Speed is associated with polarization during subjective evaluation: no tradeoff, but an effect of the ease of processing

Abstract

In human perceptual decision-making, the speed-accuracy tradeoff establishes a causal link between urgency and reduced accuracy. Less is known about how speed relates to the subjective evaluation of visual images. Here, we conducted a set of four experiments to tease apart two alternative hypotheses for the relation between speed and subjective evaluation. The hypothesis of “Speed-Polarization Tradeoff” implies that urgency causes more polarized evaluations. In contrast, the “Ease-of-Processing” hypothesis suggests that any association between speed and polarization is due to the salience of evaluation-relevant image content. The more salient the content, the easier to process, and therefore the faster and more extreme the evaluation. In each experiment, we asked participants to evaluate images on a continuous scale from − 10 to + 10 and measured their response times; in Experiments 1–3, the participants rated real-world images in terms of morality (from “very immoral,” -10, to “very moral,” +10); in Experiment 4, the participants rated food images in terms of appetitiveness (from “very disgusting,” -10, to “very attractive,” +10). In Experiments 1, 3, and 4, we used a cueing procedure to inform the participants on a trial-by-trial basis whether they could make a self-paced (SP) evaluation or whether they had to perform a time-limited (TL) evaluation within 2 s. In Experiment 2, we asked participants to rate the easiness of their SP moral evaluations. Compared to the SP conditions, the responses in the TL condition were consistently much faster, indicating that our urgency manipulation was successful. However, comparing the SP versus TL conditions, we found no significant differences in any of the evaluations. Yet, the reported ease of processing of moral evaluation covaried strongly with both the response speed and the polarization of evaluation. The overall pattern of data indicated that, while speed is associated with polarization, urgency does not cause participants to make more extreme evaluations. Instead, the association between speed and polarization reflects the ease of processing. Images that are easy to evaluate evoke faster and more extreme scores than images for which the interpretation is uncertain.

Introduction

In the study of decision-making, the well-known speed-accuracy tradeoff is arguably one of the most fundamental phenomena, connecting an increase in processing time to enhanced performance (Vargas and Lauwereyns 2021; Heitz 2014; Wickelgren 1977; Luce 1991). This is true not only for humans across a wide variety of task domains (Fiedler et al. 2021; Onagawa et al., 2019; Yamaguchi et al. 2013; Link and Tindall 1971), but even for bees (Chittka et al. 2003) and ants (Franks et al. 2009). In perception and memory tasks as well as economical choices (Kahneman 2003), it has been shown that decision-makers can adapt to time pressure, but also strategically set the amount of time and effort they invest in processing information for decision-making by weighing the cost of time versus the gain in accuracy (Boureau et al. 2015; Gluth et al. 2012; Bogacz et al., 2006; Reddi and Carpenter 2000).

When measuring the accuracy of performance based on perceptual discrimination or memory, researchers rely on objective information that can readily be verified. However, we often make decisions or judgments based on subjective evaluations (Xu et al. 2023; Ounjai et al. 2018; Margulis 2017; Kahneman 2003). Examples include decisions or judgments based on the perceived attractiveness of faces, foods, music, or works of art, to name the most commonly researched categories (Xu et al. 2020; Ounjai et al. 2018; Margulis 2017; Shimojo et al., 2003). Another type of subjective decision is based on moral evaluation, which may be shaped by cultural background and personal experience, with a complexity of processing that does not afford an easy metric of accuracy (Sudo et al. 2021; Malle 2021; Garrigan et al. 2016). For these diverse types of subjective decision-making, little is known about the relationship between speed and decision performance. The present paper investigates this relationship.

The role of speed in subjective evaluation

As a metric of performance in subjective decision-making, the nearest equivalent to accuracy would arguably be the consistency of response, measured across a population of decision-makers (e.g., the level of agreement or consensus) or measured across a sample of decisions for the same items under different conditions (e.g., the level of reproducibility) (Hornsby and Love 2020; Kahneman 2003). Evaluations would not be random, and even preference reversal would follow an internal logic that should give reproducible responses (Sénémeaud et al. 2013; Lichtenstein and Slovic 2006; Camerer and Weber 1992).

The literature on value-based decision-making suggests that evaluations are based on a sequential sampling process that retrieves information from memory (Shadlen and Shohamy 2016; Krajbich et al. 2010; Rangel et al. 2008; Weber and Johnson 2006; Busemeyer and Townsend 1993). Within this framework, default expectations might be that time constraints introduce random noise in the evaluation, leading to over- or underestimation of a “true” value. Extra time for information processing should give decision-makers the opportunity to form more precise and more stable object representations of the items under consideration (Curley et al. 2017; Wilson and Schooler 1991). This would lead to a speed-precision tradeoff, with more consistency in the evaluation of slower responses (i.e., a narrower response distribution, with a smaller standard deviation) but not necessarily a change in the mean of the evaluations.

However, given that subjective decision-making involves a complex system with parallel and interactive streams of affective processing, it is possible that speed affects the mean of evaluations. Subjective decision-making includes fast, intuitive emotional reactions and slow, controlled evaluations based on reasoning (Malle 2021; Bailey and Muldrow 2018; Cameron et al. 2017; Garrigan et al. 2016; Sokol-Hessner et al. 2012; Dane and Pratt 2007). Thus, a key question arises: Does time pressure lead to a bias towards emotionally charged information? If more polarizing information would be sampled first, then biased conclusions may occur when the process is halted early under time pressure (Lin and Jia 2023; Kruglanski and Gigerenzer 2018; Wittmann and Paulus 2008; Dror et al. 1999).

Preliminary evidence in support of an association between speed and polarization has been obtained in our laboratory, with faster responses for food images that receive highly positive or negative ratings than for food images that receive moderate ratings (Wolf et al. 2018, 2019). Also, when participants were given predictive cues indicating the valence of impending food images, we found that the responses were faster and the ratings more extreme when participants were given reliable predictions as compared to neutral or invalid predictions (Ounjai et al. 2018). Critically, however, our previous studies did not directly manipulate speed. Here, we designed a set of four experiments focusing on the role of speed in subjective evaluation by comparing speeded versus non-speeded conditions.

The present study

In the present study, we raise a hypothesis of “Speed-Polarization Tradeoff,” proposing an active, causal link such that faster responses would lead to more extreme evaluations. Particularly, time pressure would act as a stressor that reinforces the emotional charge and biases the evaluation toward fast, intuitive emotional reactions (Lin and Jia 2023; Xu et al. 2020; Bird et al. 2019; Starcke and Brand 2016; Cyders 2008; Maule et al. 2000; Dror et al. 1999). Consequently, urgency would lead to polarization in the evaluation.

The hypothesis of a speed-polarization tradeoff can be contrasted against a hypothesis of “Ease of Processing.” By this account, any connection between speed and polarization would not reflect an active, causal link that can be set strategically by the participant. Instead, the salience of evaluation-relevant content might influence the ease of processing such that images with salient content produce faster responses and more extreme ratings than images whose content is more complex or ambiguous (Andrejević et al. 2022; Quandt et al. 2022; Berberyan et al. 2021; Isham 2020; Polanía et al. 2019). Urgency, then, would have no effect on the polarization of the evaluation. Instead, the relation between speed and polarization would be governed by the ease of processing.

