contemporary Philosophers on Emotions, ed Solomon RC (Oxford Univ Press, ing. Neuroimage 19:513–531. Oxford). 35. Wager TD, et al. (2008) The Neuroimaging of Emotion, eds Lewis M, Haviland- 5. Frijda NH (2003) Emotions and hedonic experience. Well-Being: Foundations of Hedo- Jones JM, Barrett LF (Guilford, New York), Vol 3, pp 249–271. nic Psychology, eds Kahneman D, Schwarz N, Diener E (Russell Sage Foundation, 36. Olofsson JK, Nordin S, Sequeira H, Polich J (2008) Affective picture processing: An New York). integrative review of erp findings. Biol Psychol 77:247–265. 6. LeDoux J, Phelps L, Alberini C (2016) What we talk about when we talk about emo- 37. Ekman P (2016) What scientists who study emotion agree about. Persp Psychol Sci tions. Cell 167:1443–1445. 11:31–34. 7. Panksepp J (2005) Affective consciousness: Core emotional feelings in animals and 38. Ekman P, Cordaro D (2011) What is meant by calling emotions basic. Emot Rev 3:364– humans. Conscious Cogn 14:30–80. 370. 8. Tsuchiya N, Adolphs R (2007) Emotion and consciousness. Trends Cogn Sci 11:158– 39. Ekman P (1994) All emotions are basic. The Nature of Emotion, eds Ekman P, 167. Davidson R (Oxford Univ Press, New York), pp 15–19. 9. Robinson MD, Clore GL (2002) Belief and feeling: Evidence for an accessibility model 40. Angie AD, Connelly S, Waples EP, Kligyte V (2011) The influence of discrete emotions of emotional self-report. Psychol Bull 128:934–960. on judgement and decision-making: A meta-analytic review. Cogn Emot 25:1393– 10. Kovecses¨ Z (2003) Metaphor and Emotion (Cambridge Univ Press, Cambridge, UK). 1422. 11. Russell JA (1991) Culture and the categorization of emotions. Psychol Bull 110:426– 41. Chervonsky E, Hunt C (2017) Suppression and expression of emotion in social and 450. interpersonal outcomes: A meta-analysis. Emotion 17:669–683. 12. Sabini J, Silver M (2005) Why emotion names and experiences don’t neatly pair. Psy- 42. Costafreda SG, Brammer MJ, David AS, Fu CH (2008) Predictors of amygdala activa- chol Inq 16:1–10. tion during the processing of emotional stimuli: A meta-analysis of 385 pet and fmri 13. Shaver P, Schwartz J, Kirson D, O’connor C (1987) Emotion knowledge: Further explo- studies. Brain Res Rev 58:57–70. ration of a prototype approach. J Pers Soc Psychol 52:1061–1086. 43. Dawel A, O’Kearney R, McKone E, Palermo R (2012) Not just fear and sadness: 14. Clore GL, Ortony A (2013) Psychological construction in the occ model of emotion. Meta-analytic evidence of pervasive deficits for facial and vocal Emot Rev 5:335–343. expressions in psychopathy. Neurosci Biobehav Rev 36:2288–2304. 15. LeDoux JE, Brown R (2017) A higher-order theory of emotional consciousness. Proc 44. Kirby LA, Robinson JL (2015) Affective mapping: An activation likelihood estimation Natl Acad Sci USA 114:E2016–E2025. (ALE) meta-analysis. Brain Cogn, 10.1016/j.bandc.2015.04.006. 16. Keltner D, Lerner JS (2010) Emotion. Handbook of Social Psychology, eds Fiske ST, 45. Kreibig SD (2010) Autonomic nervous system activity in emotion: A review. Biol Psy- Gilbert DT, Lindzey G (Wiley, New York). chol 84:394–421. 17. Rosenberg EL (1998) Levels of analysis and the organization of affect. Rev Gen Psychol 46. Lench HC, Flores SA, Bench SW (2011) Discrete emotions predict changes in cogni- 2:247–270. tion, judgment, experience, behavior, and physiology: A meta-analysis of experimen- 18. Russell JA (2003) Core affect and the psychological construction of emotion. Psychol tal emotion elicitations. Psychol Bull 137:834–855. Rev 110:145–172. 