The effects of encoding levels on the item and source memory of fearful and disgusting stimuli

Abstract

Background

A number of studies have shown that fear and disgust, both highly negative and arousing emotions, have better memory performance compared with neutral emotions. However, previous studies found that fear and disgust had significant differences in memory performance. Although one study has examined the effect of encoding level on item memory for fear- and disgust-related words, it remains unclear whether encoding level similarly affects item memory for fearful and disgusting images, as well as source memory for emotional images. This study examined the effects of encoding level (deep vs. shallow) and emotion type (fear, disgust and neutral) on item and source memory in three experiments.

Method

Each experiment included three phases: encoding, distraction, and retrieval. In Experiment 1, 44 participants (24 in the deep encoding group, 20 in the shallow encoding group) were asked to complete the incidental encoding and retrieval of item and extrinsic source. In Experiment 2, 56 participants (29 in the deep encoding group, 27 in the shallow encoding group) were asked to complete the same task as in Experiment 1 in intentional encoding. In Experiment 3, 57 participants (27 in the deep encoding group, 30 in the shallow encoding group) were asked to complete the intentional encoding and retrieval of item and intrinsic source.

Results

Our results showed that for item memory, the emotional memory advantage, especially for fearful stimuli, was observed in the shallow encoding. For source memory, no emotional memory advantage was observed in the shallow or deep encoding conditions in Experiment 1. However, in Experiments 2 and 3, fearful stimuli enhanced both frame and color source memory compared to disgusting and neutral stimuli in the shallow encoding conditions. Notably, in Experiment 3, disgusting stimuli impaired color source memory in the deep encoding conditions.

Conclusions

Our results suggested that fear and disgust differed in terms of item and source memory performance. However, these differences were moderated by key boundary conditions, including the level of encoding and the cognitive resource demands associated with the emotional specificity of the stimuli.

Background

Fear and disgust are threat-related stimuli that elicit intense negative affect and heightened arousal [1, 2]. These emotional responses are known to activate defensive behaviors [3, 4], which facilitate individuals’ anticipation of and response to potential threats, thereby influencing their health and survival. However, even when the two types of stimuli are matched for valence, arousal, and visual characteristics (e.g., subjective distinctiveness, visual complexity, hue, brightness), individuals still show differential memory performance for fear and disgust [5, 6, 7, 8, 9]. This phenomenon suggests that different emotions such as fear and disgust may shape cognitive processes in specific ways that are tailored to the unique demands associated with each type of emotional stimulus [10, 11]. Consequently, investigating and elucidating these differences is of utmost importance to provide valuable insights into both the evolution of human survival strategies and our understanding of clinically relevant emotional disorders.

Previous studies have shown that disgusting stimuli are easier to recall [2, 5] and recognize in memory than fearful stimuli [6, 8], but other studies have shown no difference between the recognition of these two emotions [12, 13, 14]. Fearful and disgusting stimuli direct attention and perception in distinct ways [15, 16, 17, 18, 19], and they may modify the way information is encoded. For example, in perceptual processing tasks such as the lexical decision task, disgusting words exhibited higher recognition owing to the spontaneous promotion of elaborate encoding compared to fearful and neutral words [13]. Conversely, in semantic processing tasks such as the affective categorization task, the emotional memory advantage was eliminated due to the encoding of fearful and neutral words being strengthened through the ‘deep orienting task’ [13]. Therefore, it appears that encoding level can modulate item memory for fearful and disgusting words. However, other studies have explored the differences between words and images in terms of their biological relevance [20] as well as their behavior and neural processing [21, 22, 23, 24]. It remains unclear how encoding level influences memory for fearful and disgusting images and whether these effects are consistent across materials. Considering the processing differences between these two emotions, we hypothesize that the disgust memory advantage would only occur in the shallow encoding conditions.

In addition, there are also controversies regarding the differences in source memory for fear and disgust. For example, researchers adopted the R/K paradigm, in which participants were asked to judge whether an image had been presented during the encoding session, whether they could recollect details about the image (remember), whether the image felt familiar but they could not recall any details from encoding (know), or whether the image had not been presented during encoding (new) [6, 8]. In their studies, the processing of emotional stimuli primarily focused on the stimuli themselves without requiring additional attentional resources for surrounding information, so more cognitive resources were invested in stimuli details [6, 8]. Therefore, participants exhibited better memory performance based on the recollection of disgusting stimuli compared to fearful stimuli [5, 6, 8, 14]. However, other researchers using the dot-probe paradigm, in which two images were presented simultaneously, found that after the images disappeared, a black circle (dot probe) appeared on either the left or right side of the fixation cross. Participants were instructed to respond as quickly as possible by pressing the button corresponding to the location of the emotional image. The study revealed that participants had better detail memory for fearful stimuli compared to disgusting stimuli [9]. In this study, participants had to process two images simultaneously and also judge the subsequent location of stimuli, consuming more cognitive resources to process the surrounding stimuli. This task might have distracted their attention when processing the stimuli, thereby reducing the level of encoding of emotional stimuli and affecting the source processing.

