Serial dependencies for externally and self-generated stimuli

Abstract

Our senses are constantly exposed to external stimulation. Part of the sensory stimulation is produced by our own movement, like visual motion on the retina or tactile sensations from touch. Sensations caused by our movements appear attenuated. The interpretation of current stimuli is influenced by previous experiences, known as serial dependencies. Here we investigated how sensory attenuation and serial dependencies interact. In Experiment 1, we showed that temporal predictability causes sensory attenuation. In Experiment 2, we isolated temporal predictability in a visuospatial localization task. Attenuated stimuli are influenced by serial dependencies. However, the magnitude of the serial dependence effects varies, with greater effects when the certainty of the previous trial is equal to or greater than the current one. Experiment 3 examined sensory attenuation's influence on serial dependencies. Participants localized a briefly flashed stimulus after pressing a button (self-generated) or without pressing a button (externally generated). Stronger serial dependencies occurred in self-generated trials compared to externally generated ones when presented alternately but not when presented in blocks. We conclude that the relative uncertainty in stimulation between trials determines serial dependency strengths.

Introduction

Humans are not passive receivers of sensory information; we actively explore and interact with our surroundings. For instance, while walking through a forest, our senses process various external stimuli, such as the wind whistling through the leaves or the chirping of animals. Simultaneously, our movements generate additional sensory input, like the visual motion on our retinas when we move our eyes or the tactile sensations when we accidentally touch a branch. To make sense of this sensory input, our brain distinguishes between stimuli that are externally generated (from the environment) and those that are self-generated (caused by our own actions). This distinction is facilitated by a phenomenon known as sensory attenuation, where the perceived intensity of self-produced stimuli is reduced (Blakemore, Frith, & Wolpert, 1999; Hughes, Desantis, & Waszak, 2013a). Sensory attenuation helps us differentiate between sensations caused by our actions and those originating from the external world (Bays, Wolpert, & Flanagan, 2005; Bays, Flanagan, & Wolpert 2006; Blakemore et al., 1999). The classical interpretation of sensory attenuation involves a forward model architecture (Miall & Wolpert, 1996; Welniarz, Worbe, & Gallea, 2021). According to this model, an efference copy is generated concurrently with a motor command and sent to sensory areas to predict the feedback from the action. If the sensory and predicted feedback match, sensory attenuation occurs (Desmurget & Grafton, 2000; Jordan & Rumelhart, 1992; Welniarz et al., 2021). However, contrasting results have also been found in this area. For instance, Schwarz, Pfister, Kluge, Weller, & Kunde (2018) failed to find evidence of visual sensory attenuation. Current research suggests other mechanisms, such as perceptual upweighting (Thomas, Yon, de Lange, & Press, 2022) and prediction errors (Yon, Gilbert, de Lange, & Press, 2018), may also contribute to attenuation. Despite debates on the exact mechanism, all experimental data show that self-produced stimuli, although attenuated, are still detectable. Sensory attenuation is not a complete cancellation of conscious stimulus detection. 

Sensory input is not only compared against internal predictions but also against stimuli encountered in the recent past. Our perception of sensory features is influenced by features we previously detected, a phenomenon known as serial dependency (Cicchini, Mikellidou, & Burr, 2017; Cicchini, Mikellidou, & Burr, 2018; Fischer & Whitney, 2014). Serial dependencies stabilize our visual experience by reducing uncertainty and creating smooth perception (Burr & Cicchini, 2014). 