The introduction of an urgency condition also allowed us to examine the relation between speed and the variability of response across participants. A speed-precision tradeoff should imply that decisions under urgency lead to an increase in variation. However, increased polarization might reduce the variability. If so, a speed-precision tradeoff could be counteracted by any association between speed and polarization, whether it be due to an active speed-polarization tradeoff or the ease of processing.

In each experiment, we asked participants to perform subjective decision-making tasks with visual materials, always judging individual images on a continuous scale from − 10 to + 10. In Experiments 1 to 3, the visual materials were photos of real-world scenes with a moral evaluation task (similar to Experiment 3 in Sudo et al. 2021). In Experiment 4, the participants were asked to judge food images as a function of preference. In Experiments 1, 3, and 4, we manipulated the urgency of the evaluation task on a trial-by-trial basis, using cues to indicate whether the participants could make a self-paced (SP) evaluation or whether they had to perform a time-limited (TL) evaluation within 2 s. In Experiment 2, we did not manipulate urgency, but obtained the participants’ self-reported perceptions on the ease of processing for each of their self-paced moral evaluations.

Experiments

Experiment 1: real-world images, with urgency manipulation

Introduction

In Experiment 1, as an exploratory experiment, we set out to examine the relation between speed and subjective evaluation with moral evaluation of real-world images. Moral evaluation was the appropriate task for present purposes since its underlying mechanisms have been characterized in the greatest detail, with solid evidence for the existence of fast, emotional processes versus slow, deliberative processes (Malle 2021; Cameron et al. 2017; Garrigan et al. 2016; Van Bavel et al. 2015; Decety and Cacioppo 2012).

We designed the present experimental paradigm to compare the moral evaluation of real-world images under self-paced viewing versus under time pressure (See Fig. 1). We prepared a stimulus set ranging from “very moral” to “very immoral” images based on the Socio-Moral Image Database (SMID) (Crone et al. 2018). We asked participants to rate real-world images on a continuous scale from − 10 (“very immoral”) to + 10 (“very moral”). To create conditions with and without urgency, we used a cueing procedure (Reuss et al. 2015) to instruct the participants at the beginning of each trial whether they could take as long as they wanted to make the evaluation (Self-Paced condition, SP; left panel of Fig. 1) or whether they had to make the evaluation within 2 s (time-limited condition; TL, right panel of Fig. 1). In the TL condition, the trial was aborted, and an error message was given when the participant took too long to respond.

Fig. 1.

Schematic representation of the paradigm in Experiment 1. The trial structure for the self-paced condition is shown on the left; the time-limited condition is on the right. Both conditions started with a fixation cross for 1 s, followed by a time-instruction cue for 1.5 s. In self-paced trials, the participants were allowed to view the images as long as they liked before indicating their evaluation. In the time-limited trials, the participants were required to complete their evaluation within 2 s; if they failed to do so, an error message was presented. Both conditions ended with a random ITI between 1.6 s and 2 s

Figure 2 presents the alternative predictions from two hypotheses. According to the hypothesis of speed-polarization tradeoff (left panel), speed causes the moral evaluations to be more polarized, and so the ratings under time pressure in the TL condition would be more extreme (i.e., more positive ratings for moral images and more negative ratings for immoral images) than the ratings without time pressure in the SP condition. In contrast, the right panel shows an alternative, ease-of-processing hypothesis, by which there would be no differences in the mean ratings between the SP and TL conditions. According to the ease-of-processing hypothesis, speed does not cause polarization. Instead, fast responses may be correlated with extreme ratings as a function of the ease of processing of moral images. Some images would have a very salient moral content, enabling a fast response with an extreme rating, whereas other images would have a less salient moral content, requiring more time to process and tending to produce more moderate ratings.

Fig. 2.

Schematic representation of the alternative predictions from two hypotheses in Experiment 1. Hypothesis I, Speed-Polarization Tradeoff, is shown on the left; hypothesis II, Ease of Processing, is shown on the right. The hypothesis of speed-polarization tradeoff predicted that the evaluation scores would be more extreme in the time-limited condition than in the self-paced condition. The ease-of-processing hypothesis implied that the moral evaluations would remain unchanged even under urgency

Methods

Participants

Experiment 1 was conducted as an exploratory study with a sample size comparable to previous studies in our laboratory (Wolf et al. 2018, 2019; Ounjai et al. 2018). Thirty-two students at Kyushu University were recruited as participants; one participant chose to discontinue the experiment midway due to physical discomfort, while another had to be excluded because she had participated in a different experiment using similar images. Thus, the current dataset consisted of data from 30 participants: 15 males and 15 females, with a mean age of 22.67 years old, and a standard deviation of 1.81. The participants were asked to use their right hand to manipulate the joystick during the experiment though 2 participants were left-handed.

For all experiments of the present study, all participants had normal or corrected-to-normal vision. No participant reported any health issues or past or present psychological disorders. The Human Ethics Committee of the Faculty of Arts and Science at Kyushu University approved the study. Informed consent was obtained in writing from each participant before the experiment; the informed consent included a no-risk statement and rights protection. All experiments in the present study were conducted with different groups of participants, and no participant had experience with any similar experiment. Each participant received 1000 yen as monetary compensation for their participation. All experiments consisted of a single session of approximately 1 h.

Apparatus and stimuli

The visual stimuli were presented in a normally lit room on a 23.8-inch full high-definition flat-panel monitor with a 1920 × 1080 pixels display resolution. The participants were seated in front of the monitor at a comfortable distance of their own choosing (with no head or chin restraint). The manual responses were recorded using a joystick (Logitech, Switzerland; model no. 963290-0403).

All events and recordings were controlled through code written in PsychoPy version 1.90.3 (Peirce 2007, 2008). The target visual stimuli consisted of 160 real-world images selected from the SMID. The images provided content ranging from very moral (i.e., morally acceptable) to very immoral (i.e., morally offensive), depicting war scenes, weapons, accidents, landscapes, and various types of social interaction. Based on the original categorization in the moral domain of SMID, we preselected a balanced set of moral and immoral images. For confirmation, we conducted a preparatory rating experiment with 10 students in the lab, asking them to rate the preselected images with respect to their moral content from − 10 (extremely immoral) to + 10 (extremely moral). We were able to select a set with 80 moral images and 80 immoral images. During the experiment, each target stimulus was presented against a black background together with an evaluation scale underneath the image. The stimulus was scaled down by one-third and centered in the middle of the upper two-thirds of the screen; the evaluation scale was centered in the middle of the lower third of the screen.

Time-instruction cues were used to indicate on a trial-by-trial basis whether the participant should perform the moral evaluation at their own pace or within 2 s (see Fig. 1). The time-instruction cues were two monochrome icons: an icon of beach landscape with a palm tree for the self-paced (SP) condition; an icon of a sand clock with a short low beep sound for the time-limited (TL) condition. The time-instruction cues were presented centered in the middle of the upper third of the screen, with their sizes fixed at 376 × 376 pixels.