47. Murphy FC, Nimmo-Smith I, Lawrence AD (2003) Functional neuroanatomy of emo- 19. Scherer KR (2009) The dynamic architecture of emotion: Evidence for the component tions: A meta-analysis. Cogn Affect Behav Neurosci 3:207–233. process model. Cogn Emot 23:1307–1351. 48. Phan KL, Wager T, Taylor SF, Liberzon I (2002) Functional neuroanatomy of emotion: 20. Barrett LF (2006) Are emotions natural kinds? Persp Psychol Sci 1:28–58. A meta-analysis of emotion activation studies in pet and fmri. Neuroimage 16:331– 21. Goldie P (2009) Getting feelings into emotional experience in the right way. Emot Rev 348. 1:232–239. 49. Vytal K, Hamann S (2010) Neuroimaging support for discrete neural correlates of basic 22. Lazarus RS (1991) Progress on a cognitive-motivational-relational theory of emotion. emotions: A voxel-based meta-analysis. J Cogn Neurosci 22:2864–2885. Am Psychol 46:819–834. 50. Panksepp J (2007) Neurologizing the psychology of affects: How appraisal-based con- 23. Roseman IJ (1991) Appraisal determinants of discrete emotions. Cogn Emot 5:161– structivism and basic emotion theory can coexist. Persp Psychol Sci 2:281–296. 200. 51. Diener E, et al. (2010) New well-being measures: Short scales to assess flourishing and 24. Smith CA, Ellsworth PC (1985) Patterns of cognitive appraisal in emotion. J Pers Soc positive and negative feelings. Soc Indic Res 97:143–156. Psychol 48:813–838. 52. Harmon-Jones C, Bastian B, Harmon-Jones E (2016) The discrete emotions question- 25. Brooks JA, et al. (2016) The role of language in the experience and perception of naire: A new tool for measuring state self-reported emotions. PLoS One 11:83–111. emotion: A neuroimaging meta-analysis. Soc Cogn Affect Neurosci 12:169–183. 53. Watson D, Clark LA, Tellegen A (1988) Development and validation of brief measures 26. Colibazzi T, et al. (2010) Neural systems subserving valence and arousal during the of positive and negative affect: The panas scales. J Pers Soc Psychol 54:1063–1070. experience of induced emotions. Emotion 10:377–389. 54. Mikels JA, et al. (2005) Emotional category data on images from the international 27. Eerola T, Vuoskoski JK (2013) A review of studies: Approaches, affective picture system. Behav Res Methods 37:626–630. emotion models, and stimuli. Music Percept 30:307–340. 55. Gross JJ, Levenson RW (1993) Emotional suppression: Physiology, self-report, and 28. Hamann S (2012) Mapping discrete and dimensional emotions onto the brain: Con- expressive behavior. J Pers Soc Psychol 64:970–986. troversies and consensus. Trends Cogn Sci 16:458–466. 56. Keltner D, Cordaro DT (2015) Understanding multimodal emotional expressions: 29. Lindquist KA, Satpute AB, Wager TD, Weber J, Barrett LF (2016) The brain basis of pos- Recent advances in basic emotion theory. Emotion Researcher, ed Scarantino A. Avail- itive and negative affect: Evidence from a meta-analysis of the human neuroimaging able at socrates.berkeley.edu/∼keltner/publications/keltner&Cordaro%202016.pdf. literature. Cereb Cortex 26:1910–1922. 57. Russell J (1980) A circumplex of affect. J Pers Soc Psychol 36:1152–1168. 30. Mauss IB, Robinson MD (2009) Measures of emotion: A review. Cogn Emot 23:209– 58. Ferguson E, Cox T (1993) Exploratory factor analysis: A user’s guide. Int J Select Assess 237. 1:84–94. 31. Moors A, et al. (2013) Norms of valence, arousal, dominance, and age of acquisition 59. Barrett LF, Russell JA (2014) The Psychological Construction of Emotion (Guilford, for 4,300 Dutch words. Behav Res Methods 45:169–177. New York).