In the current study, we asked how both phenomena, sensory attenuation, and serial dependencies, interact. On the one hand, serial dependencies could distinguish between self- and externally generated sensations. In this view, serial dependencies for a current self-generated event should only be found if previous events were self-generated too. On the other hand, serial dependencies could be independent of sensory attenuation and smoothen current according to past experiences, irrespective of whether they were self- or externally generated. The latter case matches the different functional goals of sensory attenuation and serial dependencies. Sensory attenuation dampens stimulus intensity because of their predictability, whereas serial dependencies support stimulus interpretation. A self-generated event, for instance, the sound produced by a press on a piano key, might appear attenuated, but we still need to distinguish the sound intensity to evaluate whether the keys were stroked correctly. Temporal predictability is a confounding factor in sensory attenuation of self-generated stimuli. For instance, producing a visual stimulus by pressing a key allows for perfect prediction of when the stimulus occurs, potentially producing sensory attenuation because of temporal predictability (Hughes, Desantis, & Waszak, 2013b). Although temporal predictability has been discussed as a prerequisite for auditory sensory attenuation (Schafer & Marcus, 1973; Sowman, Kuusik, & Johnson, 2012), recent studies found sensory attenuation even when stimuli were temporally unpredictable (Bäss, Jacobsen, & Schröger, 2008; Lubinus et al., 2022). Horváth, Maess, Baess, & Tóth (2012) proposed that mere temporal contiguity between an action and its resultant stimulus, even when its occurrence is not predictable, is sufficient to elicit attenuation effects. 

In the current study, we first measured contrast sensitivity for visual stimuli while experimentally varying the factors of self-generation and temporal stimulus predictability. We hypothesized that contrast sensitivity would be lower when stimuli were temporally predictable. We then examined serial dependencies in spatial localization. Previous studies showed that for localizing objects in the periphery, the visual system takes into account the prior locations of that object (Manassi, Liberman, Kosovicheva, Zhang, & Whitney, 2018). Errors in localization were biased by the object's position in preceding trials. Cont and Zimmermann (2021) reported similar biases when the previous localization involved saccades. We hypothesized that serial dependencies will increase when the preceding stimuli are less predictable and thus less attenuated. In separate trials, subjects either generated the localization stimulus themselves by pressing a button or the stimulus appeared automatically. We measured these trials either blockwise or intermingled and asked how sensory attenuation and temporal predictability would affect serial dependencies. 

General methods

Participants

In the first experiment, we tested N = 40 participants between the age of 19 to 27 (M_Age = 22.03, SD_Age = 3.29, 11 male). In the second and third experiments, we included N = 55 participants. Demographic data for Experiment 2 was (M_Age = 23.11, SD_Age = 3.7, Range_Age. = 18–31, 17 male) and for Experiment 3 it was (M_Age = 24.25, SD_Age = 4.07, Range_Age. = 18–35, 14 male). 

Informed consent was obtained from all participants. The experimental procedures followed the ethical guidelines stated in the Declaration of Helsinki and received approval from the local ethics committee of the mathematical and natural science faculty at Heinrich Heine University, Duesseldorf. Participants were recruited from the University of Duesseldorf and received compensation in the form of participation hours or an expense allowance. 

Apparatus

In all three experiments, participants were seated in front of a computer (Dell optiplex 7000). Experiments were built in MATLAB (version 2023a; MathWorks, Natick, MA, USA) using the Psychophysics Toolbox Version 3 (Brainard, 1997; Pelli, 1997; Kleiner, Pelli, Ingling, Murray, & Broussard, 2007) and presented on a Dell P2419H Screen with a refresh rate of 60 Hz placed at 57 cm from the observer. Before the start of each experimental condition, the screen was gamma-corrected. The background color was black during all experiments (0.6 cd/m²). Data analysis for all experiments was conducted with R Studio, Version 1.4.1103. 

