The Neural Basis of The Continued Influence Effect of Misinformation

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Acta Psychologica Sinica ©2022 Chinese Psychological Society

2022, Vol. 54, No. 4, 343354 https://doi.org/10.3724/SP.J.1041.2022.00343

The neural basis of the continued influence effect of misinformation


JIN Hua1,2,3, JIA Lina2, YIN Xiaojuan2, YAN Shizhen2, WEI Shilin2, CHEN Juntao2
(1 Key Research Base of Humanities and Social Sciences of the Ministry of Education, Academy of Psychology and Behavior,
Tianjin Normal University, Tianjin 300387, China) (2 Faculty of Psychology, Tianjin Normal University, Tianjin 300387, China)
(3 Tianjin Social Science Laboratory of Students’ Mental Development and Learning, Tianjin 300387, China)

Abstract
In this study, the differences between the activation and functional connectivity conditions of related brain regions by task-fMRI
were analyzed to reveal the neural basis of the CIEM and provided more evidence for the hypothesis of mental-model-updating and
memory-retrieval-failure. The results showed that the inference scores of retraction condition were significantly higher than that of
control condition, and the CIEM exists. In the encoding phase, the activation of left middle temporal gyrus in retraction condition
was significantly weaker than that in control condition. While in the retrieval phase, the activation of middle frontal gyrus and ante-
rior cingulate gyrus was weaker in retraction, and the functional connectivity between middle frontal gyrus and precentral gyrus was
stronger in retraction. The results suggest that the above brain regions may be involved in the formation of the CIEM, and provide
evidence from the neural level that the hypothesis of mental-model-updating and memory-retrieval-failure may explain the different
phases of the CIEM formation.
Key words continued influence effect of misinformation, task-fMRI, inhibition control, functional connectivity

this phenomenon are still unclear. At the cognitive level, men-


1 Introduction
tal-model-updating and memory-retrieval-failure hypotheses
The continued-influence effect of misinformation (CIEM) try to explain this phenomenon (Ecker et al., 2015; Johnson &
refers to the phenomenon that misinformation continues to Seifert, 1994; Lewandowsky et al., 2012). According to the
influence people’s judgment and reasoning even after it has mental-model-updating hypothesis, when people read a story or
been retracted (Gordon et al., 2019; Johnson & Seifert, 1994; an event, they build a situation model or mental model struc-
Swire et al., 2017; Wahlheim et al., 2020). The CIEM tradition- tured around the causal chain of events (Graesser et al., 1994;
ally uses a paradigm in which individuals read a story that in- Rich & Zaragoza, 2020). Attempts to replace initial misinfor-
cludes misinformation and retraction and participants then re- mation with retract information threaten a model's internal
ceive a series of inference questions regarding the story (Ecker coherence, and the occurrence of the CIEM therefore is par-
et al., 2017; Rich & Zaragoza, 2016). Inference questions are ticularly due to the poorly encoded retraction or incomplete
usually related to measures taken after the occurrence of the integration into the model (Brydges et al., 2018; Gordon et al.,
previous misinformation. For example, the inference question 2019). The memory-retrieval-failure hypothesis based on
corresponding to the misinformation “fire” is “measures de- memory dual-process theory (Brydges et al., 2018; Rugg &
signed to prevent fires should be reviewed”. The CIEM is Curran, 2007), argues that when people encounter questions
measured by the extent to which participants rely on the misin- about a story, all information relevant to the story, including
formation when responding to inference questions (Johnson & both the misinformation and the retraction, can become acti-
Seifert, 1994). Inference questions are also rated on a Likert vated (Rich & Zaragoza, 2020). A strategic monitoring process
point scale (0 = completely disagree ~ 10 = completely agree), (contrast to an automatic processing, including retrieval of the
with higher scores denoting stronger reliance on the initial context details, evaluation of the output information, etc.) is
misinformation (i.e. the larger CIEM) (Ecker et al., 2020; Swire then required to determine the validity of this automatically
et al., 2017). There are also studies that manipulate the critical retrieved information (Lewandowsky et al., 2012), and to pre-
information with or without misinformation which corresponds vent the activated misinformation from influencing the reason-
to retraction or control condition. The difference in inference ing process (Brydges et al., 2018). Therefore, the failure of
scores under retraction and control conditions is compared to strategic retrieval processes and the ineffective inhibition of
indicate the CIEM (Gordon et al., 2017; Rich & Zaragoza, activated misinformation (Brydges et al., 2020) are the main
2016). This design can avoid inflating inference scores or in- reasons leading to the CIEM. Thus, the mental-model-updating
terference with additional factors caused by focusing only on hypothesis focused much more on the encoding or integration
retraction condition. information phase, while the memory-retrieval-failure hypothe-
Previous studies have proved the existence of the CIEM sis mainly focused on the retrieval information phase. Unfor-
(Ecker et al., 2015; Johnson & Seifert, 1994; Rich & Zaragoza, tunately, the traditional behavioral indicators cannot effectively
2016), but the cognitive mechanism and neural basis behind distinguish between the above two processing phases.

Received Date: May 10, 2021


JIN Hua and JIA Lina are co first authors
Corresponding author: JIN Hua, E-mail: jinhua@tjnu.edu.cn
The original article is in Chinese, translated by Eastrans and proofread by the author. The Chinese version shall always prevail in case of any discrepancy or
inconsistency between the Chinese version and its English translation.
Acta Psychologica Sinica

