Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

Memory

ISSN: (Print) (Online) Journal homepage: www.tandfonline.com/journals/pmem20

Witnesses’ susceptibility to misleading post-event


information delivered in a social media-style video

Stefanie J. Sharman, Meaghan C. Danby & Atticus D. Gray

To cite this article: Stefanie J. Sharman, Meaghan C. Danby & Atticus D. Gray (2024) Witnesses’
susceptibility to misleading post-event information delivered in a social media-style video,
Memory, 32:1, 100-110, DOI: 10.1080/09658211.2023.2294692

To link to this article: https://doi.org/10.1080/09658211.2023.2294692

© 2023 The Author(s). Published by Informa


UK Limited, trading as Taylor & Francis
Group

Published online: 26 Dec 2023.

Submit your article to this journal

Article views: 678

View related articles

View Crossmark data

This article has been awarded the Centre


for Open Science 'Open Data' badge.

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=pmem20
MEMORY
2024, VOL. 32, NO. 1, 100–110
https://doi.org/10.1080/09658211.2023.2294692

BRIEF REPORT

Witnesses’ susceptibility to misleading post-event information delivered in a


social media-style video
Stefanie J. Sharman , Meaghan C. Danby and Atticus D. Gray
Deakin University, Geelong, Australia

ABSTRACT ARTICLE HISTORY


In many criminal cases, outcomes rely on eyewitness evidence. Exposure to misleading Received 24 July 2023
information after an event reduces the accuracy of witnesses’ memories. In some Accepted 4 December 2023
circumstances, warnings about misinformation can protect witnesses. As social media is a
KEYWORDS
growing source of misleading information, this study examined the effect of misleading Misinformation effect; pre-
post-event information delivered via a social media-style video, as well as the utility of a warnings; suggestibility;
minimal versus detailed warning. Participants (N = 145) watched a video showing an social media; eyewitness
electrician stealing items from a client’s home. Next, they received one of three pre-warnings memory
regarding forthcoming misleading information: a minimal warning indicating that caution
should be taken, a detailed warning specifying the presence of misleading information, or
no warning. Participants received the misleading information via a social media video or a
standard text-based narrative. Finally, they completed a recognition test. Although delivery
method did not affect errors for misleading items, detailed warnings were only effective
against text-based misleading information. Participants were more confident about their
correct than incorrect responses for misleading items; confidence was not affected by
delivery method or warning. This experiment is the first to demonstrate people’s
susceptibility to misleading post-event information delivered in a social media-style video
using an eyewitness paradigm.

When people receive misleading or incorrect post-event charged (e.g., R v Alexander [2013]; Strass v Police
information about an event that they have experienced, [2013]). However, if witnesses view photos of innocent
the accuracy of their memories for the original event is people, these photos may act as misleading information
reduced; this phenomenon is known as the misinforma- and affect witnesses’ memories of the offender (Eisen
tion effect (see Loftus, 2005, for a review). This effect has et al., 2020).
important implications for the justice system as eyewitness Experiments investigating the misinformation effect
testimony is one of the most powerful forms of evidence have typically used a three-stage procedure (e.g., Loftus
(Williams et al., 1992). Witnesses may encounter mislead- et al., 1978; Takarangi et al., 2006). First, participants
ing information through any number of sources, including experience an event, often in the form of watching a
other witnesses, the questions asked by investigative inter- video. Next, they are given a written description of the
viewers, and media reports (e.g., Gabbert et al., 2012; event and told that this description was provided by
Goodwin et al., 2017; Sharman & Powell, 2012). One poten- another participant. This description contains key details
tial source of misleading information that has become from the original event, some of which are described incor-
more prevalent over recent years is social media. For rectly (misleading items). For example, the logo on a van
example, in a study of the spread of verified true and might be described as an “AJs electricians” logo, when in
false news stories in Twitter from 2006 to 2017, Vosoughi fact it was an “RJ’s electricians” logo. Other details are
et al. (2018) found that false news stories spread faster described without any misleading information (control
and further than true news stories, particularly for political items). For example, the logo on a van might be described
news stories. There are documented cases in which wit- as a “logo”, without any additional information. Third, par-
nesses to crimes searched for suspected offenders on ticipants are given a memory test about the original event.
social media, which led to people being identified and For example, if given a recognition test, they might be