Experiment 1—Visual sensory attenuation

Procedure—Experiment 1

In Experiment 1, participants were asked to fixate a blue square (0.55 × 0.55°) presented 6.5° left of the screen center. Next, a probe stimulus (gray square, 0.55 × 0.55°, stimulus contrast: 6.5 (Weber contrast) was presented for 500 ms at a location of 6.5° to the right of the screen center at the horizontal meridian. The Weber Contrast (L_ref − L_background)/L_background is calculated as the luminance of the reference minus the luminance of the background divided by the luminance of the background. After the probe stimulus disappeared, a reference stimulus (gray square, 0.55° × 0.55°) appeared for 500 ms at the same position as the probe stimulus. The reference stimulus varied in contrast between 4.4 and 8.6 in seven equidistant steps. Each stimulus intensity was presented 10 times. Every condition included 70 trials. Stimulus intensities were randomized throughout trials. Participants were asked to indicate which stimulus appeared brighter. Responses were given with right and left arrow keys. After entering the answer, the next trial started immediately. 

In total, participants had to undergo four conditions in Experiment 1. We varied between two factors between conditions: the factor control over the probe (self-generated vs. externally generated) and the factor predictability of the probe (predictable vs. unpredictable). The outline of conditions is shown in Figure 1A. 

Figure 1.

 

(A) Structure of conditions. We introduced four different conditions, with the visual probe being either predictable or unpredictable and self-generated or externally generated. During self-generated conditions, the visual probe appeared only after a self-generated button press. In externally generated conditions the probe appeared automatically after a certain time interval. When the probe occurred predictably, it appeared either directly after the button press (self-generated) or 1000 ms after the last trial (externally generated). In the unpredictable condition, the probe occurred unpredictably in a time interval between 1000 ms to 3000 ms. (B) Outline of one trial in Experiment 1. Participants were asked to either press a button to trigger the occurrence of a visual probe stimulus (self-generated) or to wait for the stimulus to occur automatically (externally generated). After 500 ms, a visual reference stimulus appeared at the same position, varying in contrast. Participants were asked to decide which stimulus appeared brighter.

During self-generated conditions (outlined on the left side of Figure 1A), participants were asked to press enter for the probe stimulus to occur. Contrary, in the externally generated conditions, a new trial started automatically one second after the last response (outlined on the right side of Figure 1A). As described in Lubinus et al. (2022) sensory attenuation effects decrease within a few hundred milliseconds after action execution (Aliu, Houde, & Nagarajan, 2009; Bays et al., 2005). In the externally generated trials, we thus do not expect the keypress from the last trial to influence the consecutive. In the temporal predictable conditions, the probe appeared directly after the button press (self-generated) or automatically one second after trial start (externally generated). In the temporal unpredictable condition, the probe appeared randomly between one to three seconds after the new trial start (externally generated) or button press (self-generated). The occurrence of conditions was randomized. The procedure in one trial can be seen in Figure 1B. 

Results—Experiment 1

In Experiment 1, we aimed to measure the classic effect of visual sensory attenuation. Participants were asked to judge differences in perceived contrast between two stimuli. Data were analyzed as psychometric functions (with a least square fit in R: 4.3.0, RStudio: 1.4.1103), plotting the proportion of times the reference was reported as brighter than the probe, as a function of the Weber Contrast of the reference stimulus. Distributions were fitted with cumulative Gaussian functions; the median of the curve estimated the point of subjective equality (PSE). A higher PSE represents a higher tendency of the first stimulus to be perceived as brighter. Thus lower PSEs show attenuation of the first stimulus. We calculated mean proportion of trials per stimulus level in which the reference stimulus was selected. Statistical analysis was performed in JASP 0.16.3 (Intel). Example psychometric functions for a participant can be seen in Figure 2. 

Figure 2.

 

Example of psychometric functions for one participant in the four conditions.

The four conducted conditions varied between two factors: the predictability of the probe (either predictable or unpredictable) and control over the probe (self-generated vs. externally generated). Attenuation was strongest in the predictable self-generated condition (PSE = 6.05, SD = 0.58) and lowest in the unpredictable self-generated condition (PSE = 6.48, SD = 0.5). A similar pattern was observed in the external conditions: externally predictable probes (PSE = 6.31, SD = 0.44) were more attenuated compared to externally generated unpredictable probes (PSE = 6.35, SD = 0.53). 