At the neural level, the previous studies have attempted to al. (2019) shortened the story length to 2 sentences. The first
simultaneously consider the neural basis and cognitive mecha- sentence was a description of an event and a reason, and the
nisms of the CIEM (Brydges et al., 2020; Gordon et al., 2017; second sentence presented confirmation or retraction informa-
Gordon et al., 2019). The basic logic of studies revealing the tion. Differences in story length, especially interval in misin-
CIEM with spatiotemporal features of neuron activations dur- formation and retraction information, may lead to different
ing information encoding or retrieval is as follows: if the men- integration of retraction into story contexts (Commander &
tal-model-updating hypothesis is reasonable, there is a system- Stanwyck, 1997; Wang et al., 2008) and to produce different
atic difference during the encoding or integration of retracting neural activities. In the information retrieval phase, Gordon et
and non-retracting information, and this difference should be al. (2017) adopted the classical CIEM paradigm, measuring the
associated with observable changes in brain activity. That is, CIEM by scores of inference questions. If inference scores in
compared with non-retracting information, changing traces in misinformation condition were higher than in control condition,
brain activation in certain brain regions during encoding or the expected CIEM would be demonstrated. However, in
integration retraction can provide further empirical evidence for Gordon et al. (2019), the experimental paradigm was image-
the model-updating hypothesis of the CIEM. Similarly, if the information matching. The CIEM was measured by individual
memory-retrieval-failure hypothesis is right, there may be dif- responding accuracy (reaction times and drift rates) about im-
ferences between conditions in functional activities in the brain age that either matched or mismatched with sentence 2. The
areas involving in information selection and inhibition in less accurate judgement at mismatching condition demon-
memory (e.g., hippocampus and prefrontal cortex) during re- strated that participants still failed to reject the picture contain-
trieval (Gordon et al., 2017). However, there were few relevant ing misinformation after retraction (i.e., the CIEM exists). The
studies and a great deal of differences between the results. For changing of the CIEM indicator may also alter the information
example, the results of Gordon et al. (2017) partially support retrieval process and then alter the difference in neural activity.
the mental-model-updating hypothesis of the CIEM. The dif- Moreover, previous studies have been limited to the analysis of
ferences in brain activity between retraction and control condi- brain activation localization and have not focused on functional
tion can be observed during encoding or integration phase. connectivity among activation regions. Compared with the
Critical information in the control conditions elicited more activation analysis of a single brain region, functional connec-
activities in the right precuneus (extending into the right poste- tivity analysis can better explore the interaction among brain
rior cingulate cortex and postcentral gyrus) than in the retrac- regions, so as to clarify the basis of brain network between
tion, but the brain-behavior analysis only found correlation different conditions. Therefore, this study will further consider
between neural activity and inference ratings in the control the brain functional connectivity related to the CIEM.
condition. No significant differences in brain activity emerged In addition, the possible contribution of the inhibitory com-
across the two conditions at retrieval phase. However, the ponent in the cognitive-neural mechanisms of the CIEM has
brain-behavior analysis found correlations between neural ac- been ignored in the previous studies. Failure of model-updating
tivity and inference ratings both in the retraction and the con- or information retrieval inducing the CIEM may involve inhibi-
trol conditions. Yet the findings of Gordon et al. (2019) seem to tion process. Studies have found that inhibitory control may be
support the memory-retrieval-failure hypothesis. They found implicated in reading comprehension in real time by managing
that the activity of two brain regions related to memory showed the activation of irrelevant information (Butterfuss & Kendeou,
a difference between conditions in the retrieval phase. Infer- 2020; Christopher et al., 2012). Butterfuss and Kendeou (2020)
ence problems elicited enhanced activity in the left angular found that higher-inhibition participants read the target sen-
gyrus and the bilateral precuneus in the retraction condition tences more slowly than low-inhibition participants under con-
than in the control condition, while no brain-behavior correla- ditions that did not refute or explain misinformation, suggesting
tion was found. The same is true of the findings from neu- higher-inhibition participants may have engaged in extra proc-
roelectrophysiology. Brydges and Ecker (2018)1 and Brydges, essing aimed at resolving the conflict of misinformation. The
Gordon and Ecker (2018)1 found no differences in ERP ampli- fMRI study by Masson et al. (2014) also found that misconcep-
tudes between different conditions. Brydges et al. (2020) found tions processing involved activation of brain areas related to
differences between two conditions in the FN400 at frontal inhibition control. In the CIEM paradigm, participants may
midline and the LPC at left-parietal during retrieval phase, need to remove or inhibit initial misinformation when process-
which supported the memory-retrieval-failure hypothesis. ing retraction and promote integration of new information into
It can be seen that the existing studies have not yet obtained an event mental-model. In processing inference problems, it is
a unified understanding of the theoretical hypothesis and neural also necessary to inhibit previously activated interference in-
basis of the CIEM. This situation may be related to the meth- formation, so as to extract correct information in memory.
odological differences of various studies. In terms of studies This study performed a region of interest (ROI) analysis to
using fMRI techniques, the research paradigm of Gordon et al. separate mental-model-updating and memory-retrieval-failure
(2017) is very different from that of Gordon et al. (2019). In the hypothesis of the CIEM and examine the neural basis of the
information encoding or integration phase, Gordon et al. (2017) CIEM. One class of ROIs was regions associated with semantic
used a short fictional news story consisting of 6 sentences as and error inhibition, including the left ventrolateral prefrontal
material. Sentence 2 as the conditional sentence presented mis- cortex (L_VLPFC, BA45), left dorsolateral prefrontal cortex
information or neutral information. Sentence 5 was the target (L_DLPFC, BA9), right anterior cingulate cortex (R_ACC,
sentence presenting retraction information. However, Gordon et BA32) and left middle frontal gyrus (L_MFG, BA10) (Collette
& Linden, 2002; Masson et al., 2014; Nathaniel-James et al.,
1 1997). Given that materials used in the CIEM task are short
Brydges and Ecker (2018), Brydges, Gordon and Ecker (2018) are cur-
rently preprints on OSF platforms, not for the journal publication. stories, retraction of misinformation may first (or simultane-
JIN Hua et al.: The neural basis of the continued influence effect of misinformation

