Abstract
Slow motion is a popular video editing tool used to enhance short-form videos (e.g., reels, stories, GIFs), which are commonly found in media entertainment and marketing communications. This research shows that slow motion increases the virality (e.g., likes, votes, views) of short-form videos and boosts brand liking, choice, and willingness to pay. The effect occurs because slow motion enhances the hedonic component of the viewing experience via processing fluency. By documenting how the success of slow motion is subject to moderators, this work shows marketers, entertainment producers, and everyday people how to use slow motion more effectively. Across a large-scale field data set and six experiments, the authors highlight that slow motion is especially effective when applied to short-form videos that are inherently pleasant and that involve complex movements that are difficult to perceive at regular speed. However, even simple movements benefit from slow motion when content creators zoom in on subtle movements to increase complexity. Moreover, slow motion is more effective when viewers engage in less elaborate processing. Finally, the authors show that the perceived disfluency of fast-motion editing is effective at boosting brand evaluations when viewers desire excitement.
Social media platforms have revolutionized the format of online videos and how they are consumed. TikTok, Instagram, Facebook, and YouTube, for example, show brief, entertaining, educational, and inspirational short-form video clips (i.e., “reels,” “stories,” “shorts”) looped and played with or without sound. Linked together algorithmically, these videos create a compelling and potentially infinite consumption experience. TikTok's average user, for instance, spends a shocking 95 minutes each day on the app (Chan 2022), with some posts collecting millions of views and likes. Thus, brands, influencers, and other content creators are adopting short-form video formats to position their products, brands, and personas on social media.
In a highly competitive social media space, content creators use editing tools to capture attention and increase engagement. To this end, slow motion is a popular stylistic tactic for short-form videos. Thanks to smartphones and social media apps with slow-motion features, millions of slam dunks, shuffle dances, and hair flips on TikTok and Instagram are tagged #slowmotion. Brands also use slow motion in social media posts, from models strutting on the catwalk to burger patties flipping through the air (see Web Appendix A for examples).
Although slow motion is a ubiquitous aesthetic effect, there is little scientific guidance on how this technique shapes the hedonic nature of the viewing experience. Based on how frequently marketers and everyday people use slow motion, it is reasonable to expect that it improves the aesthetic appeal of dynamic visual content. Yet, research suggests that slowing consumption experiences does not improve evaluations. Radio broadcasts, for instance, are liked more when sped up rather than played at slower or regular speed (LaBarbera and MacLachlan 1979; Pronin and Wegner 2006), which suggests that the slower presentation of visual content might be experienced as boring. In television commercials, scenes of actors eating chocolate or shampooing their hair in slow motion seem “fake” and “posed,” which lowers product evaluations (Yin, Jia, and Zheng 2021). Based on these findings, slow-motion short-form videos might lead to negative consumer responses despite their widespread usage.
We examine when and why slow motion might boost consumer evaluations by enhancing the aesthetic appeal of short-form videos. To decode the effect of slow motion on aesthetic appreciation, our investigation relies on the dominant explanatory framework in this domain: the theory of processing fluency (Reber, Schwarz, and Winkielman 2004). According to this framework, aesthetic pleasure is a function of the perceiver's processing dynamics: the more easily (i.e., fluently) viewers can process an object, the more positive the aesthetic response, because processing ease triggers mild positive affect, which is attributed to the object being processed. Because slow motion lowers the amount of information that viewers need to process per unit of time, we suggest that slow motion feels “easier to watch” (i.e., more fluent), which should enhance consumers’ aesthetic appreciation of the video.
To provide actionable recommendations on how to use slow-motion effects, we derive four theoretically and managerially relevant moderators from processing fluency theory. We highlight the type of content and the type of consumers for which slow motion is effective at increasing evaluations (see Figure 1 for the conceptual model). In terms of content, we show that slow motion only boosts evaluations when short-form videos display complex movements. Watching a complex movement (e.g., an intricate choreography) is easier and thus more pleasant in slow motion compared with regular speed. In contrast, watching simple movements (e.g., basic choreography) is easy regardless of playback speed. We also show that the effect of slow motion on consumer evaluations differs between negatively and positively valenced content. Because fluency amplifies consumer judgments (Landwehr and Eckmann 2020; Mrkva and Van Boven 2020), slow motion increases the appeal of pleasant content (e.g., ballet) but decreases the appeal of aversive content (e.g., a bloody boxing bout). In our examination of consumer characteristics, we find that slow motion has a stronger effect on evaluations for consumers in a heuristic (vs. deliberate) information processing mode. Positive affect induced by fluent processing is an incidental cue that is discounted in deliberate judgments (Koch and Forgas 2012). Finally, we highlight the moderating role of consumption goals. Among consumers seeking excitement (vs. relaxation), fast motion is effective because disfluency triggers feelings of risk (Dohle and Montoya 2017; Song and Schwarz 2009).
Conceptual Model.
Our research contains three theoretical contributions. First, we deepen scholarly understanding of how slow motion shapes hedonic experience. Most previous work in marketing and psychology examines how slow motion influences cognitive inferences about products or people (e.g., Caruso, Burns, and Converse 2016; Jia, Kim, and Lin 2020; Yin, Jia, and Zheng 2021). Despite its importance for marketplace behavior, the aesthetic consequences of slow motion have been underexamined. Second, this investigation is the first to extend the theory of processing fluency to the context of dynamic visual content. There is a rich base of evidence demonstrating how text and images can be presented fluently (see Reber, Schwarz, and Winkielman 2004), yet how people experience the processing of dynamic content has been largely overlooked. Finally, our investigation reveals that there is beauty in slowness, thus creating a situational counterweight to well-established findings that a faster-than-normal playback speed is pleasing (Pronin and Wegner 2006).
Theoretical Background
Slow Motion
Slow motion is “the action of showing film or playing back video more slowly than it was made or recorded, so that the action appears much slower than in real life.” 1 The technique is frequently used in movies, commercials, music videos, and sports broadcasts. Since the advent of smartphones and social media platforms, the use of slow motion has exploded in the digital space: 8.5 million videos on Instagram are tagged #slowmotion, TikToks tagged #slowmotion have accumulated 60.5 billion views, and more than 3,000 slow-motion GIFs are available on Giphy, a free catalog of GIFs used in social media posts and text messages.
Prior research shows slow motion shapes people's cognitive, physiological, and affective responses toward visual information (for a review, see Table 1). Cognitive inferences formed by viewers about slow-motion imagery have been studied most extensively. In particular, slow motion makes displayed action appear more deliberate and intentional because it gives the false impression that an actor had more time to premeditate (Caruso, Burns, and Converse 2016). Jurors, for instance, perceive a shooter as acting with greater intent when the surveillance video of the crime is played in slow (vs. regular) speed (Caruso, Burns, and Converse 2016), and referees give harsher penalties when fouls are replayed in slow motion (vs. regular speed; Spitz et al. 2018). In television commercials, slow-motion tactics thus backfire because behaviors that appear intentional also seem extrinsically motivated and thus insincere (Yin, Jia, and Zheng 2021). An actor's smile after a bite of dessert, for example, appears more posed (i.e., fake) in slow motion than in regular speed.
Review Table.