Procedure

The experimental session was divided into 4 blocks of 40 trials, with breaks between the blocks. Participants were asked to give their informed consent in writing and fill out a pre-questionnaire prior to the experiment. Before the actual data collection, participants were given detailed task instructions and presented with 20 practice trials to familiarize themselves with the time instructions and the operation of the joystick.

All trials started with a 1-s fixation screen, in which a fixation cross was centered at the middle of the top third of the display (see Fig. 1). Then, a time-instruction cue was presented, centered at the same position as the fixation cross, for a period of 1.5 s, to indicate whether it was an SP trial or a TL trial. In the case of a TL cue, the visual presentation was accompanied by a low beep sound to increase the sense of urgency. Following the 1.5 s presentation of the time-instruction cue, the target image was displayed together with the evaluation scale until the participant had completed their moral evaluation, or for a maximal duration of 2 s in a TL trial. Participants were asked to rate the depicted content in each image on a continuous scale from − 10 (“very immoral”) to + 10 (“very moral”) by bending the joystick in the corresponding direction and clicking the trigger on the joystick to mark the evaluation. The registered values were continuous, up to eight digits after the decimal point. The starting point for the evaluation was 0 and then the participants had to bend the joystick to the left or right.

The moral evaluation task was explained at the beginning of the experiment, but not shown in words during the trials; instead, only the evaluation scale was shown in each trial. The participants were instructed to perform their ratings by asking themselves for each image “How morally acceptable is this?” An evaluation of + 10 should reflect the most morally acceptable case; conversely, a value of -10 should reflect the complete opposite, that is, the least morally acceptable case. In a TL trial, in case the participant missed the 2 s deadline, a high beep was presented together with the message “Time out!!” displayed in red, centered at the middle of the upper third of the screen. At the end of each trial, there was an inter-trial-interval (ITI) of between 1.6 and 2 s, with a blank screen.

An experimental session consisted of 160 trials, divided into 4 conditions, with 2 levels of time control (SP, TL) and 2 levels of morality of the image content (moral, immoral). No image was presented more than once. All trials were presented in pseudorandom order and each block contained an equal number of trials for each condition. The assignment of images to SP versus TL conditions was counterbalanced across participants.

Results

The present data set included data from 30 participants. All participants completed all trials in the SP condition; on average, participants completed 94.54% of the trials in the TL condition (standard deviation, S.D. = 4.76). In the remaining 5.46% of the TL trials, the participants failed to respond within the deadline of 2 s. No participant missed more than 20% of the TL trials. The number of completed trials provided sufficient data for the present analyses.

We focused on the evaluation scores and response times as dependent measures and performed all statistical analyses using the JASP software package (JASP Team 2024). Throughout the study, α level was set at 0.05 and post-hoc testing was conducted with Bonferroni correction. The evaluation scores were defined as the ratings on a scale from − 10 (“very immoral”) to + 10 (“very moral”), as confirmed by the participants when clicking the trigger on the joystick. To be able to compare the evaluation scores for different image categories in terms of their polarization (i.e., distance from zero), we used unsigned evaluation scores (i.e., absolute values) in all analyses of the present study. The response times were defined as the time from the onset of the target image to the click on the joystick.

Analysis of average evaluation scores

Figure 3 presents the means and 95% confidence intervals for the evaluation scores (left panel) and for the response times (right panel) as a function of the morality of the image content and the time control condition. This analysis was based on the averaged data for each participant in each condition (N = 30). We conducted a two-way repeated-measures ANOVA using morality of the image content (i.e., Moral versus Immoral) and time control (i.e., SP versus TL) as within-subjects factors.

Fig. 3.

Average evaluation scores (left panel) and response times (right panel) of Experiment 1 for each participant in each condition (N = 30). The blue bars show the data with moral images; the orange bars represent the data with immoral images. The bars from the time-limited conditions are shown in a darker shade than the bars from the self-paced conditions. The error bars reflect the 95% confidence interval around the mean

There were no significant main effects for time control (F (1, 29) = 0.482, p = 0.493, η²_p = 0.016) or morality of the image content (F (1, 29) = 0.516, p = 0.478, η²_p = 0.017), but we obtained a statistically significant interaction between the two factors (F (1, 29) = 6.600, p = 0.016, η²_p = 0.185). The effect of time control for moral images (5.428 for SP versus 5.244 for TL) appeared to go in the opposite direction as compared to that for immoral images (5.345 for SP versus 5.682 for TL), but post-hoc comparisons indicated there was no significant difference in any pair of evaluation scores.

We also conducted a two-way repeated-measures ANOVA based on the response times, using morality of the image content (moral versus immoral) and time control (self-paced versus time-limited) as within-subjects factors. The results showed a significant main effect for time control (F (1, 29) = 62.065, p < 0.001, η²_p = 0.682), but not for morality of the image content (F (1, 29) = 3.206, p = 0.084, η²_p = 0.100), nor was there a statistically significant interaction between the two factors (F (1, 29) = 3.975, p = 0.056, η²_p = 0.121). Post-hoc pairwise comparisons indicated that the response times following the SP cue for moral images (2.677 s) and for immoral images (2.909 s) were both significantly longer than following the TL cue (1.240 s for moral images and 1.236 for immoral images) with p values < 0.001. It proves that our time control manipulation successfully induced participants to speed up their responses when given a deadline of 2 s. Importantly, despite this solid increase in response speed, the participants’ average evaluation scores did not change.

Median-split analysis

In previous research on preference formation with respect to food, we observed a robust association between response speed and polarized evaluation, such that extreme evaluation scores were consistently associated with fast response times (Wolf et al. 2018, 2019; Ounjai et al. 2018). To examine whether such an association between speed and polarization also exists for the moral evaluations in the present paradigm, we performed a median-split analysis, separating the fast responses from the slow responses for each participant in each condition. Figure 4 presents the averages and confidence intervals of the evaluation scores, comparing fast versus slow responses, for each of the 4 conditions.

Fig. 4.

Median-split comparisons between evaluation scores with fast responses versus slow responses in Experiment 1, separately for each condition in the 2 × 2 design with the factors time control (self-paced and time-limited) and morality (moral and immoral), based on the averaged data for each participant (N = 30). The error bars reflect the 95% confidence interval around the mean

We conducted a three-way repeated-measures ANOVA based on the averaged evaluation scores for each participant (N = 30), using morality of the image content (moral versus immoral), time control (self-paced versus time-limited), and response speed (fast versus slow) as within-subjects factors.

The results indicated that only response speed showed a significant main effect (F (1, 29) = 39.998, p < 0.001, η²_p = 0.580), whereas the F values for the other main effects were less than 1. The analysis showed a significant interaction between response speed and time control (F (1, 29) = 26.217, p < 0.001, η²_p = 0.475). Post-hoc analysis revealed that response speed had a significant effect on evaluation in the SP conditions by making them more extreme while the effect was not present in the TL conditions.

Analysis of the level of variation

The hypothesis of a speed-precision tradeoff implies that the evaluations should become less precise (i.e., more variable) in a speeded condition. To test this hypothesis, we computed the standard deviation of the evaluation scores for each participant in each condition. These participant-based standard deviations were then analyzed in a two-way repeated-measures ANOVA, using morality of the image content (moral versus immoral) and time control (self-paced versus time-limited) as within-subjects factors.

