The study aimed to explore the impact of subliminal stimuli on the conscious experience of agency reported by Linser and Goschke (2007), while circumventing post hoc data selection and adjusting prime visibility prior to the main experiments. For experiment 1, which included semantic prime stimuli, we made use of JsQuest to adjust individual thresholds of stimulus intensity for prime visibility, based on 50 preceding trials. For experiment 2, utilizing figurative prime stimuli, we included an ISI of 64ms during the visible (i.e., “prime aware”) condition. For both experiments each participant underwent visible (i.e., “prime aware”) and non-visible (i.e., “prime unaware”) conditions. Our results revealed a fundamentally different picture to the observations reported by Linser and Goschke (2007), who found that participants overestimated their agency over objectively uncontrollable stimuli when congruent masked primes were presented just before the action. The effect was even more pronounced when participants were seemingly unaware of the primes. In our study, we observed in Exp. 1 that predictive information influenced the modulation of the subjective experience of agency only if said primes were processed consciously (Fig. 2). In Exp. 2, we did not observe significant variations in control judgements based on prime visibility or congruency. Differences between the conditions were found in the response times (Fig. 4). These observed differences in response times indicate that the visible primes were successful in inducing response priming.
Linser and Goschke (2007) interpreted their findings within the context of conscious will, and proposed that subliminal primes activate representations of the upcoming stimuli, which then modulate the experience of agency. Further, the authors proposed that unconsciously perceived primes might induce an effect akin to sensory attenuation, diminishing the sensory representation of subsequent stimuli and thereby influencing participants to perceive those stimuli as effects of their own actions. However, recent approaches to explain attenuation effects incorporate a more comprehensive predictive mechanism, extending beyond the exclusive influence of self-initiated motor predictions (Dogge et al., 2019). Efficient processing of sensory input is achieved through predictive cues, which guide attention towards or away from stimuli. Sensory attenuation thus may reflect attention orienting based on predictions from both motor and non-motor sources, contributing to optimized processing (Schröger et al., 2015; Dogge et al. 2019). Indeed, several studies observed attenuation of non-motor based sensory input, if its identity and temporality was predictable for the participants (Baess et al., 2008; Lange et al., 2009; Kaiser & Schütz-Bosbach, 2018). Since this mechanism is functioning at the level of conscious awareness, consciously perceived stimuli serve as more informative predictive cues for the experience of agency than unconsciously perceived stimuli (Chennu et al., 2013). In this context, the results of our study (i.e., differences in reported control only during the visible conditions) would offer a more coherent interpretation compared to the outcomes of Linser and Goschke (2007). Moreover, it is important to note that the influence of sensory attenuation on the experience of agency is still up to debate, given that both concepts seem to encompass distinct processing levels (Dewey & Knoblich, 2014; Wolpe & Rowe, 2014; Wen et al., 2019).
Instead, the results of Linser and Goschke (2007) may be parsimoniously explained by the post-hoc selection of participants based on awareness measurements. As Shanks (2017) illustrated, forming subgroups post-hoc based on extreme cutoffs of one dimension (e.g., awareness measure) can lead to regression-to-the-mean artifacts in the subgroup's score of another variable (e.g., performance measure). These artifacts are prone to be more pronounced when using unreliable measures for awareness and performance, which could lead to a maximized effect of this artifact. With the post-hoc selection method used in Linser and Goschke (2007), participants in the “prime unaware” group may have been more aware than their awareness scores indicated, suggesting that certain individuals might actually have been aware of the primes. Consequently, the underestimation of the awareness scores may have led to an overestimation of participants' “prime unaware” performance (Shanks, 2017; Rothkirch et al., 2021). Our results of experiment 1, which subverted the post hoc selection of the data by individually adjusting prime visibility prior to the main tasks, indicated no significant influence of subliminal priming on the experience of conscious agency.