A repeated measures 2 × 2 analysis of variance (see Figure 3) showed that the predictability of the probes accounted for sensory attenuation (F(1, 39) = 6.057, p = 0.018, η² = 0.037). Control over the probe, whether externally or self-generated, did not influence performance (F(1, 39) = 0.059, p = 0.572). There was no interaction effect between the two factors (F(1, 39) = 3.152, p = 0.084). This indicates that temporal predictability of probes is responsible for sensory attenuation in the visual domain. 

Figure 3.

 

Results for Experiment 1. Weber contrast at which both visual stimuli appeared as equally intense as a function of temporal predictability of stimulus appearance. Data is shown for the externally generated (shown in red) and the self-generated (shown in green) condition. Error bars represent S.E.M.

Experiment 2—Serial dependencies and temporal predictability

Introduction—Experiment 2

In Experiment 2, we sought to find out whether deviating temporal predictability across conditions influenced serial dependencies. Participants were asked to indicate the seen probe positions of a red square (Weber contrast: 24.8). The probe appeared either predictable or unpredictable after an auditory signal. 

Procedure—Experiment 2

In Experiment 2 participants saw a black screen with a blue fixation square (0.55° × 0.55°) presented 6.5° left of the screen center. Subjects were instructed to fixate the fixation square during the whole trial, except when giving their answer. During a trial, a probe stimulus (red square, 0.55° × 0.55°, Weber contrast: 24.8) presented either at 4.5°, 5.5°, 6.5°, 7.5°, or 8.5° to the right of the screen center was flashed for 24 ms. As soon as the probe stimulus disappeared, the same red square, identical in physical properties to the probe stimulus, emerged in the lower right corner of the screen. Participants were asked to align the position of this red square with the location of the flashed probe stimulus. The red square was movable with the mouse. To confirm the response, participants clicked the mouse switch. After entering a response, the next trial started immediately with the presentation of the blue fixation square. 

We varied the temporal predictability of the probe stimuli. At the start of each trial, a tone (frequency: 1000 Hz) was presented. Three conditions were tested: predictable probe, unpredictable probe, and alternating predictable/unpredictable probe. In the predictable probe condition, the probe appeared 300 ms after the tone. In the unpredictable probe condition, the probe appeared randomly between one and three seconds after the tone. In the alternating condition, predictable and unpredictable trials were alternately presented. Conditions were presented randomly, with 231 trials per condition, excluding the first trial for analysis. The order of stimulus displacement location was predetermined to avoid extreme values following each other. Each combination of n and preceding trial probe locations occurred at least three times. 

Results—Experiment 2

For data preprocessing, we excluded trials with reaction times exceeding 10 seconds, trials in which the clicked position was more than 3 SD from the participant's mean difference in clicked and shown position, and trials in which the difference in visual angle between the shown and clicked target position exceeded 10°, because such probes would fall outside the peripheral area (see Manassi et al., 2018). We first analyzed whether serial dependencies were present in the different conditions. Serial dependencies would indicate that participants considered their responses from previous trials for the upcoming trial. We conducted linear regressions to check for serial dependencies. For each participant and condition, we fitted a linear regression with the difference between consecutive stimulus values (Preceding trial stimulus value − Current stimulus value) and the response error on the current trial (n) as the y-axis. We then averaged slopes across all participants and compared mean slope values between conditions using paired t-tests. 

Results for blocked conditions

Serial dependencies were present in both conditions. Participants relied on the probe position in preceding trials when giving their answer in current trials (predictable probe: t(54) = 6.877, p < 0.001, d = 0.93, unpredictable probe: t(54) = 6.891, p < 0.001, d = 0.93). There was no significant difference between blocked conditions of probes appearing either predictably (M = 0.169, SD = 0.18) or unpredictably (M = 0.153, SD = 0.17) as shown by a paired sample t-test (t(54) = 0.748, p = 0.457, see Figure 4A). 