There were significant main effects on the participants’ standard deviations due to time control (F (1, 29) = 13.283, p < 0.001, η²_p = 0.314) as well as morality of the image content (F (1, 29) = 12.633, p < 0.001, η²_p = 0.303). The interaction between the two factors was not significant (F (1, 29) = 1.421, p = 0.243, η²_p = 0.047). In contrast to the hypothesis of a speed-precision tradeoff, the standard deviations were smaller in the time-limited condition (average SD = 2.119) than in the self-paced condition (average SD = 2.419). Images with moral content (2.103) also generated smaller standard deviations than images with immoral content (2.435).

Correlation analysis

To gain further insights into the present data patterns as a function of response speed and the time control conditions, we performed item-based analyses, computing for each image (N = 160) the average evaluation score and response time in the SP condition versus the TL condition. Figure 5 presents the scatter plot with histograms and a linear regression line (with a 95% confidence interval) for the item-based correlation between evaluation scores in the SP condition versus the TL condition. Pearson’s r value was 0.694 (R² = 0.481; p < 0.001), indicating very strong evidence supporting a positive correlation.

Fig. 5.

Item-based scatter plot with histograms for the correlation analysis (N = 160) between the evaluation scores in the self-paced versus time-limited conditions in Experiment 1. The linear regression line is indicated with 95% confidence interval

Figure 6 presents the scatter plots with histograms and linear regression lines (with a 95% confidence interval) for the item-based correlations between the evaluation scores and the response times in the SP condition (left panel) and in the TL condition (right panel). For the SP condition, Pearson’s r value was − 0.523 (R² = 0.274; p < 0.001); for the TL condition, Pearson’s r value was − 0.291 (R² = 0.085; p < 0.001). Thus, in both conditions, items that elicited more polarized evaluation scores were processed faster.

Fig. 6.

Item-based scatter plots with histograms for the correlation analysis (N = 160) between the evaluation scores and response times in Experiment 1. The left panel shows the data in the self-paced condition; the right panel shows the data in the time-limited condition. The linear regression lines are indicated with their 95% confidence intervals

Discussion

In Experiment 1, we set out to examine the role of speed in subjective evaluation using real-world images ranging from “very moral” to “very immoral” taken from SMID, and asked participants to give ratings to reflect their evaluation on a continuous scale from − 10 to + 10, either with or without time pressure.

We found that the response speed was much faster in the time-limited condition than in the self-paced condition, proving that our manipulation of urgency was successful. However, the urgency did not affect the means of moral evaluations, with ratings for moral and immoral images that were unchanged in the time-limited versus the self-paced condition. Within conditions, however, a median-split analysis separating fast from slow responses showed that fast responses were associated with more extreme ratings than slow responses only in the SP condition. Thus, while urgency produced increased speed in responding, it also abolished the difference between mean evaluations for fast versus slow responses within the TL condition.

In line with the latter finding, the standard deviations of the moral evaluations tended to be larger in the self-paced condition than in the time-limited condition. This observation appears to run against the hypothesis of a speed-precision tradeoff. However, our measures of precision are not based on within-participant measures of consistency for the same stimuli. In the present paradigm, we opted to avoid repeated presentations, concerned that stimulus repetition might alter the evaluation process (introducing priming and/or familiarity effects). As a consequence, we were not able to measure within-person consistency in evaluation. To study within-person consistency would require a careful research design to control for effects of repetition, order, and time control, adding a level of complexity to be addressed in future investigations.

With the present paradigm, we were able to conduct an item-based analysis, examining the consistency in evaluation for the same images among different participants. The item-based analysis of the data underscored that the ratings given to the images were highly consistent across the time-limited and self-paced conditions, and that extreme ratings (“very moral” and “very immoral”) were associated with faster responses than more moderate ratings.

An additional caveat with the present study is that the joystick we used has an asymmetric handle design and weight distribution that likely implies an uneven resistance of movement across the range, with the largest resistance for initial movements. This may make moderate evaluations more motorically challenging, causing increased response times and thus introducing a confounding factor.

Taken together, the findings in Experiment 1 were compatible with the ease-of-processing hypothesis. However, we did not actually assess the ease of processing of the evaluations. Experiment 2 was designed to provide more direct evidence in support of the ease-of-processing hypothesis.

Experiment 2: real-world images, with easiness rating

Introduction

The data pattern in Experiment 1 suggested that, while speed is associated with polarization of evaluation in self-paced responses, this is not due to an active tradeoff. Following on from the exploratory investigation in Experiment 1, we pre-registered Experiments 2, 3 and 4 of the present study as confirmatory research at OSF (https://osf.io/8fjrk). In Experiment 2, we aimed to examine the ease-of-processing hypothesis by effectively tracking the participants’ self-reported easiness of the moral evaluations on a trial-by-trial basis, under self-paced viewing.

Additionally, we changed the image pool based on the SMID to create a more continuous distribution, including the middle range, and categorizing the images as “moral,” “moderate,” or “immoral.” As in Experiment 1, we asked participants to rate real-world images on a continuous scale from − 10 (“very immoral”) to + 10 (“very moral”). In every trial, immediately after performing the moral evaluation, the participants had to rate the easiness of that evaluation. To avoid confusion between the two kinds of assessments, we used horizontal response mapping for the moral evaluation and vertical response mapping for the easiness rating.

Methods

Participants

Experiments 2, 3, and 4 were designed and pre-registered in parallel following on from the exploratory Experiment (1) Before data collection, a power analysis using G*Power version 3.1 (Faul et al. 2007) was calculated to determine the appropriate sample size for a 2 × 3 repeated-measures within-factors ANOVA of 1 group with 2 measurements to achieve 95% power (1- β error probability), detect a medium effect size Cohen’s f of 0.25, and maintain a standard α error probability of 0.05, with the effect size specification as in G*Power 3.0. The indicated sample size should be 54 but was erroneously recorded as 36 due to a human error in calculation. A sensitivity analysis with the intended parameters for a total sample size of 36 indicated there was adequate power to detect a Cohen f of at least 0.309. The sample size of 36 was applied for each of the three pre-registered experiments. Accordingly, 18 males and 18 females with a mean age of 21.22 years old and a standard deviation of 3.42 were recruited at Kyushu University as participants and completed experiment (2) One participant was left-handed.

Apparatus and stimuli

The experimental set-up was the same as in Experiment 1, except that we used an upgraded version of PsychoPy 3 (version 2021.2.3) and added an eye tracker (GazePoint GP3) and a biometric kit (with the function of recording GSR and heart rate) to record the participant’s eye movements and biometric data (not analyzed in the present study). Participants could respond by joystick in the horizontal and vertical direction to give a corresponding score on the left-right and up-down scale showing on the screen. A chinrest with a forehead bar was used to fix the head location at 60 cm from the eye tracker and at 100 cm from the monitor. Participants were able to adjust the chair height for a comfortable posture to avoid movement during the experiment.