ously) require sentences comprehension and discourse integra- to the interaction between brain regions at the corresponding
tion. In order to further clarify whether the encoding process of phase, which would provide more evidence for mental-
retraction involves only inhibition or also discourse processing, model-updating and memory-retrieval-failure hypothesis.
we also defined regions associated with text (information)
2 Method
comprehension as another type of ROIs to further separate the
unique role of inhibition. Based on a meta-analysis of fMRI 2.1 Participants
imaging studies on discourse comprehension by Yang et al. 31 participants (10 males, Mage = 21.23 years, SDage = 1.77)
(2019), the regions related to text comprehension include the took part in this study. Data from six participants were removed
left inferior frontal gyrus (L_IFG, BA45/47), left middle tem- from the fMRI analyses for excessive head motion during
poral gyrus (L_MTG, BA21/22) and left medial frontal gyrus scanning (exceeded 3 mm and 1.5°). All participants had nor-
(L_MFG, BA6) etc. Similarly, besides the regions associated mal or corrected-to-normal vision. None of the participants
with inhibition in retrieval phase, the general semantic retrieval reported a history of neurological or psychiatric disorders and
processes also need to be considered. The fMRI studies related contraindication in MRI. All participants provided written in-
to general semantic retrieval have found that the left hippo- formed consent and were paid for their participation.
campus and bilateral parahippocampal gyrus generally have 2.2 Stimuli
activation in semantic memory retrieval (Hoscheidt et al., The experimental materials were adapted from the Gordon
2010). So to clarify whether neural processing is also related to et al. (2017). Gordon et al. (2017) used a total of 22 fictional
general semantic memory retrieval in retrieval phase, we also news stories. In order to avoid the understanding ambiguity
adopt the left hippocampus and bilateral parahippocampal caused by cultural differences, some stories had been modified,
gyrus as ROIs. The in-depth study on the cognitive-neural such as replacing foreign place names with “a city” or “an
mechanism of the CIEM is conducive to a clearer understand- area”. In view of the BOLD signal containing high background
ing of the causes of CIEM, which can provide more evidence noise, a certain number of event-related signals need to be su-
for existing theoretical hypotheses and further clarify the action perimposed on average to obtain the valid signal. Thus, on the
phases of corresponding hypotheses. basis of the existing materials of Gordon et al. (2017), similar
In summary, based on hypothesis of integration updating fictional news stories were written, increasing the number of
failure and memory retrieval failure, the study would adopt the materials to 40 stories. Each news story was six sentences long
classical paradigm of the CIEM and fMRI technology to inves- and two different versions were created that differed only in the
tigate the neural basis and cognitive mechanism of the CIEM content of sentence 2. The control version had no misinforma-
from views of semantic processing and interference informa- tion in sentence 2, while the retraction version contained mis-
tion inhibition. Because the fMRI experiment needs to set base- information in sentence 2. Sentence 5 in the retraction version
line conditions to achieve inter-condition comparison, and to was the retract information, while sentence 5 in control version
avoid the larger head movement caused by too long scanning was transitional sentence because no misinformation was pre-
time, this study mainly refers to the experimental paradigm of sented in previous sentences.
Gordon et al. (2017). This paradigm uses brief stories and in- The critical sentence (i.e., sentence 5) and inference ques-
cludes control conditions, which also measured the CIEM by tions were identical in both versions. Inference questions used a
typical inference problems. This paradigm thus is easy to ma- 9-point scale from “1 = strongly disagree” to “9 = strongly
nipulate variables and effectively reveal the possible cause of agree” to measure participants’ reliance on misinformation (i.e.,
CIEM. We proposed the following hypotheses based on the the CIEM). This manipulation ensured that the target sentences
previous studies and the above reviews: 1) inference scores for inter-conditions comparisons were identical, avoiding the
under retraction condition will be significantly higher than possibility that the observed difference between conditions
under control condition, suggesting the existing of CIEM in might be caused by the difference in the content and valence of
behavior; 2) with regard to neural basis-cognitive mechanism, the target sentence. Each four stories with similar themes also
mental-model-updating hypothesis and memory-retrieval-failure had the same inference problems in the 40 stories, such as the
hypothesis may explain the different phase of the CIEM. The story “fire in the building” and “fire in the cabin” of the corre-
specific manifestations are as follows: ① the CIEM is related sponding retraction was “investigation found no evidence of a
to abnormal activation in brain regions responsible for text fire”, the inference problem was “it is necessary for the fire
processing or inhibition control in the encoding phase. That is, services to be at the scene” etc. (see Appendix). It could mini-
individuals fail to effectively encode current semantic informa- mize the interference with results due to differences in target
tion or inhibit previous misinformation, resulting in incomplete sentences among different stories. Furthermore, it has been
or failed information integration and updating of mental models. found that familiarity of information can affect the CIEM
It would support the mental-model-updating hypothesis; ② the (Swire et al., 2017). Since this study mainly compared the dif-
occurrence of CIEM would also accompany with abnormal ferences in inference scores under the retraction and control
activation in brain regions responsible for semantic retrieval or condition, the familiarity of the control condition as a baseline
semantic inhibition in the retrieval phase. That is, individuals was expected to match the retraction. So 20 participants (not
would fail to effectively retrieve semantic information or in- participating in the formal fMRI scanning, 5 males, mean age =
hibit irrelevant semantic information, resulting in invalid or 22.05 ± 3.19 years) assessed the familiarity of the first two
failed information retrieval. It would support the hypothesis of sentences of all stories (1 = completely unfamiliar, 7 = fully
memory retrieval failure; ③ The brain regions related to text familiar). No significant difference was found in familiarity
comprehension, semantic retrieval and inhibition would show under control and retraction conditions, (4.13 ± 0.40 vs. 4.21 ±
changes in functional connectivity at the encoding or retrieval 0.36), t(39) = −1.05, p = 0.30. All sentences were controlled at
phase, suggesting that the occurrence of CIEM may be related 10~14 words length. For each story, half of the participants
Acta Psychologica Sinica

read its control version and the other half of participants read timing and head motion correction were performed. The struc-
its retraction version. All stories were presented pseudo-ran- tural images of each participant were registered into the aver-
domly, with each story as a trial in the experiment and 40 trials age functional images, and functional images were normalized
for each participant. Based on the previous research (Gordon et to the MNI (Montreal Neurological Institute) space using
al., 2017), the study focused on the encoding phase (i.e., proc- DARTEL in space standardization. Functional images were
essing of sentence 5) and retrieval phase (i.e., processing of the then spatially smoothed with 8-mm full width at half maximum
inference problems). (FWHM) Gaussian kernel.
2.3 fMRI task and procedure
3 Data analysis
Participants were required to practice on the computer out-
side the magnet before the formal experiment to be familiar 3.1 Behavioral analysis
with the experimental process. All practice materials were not The behavioral data were analyzed by SPSS 25. Paired-t test
appeared in the formal scans. Meanwhile, participants were was conducted for dependent variable (inference scores). If
instructed to respond to inference questions by indicating num- retraction of misinformation was not sufficient to bring ratings
bers 1 to 9 through their own habitual gestures. Participants re- back to a baseline level (control condition), then it indicated
peated gestures practice in the fMRI simulator until they met that participants partially depend on the misinformation to
the requirement before entering the fMRI chamber. During make judgment in inference process, i.e., there is a CIEM. We
gestures practice, participants were required to lie flat and used statistical p < 0.05 as the threshold.
change the number of extended fingers indicating 1 to 9 while 3.2 fMRI analysis
keeping their arms and wrists as still as possible. The formal This study mainly compared the brain activation differences
scan contained four functional runs, and there was a short rest between control and retraction conditions in the encoding and
between runs without head movement. The entire scan took retrieval phase. Brain imaging data would be analyzed at two
about 50 minutes to complete. levels: ROI-based analysis and brain-behavior correlations; and
A fast event-related design (jittered) was adopted. The be- the seed-based functional connectivity analysis with common
ginning of each trial was presented by a red exclamation mark method of task-fMRI functional connectivity analysis-psy-
(!). The first four sentences in a story were appeared (7.5 s) chophysiological interaction (PPI) (Di et al., 2021).
automatically after 0.5 s. Then 1~6 s fixation “+” was presented, 3.2.1 ROI-based analysis and correlation with behavior
followed by sentences and “+” until the end of the story. A Functional data after preprocessing were analyzed using
question mark (?) was presented on screen for 0.5 s before in- general linear model (GLM). As mentioned above, the proc-
ferences problems, followed by inference sentence 1 for 2 s. essing phase given serious attention in this study was the brain
Then a response scale (9-point) was shown on screen for 4 s BOLD signal in processing retraction (encoding phase) and
after 2~6 s “+”, followed by inference sentences 2 and 3 after inference problems (retrieval phase). At the first level, GLM
the jitter of 2 to 6s (see Figure 1). implemented in SPM 12 (http://www.fil.ion.ucl.ac.uk/spm) was
2.4 Image acquisition applied for each participant’s signal in the encoding and re-
Image data were acquired by Siemens Prisma 3.0 T mag- trieval phase, respectively. Six head motion parameters, obtained
netic resonance scanner and 64-channel head coil at Tianjin by head motion correction, were included as nuisance regressors.
Normal University. Participants were placed in the supine posi- As mentioned above, based on previous studies (Masson et
tion with spongy pads placed in the coil to fix their heads. al., 2014; Yang et al., 2019) , our ROIs included left IFG
Stimuli were presented with E-Prime 2.0 software via a mirror (BA45/47), MTG (BA21/22) and MFG (BA6), which associ-
and a projection system. Functional images were obtained with ated with information encoding processing. The following re-
an echo-planar imaging (EPI) sequence (TR = 2000 ms, TE = gions were also included: the left VLPFC (BA45, partial over-
30 ms, flip angle = 90°, field of view (FOV) = 224 mm × 224 lap with inferior frontal gyrus), left DLPFC (BA9), right ACC
mm, matrix size = 64×64, slice thickness = 3.5 mm, slice gap = (BA32) and left MFG (BA10). These regions used Brodmann
0.7 mm, number of slices = 33). A high resolution T1-weighted areas (BA) as template. Meanwhile, the BA template did not
structural imaging was acquired with the MPRAGE sequence clearly divide the hippocampus and parahippocampal gyrus, so
(TR = 2530 ms, TE = 2.98 ms, flip angle = 7°, FOV = 256 mm × the left hippocampus (37) and bilateral parahippocampal gyrus
256 mm, matrix size = 256×256, slice thickness = 1 mm). (39/40) combined Anatomical Automatic Labeling (AAL) tem-
2.5 fMRI data preprocessing plate. Masks creating were implemented in RESTplus V1.24
Imaging data were preprocessed using DPABI V4.2 soft- (Jia et al., 2019) and DPABI V4.2 software was used to extract
ware (Yan et al., 2016) implemented in Matlab R2015a. The BOLD signal from ROI across conditions. SPSS 25 was used
first 5 images of each scanning were removed for signal equi- for statistical analysis to compare the difference between each
librium and the adaptation to the fMRI environment. Then slice ROI activation signal values, with p < 0.05 as the threshold.