CONTACT Stefanie J. Sharman stefanie.sharman@deakin.edu.au School of Psychology, Deakin University, 221 Burwood Highway, Burwood, VIC
3125, Australia
© 2023 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group
This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License (http://creativecommons.org/licenses/by-
nc-nd/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited, and is not altered, transformed, or
built upon in any way. The terms on which this article has been published allow the posting of the Accepted Manuscript in a repository by the author(s) or with their consent.
MEMORY 101

asked to choose between the control item and the mis- more systematically. Overall, this literature suggests that
leading item (“Did the logo say AJ’s electricians or RJ’s elec- delivering misinformation via videos may be more dama-
tricians?”). Results show that participants’ accuracy is lower ging to witnesses’ memory accuracy (i.e., create a larger
for misleading items than control items (Sharman & misinformation effect) than delivering misinformation via
Powell, 2012; Takarangi et al., 2006). The Source Monitor- text.
ing Framework (SMF; Johnson et al., 1993; Mitchell & In general, decisions that rely on more superficial or
Johnson, 2000) can be used to explain the misinformation heuristic processing are less accurate than those that rely
effect. During the memory test, participants need to on more effortful or systematic processing (Hekkanen &
monitor the source of the remembered details and McEvoy, 2002; Lindsay & Johnson, 1989; Zaragoza &
decide whether they came from the original event or Lane, 1994). Nonetheless, people’s reliance on heuristic
from the post-event description. Often, this monitoring processing can be overcome when they are motivated to
fails and participants misattribute the misleading items process the information more effortfully. For example,
from the post-event description to the original event Sundar et al. (2021) found that participants rated fake
(Lindsay & Johnson, 1989; Rantzen & Markham, 1992). news stories as more credible – and had a larger intention
Video-sharing social media websites like TikTok, to share the story – when they were presented via videos
YouTube and Instagram are fast becoming a leading rather than audio or text delivery methods, which did not
source of information for many people (Anderson, 2015; differ from each other. However, this result did not occur
Chen et al., 2021), so establishing the effects of misinfor- for participants who rated the content of the news
mation delivery via social media videos is particularly stories as more personally important to them. Sundar
important. In the current experiment, we examined the et al. reasoned that participants who rated the stories
effects of providing participants with misleading post- more personally important were more likely to systemati-
event information in a standard text-based narrative or cally process the stories and less likely to rely on heuristics
in a social media-style video. To date, no studies have than participants who did not rate the stories as personally
examined the effects of misleading information delivered important. As a result, participants who rated the stories as
through social media videos on witnesses’ memory, more personally important did not find them more cred-
although they have examined other sources of misleading ible when presented in video format, compared to audio
information, including co-witnesses and photos (see, for or text. This finding fits with Chaiken’s (1980) proposal
example, Greene et al., 2022; Paterson & Kemp, 2006). In that people will typically rely on a heuristic strategy to
fact, we only found one study that considered if mislead- reduce the amount of cognitive effort they spend proces-
ing information delivered in video format can affect sing information; however, they will switch to a more sys-
memory reports (compared to no misinformation; Cullen tematic strategy when they think it is important to make
et al., 2021). In this study, participants watched a video an accurate judgment.
of a simulated crime; later, some participants watched a In our experiment, we expected our participants to rely
re-enactment video containing misleading items while predominantly on heuristic processing in their source
other participants did not. Participants who watched the monitoring decisions. However, we also expected them
re-enactment video reported more misleading items to make source decisions more systematically when the
than those who did not watch it, demonstrating that mis- task requirements encouraged them to do so (Lindsay &
leading information delivered in a video affects partici- Johnson, 1989). Research has shown that participants in
pants’ memory reports. misinformation experiments can be encouraged to use
Previous communications research examining the more systematic processing and adopt more stringent
credibility of fake news stories can provide us some source monitoring criteria at test (e.g., Parker et al.,
ideas about how video versus text-based misinformation 2008). Indeed, when participants are given a post-
may affect participants’ memory reports. Experimental warning – that is, they are given a warning that they
findings have shown that when people are exposed to have been exposed to misleading information after their
fake news via video or text, they rate the credibility of exposure to it and before taking the memory test – their
fake news in videos higher than fake news in text (e.g., test accuracy improves (see Blank & Launay, 2014, for a
Lee & Shin, 2022; Sundar et al., 2021; Tseng et al., 2022). review). This warning reduces the misinformation effect
This effect is likely due to participants relying on a by encouraging people to examine their memories more
“seeing is believing” heuristic when processing the misin- closely when deciding whether the characteristics con-
formation. Both the Modality–Agency–Interactivity–Navig- tained in each memory are indicative of the post-event
ability (MAIN) Model (Sundar, 2008) and the Limited description or of the original event (Eakin et al., 2003;
Capacity Model (Lang, 2000) suggest that video infor- Oeberst & Blank, 2012). Post-warnings are generally more
mation is processed more heuristically than text-based ecologically valid than pre-warnings (which are given
information because the multi-media format of video before participants are exposed to the misleading infor-
may overwhelm people’s finite information-processing mation), as investigative interviewers do not usually
resources. Text-based information, on the other hand, know whether witnesses have encountered misleading
which consists of just one medium (text) is processed information after witnessing an event – they can only
102 S. J. SHARMAN ET AL.

