In everyday processing of information, one can be engaged in focused and elaborate processing. However, given our increasingly saturated information environments, we may be engaged in more surface level processing, especially in online contexts (Boczkowski et al., 2017, see also Newman et al., 2022). Under these conditions, where cognitive resources are more distributed, the referential theory might lead us to expect a relatively smaller ITE compared to when people are focused solely on encoding a given claim. The reason for this smaller ITE is due to the reduced opportunity for semantic elaboration under these conditions (see Lozito and Mulligan, 2006; Knott and Dewhurst, 2007). To date, very little ITE research has considered the impact of distracting participants from encoding the semantic elements of the claims. While we know that semantic oriented goals can increase the magnitude of the ITE, it is possible that other goals that draw people’s attention to surface-level characteristics of content they encounter may somewhat limit semantic encoding—and referential activation—thus reducing the magnitude of the ITE.

Another possibility is that beyond a referential account of the ITE, in adding a secondary task or distracting people, we might be simply directing attention away from thinking about truth. When people focus on detecting truth at encoding, they show a smaller ITE (e.g., Brashier et al., 2020; Jalbert et al., 2020). Given the lack of research on the impact of secondary tasks, we initially expected that a secondary task may distract people in this manner by directing goal orientation away from truth and thus perhaps producing a larger ITE. This was our pre-registered hypothesis.

Borrowing an adapted ITE paradigm from another line of research (Ly et al., 2023), we examine the question about the impacts of a secondary task in a series of three experiments. Our goal was to understand how asking participants to attend to linguistic level features of a claim (drawing their attention to less semantic-based aspects of the claim) may influence their susceptibility to the ITE. In Ly et al. (2023) (N = 1,102), we examined the ITE under different processing conditions and used a paradigm inspired by the Transfer-Appropriate Processing (TAP) framework (see Morris et al., 1977; Graf and Ryan, 1990; Meier and Graf, 2000). In drawing on the TAP framework, we added an ongoing linguistic-based secondary task to the classic ITE paradigm that required participants to attend to linguistic features of each claim (e.g., count the vowels in the underlined word). In our original research we were interested in whether giving people a matching or mismatching secondary task influenced the magnitude of the ITE. In the current study, we use this secondary task to investigate how the impact of a distraction task—with a linguistic focus—influences the magnitude of the ITE compared to a no secondary task baseline control.

Adding a linguistic-based secondary task affords two opportunities. First, we can examine how adding a distraction task impacts the magnitude of the ITE compared to a control condition, where participants are simply reading. Second, because the answers to this task can be coded for accuracy, we can also examine how performance on this task influences the magnitude of the ITE. We were able to explore the answer to this second question by examining data that we already had from Ly et al. (2023). We first coded people’s responses to the linguistic task (e.g., counting vowels in the underlined word of each claim) to calculate participants’ accuracy. We then calculated the magnitude of the ITE for each participant. In these exploratory analyses of accuracy scores, a systematic finding emerged between secondary task accuracy and the magnitude of the ITE. We found that higher accuracy correlated with a larger ITE, a kind of individual difference metric capturing variation in task performance. This analysis was not included in Ly et al. (2023), because (1) there was no control condition where participants simply read claims without a secondary task and (2) these analyses were not pre-registered.¹ Nonetheless, these exploratory findings led to the current study in which our key focus was to compare the magnitude of the ITE when there is a secondary task to a control condition where no tasks are involved. Secondly, because the answers to this task can be coded for accuracy, we can examine how performance on this task influences the magnitude of the ITE.

In the ITE literature, there is to our knowledge no work that has comprehensively explored the effect of and accuracy of a secondary task as a way to capture the impact of dual processing requirements on the ITE. Existing studies have had participants make ratings at encoding, (e.g., rating interest levels or making semantic categorizations at encoding; Ladowsky-Brooks, 2010; Nadarevic and Erdfelder, 2014; Nadarevic and Aßfalg, 2017; Nadarevic et al., 2018). While we may consider these ratings as secondary tasks, these ratings encourage participants to focus on the entire claim. These tasks to not, however, interrupt reading or reduce semantic encoding of the claim. Indeed, it may be the case that in the current experiments we can index the extent to which the dual tasks impact some participants more than others (individual resilience to dual tasks) and the consequence for ITE. In everyday life, people are often consuming information and multi-tasking (Boczkowski et al., 2017; Newman et al., 2017). It is, therefore, important to consider how the ability to dual task can impact the effect of repetition on perceived truth. Further, this is an important question to address when considering the referential theory of the ITE which emphasizes the activation of semantic networks as a core underlying mechanism of the ITE. What happens when people may be under cognitive demands that could draw their attention away from deeper semantic processing to more surface level features of the content they encounter? These are important theoretical and applied questions when one considers the variation in engagement with content or the task at hand on different online or social media platforms. Our current study addresses one instantiation of “ongoing cognitive processing” when consuming information online and provides early evidence into how factors such as attentiveness or cognitive capacity may shift the impact of repetition on peoples’ truth perception.