Figure 1. Experimental flow diagram


JIN Hua et al.: The neural basis of the continued influence effect of misinformation

Analysis of brain-behavior correlations focused on the rela-


tionship between brain activities in the encoding and retrieval
phase and participants’ scores on the inference problems. Brain
imaging data were BOLD signal values extracted from each
ROI, and behavioral data were mean of the three inference
questions corresponding to each story. Correlations analysis
was implemented in SPSS 25 to examine change trends of
brain-behavior correlations, with p < 0.05 as the threshold.
3.2.2 Seed-based functional connectivity
In the PPI analysis, seed regions were selected to investigate
the functional connectivity between the seed regions and other Figure 2. Results of behavioral inference scores
brain areas of the whole brain under specific conditions. PPI
analysis included seed regions definition, time series extraction, 4.2 fMRI results
PPI modeling analysis at first-level, and comparison of func- 4.2.1 Inter-condition differences in ROI activation and
tional connections at group-level. Based on the previous results, correlation
the small-volume correction (SVC) (Steuwe et al., 2015) was Paired-t tests were conducted to analyze ROIs’ signals be-
used to find the peak coordinates of seed regions, and the SVC tween different conditions during the encoding phase. The re-
threshold was set at uncorrected p < 0.005. Each seed region sult revealed a significant activation difference between retrac-
defined by a sphere with 6 mm radius centered at the peak tion and control condition in the left MTG (BA21, 22), (BA21:
voxel to investigate the connectivity of each seed region to the t(24) = 2.34, p = 0.03, Cohen’s d = 0.47, 95% CI [0.03, 0.48];
other brain regions under different conditions. PPI analysis was BA22: t(24) = 2.26, p = 0.03, Cohen’s d = 0.45, 95% CI [0.02,
performed in SPM 12. The GLM of PPI includes: (1) main 0.48]) (see Figure 3). No significant activation differences
effect of seed region activation; (2) main effect of experimental across conditions were found in the other ROIs associated with
conditions; and (3) interaction of seed region activation and inhibitory control and discourse comprehension.
experimental conditions. Six head motion parameters were also The same analysis was performed in the retrieval phase. The
included in GLM to avoid minor impact on the signal. Psycho- comparison results (see Figure 4) showed that activations of the
physiological interaction was the variables mainly focused, following brain regions under the retraction condition was sig-
which were included into the group-level analysis. Functional nificantly weaker than that of the control condition: the left
connectivity used multiple corrections (FWE p < 0.05), and the MFG (BA10), t(24) = 2.39, p = 0.03, Cohen’s d = 0.48, 95% CI
extension threshold was 50 voxels. [0.04, 0.55]; right ACC (BA32), t(24) = 2.24, p = 0.04, Cohen’s
d = 0.45, 95% CI [0.01, 0.28]. And also the region showed a
4 Results significant marginal difference between conditions: left DLPFC
4.1 Behavior results (BA9), t(24) = 2.05, p = 0.05, Cohen’s d = 0.41, 95% CI
Analysis of inference scores found that the scores of retrac- [−0.001, 0.40]. No significant differences between conditions
tion condition (4.48 ± 1.27) was significantly higher than that were found in the VLPFC (BA45) associated with inhibition
of control condition without misinformation (3.98 ± 1.20), t(30) = and in the hippocampus and parahippocampal gyrus associated
3.94, p < 0.001, Cohen’s d = 0.71, 95% CI [0.24, 0.75] (see with general semantic information retrieval.
Figure 2). This suggested that retraction of misinformation was Correlation analysis was conducted between individual be-
not sufficient to bring inference scores back to a baseline level, havioral inference scores and beta values of brain regions
demonstrating a significant reliance on the misinformation in showing inter-condition differences. It was found that the beta
the inference judgment, namely a significant CIEM. value in the left MFG in retraction condition had marginal

Figure 3. Activation differences at left MTG in encoding phase


Acta Psychologica Sinica

Figure 4. Activation differences at left MFG, right ACC and left DLPFC in retrieval phase

significant negative correlation with the inference scores, r = −0.38, right ACC and the rest of the brain regions was compared under
p = 0.06 (see Figure 5). It was shown that when the activation the control and retraction condition at the retrieval phase. Re-
change in this region was small, the higher behavioral inference sults demonstrated that the right ACC showed decreased func-
scores meant greater reliance on misinformation, namely the tional connectivity with the bilateral Inferior Occipital Gyrus
greater CIEM. No significant correlations were found between (L/R_IOG) during control > retraction conditions (pFWE < 0.05)
behavior data and beta values in the left MTG associated with (see Table 1 and Figure 6). The reverse compare (control <
discourse comprehension, and in other brain regions associated retraction), however, failed to yield any supra-threshold activa-
with inhibition, p > 0.05. tions.