In our second experiment, control judgments based on prime visibility or congruency showed no significant variations. Conditions only differed in response times during the response priming task (Fig. 3). This finding stands in contrast to the outcomes reported by Linser and Goschke (2007). However, it aligns with the results of Huys et al. (2020), who observed that subliminal and supraliminal prime stimuli did affect response times, but not the participants’ judgements of agency, as well. These findings suggest that the processing of unconscious primes might not exert a substantial influence on constructs that require higher-order processing. This view is in good agreement with studies utilizing the continuous flash suppression (CFS) paradigm, which highlight that the relationship between visual processing of primes and other cognitive processes is complex and prone to be influenced by various experimental design factors (Handschack et al., 2023; Moors et al., 2019; Kang et al., 2011; Benthien & Hesselmann, 2021).
Through their design of the control judgment task, Linser and Goschke (2007) aimed to make participants feel that the effect stimuli were related to their keypresses. However, the effectiveness of this manipulation is debatable. While the individual ratings of participants' conscious experience of control are not reported by Linser and Goschke (2007), in our study, we found that almost 40% (experiment 1: 39.95%; experiment 2: 37.62%) of the control ratings indicated "no control" (0%), regardless of the conditions (see Figure S2).
The relatively high amounts of "no control" ratings should raise concerns about the validity of the agency measure, as well. Participants were instructed to rate their subjective feeling of control over the effect stimuli on a scale from 0% (no control) to 100% (complete control), and ratings were given twice for each condition, after blocks of 40 trials. Previous studies in visual consciousness research have shown that different awareness scales can lead to varying relationships between performance and awareness (Reddy, 2022; Dong et al., 2015). When utilizing a 0-100% scale including only a label for each endpoint, participants are prone to interpret the scale differently, leading to subjective and less comparable responses. In an attempt to develop a coherent scale for judging the subjective experience of agency, Dong et al. (2015) observed that participants consistently preferred to use a 6-point Likert scale, including clear and meaningful labels for each point, to rate their experience of agency. Future studies examining agency might benefit from utilizing this participant-generated 6-point rating scale. Moreover, the reliance on only two measures per condition may compromise the reliability of the agency measurements. Limited assessments might not adequately capture participants' true experiences and perceptions, leading to potential inconsistencies in the results. Increasing the frequency of asking about felt control would likely provide a more robust basis for drawing conclusions about cue integration and its impact on perceived agency (Takaki et al., 2014).
Additionally, it's worth considering that the subjective measurement employed in the present task may not have adequately discriminated between the assessment of subjective agency experiences and the evaluation of task accomplishments. If participants report their experience of agency even when they are not directly controlling the actions, their judgements may be prone to heuristic responses rather than genuine reports of agency (Dong et al., 2015). Recently, Reddy (2022) discussed that especially in studies employing trials where agency is ambiguous, and subject to change based on prime stimuli, participants might rely on simplified strategies such as correlation of prime and effect-stimuli, attribute substitution or cross-modal correspondence of sensory input to respond to the questions regarding their felt control. Thus, it is possible that the given answers rather reflect judgement effects (e.g., relying on information about the congruency between prime and effect stimulus) than the participants’ genuine perceptions of agency (Dong et al., 2015). Future studies might benefit from altering the experimental setup to remove uncertainty. This could be applied during the main task and during the agency measurement (e.g., using a categorical "who did it?" question instead of the ambiguous "how much control is felt?" question).
In summary, our study highlights the need to carefully consider the potential impacts of methodological factors, especially post-hoc data selection, and experimental design on the results reported by Linser and Goschke (2007). Moreover, the results of this study suggest that unconscious stimuli may have a limited impact on the conscious experience of agency. These findings challenge the assumptions of Wegner (2002)’s theory of apparent mental causation, which proposed that mental causation is inferred from observed associations between intentions and actions, potentially leading to biases in beliefs about voluntary actions and conscious experiences of agency. In contrast, our study's outcomes align more closely with broader frameworks that emphasize the role of conscious predictions in shaping the sense of agency by aligning sensory information with intentions. Further research is needed to explore the differential effects of non-consciously and consciously perceived prime stimuli on the sense of agency, as well as thoroughly investigate subjective measurements of agency.