Furthermore, fearful and disgusting stimuli are not homogeneous with respect to their demand for exploration. According to the biological cost and benefit hypothesis, exploring sources of disgust does not incur a biological cost, despite their initial appearance of danger [15]. On the contrary, such exploration reveals potential benefits, as they may ultimately be deemed consumable [15]. In contrast, fearful stimuli are often associated with imminent survival threats. Devoting excessive processing resources to threatening or fear-evoking events (e.g., predators or attackers) rather than to avoidance or defensive strategies would incur greater costs than benefits. Thus, even when processing resources are limited, individuals can quickly encode fearful stimuli to ensure survival. Therefore, we hypothesize that due to the greater urgency of fearful stimuli processing, the source memory performance of fearful stimuli may be better than disgusting stimuli in the shallow encoding conditions, whereas when individuals have sufficient cognitive resources for in-depth processing, disgusting stimuli may perform better in the deep encoding conditions.

In the current study, we aimed to investigate how differences in memory performance (item memory and source memory) between fear and disgust are influenced by the level of encoding. Therefore, this study adopted the classic encoding-retrieval paradigm, in which participants encoded images with different contexts (e.g., frame context in Experiments 1 and 2: window frame and photo frame; color context in Experiment 3: red and green), and then tested both item memory and source memory. Item memory refers to the memory of an event or fact itself [25], while source memory pertains to the memory of the context in which an event or fact was acquired [26]. Both types of memory reflect distinct cognitive processing mechanisms.

Experiment 1

Method

Participants

We calculated the required sample size using G*Power (version 3.1) [27], which yielded a total of 44 participants, assuming a medium effect size (f = 0.25 [28, 29, 30]) for a 2 × 3 mixed experimental design (F-tests, ANOVA: Repeated Measures, within-between interaction) with 95% statistical power (α = 0.05) and the number of measurements = 3. The correlation among repeated measures (0.5) and the epsilon for sphericity correction (1) were set to be the default. Overall, 44 healthy Chinese undergraduate and graduate students participated in this study as paid volunteers. Of these, 24 belonged to the deep level group (20 female, mean age = 22.42 ± 2.57), while 20 belonged to the shallow level group (17 female, mean age = 21.55 ± 2.21). Participants were randomly assigned to either the deep or the shallow encoding group. All participants were right-handed and had normal or corrected-to-normal visual acuity. Participants provided written informed consent before participating, and the study was approved by the university ethics committee.

Apparatus and stimuli

The stimuli were displayed on 19-inch Dell monitors with a screen resolution of 1024 × 768 pixels and refresh 75 HZ. Each image was presented at 300 × 205 pixels with a black background. The software package, E-prime 2.0, was used to present the stimuli and collect the data.

We selected 491 images taken from the International Affective Image System [31], the Chinese Affective Image System [32] and Baidu images database (https://image.baidu.com/, a commonly used website in China). To match the fearful and disgusting images, a pilot study (N = 10) was conducted to obtain subjective ratings of the valence, arousal degree of fear and disgust using a 9-point scale. The questions were worded as follows: Valence: “How positive or negative do you feel the image is?” with responses ranging from 1 = “very negative” to 9 = “very positive”. Arousal: “How relaxed or excited do you feel while viewing the image?” with responses ranging from 1 = “very calm” to 9 = “very excited”. Fear: “How frightened do you feel while viewing the image?” with responses ranging from 1 = “no fear” to 9 = “very frightening”. Disgust: “How disgusting do you find the image?” with responses ranging from 1 = “no disgust” to 9 = “very disgusting”. Using these rating, 50 disgust-themed (e.g., feces and vomit), 50 fear-themed (e.g., snake and angry tiger) and 50 neutral-themed images (e.g., wardrobe and book) were chosen such that the disgusting images were rated as high in disgust and low in fear, while the fearful images were rated as high in fear and low in disgust. The disgusting and fearful images did not differ significantly on valence(M_disgusting = 2.976, SD_disgusting = 0.468; M_fearful = 2.912, SD_fearful = 0.588), or arousal (M_disgusting = 6.244, SD_disgusting = 0.518; M_fearful = 6.364, SD_fearful = 0.770) (ps > 0.742), Disgusting images received significantly higher disgust ratings than fearful images (M_disgusting = 6.444, SD_disgusting = 0.606; M_fearful = 4.276, SD_fearful = 1.218) (p < 0.001), but significantly lower fear ratings than fearful images(M_disgusting = 3.746, SD_disgusting = 0.652; M_fearful = 6.290, SD_fearful = 0.090) (p < 0.001). Neutral images received lower ratings of arousal (M = 3.214, SD = 0.504), fear (M = 1.390, SD = 0.492) and disgust (M = 1.492, SD = 0.597) than fearful and disgusting images (ps < 0.001), and received higher ratings of valence than fearful and disgusting images (indicating neutral valence; M = 5.150, SD = 0.257) (ps < 0.001).

Procedures

The experiment included the encoding, distraction, and retrieval phases.

Encoding phase. In order to become familiar with the experimental task, participants completed 8 practice trials that were not included in the formal experiment. In the formal encoding phase, 99 stimuli were presented, including 33 fearful, 33 disgusting, and 33 neutral images. The stimuli were presented successively in a pseudorandom but fixed order in such a way that no more than two images of the same emotional categories or with the same frame occurred in succession [22, 33]. In each encoding trial, a fixation sign was shown for 1000 ms at the center of the screen. Then an emotional image (fearful, disgusting or neutral) with contexts (source information: window or photo frame) was presented on the left or right side of the monitor for 3000 ms, during which participants were asked to complete the deep encoding task or the shallow encoding task. In the deep encoding group, participants were asked to complete a semantic processing task, rating the arousal level of the image (1 = “very calm” to 9 = “very excited”). In the shallow encoding group, participants were asked to complete a perceptual processing task, judging the location of the images at either the left or right position on the monitor. The position of the images on the left or right side of the monitor was counterbalanced among participants.