However, research that reveals how slow motion triggers cognitive inferences (e.g., intentionality) provides little insight into how slow motion influences the aesthetic and, thus, hedonic nature of short-form videos. Two existing papers that examine affective outcomes of slow motion provide conflicting results. Slow-motion videos of ballet and tennis triggered more positive emotions when played in slow motion (vs. regular speed; Wöllner, Hammerschmidt, and Albrecht 2018). However, viewing negative videos (e.g., TV news stories about natural disasters) elicits more negative emotions when played in slow motion (vs. regular speed; Barnett and Grabe 2000). Apparently, slow motion can change the hedonic nature of the viewing experience, making it positive in some cases and negative in other cases. It has been suggested that these conflicting findings might be explained by the fact that slow-motion videos increase viewer immersion (Jung and Dubois 2022), since immersion can intensify the valence of an experience (Diehl, Zauberman, and Barasch 2016). Yet, Jung and Dubois (2022) did not experimentally manipulate valence, and other research even conflicts with their interpretation by suggesting that engagement is inherently pleasurable (De Oliveira Santini et al. 2020), implying a mere main effect of immersion instead of an intensifying effect. Thus, existing work has neither theoretically nor empirically reconciled the conflicting affective consequences of slow motion.
Slow Motion Increases Processing Fluency
Our investigation set out to examine how slow motion changes the aesthetic and thus hedonic nature of the viewing experience. We suggest that to understand how slow motion shapes the aesthetic properties of short-form videos, one needs to decode how slow motion affects the most prominent driver of aesthetic pleasure: processing fluency.
According to the hedonic fluency model (Winkielman and Cacioppo 2001), which is the dominant theoretical interpretation of fluency effects, processing ease is the causal determinant of generic mild positive affect (i.e., a mild pleasurable experience). High fluency may elicit positive affect because it is associated with progress toward successful recognition of the stimulus and error-free processing (Carver and Scheier 1990; Derryberry and Tucker 1994; Ramachandran and Hirstein 1999). High fluency may also feel good because it signals that the stimulus is familiar and thus unlikely to be harmful (Zajonc 1968). Evolutionary accounts even postulate that the relationship between fluency and positive affect is hardwired because it is beneficial for survival to detect and approach fluent situations (Thornhill and Gangestad 1993). Importantly, the positive affect that is elicited by high processing fluency serves as a heuristic cue that subsequently feeds into judgments of aesthetic appreciation (e.g., liking). Consumers’ aesthetic appreciation of stimuli is thus positively linked to how easily they find it to perceptually process and understand them.
Based on the hedonic fluency model, we propose that slow motion increases the hedonic component of the viewing experience by making movement easier to process. Humans can only perceive a limited amount of visual information per unit of time (Holcombe 2009). When action unfolds quickly, the rate of visual input risks overwhelming the visual system. Slow motion can overcome this limitation. Slowing a dynamic scene lowers the number of frames that need to be processed per unit of time. Additional processing time should make it easier to consume the visual input. Watching a short-form video in slow motion should feel easier than watching the same scene at regular speed. The incidental positive affect that is triggered by the experience of processing fluency serves as a heuristic cue to boost aesthetic liking. Moreover, as suggested by prior research, experienced processing fluency also enhances downstream consumer judgments and evaluations, such as purchase decisions (e.g., Landwehr, Labroo, and Herrmann 2011).
Boundary Conditions
We develop boundary conditions for the effect of slow motion on consumer evaluations that are derived from fluency theory (see Figure 1). The moderators help increase the effectiveness of slow motion by informing two important managerial decisions: (1) Which types of content benefit most from slow motion? and (2) Which types of consumers respond most positively to slow-motion tactics?
Content-related moderators
Our first content-related moderator of the hedonic fluency process concerns the complexity of the movements that are depicted in the video. If fluency is the key driver of the effect of slow motion on consumer evaluations, then the effect should be pronounced among short-form videos that depict complex movements that are difficult to process at regular speed. According to previous research (e.g., Landwehr, Labroo, and Herrmann 2011; Van Grinsven and Das 2016; Cox and Cox 2002), classic fluency manipulations are more effective when applied to visually complex (vs. simple) stimuli. Inherently simple stimuli are relatively easy to process regardless of how they are presented, whereas inherently complex stimuli are easier to perceive for fluent presentation modes. To illustrate, visually complex brand logos and fashion apparel benefit more from repeated exposure (a classic fluency manipulation) than relatively simple logos and fashion apparel (Cox and Cox 2002; Van Grinsven and Das 2016). In a similar vein, complex car designs, but not simple car designs, achieve higher sales when the fluency of a car's global shape is enhanced by a typical form (Landwehr, Labroo, and Herrmann 2011). Taken together, when seeing dynamically complex movements in slow motion (vs. regular speed), the viewing experience should appear more fluent, which should trigger positive affect and thus boost consumer evaluations. For dynamically simple movements in slow motion (vs. regular speed), the viewing experience is already relatively fluent due to the simple content such that there is less room left for slow motion to enhance fluency, positive affect, or liking.
Our second content-related moderator concerns the valence of the depicted content. Typically, slow-motion short-form videos on social media depict positively valenced content (e.g., people dancing). Sometimes, however, negatively valenced content is slowed down (e.g., car crashes). According to a recent extension of fluency theory, the effect of fluency on the hedonic component of the viewing experience markedly differs between pleasant and unpleasant stimuli due to two fluency processes that occur simultaneously (Landwehr and Eckmann 2020). On the one hand, fluency signals a problem-free operation of the cognitive system, which is experienced as pleasurable and, in turn, shifts judgments in a positive direction (i.e., hedonic fluency process; Winkielman et al. 2003). On the other hand, fluency triggers nonspecific activation, which amplifies judgmental tendencies (i.e., amplifying fluency process; Bornstein and D’Agostino 1994). If both processes occur simultaneously, the effect of slow motion on liking should be attenuated or even reversed for negatively valenced content. Indeed, repeated exposure increased liking of affectively positive words but had no effect on affectively negative words (Mrkva and Van Boven 2020). Other studies reveal a reversal in which negative words and aversive pictures were more unpleasant when they were presented in a more (vs. less) fluent manner (Albrecht and Carbon 2014; Grush 1976). In summary, our processing fluency explanation resolves the discrepant affective outcomes identified by previous slow-motion research (Barnett and Grabe 2000; Wöllner, Hammerschmidt, and Albrecht 2018).
Consumer-related moderators
The first consumer-related moderator of the hedonic fluency process concerns the extent to which a consumer deliberates more or less about the judgment at hand. According to the hedonic fluency model, slow-motion short-form videos should be experienced as fluent, which should, in turn, trigger mild positive affect that serves as a heuristic cue for subsequent judgments. The heuristic reliance on affect inherent in this process implies that the slow-motion effect should be especially pronounced among consumers who rely on heuristic cues when forming judgments: those who engage in less deliberate processing. Specifically, previous research suggests that people who are relatively unmotivated or unable to process a message carefully will base their judgments on incidental positive affect, which is triggered by peripheral message characteristics such as attractive-looking models, celebrities, or pleasant background music (Petty and Cacioppo 1986). In contrast, people who are motivated or able to process a message carefully will discount such heuristic cues in favor of more central message characteristics that withstand critical examination and cognitive scrutiny. Accordingly, prior research shows that the positive affect that is induced by fluent processing is discounted among consumers who process more deliberately such that the effect of manipulated fluency on judgments is attenuated. For instance, negative mood, which induces more careful processing, eliminates the influence of processing fluency on judgments of truth (Koch and Forgas 2012). Similarly, priming consumers to think abstractly (vs. concretely) weakens the effect of fluency on judgments because it causes them to focus on objective information (vs. subjective feelings; Tsai and Thomas 2011). Finally, the effect of repetition, which is a classical fluency manipulation, on liking is reduced when consumers are given ample (vs. little) time to form judgments and when consumers believe the quality of their judgments will be (vs. will not be) evaluated later on—with both inducing more deliberate processing (Kruglanski, Freund, and Bar-Tal 1996). Applied to our hypotheses, slow-motion short-form videos should be relatively more successful in boosting consumer evaluations among people who engage in less deliberate processing. That is because the less people deliberate, the more they are influenced by the positive affect that is triggered by a slower/more fluent presentation.
