Edited by: Boris C. Rodríguez-Martín, Fundación Recal, Spain
Reviewed by: Laura Nynke Van Der Laan, University of Amsterdam, Netherlands; Marília Prada, Instituto Universitario de Lisboa (ISCTE), Portugal
This article was submitted to Eating Behavior, a section of the journal Frontiers in Psychology
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
Previous research suggested a role of gaze in preference formation, not merely as an expression of preference, but also as a causal influence. According to the gaze cascade hypothesis, the longer subjects look at an item, the more likely they are to develop a preference for it. However, to date the connection between viewing and liking has been investigated predominately with self-paced viewing conditions in which the subjects were required to select certain items from simultaneously presented stimuli on the basis of perceived visual attractiveness. Such conditions might promote a default, but non-mandatory connection between viewing and liking. To explore whether the connection is separable, we examined the evaluative processing of single naturalistic food images in a 2 × 2 design, conducted completely within subjects, in which we varied both the type of exposure (self-paced versus time-controlled) and the type of evaluation (non-exclusive versus exclusive). In the self-paced exclusive evaluation, longer viewing was associated with a higher likelihood of a positive evaluation. However, in the self-paced non-exclusive evaluation, the trend reversed such that longer viewing durations were associated with lesser ratings. Furthermore, in the time-controlled tasks, both with non-exclusive and exclusive evaluation, there was no significant relationship between the viewing duration and the evaluation. The overall pattern of results was consistent for viewing times measured in terms of exposure duration (i.e., the duration of stimulus presentation on the screen) and in terms of actual gaze duration (i.e., the amount of time the subject effectively gazed at the stimulus on the screen). The data indicated that viewing does not intrinsically lead to a higher evaluation when evaluating single food images; instead, the relationship between viewing duration and evaluation depends on the type of task. We suggest that self-determination of exposure duration may be a prerequisite for any influence from viewing time on evaluative processing, regardless of whether the influence is facilitative. Moreover, the purported facilitative link between viewing and liking appears to be limited to exclusive evaluation, when only a restricted number of items can be included in a chosen set.
Tracking the gaze provides a moment-by-moment assessment of thought processes in a wide variety of contexts (
The dominant hypothesis in the literature on evaluative processing is that viewing leads to increased evaluation. As an early reference in this vein, we note the classic work by
The basic phenomenon of gradually increasing gaze likelihood on the to-be-chosen item was replicated in a variety of forced-choice tasks with a range of stimuli, including human faces (
Formally, the hypothesis of a relationship between gaze fixation and evaluative processing can be understood as an accumulator model of decision-making (
To prove that the gaze bias not only expresses but also influences preference formation,
In short, the literature on gaze bias with 2AFC tasks converges on a strong, causal relationship between viewing and preference formation. However, it is less clear to what extent this relationship is generalizable to other types of evaluative processing. Most conspicuously, the 2AFC tasks require a comparative evaluation with a direct competition between two items, such that subjects should select one item at the expense of the other. In this case, as suggested also by the fMRI data by
Another limitation of the 2AFC task, as raised by
Effectively, our primary research question was whether the purported link between gaze and liking applies to absolute evaluative processing, with single images, under conditions that require no spatial selection on the screen. The corollary of the fMRI findings by
Classic work by
To examine our hypothesis that viewing would not necessarily lead to increased liking in absolute evaluative processing, we designed a study with four types of evaluation. We varied the type of exposure. Subjects were either free to view the images at their own pace, or forced to view the images for a computer-generated duration. We also varied the nature of the type of evaluation. Subjects were either asked to give a rating (i.e., non-exclusive; with no restriction on the number of items that can receive a positive evaluation), or to pick a pre-determined number of items (exclusive; with a cap on the number of choices).
Finally, in the present study we opted to use naturalistic food images for evaluative processing for both practical and theoretical reasons. Our lab is engaged in a comprehensive research project on the evaluative processing of food as an important domain of health and consumer behavior; additionally, we were able to use a well-established database of food images (
All 78 subjects were students from Kyushu University (37 females and 41 males;
Stimuli consisted of a set of 320 naturalistic food images with a resolution of 600 × 450 pixels (96 dpi, standard Red Green Blue (sRGB) color format). The set of images was drawn from a food-pictures database for experimental research (
We created four sets of 80 pictures that showed no significant differences in any of the objective or subjective characteristics of the food-pictures database. For each participant, a different set of 80 pictures was used in each of the four evaluation tasks (see below for the definition of tasks). Thus, we ensured that subjects were never exposed to the same food image twice. The allocation of picture sets to tasks was counterbalanced; the order of food pictures was randomized within each task; and the order of the tasks was counterbalanced across subjects. Images were presented as a single stimulus on a black background.