The visual stimuli used in this study comprised a total of 180 real-world images consisting for more than two-thirds of images used in Experiment 1, while replacing and adding images carefully selected from the SMID to ensure well-balanced moral, neutral, and immoral categories, encompassing a wide range of content.

Procedure

The experimental session was divided into 4 blocks of 45 trials, with stimuli randomly presented in each trial and breaks between the blocks. Before the actual data collection, participants were given detailed task instructions and presented with 10 practice trials to familiarize themselves with the operation of the joystick.

All trials started with a 1-s fixation screen, in which a fixation cross was centered at the middle of the top third of the display. A 0.5 s blank screen was shown before the moral evaluation of the image. In the moral evaluation frame, each target stimulus was presented against a black background together with an evaluation scale underneath the image. All images were adjusted keeping the original proportions to the same height at 600 pixels with 96 dpi for horizontal and vertical resolution. The stimulus was centered in the middle of the upper two-thirds of the screen; the evaluation scale was centered in the middle of the lower third of the screen. Participants were asked to rate the depicted content in each image on a continuous left-right scale. The response mapping for the moral evaluation was counterbalanced across participants: Either from − 10 (“very immoral”) at the left to + 10 (“very moral”) at the right, or opposite, from + 10 (“very moral”) at the left to -10 (“very immoral”) at the right. The participants had to bend the joystick in the direction of their evaluation and click the trigger on the joystick to mark the evaluation.

The moral evaluation task was explained at the beginning of the experiment, but not shown in words during the trials; instead, only the moral evaluation scale with the labels of the two polar ends was shown in each trial. As in Experiment 1, the participants were instructed to perform the moral evaluation by asking themselves for each image “How morally acceptable is this?” After the moral evaluation, the participants were asked to rate the easiness of the moral evaluation on an up-down scale. For this judgment, the participants were instructed to ask themselves for each image “How easy was it to give the moral evaluation for this?” The response mapping for the easiness rating was also counterbalanced across participants: Either from 0 (“very difficult”) at the bottom to 100 (“very easy”) at the top, or opposite, from 100 (“very easy”) at the bottom to 0 (“very difficult”) at the top. The easiness rating marker started from 50, the middle of the scale. The question “How easy was it to rate the morality?” was displayed above the scaler in each trial. At the end of each trial, there was an inter-trial interval (ITI) of between 1.5 and 2.0 s, with a blank screen.

Results

Preliminary analysis to check the counterbalancing showed there were no significant group differences in any of the dependent measures, indicating that the scale directions did not impact on the participants’ performance. To gain insights into the present data patterns as a function of response speed, we performed item-based analyses, computing for each image (N = 180) the average moral evaluation, the easiness rating, and the response time of the moral evaluation across participants.

Figure 7 presents the scatter plot with histograms and linear regression line (with a 95% confidence interval) for the correlation between the moral evaluations and the response times. The Pearson’s r value was − 0.543, with R² = 0.295 (p < 0.001), confirming that the responses tended to be faster for items that elicited more polarized evaluation scores.

Fig. 7.

Item-based scatter plot with histograms for the correlation analysis (N = 720) between the moral evaluation scores and response times in Experiment 2, with linear regression line and its 95% confidence interval

Figure 8 presents the scatter plot with histograms and linear regression line (with a 95% confidence interval) for the correlation between the moral evaluations and the easiness ratings. The Pearson’s r value was 0.878, with R² = 0.771 (p < 0.001), indicating that items that elicited more polarized evaluation scores were strongly associated with higher easiness ratings.

Fig. 8.

Item-based scatter plot with histograms for the correlation analysis (N = 720) between the moral evaluation scores and the easiness rating in Experiment 2, with linear regression line and its 95% confidence interval

Figure 9 presents the scatter plot with histograms and linear regression (with a 95% confidence interval) for the correlation between response times and easiness ratings (Pearson’s r = -0.678, R² = 0.460, p < 0.001). The easier the images were to rate, the faster were the response times.

Fig. 9.

Item-based scatter plot with histograms for the correlation analysis (N = 720) between the easiness rating and response times of moral evaluation in Experiment 2, with linear regression line and its 95% confidence interval

Discussion

In Experiment 2, we set out to provide more direct evidence in favor of the ease-of-processing hypothesis by tracking the participants’ self-reported easiness of the moral evaluations on a trial-by-trial basis, under self-paced viewing. The results showed a clear correlation between the easiness ratings and the response times. Easier ratings were performed faster. Notably, the easier ratings also showed a significant correlation to the moral evaluations. The more extreme evaluation scores were associated with higher ratings of easiness. Taken together, the data in Experiments 1 and 2 are best accommodated by the ease-of-processing hypothesis to explain the association between speed and polarization of the moral evaluations.

However, Experiment 2 utilized a slightly modified stimulus set. Thus, we decided to conduct a third experiment in an effort to establish the reproducibility of the findings, and to test the time-limited condition with the stimulus set of Experiment 2.

Experiment 3: replication of experiment 1, with balanced stimulus set

Introduction

In Experiment 2, we used a more balanced stimulus set than in Experiment 1, with a more uniform distribution of images across the spectrum from very moral to very immoral. We reasoned it was necessary to replicate Experiment 1 with the balanced stimulus set to properly rule out the hypothesis of a speed-polarization tradeoff. It remained possible that we had failed to obtain evidence for a tradeoff in Experiment 1 due to the unbalanced stimulus set, if urgency would affect especially the evaluation of images in the middle range, not close to either end of the moral scale. Compared to the clearly moral or immoral images, which are associated with more stable and concentrated evaluations, the middle range of images might be more ambiguous or volatile, and therefore more susceptible to polarization of evaluation under time pressure. Then, Experiment 3 was conducted as a replication of Experiment 1, in all aspects comparable, but with a new group of participants and a balanced set of stimuli.

Methods

Participants

A total of N = 36 individuals at Kyushu University (18 males and 18 females with a mean age of 20.92 years old, and a standard deviation of 3.50) met the participation criteria outlined in the preregistration. All participants were right-handed and completed the experiment.

Stimuli, apparatus, and procedure

The target visual stimuli and apparatus were the same as those in Experiment 2. The experimental paradigm was the same as in Experiment 1 (see Fig. 1). An experimental session consisted of 180 trials, divided into four blocks of 45 trials, with two levels of time control (SP, TL) and three levels of morality of the image content (moral, neutral, immoral). No image was presented more than once, and all trials were presented in pseudorandom order to ensure that each block of 45 trials contained the same distribution of trials per condition. The assignment of images to SP versus TL conditions was counterbalanced across participants.

Results

All participants completed all trials in the SP condition. On average, the participants completed 94.78% of the trials in the TL condition (standard deviation, S.D. = 4.79). The remaining 5.32% consisted of TL trials in which the participants failed to respond within the deadline of 2 s. No participant missed more than 20% of the TL trials. The number of completed trials provided sufficient data for the present analyses.

Overall analysis

Figure 10 presents the means and 95% confidence intervals for the evaluation scores (left panel) and for the response times (right panel) as a function of the morality of the image content and the time control condition. This analysis was based on the averaged data for each participant in each condition (N = 36).

Fig. 10.