Table 1
Between-condition comparisons of functional connectivity with
right ACC as seed region
Peak MNI coordinates
Regions Voxels Peak t value
x y z
control > retraction
Left inferior occipital gyrus 166 −24.5 −91 −7 10.11
right inferior occipital gyrus 123 21 −84 −7 10.10
control < retraction
No supra-threshold activation
Note. p < 0.05 (FWE corrections), k > 50; the same below.

Figure 5. Relationship between left MFG and behavioral infer-


ence scores

4.2.2 Inter-condition differences in functional connectivity


According to the previous results, the left MFG and right
ACC related to the retrieval phase and the left MTG related to
the encoding phase were defined as seeds in functional connec-
tivity analysis. SVC obtained peak coordinates for the left
MTG (MNI: −60, −35, −11), right ACC (MNI: 0, 46, 28) and Figure 6. Functional connectivity activation with right ACC as
left MFG (MNI: −32, 63, 0). seed region (ACC = anterior cingulate cortex; IOG = inferior oc-
The difference in the functional connectivity between the cipital gyrus; L = left; R = right)
JIN Hua et al.: The neural basis of the continued influence effect of misinformation

Similarly, the difference in the functional connectivity be- coding or integration of retraction information, and changes in
tween the left MFG and the rest of the brain regions was com- activation (including functional connectivity) of the right ACC
pared under the control and retraction condition at the retrieval and left MFG during information retrieval may be the neural
phase. It also showed decreased connectivity with the bilateral basis of the CIEM. The occurrence of the CIEM is not only
IOG during control > retraction conditions; while increased related to the encoding or integration of information in the
functional connectivity with right precentral gyrus (R_PcG) for process of text reading, but also related to the process of infor-
the control < retraction conditions (see Table 2 and Figure 7). mation retrieval after reading. The mental-model-updating and
Finally, functional connectivity analysis on the left MTG re- memory-retrieval failure hypothesis may explain the different
vealed no significant results in the encoding phase. phases of the CIEM.
5.1 Neural basis of the CIEM
Table 2 The left MTG、left MFG and right ACC may involve in the
Between-condition comparisons of functional connectivity with left CIEM formation at different phases of information processing.
MFG as seed region
Only the left MTG related to text comprehension showed
Peak MNI coordinates weaker activation in the retraction condition than in the control
Regions Voxels Peak t value
x y z condition at semantic encoding phase of target sentence, sug-
gesting that the left MTG may involve in the CIEM during
control > retraction
encoding or integration phase of retraction, which is consistent
Left inferior occipital gyrus 158 −17.5 −94.5 −7 10.34 with  in hypothesis 2. The MTG is generally associated with
Right inferior occipital gyrus 101 21 −84 −7 8.95 general reading comprehension (Weber et al., 2019; Yang et al.,
control < retraction 2019). Weber et al. (2019) found that the left MTG activated
stronger during silently reading sentences than pseudo word
Right precentral gyrus 61 35 −17.5 49 8.44
lists in a language localizer task with the whole-brain analysis.
The encoding phase only involved the brain region of text
processing and had weaker activation in the retraction condi-
tion, which is consistent with the findings of contextual consis-
tency in discourse comprehension (van Moort et al., 2020). van
Moort et al. (2020) found that the text-based inconsistencies
elicited weaker neural responses, and speculated that it might
be due to the fact that the target sentence was not completely
contradictory to the preceding text. We assumed that this phase
may be more about text-based comprehension processes and
more sensitive to congruency (control) condition only. At the
same time, the less activation and inactivation of inhibi-
Figure 7. Functional connectivity activation with left MFG as tion-related regions in the retraction condition may be due to
seed region (MFG = middle frontal gyrus; PcG = precentral gyrus) the conflict or contradiction between contextual information
during processing do not reach the level of initiating inhibitory
control. The results also suggest that the processing abnormali-
5 Discussion
ties of the retraction may be related to functional changes in
Combined with the classical paradigm of the CIEM and semantic processing brain areas during discourse comprehen-
task-fMRI technology, this study discussed the possible neural sion instead of functional changes in inhibiting-related brain
basis and cognitive mechanism of the CIEM from the perspec- regions.
tive of inhibitory control and semantic processing. Behavioral The left MFG and right ACC showed similar inter-condition
results were in line with the previously researches (Gordon et differences in activation and functional connectivity during
al., 2017; Rich & Zaragoza, 2016) and hypothesis 1, which retrieval phase, suggesting that these two brain regions may be
found that inference scores in the retraction condition was sig- involved in the CIEM at the retrieval phase. These results are
nificantly higher than that in the control conditions. It indicated consistent with ② and ③ in hypothesis 2. The functional
that the misinformation was not completely retracted, and par- connectivity in general control networks is considered to be
ticipants still partial relied on misinformation for judgment (i.e. relevant to reading (Jolles et al., 2020). Yuan et al. (2021) found
a significant CIEM). The fMRI results found that there were no that bilingual language control and general cognitive control
activation differences across conditions in the brain regions had overlapping brain network (e.g., dorsal anterior cingulate
responsible for inhibition at encoding or integration phase of cortex, etc.) in the process of conflict detection and inhibition
information. There was activation difference across conditions control. The phenomenon of lower brain activity in experimen-
in the left MFG and right ACC in the information retrieval tal condition than in control may suggested a task-induced
phase and this difference was related with inference scores. The neuronal suppression (Morita et al., 2019). Most previous stud-
activation of the left MTG associated with reading comprehen- ies have found that increased activation in related regions of
sion differed under different conditions in the information en- inhibitory control was associated with successful inhibitory
coding or integration phase. Further, functional connectivity control (Hinault et al., 2019). Cohen-Gilbert et al. (2017) re-
analysis also found that the functional connectivity of the right quired 23 healthy college students to finish the task of inhibi-
ACC and the bilateral IOG, the left MFG and right PcG were tion control during fMRI scan. Results showed that decreased
different under different conditions. The results suggested that activations in the DLPFC, DMPFC (included MFG) and ACC
changes in activation of the left MTG during information en- were significantly associated with higher binge drinking in the
Acta Psychologica Sinica