To be transparent about the genesis of this line of research, we mentioned earlier that we initially found a positive relationship between participants’ task accuracy and the magnitude of the ITE in the exploratory analyses of another ITE project (Ly et al., 2023). In the current study, across three experiments, we further examined this curious finding. We followed the general method of the Ly et al. (2023) paper but with a few key changes. A key difference in our three experiments presented here is that we included a control condition where participants completed no tasks at encoding or test. By doing so, we can contextualize how the presence of a secondary task itself alters the magnitude of the ITE. We refer to the control condition as the No Task condition throughout. These experiments extend the current understanding of the referential theory of the ITE by testing how tuning people’s processing to focus on a specific feature of a claim (e.g., counting vowels in one of the words) can moderate the ITE. Further, we examined how participants’ attentiveness to a secondary task (as indexed by accuracy scores) may moderate the ITE. All our experiments were pre-registered which is available along with Supplementary material and datasets, via the following link: https://osf.io/khp45/?view_only=5cde30c1841c43d1a1d9f0c33b01b963.

Recall that in Experiment 1, we aimed to investigate two questions. Firstly, does including a secondary task impact the magnitude of the ITE? The answer is no. We found an ITE across secondary task conditions; however, there was no difference in the size of the ITE between conditions with tasks compared to no tasks. Secondly, does participants’ accuracy on the secondary task correlate with the magnitude of the ITE? The answer is yes. We found that there was a positive association between the size of the ITE and overall task accuracy—the ITE magnitude increased with increasing accuracy on the secondary item-by-item task.³

In Experiment 1, we aimed to investigate how including an ongoing secondary task affects the magnitude of the ITE. We found no differences in the size of the ITE depending on whether a task was present (No Task vs. a Task condition) or the type of task participants completed (shoebox or vowel-counting task). This finding suggests that the ITE may not be impacted when participants are engaged in dual-tasking throughout the experiment. But this experiment considered the impact of a secondary ongoing task which occurred at both encoding and retrieval. It is possible that temporal placement of the secondary task matters. Indeed, evidence suggests that processing shifts at encoding tend to influence the size of the ITE, rather than activities at test (Unkelbach, 2007; Garcia-Marques et al., 2016; Brashier et al., 2020; Jalbert et al., 2020; Newman et al., 2020). Specifically, evidence shows that variables at encoding—that place less emphasis on veracity of claims at encoding—tend to increase the size of the ITE (Unkelbach, 2007; Jalbert et al., 2020; Newman et al., 2020). Hence, replicating Experiment 1, in Experiment 2 we examined whether the temporal placement of our ongoing secondary task (at encoding or at test) influenced the general magnitude of the ITE.

We want to note that the duration of the encoding period of the No Task and Task conditions varied (e.g., with No Task lasting 10 s as claims are automatically presented and Task conditions having the initial 3s reading time prior to completing a task). Our data in the Task at Test only condition of Experiment 2 and 3 address this concern. In this Task at Test condition, participants experience the same exposure conditions (duration) as those in the control condition. Despite the matched encoding conditions, Figures 5, 6 show that the ITE in the No Task condition is always larger than the ITE in the Task at Test condition—suggesting that differences in the exposure duration do not lead to changes in the overall pattern.

Further, in line with the exploratory analysis of Ly et al. (2023)’s data, we found that as participants’ accuracy on the ongoing secondary tasks increased, so did their ITE. Assuming that performance on the tasks indicates how engaged or attentive participants were when completing tasks, these findings suggest that increased attentiveness to the secondary task may increase people’s susceptibility to the ITE. Another interpretation of this pattern is that it indexes people’s ability to dual task in this context (Kane et al., 2001; Engle, 2002, 2018). We consider these accounts further in the General Discussion. In Experiment 2, we further attempt to understand and replicate this correlational finding and manipulate the placement of the secondary ongoing task.