Our second consumer-related moderator of the hedonic fluency process examines what happens when consumers strive to experience negative affective experiences. People sometimes expose themselves to consumption experiences that involve feelings of risk, danger, and hazardousness (Andrade and Cohen 2007), which are typically associated with disfluent processing (Dohle and Montoya 2017; Zürn and Topolinski 2017). How might these consumers react to slow-motion videos that trigger mild positive affect due to their fluency? According to our theory, slow-motion short-form videos should not boost consumer evaluations among these viewers because, for them, the accompanying mild positive affect (and feelings of safety and familiarity) is misaligned with their consumption goal. Rather, these consumers should react more positively to short-form videos that play in regular speed (or even fast motion) because the resulting disfluent processing triggers desirable “aversive” feelings of risk that these consumers seek out. Indeed, previous research has shown that when consumers strive to experience negative affective states, disfluent presentation modes can lead to positive consumer evaluations. For instance, roller coasters with difficult (vs. easy) to pronounce names are more appealing because they seem more adventurous and exciting (Song and Schwarz 2009). Similarly, risk-seeking consumers evaluated products more positively when they were described in a difficult-to-read font (Pocheptsova, Labroo, and Dhar 2010). To test this consumer-related boundary condition empirically, we manipulate consumers’ consumption goal by instructing them to seek a positively valenced experience that is associated with safety (a relaxing vacation) or a relatively more negatively valenced experience that is associated with risk (an exciting vacation). We expected that, among consumers who strive to experience risk, thrill, and excitement, short-form videos in regular speed should lead to favorable evaluations due to affectively mildly negative feelings of risk that are triggered by a faster, and thus less fluent, viewing experience.
Empirical Overview
We test our hypotheses using more than 600 short-form videos spanning a wide range of topics (e.g., sports, nature, food) and examine consumers’ behavior on a GIF-sharing platform. The videos are just 1–5 seconds long and display movements repeatedly and continually in an infinite loop, which is representative of short-form videos on social media (e.g., TikTok, Instagram) and controls for presentation duration.
Seven studies examine the effect of slow motion on liking, indicators of virality (e.g., votes, views, likes), willingness to pay (WTP), brand liking, and choice. We provide evidence for the fluency account in the lab and field and via mediation and moderation. Study 1 shows that videos are liked more in slow motion (vs. regular speed) and documents that processing fluency statistically explains this finding. Studies 2a–c and 3 test content-related moderators of the effect of slow motion on liking. Consistent with fluency theory, slow motion only increases liking for content high in dynamic complexity (Study 2a). Yet, even simple scenes can benefit from slow motion if content creators zoom in on the action to increase dynamic complexity (Study 2b). Moreover, slow motion only increases liking when the underlying content is positively (vs. negatively) valenced (Study 2c). Our field data set (Study 3) conceptually replicates both content-related moderators on a GIF-sharing platform that we analyze in terms of views, votes, and ratings. Studies 4a and 4b examine consumer-related moderators of the effect of slow motion on brand liking and consequential choices. Slow-motion videos increase brand liking and choice, but the effect was particularly pronounced among consumers who engaged in less (vs. more) deliberate processing (Study 4a). Finally, Study 4b shows that relatively disfluent regular-speed videos can boost brand evaluations when consumers seek excitement. The Appendix reports means, standard deviations, and sample sizes for all conditions and all dependent measures. We also provide hyperlinks to all video stimuli (Web Appendix B). All data sets and R scripts are available on OSF (https://osf.io/jka9e/). No participants were excluded from the analyses.
Our laboratory experiments consistently manipulate playback speed within-subjects. This approach mirrors consumers’ online viewing behavior, where slow-motion and regular-speed short-form videos are watched back-to-back (e.g., in Instagram reels). Importantly, the within-subjects design is appropriate because fluency is a relative experience (Wänke and Hansen 2015). Consumers can only experience a stimulus as easy to process if this stimulus is fluent relative to a standard of comparison (e.g., a context of other stimuli). Accordingly, an experiment in Web Appendix C empirically shows that slow-motion clips were liked significantly more when presented in a context of regular-speed clips (vs. a context of slow-motion clips).
Study 1: Slow Motion Increases Liking by Enhancing Fluency
We designed Study 1 to examine the effect of slow motion (vs. regular speed) on liking (H1) and provide mediation evidence for the proposed perceptual fluency process (H2). We manipulated the speed of video clips (slow motion vs. regular speed) within-subjects and measured liking and perceptions of fluency in response to each clip. We expected that slow motion (vs. regular speed) would increase liking, and that fluency would explain this effect.
Design and Procedure
Stimuli development
We collected 14 slow-motion videos that covered a wide range of topics, including sports (e.g., a dunk in basketball), nature (e.g., a wave crashing), food (e.g., a strawberry being dipped in chocolate), and fashion (e.g., a model on a runway). We created regular-speed versions of these videos by increasing their frame rate in Photoshop. A pretest verified that we had successfully manipulated perceived speed (Web Appendix D).
Main study
The effect size was unknown, but we estimated that about 200 participants would be needed to reliably detect small-sized effects in a within-subjects design. At the end of data collection, 199 Amazon Mechanical Turk (MTurk) participants from the United States had completed the survey (68 women, 131 men; Mage = 34.27 years, SD = 11.13) in response for a small payment ($.60).
Participants rated 14 videos that were presented sequentially, in random order, and without sound. For each video, we randomly assigned participants to see either the slow-motion version or the regular-speed version. This design did not expose participants to both speed versions of a video to minimize demand. We first measured liking (“How much do you like this video clip?”; 1 = “Not at all,” and 7 = “Very much”) and then fluency (“The process of studying this video clip is …”; 1 = “difficult,” and 7 = “easy”; Graf, Mayer, and Landwehr 2018) on two consecutive pages that showed the video above the scale. This process was repeated for all 14 videos.
Results and Discussion
To test the effect of playback speed (regular = −1 vs. slow motion = 1) on liking via processing fluency, we estimated three regression equations (Muller, Judd, and Yzerbyt 2005) using the lmerTest library in R (Kuznetsova, Brockhoff, and Christensen 2017). We conducted the analysis on a disaggregated level to use the full available information. To account for the repeated measurement structure of the data (i.e., 14 evaluations per participant) and the random sampling of stimuli, we used linear mixed models that contained one random intercept per participant and one crossed random intercept per stimulus (Westfall, Kenny, and Judd 2014).