prior 3 months during negative relative to neutral inhibitory This study observed no inter-condition differences in activation
trials. Additionally, the current results are also consistent with or functional connectivity in brain regions responsible for inhi-
“the lower activation in retraction condition” found by Gordon bition, suggesting that the CIEM may be related to abnormal
et al (2017). They found that retraction elicited less activity in information encoding rather than abnormal information inte-
the right precuneus at the encoding phase, and assumed that gration at the encoding phase. It further clarifies the explana-
this activation reduction might be linked to integration difficul- tion of the “mental-model-updating” hypothesis for the CIEM.
ties. Taken together, decreased activation of the left MFG and At the level of text comprehension, however, which cognitive
right ACC may reflect failure or difficulty in inhibiting execu- component results in the encoding anomaly (e.g., the failure or
tion in this condition. The weakened functional connectivity no startup of situational model updating, etc.) requires new
between the left MFG and right ACC with the bilateral IOG in experimental design to explore.
the retraction condition may be associated with weakening of At the retrieval phase, inter-condition differences in activa-
general text information processing. The IOG was found to be tion in the brain regions for general semantic information re-
related to visual text reading (Jin et al., 2009). Jin et al. (2009) trieval such as the hippocampus were not significant, and only
found strong bilateral activation in the middle and inferior oc- inter-condition differences in activation and functional connec-
cipital lobe during story-reading and predictive inference. In tivity in the brain regions for inhibition (left MFG and right
the retraction condition, there is not only visual reading and ACC) were found. The results support the “memory-retrieval-
general judgment but also inhibition of misinformation, which failure” hypothesis and further clarify the content of “memory-
may relatively weaken the connectivity with the IOG and en- retrieval-failure”. This failure may be not a general retrieval
hance connectivity with inhibition-related regions. failure of semantic information but likely a suppression failure
Stronger functional connectivity between the left MFG and of distracting semantic information. Based on the dual-process
right PcG during information retrieval under the retraction con- theory of memory (Brydges et al., 2018; Rugg & Curran, 2007),
dition also supports the above explanation. The PcG was also the retrieval failure hypothesis supposes that both valid (retract
thought to be closely related to cognitive control, inhibition, etc. information) and invalid information (misinformation) be
(Li et al., 2006; Ragland et al., 2002). With a stop-signal task of automatically activated during retrieval process. And either
response inhibition, Li et al. (2006) found that more efficient failure of strategic retrieval or no strategic retrieval during pro-
response inhibition was associated with middle frontal and cessing will lead to the CIEM (Ecker et al., 2011; Lew-
precentral frontal cortices etc. It needs to be pointed out that, andowsky et al., 2012). From the view of inhibition, our results
although the enhancement of functional connectivity among prove that strategic retrieval may be performed in retraction,
brain regions associated with inhibition under retraction condi- because the brain areas related to inhibition were activated at
tion is in response to the need to inhibit misinformation, it may the retrieval phase. However, the inhibition failure or imperfec-
not be able to “achieve what you want”. Akkermans et al. (2018) tion reflected by the weaker activation and the increased con-
revealed that smokers had stronger functional connectivity nectivity strength in the regions for inhibition at the retrieval
among brain regions associated with inhibition, but they per- phase may result in participants’ continued reliance on the pre-
formed poorly in the task of response inhibition. vious misinformation. The negative correlation of activation
The results of this study are not completely consistent with beta values in the inhibitory-related region (MFG) with infer-
those of Gordon et al. (2017). The study of Gordon et al. (2017) ence scores also partially supports the above interpretation.
may reduce the detection power of the experimental design due To sum up, the occurrence of the CIEM may be related to
to the lower signal-to-noise ratio caused by slightly less number both the failure of encoding in the process of information com-
of participants and stimuli. They recruited 23 valid participants, prehension and the failure of retrieval in the process of infor-
and used only 11 stimuli in each condition. mation retrieval. The mental-model-updating and memory-
5.2 Implications for the CIEM cognitive hypothesis retrieval-failure hypothesis may explain the different phases of
As mentioned above, there are currently two main theories the CIEM formation. However, at present, it is necessary to
trying to explain the CIEM. One is the hypothesis of “men- further explore the more precise mechanism, such as whether
tal-model-updating”. It is believed that the occurrence of the the failures of encoding and retrieval play different and deci-
CIEM is mainly due to abnormal encoding or integration of sive roles in the occurrence of the CIEM.
retraction information at the encoding phase. The other is the 5.3 Limitations and future studies
hypothesis of “memory retrieval failure”, emphasizing anomaly Our findings provide more abundant experimental evidence
in the processing of inference problems at the retrieval phase. for clarifying the neural basis of the CIEM and refine the theo-
Thus, the activation differences at the encoding phase may be retical hypothesis of the CIEM, but this study also has the fol-
more likely to support the hypothesis of the “mental-model- lowing limitations. Firstly, referring to the paradigm of Gordon
updating”, while the differences at the retrieval phase may be et al. (2017), the baseline was set as a control condition without
more likely to support the hypothesis of “memory-retrieval- misinformation. Although the within-item design avoids the
failure”. signal pollution caused by differences in physical features and
The finding that the left MTG related to reading compre- semantics of target sentences under different conditions, it also
hension has inter-condition differences in activation at informa- has some shortcomings. The information of sentence 5 (target
tion encoding or integration supports the hypothesis of “men- sentence) and sentence 2 (condition sentence) was partially
tal-model-updating”. In situational or mental model updating, repeated in the retraction condition, while their information was
readers may try to eliminate conflict or incompatibility when not repeated in the control condition. The observed in-
the context information is inconsistent (Wang et al., 2006). ter-condition differences in brain activation may also be mixed
Therefore, it is possible that updating a mental model in retrac- with the difference in information repeatability between these
tion condition also involves the process of inhibiting control. conditions. However, the inter-condition differences in brain
JIN Hua et al.: The neural basis of the continued influence effect of misinformation