Distraction phase. After completing the encoding task, participants performed 40 calculations for 5 min before undergoing a surprise recognition test for the images in the retrieval phase.

Retrieval phase. In order to become familiar with the retrieval phase, participants completed 12 practice trials that were not included in the formal experiment. In the formal retrieval phase, 99 old images from the encoding phase without the contexts were intermixed with 17 new images from each emotion category to give a total of 150 images. The images were presented in a pseudorandom but fixed order in such a way that no more than three old or new images and no more than two images of the same emotional category occurred in succession [22, 33]. In each test trial, a fixation sign was displayed for a duration of 1000 ms at the center of the screen. Then, emotional images were displayed, during which participants were asked to judge whether the images were old (studied during the encoding phase) or new (unstudied during the encoding phase) by pressing the “F” key or “J” key. If the image was judged to be old, participants were then prompted to immediately identify the contextual frame (window or photo frame) that surrounded the image for the source memory test, using the same keypress response as before. If the image was judged to be new, the next image was immediately displayed. There was no time limit on responses to the retrieval task (Fig. 1).

Fig. 1.

Illustration of the experimental stimuli and procedure in Experiment 1. In the shallow encoding task, participants were asked to judge whether the images were on the left or right side of the monitor (Encoding phase a). In the deep encoding task, participants were asked to rate the arousal level of the images (Encoding phase b)

Date analyses

Two-way repeated-measure ANOVAs were used to analyze item and source memory accuracy separately. Item memory accuracy was calculated as the overall proportion of correct responses by dividing the sum of hits and correct rejections within each condition (fear, disgust, and neutral) by 50, which is the total number of old and new images (). Source memory accuracy was calculated by dividing the number of correct source judgments for the items correctly identified as old each condition (fear, disgust, and neutral) by 33, which is the total number of old images (). The analyses used encoding level (deep vs. shallow) as a between-subjects variable and emotion category (fear, disgust, and neutral) as a within-subjects variable. The Bonferroni adjustment was also used for multiple comparisons in post hoc tests for significant main effects of emotion, as well as for significant interaction effects between encoding level and emotion category (p < 0.05).

Results

Table 1 shows the proportion correct for the item memory and source memory in Experiment 1.

Table 1.

Proportion correct for the item memory and source memory in experiment 1

			Emotion Category
	Encoding level		Fear(SD)	Disgust(SD)	Neutral(SD)
Item memory	Deep	proportion of correct	0.843(0.091)	0.837(0.102)	0.838(0.101)
		hit rate	0.827(0.117)	0.802(0.157)	0.826(0.145)
		false alarms	0.123(0.134)	0.094(0.102)	0.137(0.084)
	Shallow	proportion of correct	0.618(0.119)	0.608(0.120)	0.545(0.106)
		hit rate	0.531(0.204)	0.512(0.223)	0.381(0.185)
		false alarms	0.213(0.225)	0.204(0.235)	0.140(0.176)
Source memory	Deep		0.381(0.129)	0.375(0.156)	0.410(0.142)
	Shallow		0.263(0.100)	0.249(0.121)	0.188(0.110)

We conducted 2 (encoding level: deep vs. shallow) ×3 (emotion category: fear, disgust vs. neutral) ANOVAs of item memory and source memory, respectively.

Item memory The results showed a significant main effect of emotion category [F(2, 84) = 5.248, p = 0.011, η_p² = 0.111], suggesting superior memory performance for fearful (p = 0.042) and disgusting (p = 0.020) compared to neutral images, no significant difference in memory performance between fearful and disgusting images (p = 1). There was also a significant main effect of encoding level [F(1, 42) = 75.990, p < 0.001, η_p² = 0.644], suggesting worse memory performance in the shallow encoding than in the deep encoding. Moreover, the interaction effect between encoding level and emotion category was significant [F(2, 84) = 4.570, p = 0.019, η_p² = 0.098]. Specifically, in the shallow encoding, participants had better memory performance for fearful (p = 0.007) and disgusting images (p < 0.001) than for neutral images, no difference in memory performance between fearful and disgusting images (p = 1). However, in the deep encoding, there was no significant difference in memory performance for fearful, disgusting and neutral images (ps = 1). In addition, for each emotion category, deep encoding produced higher performance than shallow encoding (ps < 0.001) (Fig. 2).

Fig. 2.