Figure 4.

 

Results for Experiment 2. Average slopes from the linear regression are shown for unpredictable and predictable probe conditions. (A) Blocked conditions are shown. We found no significant effects between the extent of serial dependencies in the unpredictable and predictable conditions. (B) Alternating conditions are shown. Serial dependencies in unpredictable generated trials were less present than in predictable trials. Error bars represent S.E.M.

Results for alternating conditions

We distinguished between alternating predictable and unpredictable trials. In alternating predictable trials, the preceding trial was unpredictable whereas the current trial was predictable (M = 0.198, SD = 0.17). In alternating unpredictable trials, the pattern was vice versa as the preceding trial was predictable (M = 0.14, SD = 0.17). Also, for both alternating conditions, the predictable probe condition (t(54) = 8.454, p < 0.001, d = 1.14) and the unpredictable probe condition (t(54) = 5.972, p < 0.001, d = 0.81), we were able to find serial dependencies. 

Serial dependencies were significantly stronger when the current trial was predictable compared to unpredictable (t(54) = 2.714, p = 0.009, d = 0.37, see Figure 4B). This suggests that an unpredictable probe in the preceding trial leads to stronger serial dependencies in the current predictable trial, and vice versa. 

Experiment 3—Sensory attenuation and serial dependencies

Introduction – Experiment 3

In Experiment 2, we manipulated the temporal predictability for externally generated stimuli. We found that serial dependencies were stronger if the previous trial was less predictable than the current trial. Temporal predictability also differed for self- and externally generated stimuli. The temporal occurrence of self-generated stimuli was perfectly predictable. In Experiment 3, we tested serial dependencies for self- and externally generated stimuli. Given the results of Experiment 2, we expected that if the previous trial was externally generated then the current trial self-generated serial dependencies should be stronger than vice versa. 

Procedure—Experiment 3

In Experiment 3, temporal predictability was manipulated for both a self-generated attenuated condition and the externally generated condition. We presented three conditions: a blocked self-generated condition, a blocked externally generated condition, and a condition in which self-generated and externally generated trials alternated. In the blocked self-generated condition, the probe consistently appeared immediately following a keypress. Conversely, in the externally generated condition, the probe consistently appeared 1000 ms after the trial start. In the alternated conditions, we intermixed both trial types. We presented externally-generated and self-generated trials in alternating fashion until the end of the experiment. In the offline analysis, we determined both, the influence of a self-generated stimuli in trial n-1 on externally-generated stimuli in trial n and the influence of an externally-generated stimuli in trial n-1 on self-generated stimuli in trial n. 

In Experiment 3, participants saw a black screen with a blue fixation square (0.55° × 0.55°) presented 6.5° to the left of the screen center. Subjects were instructed to fixate on the square throughout the trial, except when giving their answer. The task was identical to Experiment 2, with participants aligning the position of a red square with the location of the flashed probe. 

Results—Experiment 3

As in Experiment 2, we differentiated between blocked conditions of self-generated and externally generated probes and a condition of alternated self-generated and externally generated probes. Data preprocessing was performed identically to Experiment 2. We averaged slopes across all participants and compared mean slope values between conditions using paired t-tests. 

Results for blocked conditions

Serial dependencies were evident in both conditions, as confirmed by a one-sample t-test against zero (externally generated trials: t(54) = 4.48, d = 0.6, p < 0.001; slope self-generated: t(54) = 5.42, p < 0.001, d = 0.73). There was no significant difference in serial dependencies between self-generated (M = 0.178, SD = 0.24) and externally generated trials (M = 0.134, SD = 0.22) (t(54) = −1.92, p = 0.06), indicating that serial dependencies are not influenced by whether the trials were self-generated or externally generated (see Figure 5A). 

Figure 5.