In Experiment 2, we wanted to know whether the magnitude of the ITE changed depending on when participants were given the secondary task. Participants either completed the vowel-counting task at encoding, at test, or at encoding and at test, or were not given a task at all. We found a difference in the size of the ITE between the secondary task placement conditions. The ITE was largest in the no task condition compared to the rest of the secondary task conditions. Notably, although Experiment 1 also had a no task condition, this pattern of results only emerged in Experiment 2 (and 3). Secondly, like in the previous experiments, we found a positive correlation between task accuracy and the size of the ITE.⁵

One key goal of Experiment 2 was to investigate the extent to which the size of the ITE depends on when participants completed the item-by-item secondary tasks throughout the experiment. We found that the size of the ITE was largest when participants did not complete any secondary task (in the no task condition) compared to the other secondary task placement conditions. Although we did not find an effect of the secondary task in Experiment 1, this finding suggests that a secondary task may change the nature of the ITE itself.

Secondly, we expected but did not find that compared to when the task was given at test or at both phases, having the task at encoding would lead to a larger ITE. In Experiment 2, we predicted and found that higher accuracy was significantly correlated with a larger ITE—notably a third replication of this finding in this line of research (see Figure 4).

As this is the first time we found a significant interaction between claim repetition status and the secondary task placement condition, we ran a replication of Experiment 2 using a university participant pool sample.

In Experiment 3, we aimed to replicate Experiment 2 using a university participant pool sample. Like in Experiment 2, we found that the size of the ITE was largest in the no task condition compared to the other secondary task placement conditions. We also found that participants’ overall accuracy on the secondary task correlated with the size of the ITE.⁶

Experiment 3 was a direct replication of Experiment 2 where we aimed to investigate two questions with a university participant pool sample, (1) whether the temporal placement of the secondary task mattered and (2) whether overall task accuracy would be associated with the magnitude of the ITE. Like in Experiment 2, our university participant sample showed a significant difference in the size of the ITE depending on whether participants completed a task. Notably, the ITE was largest in the no task condition compared to the other secondary task conditions (task at encoding, task at test, task at encoding and test). Considered together, across Experiments 2 and 3, we found a systematic effect of the no task condition compared to secondary tasks, but no systematic pattern regarding the temporal placement of the secondary task on the size of the ITE. The findings regarding temporal placement of the task are at odds with research that has captured a greater change in the ITE when processing manipulations are employed at encoding (Jalbert et al., 2020). We discuss this finding and the larger ITE in the no task condition further in the section “5. General discussion.”

Lastly, as in all experiments reported here, we again found that as overall task accuracy increased, so did the size of the ITE. We performed a mini meta-analysis to further test the robustness of this pattern of results.

In two out of our three Experiments (Experiment 2 and 3), we found a significant interaction between repetition and task condition. The findings of Experiments 2 and 3 are consistent with studies that have shown a reduced ITE when semantic processing of a claim at encoding is disrupted or constrained. For example, Begg et al. (1985) tested whether there were differences between repeating part of a sentence compared to a full sentence in moderating the ITE. At encoding, participants were shown topic phrases (topic-repetition condition, e.g., “Hen’s body temperature”) or full trivia claims (verbatim repetition condition, e.g., “The temperature of a hen’s body is about 104 degrees Fahrenheit.”). Later, participants made truth ratings on the full claim. Begg et al. (1985) found a smaller ITE in the topic-repetition condition compared to the verbatim repetition condition. This finding aligns with those reported here where participants who had the task of attending to a specific word (in the conditions with a tangential task), rather than attending to the full claim (in control)—possibly restricting semantic activation—led to a smaller ITE compared to the No Task (control) condition. We know that the extent of semantic activation (partial vs. full activation) influences the size of the ITE (Begg et al., 1985; Ly et al., 2023). The findings of Begg et al. (1985) and those in the current study further support the referential theory of the ITE (Unkelbach and Rom, 2017) which emphasizes the role that reactivation of semantic networks plays in producing the ITE when people are re-exposed to content.