Proportion correct for the item memory and source memory in Experiment 1

Error bars represent standard errors. ^*p < 0.05; ^**p < 0.01; ^***p < 0.001

Source memory The main effect of emotion category was not significant [F(2, 84) = 0.647, p = 0.526, η_p² = 0.015]. There was a significant main effect of encoding level [F(1, 42) = 24.611, p < 0.001, η_p² = 0.369], suggesting worse memory performance in the shallow encoding than in the deep encoding. Moreover, the interaction effect between encoding level and emotion category was significant [F(2, 84) = 4.163, p = 0.019, η_p² = 0.090]. However, in the shallow encoding, there was no significant difference in memory performance for fearful, disgusting and neutral images (p_{fearful vs. disgusting} = 1, p_{fearful vs. neutral} = 0.090, p_{disgusting vs. neutral} = 0.140). In the deep encoding, there was no significant difference in memory performance for fearful, disgusting and neutral images (p_{fearful vs. disgusting} = 1, p_{fearful vs. neutral} = 1, p_{disgusting vs. neutral} = 0.594). In addition, participants had better memory performance for fearful, disgusting and neutral image in the deep encoding than in the shallow encoding (p_fearful = 0.002, p_disgusting = 0.005, p_neutral < 0.001) (Fig. 2).

Discussion

The Experiment 1 found that compared to neutral images, fearful and disgusting images showed stronger memory effects specifically for emotional items in the shallow encoding task. However, this enhanced memory effect was not observed in the deep encoding task. Consistent with previous finding, the results of Experiment 1 indicate that emotional and neutral tend to be amplified in the shallow encoding [34, 35, 36]. The further results suggest that threat-related information (e.g., fear and disgust) may capture more attention resources through a bottom-up attention capture mechanism [37], subsequently exhibiting enhanced memory consolidation in the shallow encoding task. However, the top-down cognitive processing approach results in elaborate processing of all emotional stimuli [38], which, in turn, diminishes the advantages associated with the specific processing of emotional memory in the deep encoding task.

However, for source memory, regardless of whether it was in the deep or shallow encoding task, we did not observe a significant advantage in emotional memory. This result may have been influenced by the incidental and intentional encoding strategies [39, 40]. In Experiment 1, we employed an incidental encoding task and instructed participants to judge the extrinsic source of the information, involving between-item associations (e.g., the context in which the car or phone was seen) [41, 42]. It is plausible that emotional stimuli automatically attract more attentional resources, thereby potentially overshadowing the contextual information associated with these emotions. Consequently, in Experiment 2, we shifted our focus from incidental memory to intentional memory in order to further examine the impact of encoding levels on item memory and source memory for emotional stimuli.

Experiment 2

Methods

Participants

We calculated the required sample size using G*Power (version 3.1) [27], which yielded a total of 44 participants, assuming a medium effect size (f = 0.25 [28, 29, 30]) for a 2 × 3 mixed experimental design (F-tests, ANOVA: Repeated Measures, within-between interaction) with 95% statistical power (α = 0.05) and the number of measurements = 3. The correlation among repeated measures (0.5) and the epsilon for sphericity correction (1) were set to be the default. Overall, 59 healthy Chinese undergraduate and graduate students participated in this study as paid volunteers. To ensure the validity of the data, three participants (one from the deep encoding group and two from the shallow encoding group) who reported verbally that they could not remember the surrounding frames during the encoding phase were excluded from the analysis after the experiment. The final sample consisted of 29 participants in the deep encoding group (20 female, mean age = 22.38 ± 3.59) and 27 participants in the shallow encoding group (20 female, mean age = 21.48 ± 2.69). Participants were randomly assigned to either the deep or the shallow encoding group. All participants were right-handed and had normal or corrected-to-normal visual acuity. They provided written informed consent prior to participation, and the study was approved by the university ethics committee.

Apparatus and stimuli

The apparatus and materials used in Experiment 2 were the same as those adopted in Experiment 1.

Procedures

The procedures in Experiment 2 were identical to those in Experiment 1, except that participants were instructed to remember both the images and their associated contexts (window or photo frame) during the encoding phase.

Data analysis

Data analyses were identical to those in Experiment 1.

Results

Table 2 shows the proportion correct for the item memory and source memory in Experiment 2.

Table 2.

Proportion correct for the item memory and source memory in experiment 2

			Emotion Category
	Encoding level		Fear(SD)	Disgust(SD)	Neutral(SD)
Item memory	Deep	proportion of correct	0.814(0.133)	0.822(0.112)	0.807(0.119)
		hit rate	0.778(0.160)	0.780(0.120)	0.767(0.158)
		false alarms	0.113(0.150)	0.095(0.102)	0.113(0.110)
	Shallow	proportion of correct	0.653(0.124)	0.653(0.128)	0.584(0.116)
		hit rate	0.590(0.209)	0.567(0.223)	0.444(0.172)
		false alarms	0.223(0.225)	0.180(0.234)	0.149(0.115)
Source memory	Deep		0.396(0.125)	0.414(0.121)	0.376(0.138)
	Shallow		0.375(0.142)	0.301(0.131)	0.232(0.112)

Item memory The results showed a significant main effect of emotion category [F(2, 108) = 7.256, p = 0.001, η_p² = 0.118], suggesting better memory performance to fearful (p = 0.019) and disgusting (p < 0.001) than neutral images, no difference in memory performance between fearful and disgusting images (p = 1). There was also a significant main effect of encoding level [F(1, 54) = 39.211, p < 0.001, η_p² = 0.421], suggesting better memory performance in the deep encoding than in the shallow encoding. Moreover, the interaction effect between encoding level and emotion category was significant [F(2, 108) = 3.776, p = 0.026, η_p² = 0.065]. Specifically, in the shallow encoding, participants had better memory performance for fearful (p = 0.002) and disgusting images (p < 0.001) compared to neutral images, no difference in memory performance between fearful and disgusting images (p = 1). However, in the deep encoding, there was no significant difference in memory performance for fearful, disgusting and neutral images (p_{fearful vs. disgusting} = 1, p_{fearful vs. neutral} = 1, p_{disgusting vs. neutral} = 0.945). In addition, for each emotion category, deep encoding produced higher performance than shallow encoding (ps < 0.001) (Fig. 3).