Average evaluation scores (left panel) and response times (right panel) for each participant in each condition (N = 36) of Experiment 3. The blue bars show the data with moral images; the gray bars show the data with neutral images; the orange bars represent the data with immoral images. The bars from the time-limited conditions are shown in a darker shade than the bars from the self-paced conditions. The error bars reflect the 95% confidence interval around the mean

We conducted a repeated measures ANOVA on the evaluation scores with the within-subjects factors time control (i.e., SP versus TL) and the morality of the image content (i.e., moral versus neutral versus immoral). There was a significant main effect of the morality of the image content (F (2, 70) = 120.078, p < 0.001, η²_p = 0.774). Post-hoc pairwise comparisons showed significant differences at p < 0.001 among all three types of images. The effect of time control was not significant, with F < 1. A statistically significant interaction was obtained between the two factors (F (2, 70) = 11.439, p < 0.001, η²_p = 0.246). Post-hoc pairwise comparisons showed significant differences as a function of time control for moral images (higher evaluation score with the SP cue than the TL cue) and neutral images (lower evaluation score with the SP cue than the TL cue), but not for immoral images.

We also conducted a repeated-measures ANOVA based on the response times, using the same factors as used in the statistical analysis of the evaluation score. The results showed significant main effects for time control (F (1, 35) = 65.278, p < 0.001, η²_p = 0.651) and morality of the image content (F (2, 70) = 29.670, p < 0.001, η²_p = 0.459). Post-hoc pairwise comparisons showed that neutral stimuli produced significantly longer response times than moral as well as immoral stimuli. There was also a statistically significant interaction between the two factors (F (2, 70) = 22.332, p < 0.001, η²_p = 0.390). Post-hoc pairwise comparisons showed the pattern of differences among the image categories was due to the SP condition, whereas in the TL condition, there were no significant differences.

Median-split analysis

Figure 11 presents the averages and 95% confidence intervals of the evaluation scores, comparing fast versus slow responses for each of the 2 × 3 conditions. As in Experiment 1, we also conducted the median split by response time on the corresponding condition and analyzed within-subjects effects of a three-way repeated-measures ANOVA based on the averaged evaluation scores for each participant (N = 36), using morality of the image content, time control, and response speed (fast versus slow).

Fig. 11.

Median-split comparisons between evaluation scores with fast responses versus slow responses in Experiment 3, separately for each condition in the 2 × 3 design with the factors time control (self-paced versus time-limited) and morality (moral versus neutral versus immoral), based on the averaged data for each participant (N = 36). The error bars reflect the 95% confidence interval around the mean

Again, to be able to compare the evaluation scores for moral and immoral images directly, we inversed the sign for the data with immoral evaluation. Significant main effects were obtained with morality of the image content (F (2, 70) = 119.746, p < 0.001, η²_p = 0.774) and response speed (F (1, 35) = 70.682, p < 0.001, η²_p = 0.669), but not time control (F < 1). Within conditions, fast responses consistently produced more extreme evaluation scores (5.282) than slow responses (4.448), but the instruction to perform time-limited evaluations did not induce polarization. As already established by the 2 × 2 repeated measures ANOVA without the median split, there was a significant interaction between the morality of the image content and time control (F (2, 70) = 11.390, p < 0.001, η²_p = 0.246). No other interactions were statistically significant (all F values < 2).

Analysis of the level of variation

As in Experiment 1, we computed the standard deviation of the evaluation scores for each participant in each condition. These participant-based standard deviations were then analyzed in a two-way repeated-measures ANOVA, using morality of the image content (moral, neutral, or immoral) and time control (self-paced versus time-limited) as within-subjects factors.

There was a significant main effect of the factor time control (F (1, 35) = 9.674, p = 0.004, η²_p = 0.217), with larger standard deviations in the SP condition than in the TL condition. There was also a significant main effect on the participants’ standard deviations from the morality of the image content (F (2, 70) = 5.573, p = 0.006, η²_p = 0.137). Post-hoc pairwise comparisons showed significant differences only with neutral images, producing smaller standard deviations than immoral images as well as moral images. The interaction between morality of the image content and time control was also significant (F (2, 70) = 3.413, p = 0.039, η²_p = 0.089). Post-hoc pairwise comparisons showed that for moral images the standard deviations were larger in the SP condition than in the TL condition, but there was no difference as function of time control for immoral and neutral images.

Correlation analysis

Figure 12 presents the scatter plot with histograms and a linear regression line (with 95% confidence interval) for the item-based correlation between evaluation scores in the SP condition versus the TL condition. Pearson’s r value was 0.886 (R² = 0.784; p < 0.001). As with the results in Experiment 1, the evaluation scores for the real-world images were very similar across the SP and TL conditions.

Fig. 12.

Item-based scatter plot with histograms and linear regression line for the correlation analysis (N = 180) between the evaluation scores in the self-paced versus time-limited conditions in Experiment 3

Figure 13 presents the scatter plots with histograms and linear regression lines (with a 95% confidence interval) for the item-based correlations between the evaluation scores and the response times in the SP condition (left panel) and in the TL condition (right panel). For the SP condition, Pearson’s r value was − 0.549 (R² = 0.302; p < 0.001). For the TL condition, Pearson’s r value at -0.658 (R² = 0.433; p < 0.001). Thus, in both cases, the significant negative correlation confirmed the association between fast responses and more polarized evaluation scores.

Fig. 13.

Item-based scatter plots with histograms for the correlation analysis (N = 180) between the evaluation scores and response times in Experiment 3. The left panel shows the data in the self-paced condition; the right panel shows the data in the time-limited condition. The trends are indicated with smooth regression lines and their 95% confidence intervals

Discussion

In Experiment 3, we introduced neutral images and aimed to replicate the findings of Experiment 1. The response speed was again much faster in the time-limited condition than in the self-paced condition; the greatest difference in response speed as a function of urgency was obtained with the neutral images. However, despite the successful speed manipulation, overall urgency did not affect the means of the evaluation scores. Zooming in on the type of image, however, a relatively small but significant difference was obtained for the moral images, with lower evaluation scores in the time-limited condition than in the self-paced condition, that is, in the direction opposite to the hypothesis of a speed-polarization tradeoff. For neutral images, the evaluation scores increased under time pressure.

Within conditions, a median-split analysis separating fast from slow responses for all image categories showed that fast responses were associated with more extreme ratings than slow responses. Item-based analysis gave solid evidence that the evaluation scores remained consistent regardless of the time-control condition. Yet, images that were rated to have extreme moral or immoral content generated quicker responses than images with moderate ratings.

As in Experiment 1, the standard deviations of the moral evaluations tended to be larger in the self-paced condition than in the time-limited condition, in contrast to the hypothesis of a speed-precision tradeoff. Again, this observation is subject to the caveat that we did not measure within-person consistency in evaluation.

Taken together, the data pattern in Experiment 3 is consistent with the ease-of-processing hypothesis but not the hypothesis of a speed-polarization tradeoff.