While we have shown that a secondary task can reduce the magnitude of the ITE, future research might examine the degree to which a secondary task reduces the ITE based on how much it hinders semantic processing of the claim. In the current study, we drew people’s attention to linguistic features of a claim, but we may have created only minor interference in processing meaning. Other secondary tasks that create more semantic interference may reduce the magnitude of the ITE further (see semantic interference in the misinformation effect: Loftus, 2005 and Frenda et al., 2011 for a review). For instance, asking people to remember (tangential) words that they hear while encountering the trivia claims may interrupt semantic processing to a greater extent than asking them to subsequently count vowels of a target word. Indeed, in our everyday experience, listening to content and reading something else is also a dual task we engage in regularly—listening to music or podcasts while scrolling news or social media online.

In our exploratory analysis of the Pilot Experiment, we found a significant positive correlation between participants’ overall task accuracy scores and the size of the ITE. In all three experiments reported here, we observed this same data pattern. This finding is consistent with research that has found that manipulations that encourage deeper processing, or people who are more likely to process information on a more elaborate level, demonstrated a larger ITE (Dechêne et al., 2010; Newman et al., 2020). However, the evidence surrounding processing style and the ITE is rather mixed in the literature. For example, Garcia-Marques et al. (2016) found that inducing System 2-like processing (more analytic processing compared to System 1) reduced the ITE. In our current study, although we found no differences in the ITE depending on task type (vowel-counting vs. shoebox; replicating our existing work; Ly et al., 2023), we do find mixed results regarding the impact of a secondary task on the ITE. However, we want to note that what does emerge as a consistent pattern across experiments is that there is a correlation between secondary task performance and the magnitude of the ITE. Given that secondary task performance is a potential indicator or index of participants’ level of attentiveness, one likely account of the patterns observed in the current study is that high performing participants who are paying more attention to the task at hand are better able to engage with the content of the claims. This deeper engagement with the content of the claims may work to facilitate conceptual processing of its semantic elements, thereby ensuring adequate references are activated in the memory network—which the referential theory of the ITE suggests is key for repetition-induced fluency at later judgments.

In further examining participants’ task accuracy scores, in Experiments 1 and 2, we also found that high and low accuracy groups show a systematic difference in the size of the ITE compared to the No Task conditions. In short, when splitting participants in all the task conditions by high vs. low accuracy, those in the high accuracy group show a similar ITE to the No Task condition. This finding suggests that the correlation may be driven by the low accuracy participants who show a significantly smaller ITE when compared to the No Task condition and points to another possible account—cognitive capacity. A similar pattern is present in Experiment 3, but is less pronounced and non-significant (see further analyses, p. S13 – S16). Considering task accuracy scores as an index of secondary task processing capacity, this cognitive capacity explanation is consistent with working memory research that shows that ongoing tasks may have more of a cognitive cost to some participants than others (Engle, 2002, 2018). In general, performance of those with high working memory capacity is relatively less impacted by ongoing cognitive tasks (i.e., task switching, increased cognitive load/task difficulty; Kane and Engle, 2000; Kane et al., 2001). We found some evidence supporting this idea—those who had high accuracy on the vowel counting task had an ITE magnitude similar to those in the control condition. That is, the ongoing secondary task did not reduce the ITE relative to the No Task condition for everyone—arguably, for those who had the capacity, there was little impact (see Supplementary Table 5). Existing research on cognitive capacity/executive function shows no correlation with the magnitude of the ITE (De keersmaecker et al., 2020). However, in that research, there was no ongoing secondary task—which may be a boundary condition to which one may observe such effects. Further research could illuminate the role of cognitive capacity and the effects we have observed here. For example, further research might include external measures of participants’ processing capacity and flexibility from the working memory literature such as the Operation Span Task (OSPAN; Unsworth et al., 2005), the Wisconsin Card Sorting (WCST) (Greve, 2001), a psychological refractory period paradigm (PRP) task (Pashler, 1991; Giesbrecht et al., 2001) or a task-switching paradigm.

Another possible explanation of our findings is that the correlation between secondary task and the ITE is driven by variations of attentiveness on the task. This raises questions surrounding data quality in online experiments—a common mode of data collection in the ITE literature. Indeed, a meta-analysis by Henderson et al. (2021) revealed that around 27% of ITE studies were conducted online. The consistently significant positive correlation between accuracy and the magnitude of the ITE suggests that there is variation in participants’ performance online and that this can have an impact on the magnitude of the ITE. Considered together, it is possible that depending on the task at hand, and perhaps even levels of interest and engagement, online studies may be capturing a smaller ITE. In future research, factors such as attentiveness and cognitive capacity may be worth considering, particularly in online environments where researchers are less able to control what participants are doing whilst they are completing the study.