Fig. 3.

Proportion correct for the item memory and source memory in Experiment 2

Error bars represent standard errors. ^*p < 0.05; ^**p < 0.01; ^***p < 0.001

Source memory The main effect of emotion category was significant [F(2, 108) = 10.145, p < 0.001, η_p² = 0.158], suggesting better memory performance to fearful (p < 0.001) and disgusting images (p = 0.011) than neutral images, no significant difference in memory performance between fearful and disgusting images (p = 0.491). There was also a significant main effect of encoding level [F(1, 54) = 11.975, p = 0.001, η_p² = 0.182], suggesting better memory performance in the deep encoding than in the shallow encoding. Moreover, the interaction effect between encoding level and emotion category was significant [F(2, 108) = 5.236, p = 0.009, η_p² = 0.088]. Specifically, in the shallow encoding condition, participants had better memory performance for fearful images than disgusting (p = 0.018) and neutral images (p < 0.001), and better memory performance for disgusting images than neutral images (p = 0.026), whereas there was no significant difference in memory performance for fearful, disgusting and neutral images in the deep encoding (p_{fearful vs. disgusting} = 1, p_{fearful vs. neutral} = 1, p_{disgusting vs. neutral} = 0.387). In addition, participants showed better memory performance for disgusting and neutral images in the deep encoding than in the shallow encoding (p_disgusting = 0.002, p_neutral < 0.001), but memory performance for fearful images did not show a significant difference between shallow and deep encoding (p = 0.281) (Fig. 3).

Discussion

Consistent with the results of Experiment 1, the results of Experiment 2 also indicated that the level of encoding affected item memory performance. Specifically, emotional memories (fear and disgust) were remembered better than neutral stimuli only in the shallow encoding tasks. Overall memory was strong, but the enhancing effects of emotion disappeared in the deep encoding tasks. However, the disgust memory enhancement effect was still not found in the experiment. This result may be attributed to item memory, which primarily relies on automatic processes that convey a sense of prior exposure to an event without requiring detailed encoding [43, 44], thus facilitating relatively easy retrieval. Additionally, fear and disgust are both negative stimuli that can elicit strong emotional responses and may share similar processing mechanisms in memory. Therefore, it was difficult to show differences in their memory performance.

For source memory, the results indicated that fearful emotions enhanced the memorization of contextual information. That may be due to the fact that fear triggered an expansion of the visual field and accelerated eye movements [19], physiological responses that enhanced the detection of potential threats in the environment. Fear-related stimuli further amplified the perception of imminent danger, prompting heightened attention to the surroundings [45]. This increased attentional focus helped ensure safety and facilitated the processing of contextual information. However, in the deep encoding task, we did not observe a source memory advantage for disgusting stimuli. This may be due to perceptual differences between fearful and disgusting stimuli. Unlike fear, which typically broadens visual attention, disgust tends to induce visual narrowing [12, 19]. This narrowing may cause individuals to focus more on intrinsic source details while neglecting surrounding background information. Intrinsic source details refer to features that are integral to an item (i.e., within-item features), such as perceptual properties (e.g., the color of an image or font) [41, 42, 46]. Given these findings, in Experiment 3, we aimed to investigate how the level of encoding affects item and intrinsic source memory for emotional stimuli.

Experiment 3

Methods

Participants

We calculated the required sample size using G*Power (version 3.1) [27], which yielded a total of 44 participants, assuming a medium effect size (f = 0.25 [28, 29, 30]) for a 2 × 3 mixed experimental design (F-tests, ANOVA: Repeated Measures, within-between interaction) with 95% statistical power (α = 0.05) and the number of measurements = 3. The correlation among repeated measures (0.5) and the epsilon for sphericity correction (1) were set to be the default. Overall, 59 healthy Chinese undergraduate and graduate students participated in this study as paid volunteers. However, two participants were excluded due to data collection problems. The final sample consisted of 27 participants in the deep encoding group (24 female, mean age = 21.07 ± 1.73) and 30 participants in the shallow encoding group (28 female, mean age = 22.17 ± 2.12). Participants were randomly assigned to either the deep or the shallow encoding group. All participants were right-handed and had normal or corrected-to-normal visual acuity. They provided written informed consent prior to participation, and the study was approved by the university ethics committee.

Apparatus and stimuli

The stimuli were as described in Experiment 1, except that the frame contexts were replaced by color contexts (e.g., red or green color overlays applied to the images). Specifically, we converted 150 images to grayscale in Photoshop, and among them, 99 images from the encoding phase were modified to either red or green. Using the Colorize option in the Hue/Saturation settings, adjustments were made to the Hue, Saturation, and Brightness channels, resulting in two versions of each image: one in red and one in green. During the encoding phase, the red or green images were presented, while the grayscale versions were shown during the retrieval phase.