The first model regressed liking on speed (fixed effect). Consistent with H1, this model revealed a total effect of speed (b = .092, p < .001). Slow-motion videos were liked more than regular-speed videos. The second model regressed fluency on speed (fixed effect). Slow-motion videos were perceived as easier to process than regular-speed videos (b = .121, p < .001). The third model regressed liking on speed and fluency as fixed effects. It showed a positive effect of fluency on liking (b = .418, p < .001) and a marginally significant direct effect of speed on liking (b = .042, p = .056). In support of H2, a Sobel test (Krull and MacKinnon 1999) indicated a significant indirect effect of speed on liking via processing fluency (z = 5.76, p < .001).
Taken together, Study 1 shows that slow-motion videos were liked more than regular-speed videos (H1) and demonstrates that fluency perceptions statistically explained this effect (H2). We conducted an additional study, reported in Web Appendix E, in which we pit the fluency mechanism against several alternative explanations for the effect of speed on liking (e.g., perceived level of visual details, arousal). Fluency emerged as a statistically stronger explanation than these alternative accounts. In addition, this study establishes a serial mediation, with fluency being the first mediator and positive affect being a subsequent second mediator, which supports the hedonic fluency model (Winkielman and Cacioppo 2001).
Study 2a: Content-Related Moderators—The Role of Dynamic Complexity
Study 2a examined our first content-related moderator of dynamic complexity. We manipulated the speed of video clips (slow motion vs. regular speed) within-subjects and measured the dynamic complexity of the movements depicted in each video via a newly developed algorithm. The effect of slow motion on liking should be accentuated (vs. attenuated) among videos that showed relatively complex (vs. simple) movements (H3).
Design and Procedure
Dynamic complexity
To measure dynamic complexity objectively, we created an algorithm that capitalizes on the fact that dynamic stimuli consist of a collection of static images that are played in such quick succession that viewers interpret them as continuous movement. The algorithm quantifies the amount of change between these individual images independent of all other image properties. To this end, we adapt an existing algorithm implemented in the statistical software R that was originally developed to measure visual typicality (Mayer 2020).
To create our dynamic complexity measure, we took four steps. First, we decompose a video file into individual images (i.e., frames). Second, we compute for each image how similar this image is to the average image across all frames of the video (i.e., a morph of all images) using R's imagefluency package (Mayer 2020). This package computes the pixel-wise correlation between all pixels of the individual image and the average representation as a measure of similarity. Thus, each image is assigned one correlation as a measure of its similarity to the average representation. Third, we compute the mean across all individual correlations following the idea that high correlations indicate a low amount of change over time and hence low dynamic complexity. Fourth, to facilitate interpretation, we subtract the average correlation from 1 such that higher values indicate higher dynamic complexity (range of measure: 0–1). Our algorithmic measure correlates with subjective perceptions of dynamic complexity, attesting to the validity of our operationalization (Web Appendix F).
Stimuli development
To maximize variation in dynamic complexity, we computed the dynamic complexity of 100 slow-motion videos (Mcomplexity = .21, SD = .16) and then selected the ten least complex videos (range: <.01–.02), ten videos of average complexity (range: .19–.23), and the ten most complex videos (range: .44–.73) for the main study. We again created regular-speed versions for each video by increasing the frame rate in Photoshop. A pretest verified our speed manipulation and confirmed that dynamic complexity was unrelated to speed perceptions. Participants are thus sensitive to the speed of movement, independent of how complex the movement is (Web Appendix G).
Main study
Study 2a manipulated speed (slow motion vs. regular speed) within-subjects and measured dynamic complexity (continuous). We recruited 212 MTurk participants from the United States (77 women, 135 men; Mage = 33.58 years, SD = 8.53) that were compensated with a small payment ($.50). Out of 30 videos, participants were randomly assigned to a subset of 15. They rated the 15 videos sequentially, in random order. As in Study 1, we randomly assigned participants to either the slow-motion version or the regular-speed version of each video. In response to each video, we measured liking using the scale from Study 1 (“How much do you like this video clip?”; 1 = “Not at all,” and 7 = “Very much”).
Results and Discussion
To test the predicted interactive effect of speed and dynamic complexity on liking, we conducted the analysis as in Study 1 on a disaggregated level such that all 15 evaluations per participant are considered. Our linear mixed model included liking as the dependent variable, and complexity (continuous; centered), the effect-coded speed manipulation (slow motion = 1 vs. regular speed = −1), and their interaction as fixed effects. As in Study 1, we again included one random intercept per participant, and one crossed random intercept per stimulus. The model revealed the predicted interaction between dynamic complexity and speed (b = .936, p < .001). There were no main effects of complexity (b = .614, p = .362) or speed (b = .010, p = .707).
We dissected the interaction by identifying the regions of complexity beyond which speed had an effect on liking (Figure 2). In support of H3, videos with complexity scores greater than .29 were liked more in slow motion (vs. regular speed). At the bottom of the distribution, videos with complexity scores below .19 were liked less in slow motion (vs. regular speed).
Effect of Speed and Dynamic Complexity on Liking.
The results of Study 2a confirm the idea that the effect of slow motion on liking depends on the dynamic complexity of the underlying content (H3). Slow motion increased liking for videos that showed complex movements, but not videos that showed simple movements. Our results even go beyond our initial hypothesis. The crossover interaction indicates that, for simple movements, slow motion is less preferred than regular speed. Watching simple movements in slow motion might have been boring.
Study 2b: Content-Related Moderators—Manipulating Dynamic Complexity Through Zooming In on a Movement
Study 2b examined the interaction effect of speed and dynamic complexity on liking (H3) in a more internally valid manner. Study 2a left open the possibility that dynamically complex videos are liked more in slow motion because they involve different subject matters. To rule out this concern, we manipulate complexity by zooming in on a movement, which increases the degree of change in the video clip and, thereby, dynamic complexity. As in Study 2a, the effect of slow motion on liking should be reduced (vs. accentuated) among videos that showed simple (i.e., “zoomed out”) as compared with complex (i.e., “zoomed in”) movements. A second goal of Study 2b was to provide additional evidence for the fluency process. We measured fluency and examined whether it statistically explains the joint effect of complexity and speed on liking.
Design and Procedure
Stimuli development
We collected five slow-motion videos and created regular-speed versions by increasing the frame rate. To manipulate complexity, we simulated the effect of zooming in by cropping out the static background surrounding the dynamic part of the scene (see Figure 3 for an exemplary illustration). To illustrate, in a video showing a cross-country skier, we cropped out the static nature background surrounding the movement. We then enlarged the video to match the size of the original version. Removing the static background increases the degree of movement and, thereby, dynamic complexity. We used our algorithm to validate this manipulation. As intended, the cropped videos (Mcropped = .23, SD = .11) were significantly more complex than the noncropped videos (Mnoncropped = .12, SD = .08; one-sided t-test: t(8) = 1.947, p = .044, d = 1.23).
Exemplary Illustration of the Effect of Zooming In on Complexity.
Main study
Study 2b manipulated complexity (simple vs. complex) between-subjects and manipulated speed (slow motion vs. regular speed) within-subjects. We recruited 296 MTurk participants from the United States (205 women, 91 males; Mage = 34.57 years, SD = 11.82) who completed the study for a small monetary compensation ($.75). The participants were randomly assigned to evaluate either the more simple (i.e., noncropped) or the more complex (i.e., cropped) versions of the videos. Our stimulus pool consisted of only five videos. To provide more stimulus replicates and ultimately increase power, we exposed participants to both the slow-motion version and the regular-speed version of each video. Thus, we manipulate speed within each stimulus rather than between stimuli (as in Studies 1 and 2a). The participants saw the slow-motion and the regular-speed version of the same video back-to-back (ten videos in total). Below each video, they indicated liking with the scale used in all previous studies. For each video, we randomized whether participants first evaluated the slow version or the regular-speed version. We also randomized the order of the five videos.