E7908 | www.pnas.org/cgi/doi/10.1073/pnas.1702247114 Cowen and Keltner Downloaded by guest on October 2, 2021 Downloaded by guest on October 2, 2021 9 aisnR 19)Prigafciesae esetvsfo erpyhlg and neuropsychology from Perspectives space: affective Parsing (1993) and RJ Evidence Davidson affect: approach-related 89. an is Anger (2009) E risk. Harmon-Jones and CS, anger, Fear, Carver (2001) D 88. Keltner JS, Lerner 87. (1987) NH Frijda 86. facial via communication for space semantic the of Dimensionality (1966) CE Osgood 77. (1974) J A Russell expressions? A, facial Mehrabian from emotion 76. of recognition universal there Is (1994) JA Russell . anxiety,and 75. fear, of model Expectancy (1991) S Reiss 74. con- to corresponding expressions facial of frequency High (2003) AB Cohen P, Rozin 73. (2009) D Keltner 72. (1872/1998) emotions? C positive are Darwin emotions. good positive What 71. (1998) discrete BL of Fredrickson science Auto- a 70. Toward good: : Feeling Beyond (2011) al. EF et MN, Perea Shiota SE, Moser 69. analysis. WH, Yeung factor SL, in size Neufeld Sample MN, (1999) Shiota S 68. Hong S, Zhang KF, Widaman RC, MacCallum 67. 4 art F(03 scooia osrcin h awna praht h science the to approach Darwinian The construction: same Psychological the (2013) at LF sad Barrett and happy feel 94. people Can (2001) JT Cacioppo AP, McGraw JT, com- Larsen meta-analysis A emotions: 93. mixed Eliciting (2015) S Kellett P, Totterdell emotion. R, Berrios of theories 92. appraisal cognitive develop- of implications Theoretical Some (1991) attachment: PC romantic Ellsworth Adult 91. (2000) PR Shaver RC, Fraley 90. expressions. facial of perception Categorical (1992) Hampshire. JJ New Magee of NL, rocks Etcoff the 85. and emotions Basic (2014) PC Ellsworth 84. dis- model biobehavioral a orient? orientation sexual does What compo- (2003) principal LM into Diamond variables statistical 79. of complex a of Analysis (1933) H Hotelling 78. complexity. Emotional (2008) emotion. LF of Barrett cause KA, candidate a Lindquist as 66. appraisal constructive Automatic (2010) A Moors 65. 3 atnLD itnG(08 iulzn aauigt-sne. using data Visualizing (2008) G Hinton LVD, Maaten 83. integrative An affect: of model circumplex The (2005) BS Peterson JA, emotion. Russell aesthetic J, and Posner spiritual moral, 82. a awe, (2003)Approaching J Haidt interest. D, of Keltner face The 81. (1993) J Reeve 80. translation. of indeterminacy for reasons the On (1970) WV Quine that expressions 64. Facial tie (2008) M The Frank M, (2005) O’Sullivan MN, JJ Shiota D, Gross Keltner D, FH, Matsumoto Wilhelm 63. L, McCarter RW, Levenson IB, sub- Mauss and electromyography 62. facial between Relationships (1980) G Schwartz S, experience. Brown emotional of 61. signs Facial (1980) S Ancoli WV, Freisen P, Ekman 60. oe n Keltner and Cowen psychophysiology. implications. 44:227–240. expressions. MA). Cambridge, studies. cross-cultural the of review 153. expres- facial occurring naturally of analysis Americans. an of in sions and concentration, fusion, York). New Press, Univ Psychol Am emotions. positive five in responding system nervous nomic Meth 513–530. Psychol pp York), New (Guilford, LF Barrett JM, Haviland-Jones M, Lewis femotion. of time? measures. and types, models, paring Vol York), New (Wiley, KT 143–161. Strongman pp ed 1, Emotion, on Studies of Review International questions. unanswered 154. and controversies, emerging ments, 21–26. 