activation of target sentences observed here are unlikely to be of this study has received the help and guidance of Wang
caused by the different information repeatability between these Zhengguang, Zhao Guang, Zhang Peng, Li Chengshi, Zhang
target sentences and corresponding condition sentences. The Qihan, Wang Qiang, Xu Zhansheng, Liu Ting, Liang Ziping, Shao
information enhancement effect of vocabulary or sentence Mengling, Zhu Ziliang, Zhang Yuping, Jin Xiaokang, Wang
repetition primarily elicited increased activity in relevant brain Chao, etc. Thank you together. At the same time, we are very
regions (Wang et al., 2016; Weber et al., 2016), whereas our grateful to all the participants for their support to this study.
study found decreased activation in certain regions in the re-
References
traction condition. In any case, it is necessary for future re-
search to further improve the material setting to directly ex- Akkermans, S. E. A., Luijten, M., van Rooij, D., Franken, I. H. A., &
clude the above possibilities. Buitelaar, J. K. (2018). Putamen functional connectivity during in-
Secondly, the study mainly focused on inhibition-related hibitory control in smokers and non-smokers. Addiction Biology,
brain regions, while the processing of interference (misinfor- 23(1), 359−368.
mation) inhibition may also involve the monitoring and resolv- Brydges, C., & Ecker, U. K. H. (2018, July 17). Updating misinforma-
tion in memory after correction: An event-related potentials (ERP)
ing of conflict information. For example, meta-analysis by Li et
study. OSF, https://doi.org/10.17605/OSF.IO/VKZXM
al. (2017) found that conflict monitoring and conflict resolution
Brydges, C., Gordon, A., & Ecker, U. K. H. (2018, July 18). Exploring
involved a combination of domain-general and domain-specific
the electrophysiological correlates of the continued influence effect
cognitive control network. There is a conflict between the re- of misinformation. PsyArXiv, https://doi.org/10.31234/osf.io/4zjyh
traction information and the initial misinformation in the cur- Brydges, C. R., Gignac, G. E., & Ecker, U. K. H. (2018). Working memory
rent paradigm, which may involve different phases of conflict capacity, short-term memory capacity, and the continued influence
processing (conflict monitoring and conflict resolution). The effect: A latent-variable analysis. Intelligence, 69, 117−122.
brain regions included in this study only partially overlap with Brydges, C. R., Gordon, A., & Ecker, U. K. H. (2020). Electrophysio-
those discovered by Li et al. (2017) (such as DLPFC, ACC, logical correlates of the continued influence effect of misinforma-
etc.). Future study could consider other cognitive control re- tion: An exploratory study. Journal of Cognitive Psychology, 32(8),
gions, such as the superior and inferior parietal cortex in the 771−784.
fronto-parietal network, to reveal the neural basis of the CIEM Butterfuss, R., & Kendeou, P. (2020). Reducing interference from
more comprehensively. In addition, some studies have found misconceptions: The role of inhibition in knowledge revision.
Journal of Educational Psychology, 112(4), 782−794.
that cognitive ability affects retraction effects (De Keers-
Christopher, M. E., Miyake, A., Keenan, J. M., Pennington, B., Defries,
maecker & Roets, 2017), while inhibition ability predicts read-
J. C., Wadsworth, S. J., Willcutt, E., & Olson, R. K. (2012). Pre-
ing ability (Doyle et al., 2018). In the future, the relationship dicting word reading and comprehension with executive function
between inhibitory control and the CIEM can be proved more and speed measures across development: A latent variable analysis.
directly by examining the differences in neural basis of indi- Journal of Experimental Psychology: General, 141(3), 470−488.
viduals with different inhibitory control abilities in the CIEM. Cohen-Gilbert, J. E., Nickerson, L. D., Sneider, J. T., Oot, E. N., Serai-
Finally, this study focused on the general process of the kas, A. M., Rohan, M. L., & Silveri, M. M. (2017). College binge
neural basis of CIEM, and adopted a brief retraction. However, drinking associated with decreased frontal activation to negative
a large number of the CIEM behavioral studies have explored emotional distractors during inhibitory control. Frontiers in Psy-
more retraction methods such as detailed explanation (Ecker et chology, 8, 1650.
al., 2020; Swire et al., 2017) and information sources (Ecker & Collette, F., & van der Linden, M. (2002). Brain imaging of the central
Antonio, 2021), etc. The future study could explore whether the executive component of working memory. Neuroscience and
Biobehavioral Reviews, 26(2), 105−125.
effectiveness of retraction methods is related to neural activity
Commander, N. E., & Stanwyck, D. J. (1997). Illusion of knowing in
during information processing or retrieval.
adult readers: Effects of reading skill and passage length. Contem-
6 Conclusions porary Educational Psychology, 22(1), 39−52.
De Keersmaecker, J., & Roets, A. (2017). “Fake news”: Incorrect, but
This study investigated the brain activation characteristics of hard to correct. The role of cognitive ability on the impact of false
the CIEM in brain regions related to semantic comprehension information on social impressions. Intelligence, 65, 107−110.
and inhibition, and its relationship with behavioral indicators Di, X., Zhang, Z., & Biswal, B. B. (2021). Understanding psycho-
and changes in the strength of brain function connectivity by physiological interaction and its relations to beta series correlation.
task-fMRI. The results showed that the occurrence of CIEM Brain Imaging and Behavior, 15(2), 958−973.
was not only related to the weakened activation of the left Doyle, C., Smeaton, A. F., Roche, R. A. P., & Boran, L. (2018). Inhibi-
MTG during information encoding, but also to the weakened tion and updating, but not switching, predict developmental dys-
activation of the left MFG and right ACC during information lexia and individual variation in reading ability. Frontiers in Psy-
chology, 9, 795−795.
retrieval, the weakened functional connectivity between the left
Ecker, U. K. H., & Antonio, L. M. (2021). Can you believe it? An in-
MFG (right ACC) and IOG, and the enhanced functional con-
vestigation into the impact of retraction source credibility on the
nectivity between the left MFG and right PcG. The findings continued influence effect. Memory & Cognition, 49(4), 631−644.
suggest that the occurrence of CIEM may be related to both the Ecker, U. K. H., Hogan, J. L., & Lewandowsky, S. (2017). Reminders
failure of semantic encoding during information comprehen- and repetition of misinformation: Helping or hindering its retrac-
sion and the failure of inhibition of misinformation retrieval tion?. Journal of Applied Research in Memory and Cognition, 6(2),
during information retrieval. The hypothesis of men- 185−192.
tal-model-updating and memory-retrieval-failure may explain Ecker, U. K. H., Lewandowsky, S., Cheung, C. S. C., & Maybery, M. T.
the different phases of the CIEM formation. (2015). He did it! she did it! no, she did not! multiple causal expla-
nations and the continued influence of misinformation. Journal of
Acknowledgements: The data collection and analysis process Memory and Language, 85, 101−115.
Acta Psychologica Sinica