Next, we measured fluency. We presented all ten videos again using the procedure described previously. Below each video, participants completed the fluency item from Study 1 (“The process of studying this video clip is …”; 1 = “easy,” and 7 = “difficult”). We collected the fluency measure after measuring liking to reduce demand.
Results and Discussion
We expect that the total effect of speed (regular = −1, slow motion = 1) on liking and the indirect effect through fluency would vary across the complexity conditions (simple = −1, complex = 1). Thus, we tested for moderated mediation (Muller, Judd, and Yzerbyt 2005) using three linear mixed models with the respective predictors as fixed effects and a random intercept per participant, and a crossed random intercept per stimulus as random effects (see Table 2 for a summary of the results).
Mixed Linear Regression Models for Moderated Mediation.
*p < .05.
**p < .01.
***p <.001.
Notes: X = independent variable, MO = moderator, ME = mediator.
Model 1 indicates a total effect of speed on liking (b = .144, p < .001) that is moderated by complexity (b = .074, p = .005; when complexity is high, speed has a stronger effect on liking [b = .218, p < .001] than when complexity is low [b = .071, p = .056]). Model 2 shows an effect of speed on fluency (b = .476, p < .001) that is also moderated by complexity (b = .165, p < .001; when complexity is high, speed has a stronger effect on fluency [b = .641, p < .001] than when complexity is low [b = .311, p < .001]). Finally, Model 3 indicates that there is a significant effect of fluency on liking (b = .244, p < .001), which is not moderated by complexity (b = .029, p = .133). In this model, the direct effect of speed on liking is no longer significant when controlling for fluency (b = .024, p = .382), and a Sobel test (Krull and MacKinnon 1999) indicates a significant indirect effect (Sobel z = 10.606, p < .001). Most importantly, the interaction between speed and complexity is no longer significant when controlling for the significant indirect effect through fluency (b = .020, p = .468; Sobel z = 6.101, p < .001). These results suggest that both the main effect of speed on liking and the combined effect of speed and complexity on liking are fully mediated by fluency.
Study 2b conceptually replicates and extends Study 2a. We manipulated complexity by zooming in on the movement in a scene. In support of H3, we find that when the movement was complex (i.e., zoomed in), slow motion was liked more than regular speed. When the movement was simple (i.e., zoomed out), however, this effect was attenuated. These results cannot be attributed to differences in the type of content, as our zooming manipulation held content constant. Study 2b also provides a direct test of the fluency account via moderated mediation. Taken together, slow motion only increases liking when it facilitates processing.
Study 2c: Content-Related Moderators—The Role of Stimulus Valence
Study 2c tested our second content-related moderator of stimulus valence. We manipulated the speed of video clips (slow motion vs. regular speed) within-subjects and manipulated between-subjects whether the underlying content was positively valenced or negatively valenced. We expected that the effect of slow motion on liking would be reduced or even reversed among videos that involved relatively unpleasant content (H4). A second goal of Study 2c was to show that slow motion influences consequential choices.
Design and Procedure
Stimuli development
We collected four slow-motion videos that depicted positively valenced content (e.g., puppies playing) and four slow-motion videos that depicted negatively valenced content (e.g., a man spitting). A validation study confirmed that the video stimuli significantly differed in terms of perceived valence (see Web Appendix H). We also ensured that dynamic complexity was equal across the valence conditions (Mnegative = .25, SD = .10 vs. Mpositive = .24, SD = .13; t(6) = .134, p = .898, d = .09). We again created regular-speed versions by increasing the frame rate. A pretest verified that our manipulation altered perceived speed independent of valence (Web Appendix I). Thus, we rule out that differences in complexity or perceived speed might account for our findings.
Main study
Study 2c manipulated valence (negative vs. positive) between-subjects and speed (slow motion vs. regular speed) within-subjects. We recruited 401 participants from the United Kingdom on Prolific Academic (256 women, 140 men, 5 nonbinary participants; Mage = 35.62 years, SD = 13.22; payment = £.63). The participants were randomly assigned to either the negative or the positive valence condition. We manipulated speed using the procedure from Study 2b such that participants saw the slow-motion and the regular-speed version of the same video back-to-back (eight videos in total). Below each video, they indicated liking with the scale used in all previous studies. For each video, we randomized whether participants first evaluated the slow-motion version or the regular-speed version. We also randomized the order of the four videos.
Next, we measured consequential choices. Participants saw screenshots of the four videos, below which they indicated whether they would want to rewatch this video in slow motion (1) or regular speed (0; binary scale). At the end of the study, one video was randomly selected and played at the participant's preferred speed. The total number of videos that participants chose to rewatch in slow motion served as our dependent measure (M = 1.79, SD = 1.35).
Results and Discussion
Liking
As in our previous analyses, we used a linear mixed model with a random intercept per participant and a crossed random intercept per stimulus to account for the repeated measurement structure and the random sampling of stimuli. Liking was the dependent variable, and playback speed (b = .145, p < .001), valence (b = 1.890, p = .001), and the interaction of speed and valence (b = .410, p < .001) were included as fixed effects. In line with the strong moderating effect of valence (H4), we observe that the effect of speed on liking is positive for pleasant videos (b = .554, p < .001) but negative for aversive videos (b = −.265, p < .001).
Choice
Because our dependent measure was a count variable, we conducted a Poisson regression. When regressing slow-motion choice on valence (0 = negative vs. 1 = positive), we detected a positive effect of valence; B = 1.179, z = 13.37, p < .001. On average, participants selected slow motion as the playback speed for 2.73 (out of 4) videos in the positive valence condition. In contrast, in the negative valence condition, participants only wanted to watch an average of .84 (out of 4) videos in slow motion.
The results of Study 2c attest to the idea that the effect of slow motion on liking depends on the valence of the underlying content (H4). Slow motion increased the aesthetic appeal of pleasant videos but decreased the appeal of unpleasant videos, presumably because fluency has hedonic and amplifying effects on evaluative judgments (Landwehr and Eckmann 2020). Study 2c also shows that slow motion unfolds effects on consequential choices.
Study 3: Content-Related Moderators—Field Evidence from a GIF-Sharing Platform
Study 3 examined the main effect of slow motion (H1) as well as the two content-related moderators of dynamic complexity (H3) and stimulus valence (H4) in an externally valid fashion. To this end, we analyzed the viewing, voting, and rating behavior of users on a GIF-sharing platform that contained slow-motion and regular-speed GIFs. We measured the dynamic complexity of each GIF with our algorithm and the valence of each GIF by asking independent coders to provide ratings. Conceptually replicating Studies 2a–c, slow motion (vs. regular speed) should boost the virality of GIFs that involve content that is relatively high (vs. low) in dynamic complexity (H3) and relatively pleasant (vs. unpleasant; H4).