2605. desire. sexual and love romantic tinguishing nents. Rev Emot 183. praht fetv ersine ontv eeomn,adpsychopathology. and development, cognitive Psychopathol Dev , affective to approach Emot Cogn 211–234. pp 3, Vol York), New ford, emotion. of physiology. and behavior, experience, 175–190. emotion among coherence binds? imagery. affective during experience jective Psychol Soc esScPsychol Soc Pers J dcPsychol Educ J 2:139–156. npress. in , 17:297–314. 39:1125–1134. d ei ,Hvln-oe M art F(Guil- LF Barrett JM, Haviland-Jones M, Lewis eds Emotions, of Handbook mtRev Emot cn Psychol J Scand sco Bull Psychol 4:84–99. h Emotions The ont eGood be to Born 17:715–734. Neuropsychology Emotion 24:417–441. h xrsino h mtosi a n Animals and Man in Emotions the of Expression The 5(4):379–389. 81:684–696. 135:183–204. 7:1–30. 3:68–75. nApoc oEvrnetlPsychology Environmental to Approach An CmrdeUi rs,Cmrde UK). Cambridge, Press, Univ (Cambridge Nro,NwYork). New (Norton, oi Emot Motiv sco Bull Psychol 7:464–475. rn Psychol Front ilPsychol Biol sco Rev Psychol 17:353–375. 115:102–141. esScPsychol Soc Pers J 6:428. eds Emotions, of Handbook 11:49–62. 110:173–192. e e Psychol Gen Rev ahLanRes Learn Mach J Emotion lnPyhlRev Psychol Clin e e Psychol Gen Rev 81:146–159. Philos J 11:1368–1378. mtRev Emot Emotion 2:300–319. MTPress, (MIT Personal J Cognition (Oxford 11:141– 9:2579– 67:178– 4:132– 6: 5: 1.Sne ,e l 20)Eptyfrpi novsteafciebtntsnoycompo- sensory not but affective the involves pain for (2004) disgust. al. on et T, perspective Singer A (1987) 111. The AE Fallon decisions: P, of Rozin outcomes 110. the to reactions Emotional (1998) al. et M, reward. Parsing Zeelenberg (2003) TE emotion. 109. Robinson of KC, Expression Berridge (2016) 108. G McNeil DC, Cordaro DA, Sauter J, Tracy D, Keltner 107. state. emotional an as Anxiety (1972) CD Spielberger mapping 106. A hypothesis: triad CAD The (1999) J Haidt S, Imada L, Lowery P, humor. Rozin and 104. Exhilaration (1993) appraisal an W Exploring art: Ruch visual in 103. interest and appraisals Cognitive (2005) emotion? PJ “basic” Silvia a love 102. Is (1996) S Wu HJ, Morgan PR, Shaver 101. emotion. and Culture (1997) KR Scherer NH, Frijda B, Mesquita 100. 1.VlnieC (2015) CW Valentine 112. expres- facial A (2012) AP Burgess PJ, Corr AD, Pickering SL, Inchley-Mort AM, Perkins 105. 2.SihC,LzrsR 19)Eoinadadaptation. and Emotion (1990) RS emotions. Lazarus CA, Intergroup Smith (2008) 128. and DM appraisal, emotion, Mackie among ER, Relations Smith (1989) E 127. Schure Ter P, Kuipers NH, Frijda 126. scale. life with satisfaction The (1985) S Griffin RJ, Larsen RA, Emmons ED, Diener 123. (2010) GA Bonanno 122. theory. facets. Affect two of (1984) tale SS A Tomkins pride: of 121. structure psychological The (2007) RW Robins JL, Tracy 120. interest. of structure appraisal the Exploring interesting? is What (2005) PJ Silvia 118. empathy of model A horror: graphic to Reacting (1990) C Heidel J, Stiff approach. R, interpersonal Tamborini An Guilt: 116. (1994) TF Heatherton AM, Stillwell RF, Baumeister 115. experiences The down: and up Leveling 114. (2009) R Pieters M, Zeelenberg N, Ven de Van 113. 