Ecker, U. K. H., Lewandowsky, S., Swire, B., & Chang, D. (2011). ling test. Neuropsychologia, 35(4), 559−566.
Correcting false information in memory: Manipulating the strength Ragland, J. D., Turetsky, B. I., Gur, R. C., Gunning-Dixon, F., Turner,
of misinformation encoding and its retraction. Psychonomic Bulle- T., Schroeder, L., Chan, R., & Gur, R. E. (2002). Working memory
tin & Review, 18(3), 570−578. for complex figures: An fMRI comparison of letter and fractal
Ecker, U. K. H., O’Reilly, Z., Reid, J. S., & Chang, E. P. (2020). The n-back tasks. Neuropsychology, 16(3), 370−379.
effectiveness of short-format refutational fact-checks. British Jour- Rich, P. R., & Zaragoza, M. S. (2016). The continued influence of
nal of Psychology, 111(1), 36−54. implied and explicitly stated misinformation in news reports. Jour-
Gordon, A., Brooks, J. C. W., Quadflieg, S., Ecker, U. K. H., & nal of Experimental Psychology: Learning, Memory, and Cognition,
Lewandowsky, S. (2017). Exploring the neural substrates of misin- 42(1), 62−74.
formation processing. Neuropsychologia, 106, 216−224. Rich, P. R., & Zaragoza, M. S. (2020). Correcting misinformation in
Gordon, A., Quadflieg, S., Brooks, J. C. W., Ecker, U. K. H., & news stories: An investigation of correction timing and correction
Lewandowsky, S. (2019). Keeping track of ‘alternative facts’: The durability. Journal of Applied Research in Memory and Cognition,
neural correlates of processing misinformation corrections. Neuro- 9(3), 310−322.
Image, 193, 46−56. Rugg, M. D., & Curran, T. (2007). Event-related potentials and recog-
Graesser, A. C., Singer, M., & Trabasso, T. (1994). Constructing infer- nition memory. Trends in Cognitive Sciences, 11(6), 251−257.
ences during narrative text comprehension. Psychological Review, Steuwe, C., Daniels, J. K., Frewen, P. A., Densmore, M., Theberge, J.,
101(3), 371−395. & Lanius, R. A. (2015). Effect of direct eye contact in women with
Hinault, T., Larcher, K., Zazubovits, N., Gotman, J., & Dagher, A. PTSD related to interpersonal trauma: Psychophysiological interac-
(2019). Spatio-temporal patterns of cognitive control revealed with tion analysis of connectivity of an innate alarm system. Psychiatry
simultaneous electroencephalography and functional magnetic res- Research-Neuroimaging, 232(2), 162−167.
onance imaging. Human Brain Mapping, 40(1), 80−97. Swire, B., Ecker, U. K. H., & Lewandowsky, S. (2017). The role of
Hoscheidt, S. M., Nadel, L., Payne, J., & Ryan, L. (2010). Hippocam- familiarity in correcting inaccurate information. Journal of Experi-
pal activation during retrieval of spatial context from episodic and mental Psychology Learning Memory and Cognition, 43(12), 1948−1961.
semantic memory. Behavioural Brain Research, 212(2), 121−132. van Moort, M. L., Jolles, D. D., Koornneef, A., & van den Broek, P.
Jia, X. Z., Wang, J., Sun, H. Y., Zhang, H., Liao, W., Wang, Z., Yan, C. (2020). What you read versus what you know: Neural correlates of
G., Song, X. W., & Zang, Y. F. (2019). RESTplus: An improved accessing context information and background knowledge in con-
toolkit for resting-state functional magnetic resonance imaging data structing a mental representation during reading. Journal of Ex-
processing. Science Bulletin, 64(14), 953−954. perimental Psychology: General, 149(11), 2084−2101.
Jin, H., Liu, H. L., Mo, L., Fang, S. Y., Zhang, J. X., & Lin, C. D. Wahlheim, C. N., Alexander, T. R., & Peske, C. D. (2020). Reminders
(2009). Involvement of the left inferior frontal gyrus in predictive of everyday misinformation statements can enhance memory for
inference making. International Journal of Psychophysiology, 71(2), and beliefs in corrections of those statements in the short term.
142−148. Psychological Science, 31(10), 1325−1339.
Johnson, H. M., & Seifert, C. M. (1994). Sources of the continued Wang, R. M., Mo, L., Jia, D. M., Leng, Y., & Li, L. (2006). Mechanism
influence effect: When misinformation in memory affects later in- of constructing and updating situation models in text-reading. Acta
ferences. Journal of Experimental Psychology: Learning, Memory, Psychologica Sinica, 38(1), 30−40.
and Cognition, 20(6), 1420−1436. Wang, R. M., Mo, L., Li, L., & Jin, H. (2008). Mechanism of the coor-
Jolles, D. D., Mennigen, E., Gupta, M. W., Hegarty, C. E., Bearden, C. dinating integration in text-reading. Acta Psychologica Sinica,
E., & Karlsgodt, K. H. (2020). Relationships between intrinsic 40(11), 1165−1177.
functional connectivity, cognitive control, and reading achievement Wang, W. C., Brashier, N. M., Wing, E. A., Marsh, E. J., & Cabeza, R.
across development. NeuroImage, 221, 117202. (2016). On known unknowns: Fluency and the neural mechanisms
Lewandowsky, S., Ecker, U. K. H., Seifert, C. M., Schwarz, N., & of illusory truth. Journal of Cognitive Neuroscience, 28(5), 739−746.
Cook, J. (2012). Misinformation and its correction. Psychological Weber, K., Christiansen, M. H., Petersson, K. M., Indefrey, P., &
Science in the Public Interest, 13(3), 106−131. Hagoort, P. (2016). fMRI syntactic and lexical repetition effects re-
Li, C. R., Huang, C., Constable, R. T., & Sinha, R. (2006). Imaging veal the initial stages of learning a new language. The Journal of
response inhibition in a stop-signal task: Neural correlates inde- Neuroscience, 36(26), 6872−6880.
pendent of signal monitoring and post-response processing. The Weber, K., Micheli, C., Ruigendijk, E., & Rieger, J. W. (2019). Sen-
Journal of Neuroscience, 26(1), 186−192. tence processing is modulated by the current linguistic environment
Li, Q., Yang, G., Li, Z., Qi, Y., Cole, M. W., & Liu, X. (2017). Conflict and a priori information: An fMRI study. Brain and Behavior, 9(7),
detection and resolution rely on a combination of common and dis- e01308.
tinct cognitive control networks. Neuroscience & Biobehavioral Yan, C. G., Wang, X. D., Zuo, X. N., & Zang, Y. F. (2016). Dpabi: Data
Reviews, 83, 123−131. processing & analysis for (resting-state) brain imaging. Neuroin-
Masson, S., Potvin, P., Riopel, M., & Foisy, L. M. B. (2014). Differ- formatics, 14(3), 339−351.
ences in brain activation between novices and experts in science Yang, X. H., Li, H. J., Lin, N., Zhang, X. P., Wang, Y. S., Zhang, Y.,
during a task involving a common misconception in electricity. Zhan, Q., Zuo, X. N., & Yang, Y. F. (2019). Uncovering cortical ac-
Mind Brain & Education, 8(1), 44−55. tivations of discourse comprehension and their overlaps with com-
Morita, T., Asada, M., & Naito, E. (2019). Developmental changes in mon large-scale neural networks. NeuroImage, 203, 116200.
task-induced brain deactivation in humans revealed by a motor task. Yuan, Q. M., Ma, F. Y., Zhang, M., Chen, M., Zhang, Z. Q., Wu, J. J.,
Developmental Neurobiology, 79(6), 536−558. Lu, C. M., & Guo, T. M. (2021). Neural interaction between lan-
Nathaniel-James, D. A., Fletcher, P., & Frith, C. D. (1997). The func- guage control and cognitive control: Evidence from cross-task ad-
tional anatomy of verbal initiation and suppression using the hay- aptation. Behavioural Brain Research, 401, 113086.
JIN Hua et al.: The neural basis of the continued influence effect of misinformation

Appendix:
Sample of fMRI experimental materials
Control condition Retraction condition
1=A new office building suddenly collapsed today
2 = Bystanders said the collapse occurred at 11: 00 in the morning 2 = Bystanders said there was a fire on the lower floor
3=The building was completed only two months ago
4 = The surrounding area has been banned from public use
5= Firefighters at the scene said there was no evidence of fire
6=The person in charge of the building has been contacted for explanation
Inference problems: 1. Fire prevention measures should be reviewed
2. The fire brigade must arrive at the scene as soon as possible
3. The fire alarm has probably been activated

You might also like