Data and Modeling Approach
Using the rvest package for R (Wickham 2019), we accessed the website of the GIF-sharing platform on September 21, 2020, in a two-step approach. First, we accessed all 237 GIFs that were tagged “slow-motion.” Second, we accessed 300 additional GIFs by using the random search function of the website to obtain a random sample of control GIFs. We sampled slightly more control GIFs to allow for exclusions. After inspecting the data, we removed 23 GIFs (22 were duplicates, and one was low quality), leaving us with 514 GIFs for analysis (236 slow-motion GIFs and 278 regular-speed control GIFs). For these GIFs, we saved the information shown in Table 3. Table A1 in Web Appendix J examines whether slow-motion GIFs and regular-speed control GIFs differ on any of those variables.
Description of Variables and Descriptive Statistics.
Dynamic complexity
We measured the objective dynamic complexity of each GIF using our newly developed algorithmic measure (for a description of the algorithm, see Study 2a; for the validation of the algorithmic measure, see Web Appendix F). The dynamic complexity of the 514 GIFs varied greatly (range = <.01–.75; M = .27, SD = .17).
Stimulus valence
To measure valence, we asked two independent raters who were blind to the research hypothesis to code the GIFs. The raters received links to the GIFs that contained no information on views, votes, likes, or tags and coded the GIF content on a scale ranging from 1 (“definitely intended to elicit negative emotions”) to 7 (“definitely intended to elicit positive emotions”). The level of agreement between the raters was high (r(514) = .545, p < .001). We thus averaged their scores to form an index of stimulus valence (α = .67).
Composite liking
The website provides three variables that relate to GIF attractiveness: rating, votes, and views. These variables are mutually dependent in complex ways (e.g., the number of votes being contingent on the number of views). Research suggests that the interaction of valence and volume of online ratings is most predictive of downstream consumer preferences such as sales and choices (Kostyra et al. 2016). Accordingly, building a composite of valence and volume to predict global consumer preferences is recommended (Rosario et al. 2016). We follow these insights by adopting a three-step procedure. First, we build the ratio between votes and views to measure volume (votes/views; Mukhopadhyay and Chung 2016). Second, we scale this new “volume” variable and the rating variable, which captures valence, by dividing both variables by their standard deviation, thereby setting the value range to the interval 0–1. Third, we multiply the two scaled variables to build our dependent measure: composite liking. Web Appendix J reports analyses separately for the variables views, votes, ratings, and volume (votes/views) and also includes a model in which the valence and volume measures are added.
Results and Discussion
We conducted one linear regression that regressed composite liking on presentation speed (slow motion = 1 vs. random category = −1), dynamic complexity, stimulus valence, and their interactions (Table 4, Model 1). The regression revealed a positive significant main effect of speed (p < .001), indicating that slow-motion GIFs score higher on our composite liking measure than regular-speed GIFs (H1). There were also significant main effects of dynamic complexity (p = .017) and valence (p < .001). Most importantly, we detected significant interactions between speed and complexity (p = .022; H3) and speed and valence (p = .004; H4). Slow motion (vs. random category) had a significant and positive effect among GIFs with complexity scores greater than .08 (Figure 4, Panel A) and valence scores greater than 3.93 (Figure 4, Panel B). The interaction term for complexity remained (marginally) significant when controlling for tags (p = .022; Model 2), quality (p = .096; Model 3), or both (p = .091; Model 4). The interaction term for valence remains significant in the model controlling for tags (p = .013), is insignificant in the model controlling for quality and tags (p = .318), and is closer to significance when controlling for quality only (p = .198).
Effect of Speed and Complexity on Composite Liking (Panel A) and Effect of Speed and Valence on Composite Liking (Panel B).
Linear Regression Models.
p < .10.
*p < .05.
**p < .01.
***p <.001.
Notes: Binary variables are effect coded; continuous variables are centered. Unstandardized regression coefficients, standard errors in parentheses.
We analyzed the behavior of users on a GIF-sharing website to provide an externally valid test of our moderators of dynamic complexity and stimulus valence. The results conceptually replicate Studies 2a–c and provide indirect support for the fluency mechanism. Across 514 GIFs, slow motion only increases liking for content high in dynamic complexity (H3). Arguably, because simple dynamic stimuli are already easy to process at regular speed. What is more, slow motion only had a positive effect on liking for pleasant content (H4). Study 3 also shows that slow motion influences key indicators of virality (votes, views, and ratings).
Study 4a: Consumer-Related Moderators—The Role of Processing Style
Studies 4a and 4b examine how slow motion shapes marketing-relevant outcomes such as incentive-compatible choices and WTP. In addition, these studies explore consumer-related moderators of the fluency process to highlight which target audiences are most susceptible to the influence of slow-motion imagery. Study 4a tested the moderator of processing style. Using an affective priming paradigm, we paired brand logos with short-form videos in slow-motion (vs. regular-speed). We then measured consumers’ processing style (more vs. less deliberate). Consistent with H5, we expected that the effect of slow motion on brand liking would be reduced (vs. accentuated) among viewers who engaged in more (vs. less) deliberation.
Design and Procedure
Study 4a manipulated speed (slow motion vs. regular) within-subjects and measured processing style (more vs. less deliberate) as a continuous independent variable. To provide enough power for an attenuated interaction hypothesis, we recruited 403 participants from the United States on MTurk (236 women, 163 men, 4 nonbinary participants; Mage = 41.93 years, SD = 12.84; payment = $.60).
Participants saw the logos of two sports apparel brands (Atalasport and Superga). A pretest confirmed that these logos were considered equally attractive and unfamiliar (Web Appendix K). Next, participants completed an affective priming procedure (Fazio 2001) in which the brand logos (neutral stimuli) were paired with basketball videos (positive stimuli). To manipulate speed within-subjects, we paired one logo with three slow-motion videos and the other logo with three regular-speed videos (see Web Appendix L for the speed pretest). We presented the logo and video (looped) on the same page. After five seconds, the survey moved to the next video. After the three exposures, we measured brand liking (“How much do you like this brand?”; 1 = “not at all,” and 7 = “very much”). This process was repeated for the other logo and speed condition. We counterbalanced which videos played in slow motion, the speed condition, and which brand was paired with the slow-motion videos.
We next measured incentive-compatible choices. Participants learned that we would raffle off gift certificates for the sports apparel brands that they had seen in the study. They then indicated which brand they would want to receive a gift certificate for if they were selected as a winner (1 = “Definitely Atalasport,” and 6 = “Definitely Superga”). We carried out the raffle such that one participant received a gift voucher for the brand of their choosing. Finally, we measured participants’ processing style with a self-constructed scale. On a seven-point Likert scale, participants indicated whether they based their evaluation of the brands on “spontaneous gut feelings” (1) or “deliberate considerations and thoughts” (7). Study 4a also measured two other individual differences (faith in intuition and product involvement), both of which did not moderate the effect of speed on brand liking (see Web Appendix M).
Results and Discussion
Brand liking
We examined whether any of the order variables (order of videos, speed condition, or brand) interacted with our predictors. Because we found no interactive effects (Web Appendix N), we ignore these variables going forward.
We conducted a linear mixed model with one random intercept per participant to account for the repeated measurement. The model included brand liking as the dependent variable and fixed effects for speed (1 = slow motion vs. −1 = regular speed; within-subjects), processing style (z-standardized; between-subjects), and their interaction. Conceptually replicating our previous findings, we detected a main effect of speed. Brand liking was significantly higher when the brand had been paired with slow-motion videos (b = .375, p < .001). There was also a negative main effect of processing style (b = −.161, p = .005). Most importantly, we detected an interaction between speed and processing style (b = −.152, p < .001). Simple slopes analyses revealed that the effect of speed on brand liking was accentuated among participants who had deliberated less (−1 SD; b = .527, p < .001) and attenuated among participants who had deliberated more (+1 SD; b = .222, p <.001). A floodlight analysis suggests that the effect of speed on liking disappeared among those who scored above 5.23 on the processing style scale.