Procedures

The procedures in Experiment 3 were similar to those in Experiment 2. The key difference was that participants were instructed to remember both the images and their associated contexts (red or green) during the encoding phase. During the retrieval phase, 99 old images from the encoding phase (presented in grayscale and without color contexts) were mixed with 17 new grayscale images from each emotion category to give a total of 150 images. Then participants were first asked to judge whether the image was old or new by pressing either the “F” key or the “J” key. If the image was judged to be old, participants were then asked to immediately indicate the associated context by pressing the “F” key for red and the “J” key for green. If the image was judged to be new, the next image was immediately displayed. There was no time limit for responses during the retrieval task (Fig. 4).

Fig. 4.

Illustration of the experimental stimuli and procedure in Experiment 3. In the shallow encoding task, participants were asked to judge whether the images were on the left or right side of the monitor (Encoding phase a). In the deep encoding task, participants were asked to rate the arousal level of the images (Encoding phase b)

Data analysis

Data analyses were identical to those in Experiment 1.

Results

Table 3 shows the proportion correct for the item memory and source memory in Experiment 3.

Table 3.

Proportion correct for the item memory and source memory in experiment 3

			Emotion Category
	Encoding level		Fear(SD)	Disgust(SD)	Neutral(SD)
Item memory	Deep	proportion of correct	0.825(0.112)	0.802(0.093)	0.803(0.127)
		hit rate	0.808(0.165)	0.759(0.124)	0.797(0.181)
		false alarms	0.139(0.092)	0.107(0.086)	0.185(0.099)
	Shallow	proportion of correct	0.739(0.102)	0.690(0.088)	0.649(0.116)
		hit rate	0.701(0.143)	0.759(0.124)	0.797(0.181)
		false alarms	0.179(0.147)	0.135(0.114)	0.208(0.117)
Source memory	Deep		0.446(0.141)	0.380(0.106)	0.460(0.154)
	Shallow		0.438(0.150)	0.348(0.114)	0.354(0.170)

Item memory The results showed a significant main effect of emotion category [F(2, 110) = 13.636, p < 0.001, η_p² = 0.199], suggesting better memory performance to fearful than disgusting (p < 0.001) and neutral images (p < 0.001), no significant memory performance between disgusting and neutral images (p = 0.287). There was also a significant main effect of encoding level [F(1, 55) = 21.428, p < 0.001, η_p² = 0.280], suggesting worse memory performance in the shallow than in the deep encoding. Moreover, the interaction effect between encoding level and emotion category was significant [F(2, 110) = 4.881, p = 0.009, η_p² = 0.082]. Specifically, in the shallow encoding condition, participants had better memory performance for fearful images than disgusting (p = 0.001) and neutral images (p < 0.001), and better memory performance for disgusting than neutral images (p = 0.045), whereas there was no significant difference in memory performance for fearful, disgusting and neutral images in the deep encoding (p_{fearful vs. disgusting} = 0.291, p_{fearful vs. neutral} = 0.543, p_{disgusting vs. neutral} = 1). In addition, for each emotion category, deep encoding produced higher performance than shallow encoding (p_fearful = 0.003, p_disgusting < 0.001, p_neutral < 0.001) (Fig. 5).

Fig. 5.

Proportion correct for the item memory and source memory in Experiment 3

Error bars represent standard errors. ^*p < 0.05; ^**p < 0.01; ^***p < 0.001

Source memory The main effect of emotion category was significant [F(2, 110) = 8.609, p < 0.001, η_p² = 0.135], suggesting better memory performance to fearful images than disgusting images (p < 0.001), neither significant difference in memory performance between disgusting and neutral images (p = 0.080) nor fearful and neutral images (p = 0.337). There was not significant main effect encoding level [F(1, 55) = 2.569, p = 0.115, η_p² = 0.045]. There was also significant interaction effect between encoding level and emotion category [F(2, 110) = 3.707, p = 0.034, η_p² = 0.063]. Specifically, in the deep encoding, participants had better memory performance for fearful (p = 0.013) and neutral images (p = 0.015) than for disgusting images, and no difference in memory performance between fearful and neutral images (p = 1), whereas in the shallow encoding, participants had better memory performance for fearful images than disgusting (p < 0.001) and neutral images (p = 0.020), and no difference in memory performance between disgusting and neutral images (p = 1). In addition, participants showed better memory performance for neutral images in the deep encoding compared to those in the shallow encoding (p = 0.017), but the memory performance for both fearful images and disgusting images showed no significant difference between shallow and deep encoding (p_fearful = 0.824, p_disgusting = 0.277) (Fig. 5).

Discussion

The results of Experiment 3 also indicated that the emotion memory enhancement effect was observed in the shallow encoding task rather than in the deep encoding task. Contrary to expectations, the results revealed fear’s memory-enhancing effects rather than those of disgust, which was inconsistent with previous findings [2, 5, 6, 8, 9]. This result may be attributed to the use of artificially colored scenes, which allowed participants to mentally segregate color and scene information rather than integrating them as parts of a single object [41]. Such manipulation increased task difficulty, thereby influencing the allocation of cognitive resources during encoding. When cognitive resources were insufficiently allocated, the processing of disgusting stimuli was disproportionately affected.