 

Results for Experiment 3. Average slopes from the linear regression are shown for externally and self-generated conditions. (A) Results for blocked condition. We found no significant effects between the strength of serial dependencies conditions. (B) Results for alternating condition. Serial dependencies in externally generated trials were less present than in self-generated trials. Error bars represent S.E.M.

Results for alternating condition

In the alternating condition, serial dependencies were present in both cases (externally generated: t(54) = 4.4, p < 0.001, d = 0.59; slope alternately self-generated: t(54) = 5.57, p < 0.001, d = 0.75). The strength of serial dependencies was significantly higher in the self-generated condition (M = 0.183, SD = 0.24) compared to the externally generated (M = 0.124, SD = 0.21), (t(54) = −2.92, p = 0.005, d = −0.39). This indicates that when the preceding trial was more predictable, information from the preceding trial is given more weight when responding in the current trial. Participants relied less on information from preceding trials when the preceding trial was self-generated (see Figure 5B). 

Discussion

In this study, we aimed to determine whether self-generated stimuli, whose intensity appears attenuated, affect the perception of subsequently presented stimuli. Additionally, we investigated the effect of temporal predictability, an inherent confound associated with self-generation of stimuli (Lubinus et al., 2022). Being the agent of an action implies precise knowledge of when the action will be initiated (Hughes et al., 2013b). Predicting the occurrence of a stimulation in time might be sufficient to evoke sensory attenuation. Consequently, we also measured the putative effect of temporal predictability on serial dependencies. 

In Experiment 1, we found sensory attenuation for visual stimuli as a consequence of temporal stimulus predictability. Stimuli appeared either immediately after a button press (predictable, self-generated) or after a random time interval (unpredictable, externally generated). Predictable stimuli were perceived as less intense, suggesting that temporal predictability allows the nervous system to anticipate the sensory consequences of an action, leading to sensory attenuation. Interestingly, self-generation itself, independent of predictability, did not significantly affect sensory attenuation. This partially contradicts previous research and highlights the complexity of processing self-generated stimuli. The role of temporal predictability in sensory attenuation has been frequently discussed in the literature. Klaffehn, Houde, and Nagarajan (2019) found that sensory attenuation occurred for self-generated and externally generated stimuli when both were identically temporal predictable. Harrison et al. (2021) concluded that temporal predictability is not necessary but reinforcing for the effect of sensory attenuation. 

In Experiment 2, an acoustic cue either predicted or did not predict the temporal stimulus appearance. We measured serial dependencies in a visual localization task when the stimulus was temporally predictable or unpredictable. In contrast to Experiment 1, we used a brighter color stimulus to facilitate easier localization and improve overall task performance. Serial dependencies refer to the tendency of individuals to be influenced by previously encountered stimuli when reproducing or classifying new stimuli (Cicchini et al., 2017; Cicchini et al., 2018). Recent studies have shown serial dependencies for various perceptual features (Cicchini et al., 2018; Cicchini, Benedetto, & Burr, 2021; Fischer & Whitney, 2014; Kiyonaga, Scimeca, Bliss, & Whitney, 2017; Liberman, Fischer, & Whitney, 2014; Manassi et al., 2018; Murai & Whitney, 2021). 

Our findings in Experiment 2 indicate that serial dependencies persisted in both predictable and unpredictable conditions. Bliss, Sun, and D'Esposito (2017) demonstrated that the length of the interval between consecutive trials significantly influences the magnitude of serial dependence. Attractive serial dependencies increased the longer the temporal interval between trials was. In our unpredictable trials, participants experienced a delay of one to three seconds before the probe appeared, resulting in a variable time interval between the current and preceding trials. In our self-generated trials, no such delay was presented. It could be argued that the stronger serial dependencies in cases in which the current trial is temporally more predictable than the previous trial result from the longer inter-trial interval. However, in that case also significant differences between the blocked conditions should be observed. Serial dependencies in blocks of unpredictable trials—because of their longer intertrial intervals—should be significantly higher than in blocks of unpredictable trials. However, this was not observed, neither in Experiment 2 nor 3. Our results can only be explained by considering how sensory attenuation affects the perception of subsequent stimuli. Specifically, when sensory attenuation occurs to a stimulus as a result of high predictability, its serial influence on the perception of a subsequent stimulus is less, especially when the subsequent stimulus is also not similarly attenuated. 