2.TayJ,MtuooD(08 h pnaeu xrsino rd n hm:Evi- : and pride of expression spontaneous The analysis (2008) evolutionary D An Matsumoto JL, : Tracy (2010) 125. E Simon-Thomas D, Keltner JL, Goetz 124. triggers, Content, Nostalgia: (2006) C Routledge J, Arndt C, Sedikides T, Wildschut 119. motivation. intrinsic in distinction interest–enjoyment The (1989) J Reeve 117. 9 art F(04 elnso od?udrtnigtecneti efrpr ratings self-report in content the understanding words? or Feelings (2004) LF appraisals Barrett How emotions: 99. situation–different Same (2007) JJ Gross I, Mauss per- M, constructionist Siemer psychological the 98. from viewed emotions Mixed (2017) JA brain. Russell the in emotion 97. of nature are the states Decoding (2016) emotional KS LaBar of PA, biomarkers Kragel neural 96. Multivariate (2015) KS LaBar PA, Kragel 95. feprecdemotion. experienced of emotions. our shape spective. Sci distinct. categorically et fpain. of nents Processes disappointment. Decision and Human Behav of nizational experience the in thought counterfactual of role York), New (Guilford, JM Haviland M, 467–482. Lewis pp LF, Ed, Barrett 4th eds Emotions, 23–49. of pp Handbook 1, Vol York), New (Academic, CD Spielberger ed Research, and ory anxiety. codes for sion moral three and disgust) anger, (contempt, divinity). autonomy, emotions (community, moral three between 605–616. pp York), New (Guilford, JM Haviland emotions. aesthetic of theory 3:81–96. 255–297. pp 2, Vol Oxford), Psychology Cultural ha 18)Fc h es n ertefc:Aia n oilfasa proto- as social and Animal face: emotion. the of fear analyses and evolutionary beast for types the Face (1986) A Ohman UK). Abingdon, dPri A(ulod e ok,p 609–637. pp York), New (Guilford, LA Pervin ed 428–439. Research, pp and Ed, ory 3rd York), New (Guilford, LF Barrett JM, Haviland-Jones M, Lewis readiness. action emotional displays. 11660. nonverbal innate biologically for dence review. empirical and Assess Pers Loss After Life About Us Tells 163–195. pp UK), Hove, (Psychology, Psychol Soc Pers J functions. tion Emotion behavior. emotional and Bull Psychol envy. malicious and benign of ¨ 20:444–455. 5:89–102. 13:83–103. mtRev Emot esnlt o Psychol Soc Personality J 49:71–75. 115:243–267. Science esScPsychol Soc Pers J 92:506–525. 9:111–117. h te ieo ans:Wa h e cec fBereavement of Science New the What Sadness: of Side Other The d er W ae R aawtiT Ofr nvPress, Univ (Oxford TS Saraswathi PR, Dasen JW, Berry eds , Emotion o onAfc Neurosci Affect Cogn Soc sco Bull Psychol 303:1157–1162. PNAS omnRes Commun esScPsychol Soc Pers J h xeietlPyhlg fBeauty of Psychology Experimental The esScPsychol Soc Pers J BscBos e York). New Books, (Basic miia tde Arts Studies Empirical 7:592–600. Emotion | esScPsychol Soc Pers J d cee R ka P Ekman KR, Scherer eds Emotion, to Approaches 136:351–374. ulse nieSpebr5 2017 5, September online Published 102:910–924. 91:975–993. 17:616–640. 9:419–429. 87:266–281. 75:117–141. Psychophysiology d ei M, Lewis eds Emotions, of Handbook 57:212–228. 10:1437–1448. rnsNeurosci Trends rcNt cdSiUSA Sci Acad Natl Proc sco Rev Psychol 76:574–586. 23:119–133. nit:CretTed nThe- in Trends Current Anxiety: adoko esnlt:The- Personality: of Handbook eds Emotions, of Handbook 23:123–145. esnlRelationships Personal 94:23–41. adoko Cross- of Handbook 26:507–513. rnsCogn Trends (Routledge, 105:11655– Motivation | E7909 Orga- Emo- J

PSYCHOLOGICAL AND PNAS PLUS COGNITIVE SCIENCES