The visual stimulus was presented on a 23.8-inch full high definition (FHD) flat-panel-monitor, with a display resolution of 1920 × 1080 pixels. The subjects were seated approximately 65 cm from the monitor. To minimize head movement a chin-rest with a forehead-support was used. A keyboard set up was used to record the subjects’ responses.
In the initial phase of the experiment, for 34 subjects, we recorded only manual (keyboard) responses due to logistic limitations. However, to establish the amount of time actually spent gazing at each stimulus, we reckoned it was necessary to add an eye-tracking device to the experimental set-up. For the next 44 subjects, we were able to record manual (keyboard) responses as well as gaze position using Eye Tribe, an eye-tracking device at 60 Hz sampling rate (The Eye Tribe Aps, Denmark); a system with sufficient reliability for present purposes (
Before the start of a session with eye tracking, the subject was asked to follow a dot on the screen for a 12-point calibration. After the calibration, the gaze coordinates were calculated through Eye Tribe with an average accuracy of less than 0.5° visual angle on a 24-inch display. To prevent heat buildup a small universal serial bus (USB) fan was used. All events and recordings were controlled through code written in Psychopy (version 1.84.2); for reference, see
To compute actual viewing time (time with eye position on the displayed naturalistic food image) raw data were filtered. First, eye positions beyond the presentation area were removed. Second, detected and recorded eye blinks were also removed from the amount of actual viewing time if they lasted longer than 50 ms. Finally, the obtained data were plotted using custom software, and statistical analyses were conducted.
One experimental session consisted of four different evaluation tasks: two different types of decision (non-exclusive versus exclusive) performed under two different types of exposure duration (self-paced versus time-controlled). Each subject was asked to participate in each of the four evaluation tasks.
For all tasks, the subjects were asked to base their decisions on how much they liked the food images. The subjects were instructed to evaluate the appetitive appeal of the food images; this type of evaluative processing presumably involves a combination of individual food preferences and the esthetic properties of the images.
In the non-exclusive evaluation tasks, the subjects were asked to give each food image a rating from 1 (“
In the self-paced evaluation tasks, the subjects could determine the length of time they viewed the images. In contrast, in the time-controlled evaluation tasks, the exposure duration was computer-generated.
The four evaluation tasks were completed in counterbalanced order. The entire session lasted approximately 50 min. Prior to each experiment an informed consent, established according to ethical principles, was obtained. Each subject completed a questionnaire designed to assess the reported health state (e.g., eyesight, amount of hours of sleep the night before the experiment, medication intake, and food allergies occurrence). Before each task, the experimenter provided the appropriate instructions. Additionally, for sessions with eye tracking, the eye tracker was calibrated and instructions to look at the displayed food image were given. In between tasks, there was a brief pause, while the experimenter set up the computer to run the next task.
The trial structure in the self-paced non-exclusive evaluation. The structure was the same in the three other evaluation tasks except for the following critical differences. The duration of the exposure frame (second row) was either self-paced or time-controlled. In the self-paced evaluation tasks, the subject had to press the space bar to proceed to the response frame. In the time-controlled evaluation tasks, the response frame replaced the exposure frame automatically after a computer-generated duration. In the exclusive evaluation tasks, the response frame presented the question, “Would you like to add this image to your food basket?” and gave the options “Yes” or “No.” In the exclusive evaluation tasks, the feedback frame indicated the number of items added to the basket so far.
This task was the same as the SPN, except for the exposure duration of the food images. Again, exposure time was defined as the length of time during which the food image was displayed on the screen. However, the food image on the stimulus screen was displayed for a pseudo-randomly chosen duration between 1 and 8 s, and was then automatically replaced by the response screen. Thus, in this task the subject had no control over the exposure time. We used only integer values for the exposure duration, ensuring each value between 1 and 8 was used 10 times over the course of 80 trials.
The question in the SPE was “Would you like to add this food image to your basket?” To make a choice, subjects were required to answer by pressing one of two specific keys: N (for “No”) or Y (for “Yes”). The aim of this task was to impose a limitation on the number of items that could receive a positive evaluation. For this purpose, the experimenter instructed all subjects to select a maximum of 15 food images. To confirm the choice and update the number of selected food images, a 1-sec feedback was presented after every decision (e.g., for the first selected image: “
In all other respects, this task was the same as the SPN. Exposure time was defined as the length of time during which the food image was displayed on the screen; the food image was removed at the time when the subject pressed the spacebar to move on to the response screen.