Experiment 4: food images, with urgency manipulation

Introduction

The previous experiments provided a solid pattern of findings favoring the ease-of-processing hypothesis to explain the association between speed and polarization in moral evaluation. To test whether the findings can be generalized to other types of subjective evaluation, we designed Experiment 4 as a conceptual replication, using a completely different type of evaluative processing based on preference, particularly with respect to food.

Similar to Experiment 3, we used a balanced distribution of “appetitive”, “neutral”, and “aversive” images based on the Things object concept and object image database (THINGS database) (Hebart et al., 2019) and the FoodCast research image database (FRIDa) (Foroni et al. 2013). We asked participants to rate food images on a continuous scale from − 10 (“very disgusting”) to + 10 (“very attractive”) following the self-paced or the time-limited instruction cues.

Methods

Participants

A total of N = 36 individuals at Kyushu University (21 males and 15 females with a mean age of 21.25 years old and a standard deviation of 2.37) met the participation criteria outlined in the pre-registration. All participants were right-handed and completed the experiment.

Apparatus, stimuli and procedure

The apparatus and experimental paradigm were the same as in Experiment 3. An experimental session consisted of 180 trials, divided into four blocks of 45 trials, with two levels of time control (SP, TL) and three levels of appetitiveness of the image content (appetitive, neutral, aversive). No image was presented more than once, and all trials were presented in pseudorandom order to ensure that each block of 45 trials contained the same distribution of trials per condition. The assignment of images to SP versus TL conditions was counterbalanced across participants.

The target visual stimuli consisted of 180 food images selected from THINGS database (Hebart et al., 2019) and FRIDa (Foroni et al. 2013). The categories of appetitive and aversive were consistent with FRIDa used in research of evaluative processing (Ounjai et al. 2018), which included strongly aversive and appetitive food images. To complete the food image database distribution, we supplemented neutral and slightly aversive images based on the original categorization in the food domain of the THINGS database. The images provided content ranging from appetitive to aversive, depicting desserts, gourmet meals, fruit, vegetables, indulgent foods (e.g., chocolates, cakes, ice creams), unadorned foods (e.g., plain white rice, unseasoned vegetables), and spoiled or rotten food. We preselected a balanced set of neutral images and slightly aversive images. For confirmation, we conducted a preparatory rating experiment with ten students in the lab, asking them to rate the preselected images with respect to their preference from − 10 (extremely disgusting) to + 10 (extremely attractive). The participants were instructed to perform the food evaluation by asking themselves for each image “How attractive is this food?” Each target stimulus was reformatted for the experiment without changing the aspect ratio to maintain the same height at 600 pixels with 96 dpi for horizontal and vertical resolution.

Results

The present data set included data from 36 participants, each presented with a total of 180 trials corresponding to the 2 × 3 design with the factors of time control (SP versus TL) and Food categories (appetitive versus neutral versus aversive). All participants completed all trials in the SP condition; on average, all participants completed 94.04% of the trials in the TL condition (standard deviation, S.D. = 3.04). In the remaining 5.96% of the TL trials, the participants failed to respond within the deadline of 2 s. No participant missed more than 20% of the TL trials. The number of completed trials provided sufficient data for the present analyses.

We conducted the same statistical methods as in Experiment 3 to analyze the data, focusing on the evaluation scores and response times as dependent measures.

Overall analysis

Figure 14 presents the means and 95% confidence intervals for the evaluation scores (left panel) and response times (right panel) as a function of the food categories and the time control condition. This analysis was based on the averaged data for each participant in each condition (N = 36).

Fig. 14.

Average evaluation scores (left panel) and response times (right panel) for each participant in each condition (N = 36) of Experiment 4. The blue bars show the data with appetitive images; the gray bars show the data with neutral images; the orange bars represent the data with aversive images. The bars from the TL conditions are shown in a darker shade than those from the SP conditions. The error bars reflect the 95% confidence interval around the mean

Concerning the evaluation scores, repeated measures ANOVA was performed to analyze the within-subjects effect of the time control (i.e., SP versus TL) and the categories of the food image content (i.e., appetitive versus neutral versus aversive). There was a significant main effect of the categories of the food image content (F (2, 70) = 95.820, p < 0.001, η²_p = 0.732). Post-hoc pairwise comparisons showed significant differences among all three categories, with neutral images receiving the least polarized scores and aversive images receiving the most polarized scores. The effect of time control was not significant (F (1, 35) = 1.486, p = 0.231, η²_p = 0.041), nor was there a statistically significant interaction between the two factors (F (2, 70) = 1.549, p = 0.220, η²_p = 0.042).

The results of repeated-measures ANOVA on the response times showed significant main effects for time control (F (1, 35) = 114.211, p < 0.001, η²_p = 0.765) and food categories (F (2, 70) = 31.581, p < 0.001, η²_p = 0.474). Post-hoc pairwise comparisons indicated that the response times for the three categories of food images were all significantly different from each other. There was also a statistically significant interaction between the two factors (F (2, 70) = 12.312, p < 0.001, η²_p = 0.260). Post-hoc pairwise comparisons proved that the pattern of differences among image categories was due to the SP condition, whereas in the TL condition there were no such differences.

Median-split analysis

As in Experiment 3, we also conducted a median-split analysis by response time. Figure 15 presents the averages and confidence intervals of the evaluation scores, comparing fast versus slow responses for each of the 2 × 3 conditions. The three-way within-subjects repeated-measures ANOVA was based on the averaged evaluation scores for each participant (N = 36), using food categories, time control, and response speed (fast versus slow).

Fig. 15.

Median-split comparisons between evaluation scores with fast responses versus slow responses, separately for each condition in the 2 × 3 design with the factors time control (self-paced versus time-limited) and food categories (appetitive versus neutral versus aversive) in Experiment 4, based on the averaged data for each participant (N = 36). The error bars reflect the 95% confidence interval around the mean

Significant main effects were obtained with food categories (F (2, 70) = 95.643, p < 0.001, η²_p = 0.732) and response speed (F (1, 35) = 156.090, p < 0.001, η²_p = 0.817), but the effect of time control was not significant (F (1, 35) = 1.245, p = 0.272, η²_p = 0.034).

There was a significant interaction between food categories and response speed (F (2, 70) = 17.252, p < 0.001, η²_p = 0.330) and also a significant three-way interaction (F (2, 70) = 3.578, p = 0.033, η²_p = 0.093. Overall, regardless of the interactions, fast responses were consistently associated with more polarized scores within conditions, but time-limited cues did not affect the evaluation scores differently as compared to self-paced cues.

Analysis of the level of variation

As in Experiments 1 and 3, we computed the standard deviation of the evaluation scores for each participant in each condition. These participant-based standard deviations were then analyzed in a two-way repeated-measures ANOVA, using food categories (appetitive versus neutral versus aversive) and time control (self-paced versus time-limited) as within-subjects factors.

There was a significant main effect on the participants’ standard deviations from the food categories (F (2, 70) = 62.980, p < 0.001, η²_p = 0.643). Post-hoc pairwise comparisons showed significant differences among all three categories, with the largest standard deviations for aversive food images (average SD = 2.935) and the smallest for neutral food images (average SD = 1.949; for appetitive food images, average SD = 2.327). The factor time control also produced a significant main effect (F (1, 35) = 7.246, p = 0.011, η²_p = 0.172) with larger standard deviations in the self-paced condition (average SD = 2.478) than in the time-limited condition (average SD = 2.329). There was no interaction between the two factors (F (2, 70) = 1.894, p = 0.158, η²_p = 0.051).