Brand choice
We first recoded this variable such that lower values indicated choice of the brand that was paired with slow-motion videos. A one-sample t-test showed that this variable (M = 2.95, SD = 1.38) was significantly lower than the scale midpoint of 3.5 (t(402) = 8.093, p < .001), which suggests a clear preference for the brand that was paired with slow-motion videos. Indeed, 279 (69.2%) out of 403 participants chose the “slow-motion brand.” Processing style was not associated with brand choice (r = .027, p = .588).
In support of H1, Study 4a shows that pairing a brand logo with slow-motion (vs. regular-speed) videos boosted brand liking and increased incentive-compatible choice. Importantly, this effect was moderated by the viewers’ processing style (H5). Among consumers who engaged in less (vs. more) deliberation while forming judgments, the effect of slow motion on brand liking was accentuated (vs. reduced). This suggests that slow motion is especially suitable for brands when consumers’ motivation, ability, or opportunity to process information is low.
Study 4b: Consumer-Related Moderators—The Role of Consumption Goal
Study 4b examined the consumer-related moderator of consumption goal (H6). Disfluent stimuli can evoke impressions of harmfulness, danger, and riskiness (Song and Schwarz 2009). Consumers who seek out such states should consider a disfluent watching experience appealing because it conveys a desirable sense of riskiness and danger. To test this idea, Study 4b prompted consumers to seek an experience that is exciting (vs. relaxing). We expected that a regular-speed video clip would increase brand interest among consumers who seek excitement (H6). A secondary goal of Study 4b was to provide between-subjects evidence for the effect of speed on evaluations.
Design and Procedure
Study 4b manipulated video speed (slow motion vs. regular) and the consumption goal of consumers (excitement vs. relaxation) in a 2 × 2 between-subjects design. To provide enough power for an attenuated interaction hypothesis, we recruited 600 participants from the United States on MTurk (341 women, 254 men, 5 nonbinary participants; Mage = 40.52 years, SD = 12.69; payment = $.40).
Participants learned that they would watch a promotional video for a hotel brand. Next, they were randomly assigned to one of the consumption goal conditions. Participants were asked to imagine that they are looking for a vacation that is “exciting and thrilling” (excitement condition) or “relaxing and tranquil” (relaxation condition). A validation study confirmed that this manipulation reliably changed participants’ consumption goal (Web Appendix O).
Participants then watched a promotional video showing a skier carving down a mountain. In the exciting condition, the video slogan read, “Experience the thrill of racing down some of the world's steepest skiing slopes,” while in the relaxing condition, the slogan read, “Experience the beauty of serene mountain panoramas.” Participants were randomly assigned to watch this video in one of two speed conditions. To provide between-subjects evidence for the effect of speed but still manipulate speed within-subjects (which is necessary to trigger fluency effects; see Web Appendix C), we varied playback speed within the video, which is a stylistic technique that is frequently utilized in short-form videos. Specifically, in the slow-motion condition, the video played in regular speed for the first 1.5 seconds and then continued to play in slow motion. In the regular-speed condition, the video played in slow motion for the first 1.5 seconds and then continued to play in regular speed. After watching the video, participants indicated how interested they were in staying at the hotel (1 = “not at all,” and 5 = “a great deal”) and how much they would be willing to pay for a one-night stay (slider scale ranging from $50 to $300).
Results and Discussion
Brand interest
A 2 × 2 analysis of variance examined the effects of speed (slow motion vs. regular), consumption goal (excitement vs. relaxation), and their interaction on brand interest. We found a main effect of speed indicating that interest was higher in the slow-motion condition (M = 2.91, SD = 1.12) compared with the regular-speed condition (M = 2.60, SD = 1.08; F(1, 596) = 12.858, p < .001, η2 = .020). There was also a main effect of consumption goal, indicating that interest was higher in the exciting condition (M = 3.03, SD = 1.01) than in the relaxing condition (M = 2.48, SD = 1.13; F(1, 596) = 40.462, p < .001, η2 = .062). Most importantly, the interaction was significant (F(1, 596) = 6.259, p = .013, η2 = .010). Among relaxation-seeking participants, brand interest was higher in the slow-motion condition (M = 2.75, SD = 1.20) than in the regular-speed condition (M = 2.22, SD = 1.00; F(1, 596) = 18.530, p < .001, η2 = .030). Among excitement-seeking participants, however, interest was equal across the speed conditions (F = .588, p = .444).
Willingness to pay
We repeated the previous model with WTP as the dependent variable. There was a main effect of speed, indicating that WTP was higher in the slow-motion condition (M = 152.70, SD = 56.80) versus the regular-speed condition (M = 140.33, SD = 51.68; F(1, 596) = 7.900, p = .005, η2 = .013). We also found a main effect of consumption goal (F(1, 596) = 7.458, p = .007, η2 = .012). Excitement-seeking participants were willing to pay more (M = 152.50, SD = 56.54) than relaxation-seeking participants (M = 140.49, SD = 52.00). The interaction between speed and consumption goal was nonsignificant (F(1, 596) = .044, p = .833, η2 < .001).
Study 4b attests to the moderating role of consumption goal (H6). Among excitement-seeking consumers, regular-speed videos can increase brand evaluations. That is because the watching experience is more disfluent at regular speed, which conveys a sense of riskiness and danger that helps build excitement. Moreover, Study 4b provides between-subjects evidence for the effect of speed on product evaluations by varying the playback speed within one and the same video. Finally, we show that slow motion increases WTP, albeit this effect was not moderated by consumption goal. As a downstream consequence, WTP is somewhat removed from gut-level “liking” because it is also shaped by factors like frugality and income. These unobserved factors may have weakened the influence of our moderator.
Internal Meta-Analysis
To test the overall validity of H1 (slow motion increases liking), we performed a single-paper meta-analysis (Grewal, Puccinelli, and Monroe 2018) on the five studies that included within-subjects manipulations of speed and continuous dependent variables (Studies 1, 2a, 2b, 2c, and 4a). Studies 3 and 4b were excluded because they manipulated speed between subjects. We included the main effect of speed on liking from Study 1 and the effect of speed on liking in the high-complexity (Study 2b) and positive-valence (Study 2c) conditions, respectively. Because Studies 2a and 4a involved mean-centered continuous moderators, we included the main effect of speed on liking (Study 2a) and on brand interest (Study 4a), respectively. The Hedges’ g of the slow motion effect is .38 (z = 2.93, p = .003), indicating a significant effect size. Moreover, Cochran's Q test of heterogeneity shows that the effect sizes across studies are heterogenous (Q(5) = 91.08, p < .001). Finally, we detect no evidence of publication bias (z = .98, p = .327).