In this task, the required choice was the same as in the SPE. With respect to the timing of the exposure duration, the task was the same as the time-controlled non-exclusive evaluation (TCN).
Preliminary analyses showed that the data pattern for the 34 subjects with only manual responses replicated the data pattern for the 44 subjects with manual responses plus eye tracking (see
Average exposure time of food images rated from 1 (“not like at all”) to 5 (“like very much”) in the self-paced non-exclusive evaluation (
To gain further insights in the observed relationship between rating and exposure time, we analyzed the polynomial contrasts. The linear contrast,
Average actual viewing time of naturalistic food images rated from 1 (“not like at all”) to 5 (“like very much”) in the self-paced non-exclusive evaluation (
To gain further insights in the observed relationship between rating and actual viewing time, we analyzed the polynomial contrasts. The linear contrast,
Average exposure time of food images rated from 1 (“not like at all”) to 5 (“like very much”) in the time-controlled non-exclusive evaluation (
Average actual viewing time of naturalistic food images rated from 1 (“not like at all”) to 5 (“like very much”) in the time-controlled non-exclusive evaluation (
Average exposure time of rejected (NO) and included (YES) naturalistic food images in the self-paced exclusive evaluation (
Average actual viewing time of rejected (NO) and included (YES) images in the self-paced exclusive evaluation (
Average exposure time of rejected (NO) and included (YES) naturalistic food images in the time-controlled exclusive evaluation (
Average actual viewing time of rejected (NO) and included (YES) images in the time-controlled exclusive evaluation (
The goal of our research was to gain insight into the role of viewing in the evaluative processing of food images. In our experimental paradigm, we varied both the type of exposure (self-paced versus time-controlled) and the evaluative task (non-exclusive versus exclusive). A common notion in the literature on evaluative decision-making is that the longer a participant looks at an item, the more likely (s)he is to develop a preference for it. However, this proposal has been derived on the basis of 2AFC evaluation tasks in which subjects were required to indicate a preference for one of two items presented simultaneously. This literature, then, converges on the proposal that viewing leads to
With our single-image paradigm, we found that the relationship between viewing and evaluative processing depended on the type of task. In the SPE, we obtained a trend similar to the 2AFC findings, with longer viewing leading to a higher likelihood of a positive evaluation (i.e., inclusion in a virtual basket). In the time-controlled tasks, both with non-exclusive and exclusive evaluation, there was no significant relationship between the viewing duration and the evaluation. Moreover, in the SPN, when subjects were asked to give a rating from 1 to 5 (with no limitation on the number of positive evaluations), we obtained a significant relationship in the opposite direction, such that longer viewing durations were associated with lower ratings. The pattern of results was the same for viewing times measured in terms of exposure duration (i.e., the duration of stimulus presentation on the screen) and in terms of actual gaze duration (i.e., the amount of time the subject effectively gazed at the stimulus on the screen).
The present evidence indicates that viewing does not intrinsically lead to increased liking when evaluating single food images. Given that the controlled exposure durations did not influence the evaluation, and that the purported relationship between viewing and liking even went in the opposite direction in the SPN, we can firmly conclude that there is no mandatory connection between gaze and increase in absolute value. Our data, then, taken together with the robust findings from 2AFC paradigms, point to a critical distinction between absolute and relative evaluative processing. Previous research had established that the comparative mode, in which multiple items are considered simultaneously, crucially relies on the gaze for fixation-dependent accumulation of relative value. While making direct comparisons, the gaze may express and influence the competition or the selection among multiple items, as according to the proposal by
With the comparison of the self-paced versus time-controlled tasks, we aimed to examine the role of self-determination in the relationship between viewing and evaluation. Importantly, by artificially imposing pseudo-random exposure durations between 1 and 8 s, our time-controlled tasks did not introduce a deadline or any urgency in the decision-making. Rather, the imposed durations were likely to mismatch with the subjects’ natural time course of evaluative processing, with durations that could be either longer or shorter than the subjects would have chosen, if given the opportunity to determine the viewing time. In this sense, our present examination with respect to self-determination cannot easily be mapped onto conventional approaches of the speed-accuracy trade-off (
Critically, we found that, without the possibility of self-determination, there was no connection between viewing time and evaluative processing. This was true in both the TCN (with ratings from 1 to 5) and the TCE (with the categorical rejection or inclusion in a virtual basket). Here, it is important to note that, in the time-controlled tasks, not only the exposure durations, but also the actual gaze durations on the images showed no connection to the evaluations. We suggest that, without control over the exposure duration, subjects may disengage their covert attention from the overt point of gaze fixation. Put differently, their mind may not necessarily be on the screen, even if their gaze is directed there. By this interpretation, the lack of a relationship between viewing and liking in the time-controlled tasks is due to a disconnection between visual processing and decision-making. For instance, if a displayed image remains on the monitor even though its value has already been formed, the internal processing may no longer have the same focus as overt attention.