Source memory indicated that fear (vs. disgust and neutral) enhanced color source recognition in the shallow encoding task, whereas fear and neutral (vs. disgust) enhanced color source recognition in the deep encoding tasks. These differences in source memory may be attributed to the survival relevance of fearful stimuli. When cognitive resources are limited, stimuli related to survival are more likely to be prioritized for processing [47]. During memory formation, the brain preferentially processes and stores survival-relevant stimuli [37], ensuring accurate recognition of environmental information in the face of threats. This defensive response facilitates individuals’ ability to maintain vigilance and take appropriate action in threatening situations. In contrast, the uncertain and ambiguous nature of disgusting stimuli (i.e., whether they pose a threat to health or survival) [9] directs individuals’ attention toward the content of the stimulus itself during processing. Notably, colors overlay may increase the difficulty of the encoding task, requiring greater cognitive resource allocation to discriminate threatening stimulus attributes, potentially at the expense of source memory encoding.

General discussion

Our study revealed that encoding level had differential effects on item and source memory for emotional stimuli. As predicted, the encoding level (shallow vs. deep) modulated the emotional effect on item and source memory performance. For item memory, the emotional memory enhancement effect was observed in the shallow encoding condition but not in the deep encoding condition, with fear’s memory-enhancing effects specifically demonstrated in Experiment 3. For source memory, Experiment 1 showed no significant effects of fearful, disgusting, and neutral emotions on source memory, regardless of whether the encoding task was shallow or deep. However, the influence of encoding level was evident in Experiments 2 and 3. Specifically, in these experiments, fearful stimuli enhanced source memory (e.g., frame and color) in the shallow encoding task. Notably, in Experiment 3, the disgusting stimuli were found to impair color source memory.

In the current study, item memory for emotional stimuli could be modulated by the level of encoding. Specifically, for item memory, the emotional memory enhancement effect was observed only in the shallow encoding condition, whereas the difference in memory performance disappeared in the deep encoding task. This finding is in line with previous studies [13, 34]. These results may involve different memory mechanisms for shallow and deep encoding [48]. Shallow encoding relies more on sensation and attention [48, 49, 50, 51]. The emotionality hypothesis suggests that emotional stimuli preferentially draw our attention and receive preferential processing in our cognitive processes [52]. Fear and disgust, two highly arousing and negative emotions that may threaten individual survival, are more likely to attract attention than neutral emotions [53, 54, 55]. The sensory specificity and vividness produced by the repetition of fearful and disgusting stimuli, as well as the privileging of negative stimulus memories, leads to their retention in memory [56]. Moreover, the amygdala-medial temporal lobe network, particularly the hippocampus, modulates emotional arousal and automates memory [57, 58, 59]. Consequently, individuals automatically engage in additional processing of negative emotions associated with these memories. In contrast, deep encoding relies on semantic and conceptual processing [50, 51, 60], in which individuals are guided to process information in a more detailed and meaningful way. This leads individuals to focus more on understanding and remembering the content of the information itself rather than the emotional features of the information. Such deep processing may diminish the relative importance of emotional factors, and thus the emotional memory enhancement effect may be diminished by more complex semantic processing. Thus, the effect of emotional stimuli is more pronounced in the shallow encoding situations. Our findings indicate that when all stimuli (fearful, disgusting and neutral) are subjected to the cognitive effort elicited by an evaluative instruction, subsequent recognition memory for all stimuli is strong and seems to outweigh the presence of emotion enhancement.

Surprisingly, however, we did not observe the item memory advantage for disgusting stimuli, which was inconsistent with previous studies [2, 5, 6, 14, 15]. In contrast, the fearful emotional memory advantage was observed compared to disgusting and neutral stimuli in Experiment 3. This discrepancy may be due to the greater level of uncertainty and ambiguity associated with disgust compared to fearful stimuli [9]. Furthermore, color overlay may have increased task difficulty, thereby influencing the allocation of cognitive resources during encoding [41]. When cognitive resources were insufficiently allocated, the processing of disgusting stimuli may be disproportionately affected. In contrast, fearful stimuli, which were more directly relevant to survival, elicit rapid and automatic negative effects that threaten survival [15]. The threat information conveyed by fearful stimuli is perceived as more urgent, facilitating automatic processing by the brain [9], and particularly during earlier cognitive stages [53, 61, 62]. Thus, when cognitive resources were relatively limited, individuals prioritize the processing of fearful stimuli.

In the present study, in Experiment 1, we did not observe significant differences in source memory across different emotional stimuli. According to the attentional narrowing theory, emotional stimuli automatically captured more attention, causing individuals to allocate more cognitive resources to processing the central aspects of the stimuli while neglecting the processing of background information [63]. This attentional bias may explain the absence of specific emotional source memory advantages in Experiment 1. However, in Experiments 2 and 3, participants were explicitly instructed to encode both the central aspects of the stimuli and the background information. This procedural change likely prompted individuals to consciously allocate cognitive resources to processing background information. Consequently, in the shallow encoding conditions, we observed a fear-related source memory advantage in Experiments 2 and 3, consistent with our predictions. One possible explanation for this finding is that differences in the evolutionary properties and perception of fearful and disgusting stimuli influence source memory. Fear typically involves an imminent threat to the individual, requiring urgent and rapid withdrawal responses rather than prolonged cognitive thought [15, 64]. It enhances the perception and processing of potential threat-related information in the environment [15, 45]. Fear may also activate neural pathways related to memory and attention [53, 62, 65], heightening sensitive and alertness to fear-related stimuli [66], thereby enhancing memory performance. Moreover, according to the principle of least effort, fearful stimuli have already been thoroughly processed in earlier cognitive stages [67], reducing the need for repeated of the item itself during the later encoding stage. As a result, individuals may allocate more attentional resources to processing source information. However, during the deep encoding, this processing advantage may extend to other emotional stimuli, potentially diminishing differences in source memory across emotions. Therefore, fearful stimuli show a stronger source memory advantage in the shallow encoding conditions. This finding further demonstrates that fear and disgust not only activate distinct sensory representations but also influence subsequent cognitive processes and behaviors in a manner consistent with their evolutionary purpose.