General Discussion
Our investigation reveals that there is beauty in slowness. Across seven studies, more than 600 short-form videos, and various judgments (e.g., liking, WTP, choice), we demonstrate that slow motion increases the hedonic component of the viewing experience by making the content subjectively easier to process (i.e., more fluent). Our initial study examines the basic effect of speed (slow motion vs. regular) on liking and provides mediation evidence for the fluency process (Study 1). The remaining studies test boundary conditions that were derived from fluency theory. First, we illustrate the role of two content-related moderators—dynamic complexity and valence—which we test in three lab studies and a large-scale field data set that analyzes the behavior of consumers on a GIF-sharing website. Slow motion only boosts liking when the underlying content involves relatively complex (vs. simple) movements (Studies 2a and 3). Yet, even simple movements can benefit from slow motion when one zooms in on the action (Study 2b). Consistent with a recent extension of fluency theory that suggests that fluency also amplifies experiences, negatively valenced content fails to benefit from slow motion, whereas positively valenced content does (Studies 2c and 3). Second, we highlight the role of two consumer-related moderators—processing style and consumption goal—that inform targeting decisions as to which audiences respond most positively to slow motion. We first show that slow motion is more effective at increasing brand liking among consumers that engage in less deliberate processing (Study 4a). Our final study highlights that among consumers seeking excitement, also relatively disfluent regular-speed videos can boost brand evaluations (Study 4b).
Theoretical Contributions
Short-form videos, such as GIFs, reels, and stories, play an increasingly important role in marketing communications and beyond. However, there is as yet no research that examines how marketers can utilize these videos to boost consumer evaluations. By examining how content creators can use slow motion, we shed light on a new medium and make important theoretical contributions to the following fields of research.
Our work provides a deeper understanding of how slow motion shapes viewing experiences. Most previous work has studied how slow motion influences cognitive inferences (e.g., Caruso, Burns, and Converse 2016; Yin, Jia, and Zheng 2021). The aesthetic consequences of slow motion, however, are less understood. By linking slow motion to an important driver of aesthetic pleasure—fluency—we fill this theoretical gap. The fluency account and our evidence in Studies 2c and 3 theoretically and empirically reconcile conflicting previous findings in the slow-motion literature (Barnett and Grabe 2000; Wöllner, Hammerschmidt, and Albrecht 2018). Fluency theory also allows us to develop novel moderators that offer practical insights on how to effectively implement slow-motion imagery. Lastly, we expand beyond research on ads, movie scenes, and sports replays.
This article is the first investigation of the role processing fluency plays in the consumption of dynamic visual content. Fluency research has traditionally focused on linguistic stimuli (e.g., text) and static visual stimuli (e.g., pictures) such as product labels (Gmuer, Siegrist, and Dohle 2014), brand logos (Janiszewski and Meyvis 2001), print advertisements (Leonhardt, Catlin, and Pirouz 2015), and stock names (Alter and Oppenheimer 2016). Consequently, there is a rich base of evidence demonstrating how text and images can be presented fluently. Our work thereby lays a foundation for future research to extend the theory of processing fluency to the context of dynamic visual content.
Lastly, we contribute to the field of empirical aesthetics by introducing a new construct: dynamic complexity. In static images, objective complexity is reflected in the amount of variation and unpredictability at the level of the color values of individual pixels (Donderi 2006). The size of the compressed image file reflects the aforementioned characteristics and is thus frequently used to measure visual complexity in diverse marketing contexts (e.g., Landwehr, Labroo, and Herrmann 2011; Pieters, Wedel, and Batra 2010). However, this measure is not applicable to dynamic stimuli because it confounds the complexity of the scene (e.g., the number of objects, patterns) and the complexity of the dynamic movement over time. Our work disentangles these sources of complexity theoretically and empirically and provides a first methodological tool that allows researchers to quantify an important property of dynamic visual stimuli.
Managerial Implications
Our inquiry provides insights for anyone who wants to use slow motion by informing two important strategic decisions. We highlight which type of underlying content benefits more from slow motion, and we identify which types of consumers respond more positively to slow motion.
Target content
First, to benefit from a slower presentation, the underlying content needs to depict movements that are sufficiently complex. Marketers should avoid slowing content that involves simple or subtle movements, lest they risk boring consumers. Yet, by applying a second visual tool, this limitation can be overcome. Even a simple scene can appear complex if content creators zoom in on subtle movements. For instance, slow-motion footage of a chef slicing a steak might be boring, but a close-up of the knife cutting through the meat in slow motion should be aesthetically pleasing. Second, we show that slow motion creates opposite effects for aversive content because fluency also amplifies evaluative judgments. Marketers can thus accentuate aversive videos by using slow motion. Drunk driving prevention campaigns should become more effective when footage of the car crash and people getting injured is slowed down. Similarly, charities may induce stronger negative feelings when showing natural disasters and destroyed homes in slow motion rather than regular speed.
Target consumers
Fluency predominantly shapes intuitive responses to products. Slow-motion tactics are thus more effective when consumers’ motivation, ability, or opportunity to elaborate is low. Marketers should therefore reserve slow-motion effects for consumers that form gut-level product judgments. In that way, slow motion should be more successful in boosting product liking and choice among consumers who are relatively less knowledgeable in a product domain. Similarly, slow motion should also be more effective when promoting low-involvement (e.g., USB sticks) rather than high-involvement (e.g., laptops) products. Finally, we show that slow motion is less effective among target audiences who seek out affective states that are associated with disfluency, such as feelings of riskiness, novelty, and danger. Among consumers who seek excitement, or consumers who are generally risk- and sensation-seeking, faster-than-normal playback speeds should be particularly effective at increasing consumers’ evaluations because the subjective experience of disfluency is aligned with their consumption goal. In summary, marketers who cater to audiences that seek out brands or experiences that appear adventurous, thrilling, or exciting should accelerate playback speeds.
Limitations and Future Research
Our work reveals several directions for future research. First, the field should examine additional moderators around fluency effects induced by presentation speed. We find no support for the idea that slow motion is more effective for hedonic (vs. utilitarian) products or consumers scoring high (vs. low) in faith in intuition or low (vs. high) in product involvement (see Web Appendices M and P). Second, we examine the effect of presentation speed on aesthetic liking for visual stimuli. It would be interesting to explore whether speed unfolds similar effects on other sensory modalities, such as sound. Visual stimuli are often more complex than acoustic stimuli. Slow motion might therefore help visual stimuli (e.g., a basketball video) but harm comparable acoustic stimuli (e.g., a basketball radio broadcast). Finally, dynamic stimuli are a blind spot within the field of empirical aesthetics. We hope that our findings inspire future research to operationalize other properties of movement. For instance, it would be promising to develop objective measures of movement symmetry and prototypicality. Linking such constructs to fluency and aesthetic liking would further our understanding of what makes movement beautiful.
Supplemental Material
sj-pdf-1-mrj-10.1177_00222437231179187 - Supplemental material for The Art of Slowness: Slow Motion Enhances Consumer Evaluations by Increasing Processing Fluency
Supplemental material, sj-pdf-1-mrj-10.1177_00222437231179187 for The Art of Slowness: Slow Motion Enhances Consumer Evaluations by Increasing Processing Fluency by Anika Stuppy, Jan R. Landwehr and A. Peter McGraw in Journal of Marketing Research
Footnotes
Appendix
Table A1 contains means and standard deviations per cell across all studies reported in the main text. For Studies 2a, 3, and 4a, we report estimated means and their 95% confidence intervals one standard deviation above/below the mean of the respective continuous moderator.
Coeditor
Maureen Morrin
Associate Editor
Dhruv Grewal
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