In addition to the self-determination of viewing time, we found that the type of evaluation had a critical impact on the relationship between viewing and evaluative processing. In fact, we obtained opposite trends in the SPN versus the exclusive evaluation. In the SPE, the likelihood of inclusion increased with longer viewing times, representing effectively a conceptual replication of the well-known gaze cascade phenomenon in 2AFC evaluation tasks (
Notably, in 2AFC evaluation tasks the choice for one option implies choosing against the other, that is, the 2AFC induces a zero-sum competition. In such a situation, the underlying mechanism likely involves a cumulative process, gradually building commitment to a choice. This kind of decision process may be regarded as confirmatory, in the preparation of action. The decision is set in motion only when sufficient evidence has been accumulated, exceeding an internal threshold, or passing a point of no return. The zero-sum competition is a key feature of this type of decision process; the underlying neural computations are expressed as likelihood ratios between mutually exclusive alternatives (
Here, we propose that similar computational mechanisms, and a similar link between gaze and preference formation, apply to the present exclusive evaluation task, in which subjects were required to decide categorically for each particular food image whether they wanted to include it, yes or no, in the limited virtual basket of maximally 15 images. This paradigm involves a type of serial decision-making in which alternatives are considered one by one instead of being compared simultaneously (e.g.,
In contrast, in the non-exclusive evaluation, there was no limitation on the number of positive evaluations. The rating given to any image should have no implications for the other ratings. Subjects were always free to give any image the maximum score of five. In this situation, there was no critical need of confirmation, and no sense of a crucial point of no return. With no reason to hesitate, when an image looked good at first glance, the subject might as well give it a five right away. Our data show that particularly the highest rating of five was associated with brief viewing (less than 2 s actual viewing time) in the SPN. Such rapid evaluative processing may have been enabled by the nature of the decision, which implied no cost. There was no risk or potential loss of opportunity associated with giving a positive evaluation.
The findings with the present evaluation tasks also raise several important questions with respect to the underlying mechanisms that must be addressed in future research. Particularly, in the present paradigm we did not control for the initial preference with respect to the food images. In this sense, we cannot dissociate the evaluative processing in terms of
Another concern with the present paradigm is that we instructed subjects to evaluate the attractiveness of the food images. On the response screen we labeled the to-be-evaluated stimulus explicitly as “food image.” Consequently, we cannot dissociate to what extent the evaluative processing was influenced by intrinsic food-related characteristics versus esthetic properties of the image. Although all subjects engaged in evaluative processing, it is possible that some subjects weighed primarily the esthetic dimension of the images, while other subjects may have based their evaluations more strongly on intrinsic properties of the food items. Food images likely represent a type of stimulus with distinctive features as compared to other visual images. The attractiveness of food images may inherently be a complex property, determined by visual as well as a range of non-visual characteristics. It will require an extensive research program to fully understand the complexity of the attractiveness of food images. The present study offers a suitable paradigm for such a research program. As notes for departure, we observe that the self-determination of viewing time and the type of decision task critically influence the relationship between viewing and absolute evaluation of food images.
This study was carried out in accordance with the Declaration of Helsinki and the ethics guidelines of the Graduate School of Systems Life Sciences, Kyushu University. All subjects gave written informed consent (regarding the purpose of the research, expected duration, procedures, and confidentiality).
All authors contributed to the design of the study. AW conducted the data collection for the study, analyzed the behavior and eye-tracking data, and prepared all figures. AW and KO programmed the experiments. AW and JL wrote the manuscript. All authors reviewed and approved the manuscript.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
We thank Jens Blechert for generously making the picture database (
The Supplementary Material for this article can be found online at:
Complete data set for the present study, organized in 8 Data Sheets: SPN_exp, Self-Paced Non-Exclusive Evaluation exposure duration; SPN_gaz, Self-Paced Non-Exclusive Evaluation actual viewing time; TCN_exp, Time-Controlled Non-Exclusive Evaluation exposure duration; TCN_gaz, Time-Controlled Non-Exclusive Evaluation actual viewing time; SPE_exp, Self-Paced Exclusive Evaluation exposure duration; SPE_gaz, Self-Paced Exclusive Evaluation actual viewing time; TCE_exp, Time-Controlled Exclusive Evaluation exposure duration; TCE_gaz, Time-Controlled Exclusive Evaluation actual viewing time.