However, it is noteworthy that the present study did not observe the expected disgust-related source memory advantage. Instead, in the deep encoding conditions in Experiment 3, disgusting stimuli appeared to impair source memory performance compared to neutral stimuli. This finding contradicted our initial hypothesis and previous research [6, 8]. A potential explanation for this discrepancy may be attributed to the increased task difficulty in the current study. In prior studies, participants were typically not required to simultaneously encode additional source information, which allowed sufficient cognitive resources to be devoted to the detailed processing of item stimuli. Furthermore, the retrieval phase in previous studies often did not involve specific source judgments, thereby reducing the overall cognitive demand on source memory tasks [6, 8]. In contrast, participants in the present study were required to simultaneously encode both the item stimuli and additional source information, which likely resulted in competition for cognitive resources between item and source encoding. Given the inherent uncertainty and ambiguity of disgusting stimuli, these stimuli demand greater cognitive resources for detailed processing. Consequently, when cognitive resources are limited, individuals may prioritize the processing of item information over source information. This prioritization negatively impacted source memory performance, particularly for specific attributes such as frame or color. As task difficulty increased, this competition for resources may even suppress disgust-related source memory performance.

There were still some limitations to the current study. First, the imbalance in the sample size of male and female participants in this study may have influenced the results. Evidence suggested that males and females employed different strategies during emotional processing, and sex differences existed in the neural activation patterns underlying emotion processing [68]. Therefore, achieving a more balanced sample in future studies would help control for the potential influence of sex on the memory retrieval of emotional stimuli. Second, although we controlled for the valence, arousal, and specific levels of fear and disgust in the stimuli to minimize their potential impact on memory performance, the present study did not account for potential differences in distinctiveness and organisation between fearful and disgusting stimuli. These differences could have potentially explained the observed disparities in memory performance [5, 6]. Future research should examine whether the differences in memory performance between fearful and disgusting images persist when the distinctiveness and organizational properties of the stimuli are adequately controlled. Third, the present study only included threat-related stimuli (fear and disgust) and did not examine positive emotional stimuli that may also possess adaptive significance. This omission limits our ability to determine whether the observed memory enhancements are specific to threat-related stimuli with evolutionary significance. It is also unclear whether similar memory advantages would be found for other highly motivational stimuli, such as positive emotional stimuli that are also important for survival and adaptation. Therefore, the present findings should be interpreted with caution, and future studies are needed to include positive, adaptively significant stimuli to further validate these effects. Finally, the present study focused solely on analyzing correct responses during the retrieval phase, without considering the potential impact of attentional awareness during the encoding phase on fear and disgust retrieval. The adoption of eye-movement techniques could provide insight into the temporal accuracy of processing emotional stimuli, allowing for a better understanding of the perceptual and cognitive processes underlying these two emotions. Therefore, future research should explore the differential effects of fear and disgust learning on eye-tracking to further our understanding.

Conclusion

The present study represents the first comparative analysis of the effects of encoding level on item- and source memory for fear and disgust. Our results revealed that for item memory, the enhancement effect of fear and disgust was only observed in the shallow encoding (vs. in the deep encoding). For source memory, fearful stimuli enhanced the source memory (frame and color) in the shallow encoding, whereas disgusting stimuli impaired the color source memory in the deep encoding. These findings suggest that memory performance for fear and disgust was modulated by emotional specificity and cognitive resource allocation. They enhance our understanding of how different emotion types of information affect memory and clarify the distinct cognitive mechanisms underlying item memory and source memory for negative emotions.

Author contributions

Yuan Yuan: Conceptualization, Data curation, Formal analysis, Writing – original draft, Writing – review & editing. Yingjie Jiang: Conceptualization, Formal analysis, Writing – original draft, Writing – review & editing. Lili Guan: Writing – original draft. Yingting Long: Writing – review & editing. Yanlin Guo: Writing – review & editing. Xiaojing Zhang: Writing – review & editing.

Funding

This work was supported by the National Natural Science Foundation of China (32271095).

This work was supported by the Natural Science Foundation of Jilin Province (20230101149JC).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

The research was examined and approved by the Northeast Normal University ethics committee (Approval No. 2020038). Accordance: The work described has been carried out in accordance with the ethical standards of the institutional research committee and/or the national research committee, and with the 1964 Helsinki declaration for experiments involving human participants. Informed consent was obtained from all individual participants included in the study. This article does not contain any studies involving animals.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.