Correlation analysis

Figure 16 presents the scatter plot with histograms and a linear regression line (with 95% confidence interval) for the item-based correlation between evaluation scores in the SP condition versus the TL condition. Pearson’s r value was 0.867 (R² = 0.752; p < 0.001). As with the results of moral evaluation, the scores for food images were very similar across the SP and TL conditions.

Fig. 16.

Item-based scatter plot with histograms and linear regression line for the correlation analysis (N = 180) between the evaluation scores in the self-paced versus time-limited conditions of Experiment 4

Figure 17 presents the scatter plots with histograms and linear regression lines (with 95% confidence interval) for the item-based correlations between the evaluation scores and the response times in the SP condition (left panel) and in the TL condition (right panel). In both cases, the linear trends were significant, with a Pearson’s r value of -0.508 in the SP condition (R² = 0.258; p < 0.001) and − 0.674 in the TL condition (R² = 0.454; p < 0.001). Once again, faster responses were associated with more polarized evaluations in both time control conditions.

Fig. 17.

Item-based scatter plots with histograms for the correlation analysis (N = 180) between the evaluation scores and response times in Experiment 4. The left panel shows the data in the self-paced condition; the right panel shows the data in the time-limited condition. The trends are indicated with smooth regression lines and their 95% confidence intervals

Discussion

In Experiment 4, we used food images in a preference evaluation task for a conceptual replication of our findings on the association between speed and polarization in moral decision-making. Consistent with our previous findings, the response speed was much faster in the time-limited condition than in the self-paced condition also when evaluating the appetitiveness of food images, proving that our urgency manipulation was very effective. However, despite the significant time pressure, the preference evaluations remained unchanged in the time-limited condition versus the self-paced condition. Thus, urgency does not lead to polarization in the preference evaluation of food images.

Within conditions, a median-split analysis separating fast from slow responses for the food images showed that fast responses were consistently associated with more polarized ratings than slow responses. This was also confirmed in the item-based correlation analysis. In all these respects, the results of Experiment 4 matched with those of Experiment 3, proving that the findings with moral evaluation are generalizable to an entirely different kind of subjective evaluation, based on food preference.

There was also no evidence in favor of a speed-precision tradeoff. Indeed, the standard deviations of the preference evaluations tended to be larger in the self-paced condition than in the time-limited condition.

General discussion

In this study, we set out to examine the relationship between response speed and subjective evaluation with moral evaluations of images depicting real-world scenes, and preference evaluations of food images. We used images with very polarized content as well as neutral content taken from the SMID (Crone et al. 2018), THINGS database (Hebart et al., 2019) and FRIDa (Foroni et al. 2013). To contrast the hypotheses of a speed-polarization tradeoff versus the ease of processing, we used a cueing procedure to manipulate the urgency of response in Experiments 1, 3, and 4; we asked participants to evaluate the images on a continuous scale from − 10 to + 10, either with or without time pressure. In Experiment 2, we asked the participants also to give a self-reported measure of the ease of processing for each image. Across the four experiments, we consistently found that extreme evaluation scores were associated with relatively fast responses. Yet, urgency via time pressure made response speed much faster without nudging the subjective evaluations to polarization. The findings rule out a speed-polarization tradeoff, and instead are better explained by an account in terms of the ease of processing.

To examine the possibility of a speed-precision tradeoff, we also examined the levels of variation in SP versus TL conditions in Experiments 1, 3, and 4. Overall, the evidence suggested that time pressure did not cause more noise in the evaluations. Indeed, in Experiments 1, 3, and 4, the data pointed to significant effects in the opposite direction. Here, we note the caveat that our present tests of precision or consistency are based on measurements across conditions. Future investigations should test for consistency by multiple repetitions of judgment for the same items with the same participants, while carefully controlling for factors such as familiarity and repetition priming.

Another consideration with respect to the notion of a speed-precision tradeoff is that in the present study we used only one level of time pressure, requiring the participants to respond within 2 s in the time-limited condition. While this level of time pressure was strong enough to more than double the speed of responding, the item-based correlations indicated that 2 s was in most cases long enough for the participants to achieve very similar subjective evaluations across the self-paced and time-limited conditions. Future research (e.g., using tachistoscopic presentations) should examine how urgency affects the moral evaluation when the given viewing time falls below the minimal amount needed to complete the normal categorization.

Speed does not cause polarization, but it remains possible that, with insufficient time to process the contents of the images, the normal association between fast responses and extreme ratings would break down. People may resort to relatively safe choices under urgency to accommodate the incomplete information processing for decision-making (Wu et al. 2022; Maule et al. 2000; Dror et al. 1999). Accordingly, one hypothesis would be that, if participants are not given enough time to recognize the image content, they may respond with moderate or neutral ratings. This view is compatible with the ease-of-processing hypothesis: With additional time pressure, it would become more difficult to perform the evaluations, possibly leading to reduced polarization. Other ways to manipulate the ease of processing would be to either degrade the visual images or to add a dual-task requirement. In all such cases, the more difficult processing should, if it has any effect at all, lead to slowed, more moderate evaluations rather than polarization.

As acknowledged in the Methods section of Experiment 1, one technical limitation in the present study is our usage of a joystick that might have an unbalanced resistance across the movement range. While it is unlikely to have impacted the comparisons between SP and TL conditions, it potentially introduced a confounding factor within conditions. Future work should address this issue by implementing different response modes (e.g., using a gamepad with dual sticks, or switching from a continuous categorization to a binary forced choice).

In the present study, we tracked commonalities in evaluation mechanisms across very different tasks, with moral evaluation versus food preference estimation. The findings corroborate our recent conceptual replication of the mechanisms of confirmation bias in subjective judgement (Ma and Lauwereyns 2024; studying moral evaluation with a paradigm modeled after Ounjai et al. 2018; on food preference estimation; see also Levari et al. 2018; for a similar conceptual replication across very different judgment tasks). More generally, there may be a need in the literature to combine lines of research that are traditionally kept separate (e.g., perceptual decision-making versus value-based decision-making).

In sum, the present paradigm established that response speed is associated with, but does not cause, polarization in the moral evaluation of real-world images and in the judgment of food preference. The data did not reflect a speed-polarization tradeoff but suggested a critical role for the ease of processing.

Author contributions

M.C. and J.L. contributed to the design of the study. M.C. programmed the experiments, analyzed the behavior, and prepared all figures. M.C. and J.Y. conducted the data collection for the study. M.C. and J.L. wrote the manuscript. All authors reviewed and approved the manuscript.

Data availability

Following on from the exploratory investigation in Experiment 1, we pre-registered Experiments 2, 3 and 4 of the present study as confirmatory research at OSF (https://osf.io/8fjrk). The datasets generated by the experiment and analyzed in the present study are available in the OSF repository (https://osf.io/fuvzt).

Declarations

Competing interests

The authors declare no competing interests.