We then measured serial dependencies between self-generated and externally generated stimuli, both blockwise and in alternation. When testing only self-generated or only externally generated stimuli, strong serial dependencies were statistically indistinguishable between both session types. This suggests that self-generating a stimulus does not reduce serial dependencies when the immediately following stimulus is also self-generated. However, when self-generated and externally generated stimuli alternated within a session, there was a significant decrease in serial dependencies. 

According to the hypothesis that serial dependencies aim to reduce uncertainty, (Ceylan, Herzog, & Pascucci, 2021; Kim & Alais, 2021; Schlichting, Fritz, & Zimmermann, 2023; Van Bergen & Jehee, 2019) one might expect strong serial dependencies if the previous trial was self-generated and the current one externally generated. Because self-generated stimuli are temporally highly predictable, they should be more certain and thus influence the perception of externally produced stimuli, which are less predictable. However, we found the opposite result. When the previous trial was self-generated and the current one externally generated, serial dependencies were weaker than if the trial order was reversed. This result can be explained by invoking sensory attenuation. Experiment 1 showed that self-generated stimuli appear attenuated because of their high temporal predictability and are therefore perceptually less certain than externally produced stimuli. Serial dependencies increase the higher the relative difference in temporal uncertainty between the previous and the current trial. However, serial dependencies in purely self-generated sessions were strong because all stimuli had the same uncertainty. Our study suggests that serial dependencies were stronger when stimuli in a previous trial were more certain than those in the current trial. It is the relative uncertainty between two consecutive trials that determines if the interpretation of a current stimulus takes into account stimulation observed in the previous trial. The higher the relative certainty of the previous stimulation compared to the current trial, the higher its influence on current perception. 

A corollary of this reasoning is that serial dependencies should similarly be modulated if stimuli are only externally generated, but their temporal predictability is varied. In Experiment 1, we have shown that temporal predictability induces sensory attenuation, and Experiment 2 revealed that serial dependencies are stronger when the previous trial was not attenuated (potentially more certain) compared to when the previous trial was attenuated (potentially less certain) because of self-generation. However, serial dependencies were still present even in the latter case. In Experiment 3, stimuli in the externally produced condition appeared consistently 1000 ms after trial start. However, the estimate of a 1000 ms interval is still more uncertain than the prediction in the self-produced condition, where continuous internal information about the appearance of the stimulus is available. 

Conclusions

If the perception of a previous trial is attenuated, the information from that trial has a diminished influence on the current trial, resulting in weaker serial dependencies. Conversely, when both the previous and current trials are subjected to sensory attenuation, the effects of past events become more pronounced, strengthening serial dependencies. Serial dependencies do not distinguish if both, the previous and the current trial are self-generated or not. Instead, the strength of serial dependencies is modulated by the relative uncertainty of the previous and the current stimulus. Our findings demonstrate that sensory attenuation and serial dependencies affect stimulus perception independently of each other. The independence reflects their different functional goals. Sensory attenuation dampens the perception of predicted stimuli and serial dependencies support stimulus interpretation. 

Acknowledgments

Supported by the European Research Council (project moreSense, grant agreement 757184) and by the Deutsche Forschungsgemeinschaft (DFG, ZI 1456). 

Commercial relationships: none. 

Corresponding author: Clara Fritz. 

Email: [e-mail]. 

Address: Institute for Experimental Psychology, Heinrich Heine University Düsseldorf, Düsseldorf, Germany.