warn witnesses before they are interviewed (see also Ech- should make more errors for misleading items than partici-
terhoff et al., 2005). It is usually only in experimental pants in the text condition. We predicted that, for both
designs that researchers are aware that participants will conditions, the minimal warning should perform similarly
encounter misleading information and can warn them in to no warning as the minimal warning does not specify
advance about it. that misleading information is present or provide an expla-
Social media videos present one of the few real-world nation about why it may be there. In contrast, the detailed
situations in which pre-warnings are given: social media warning should encourage participants to process the
videos sometimes appear with warnings about the vera- post-event information systematically, which – in turn –
city of their content and these warnings appear at the should enhance participants’ ability to detect differences
beginning of the videos. Research examining the effective- between the original event and the post-event infor-
ness of pre-warnings given before text-based misinforma- mation (Tousignant et al., 1986). However, we might
tion suggests that they reduce the misinformation effect expect that our video participants should find it more
(e.g., Chambers & Zaragoza, 2001; Greene et al., 1982; Kar- difficult to detect discrepancies (compared to our text par-
anian et al., 2020). Pre-warnings may affect both the ticipants) as our video participants were not in control of
encoding and retrieval of information. First, at encoding, the speed at which they watched the post-event infor-
pre-warnings may encourage participants to scrutinise mation – it was shown at one set pace. In comparison,
the post-event information more carefully for details that our text participants could slow down their reading,
are different from the original event (Bailey et al., 2021; which should provide them with a greater opportunity
Greene et al., 1982; Karanian et al., 2020). For example, to detect discrepancies between the original event and
Greene et al. (1982) found that participants who were the post-event narrative (Greene et al., 1982; Tousignant
pre-warned read a text-based event description containing et al., 1986). Therefore, we also predicted that the detailed
misleading information more slowly than participants who warning should be less effective in helping participants to
were not pre-warned. Slowing down their reading time avoid making errors when they are given misleading infor-
may have provided participants with a greater opportunity mation in a video compared to a text-based narrative. In
to detect discrepancies between the original event and the other words, we expected our video participants to make
post-event description, thus reducing the misinformation more errors than our text participants when they were
effect (Tousignant et al., 1986). Second, at retrieval, pre- given a detailed warning.
warnings may encourage participants to make slower Finally, we explored whether participants’ confidence in
and more deliberate source monitoring decisions, which their responses on the memory test were affected by the
may increase the accuracy of those decisions (Blank & delivery method and the level of warning that participants
Launay, 2014; Karanian et al., 2020). received. We predicted that participants should show the
In our study, we exposed participants to one of two “signature pattern of misinformation” (p. 417, Higham
delivery methods for misleading information (a text- et al., 2017); that is, in addition to making more errors
based narrative or a social media-style video) and tested about misleading items, they should be more confident
the effects of three types of pre-warning on participants’ about their responses for these items (see also Tomes &
memory reports: no warning, a minimal warning, or a Katz, 2000; Weingardt et al., 1994). We also predicted
detailed warning. The minimal warning was based on that a detailed warning should reduce participants’ confi-
the warning applied to TikTok videos that fact-checkers dence for misleading items whereas the minimal
are unable to verify as correct or not: “Caution: Video warning should have no effect (see Higham et al., 2017
flagged for unverified content”. In contrast, our detailed for similar results). We made no prediction about delivery
warning stated that misleading information was present method on confidence or the interaction between delivery
in the post-event information, discredited the source of method and warning.
the misleading information, and contained an explanation
about why misleading information was included in our
experiment. Our detailed warning was based on Blank
and Launay’s (2014) meta-analysis on post-warning Method
studies. It was appropriate to use the wording from post-
Participants
warnings as the pre-warning in our experiment as the
main difference between the two is the timing of the One hundred and forty-five participants completed the
warning (see also Karanian et al., 2020). experiment. Forty-two percent of participants were aged
We expected participants to rely predominantly on 18–29 years, 29% were aged 30–39 years, 19% were
heuristic processing when monitoring the source of their aged 40–49 years and 10% were aged 50 years and over.
memories at test, regardless of whether they were One hundred participants identified as female, 41 ident-
exposed to misleading post-event information in a video ified as male and 4 participants identified as non-binary.
or in text. However, because video-based information is A posthoc power analysis revealed that the power we
processed more shallowly than text-based information, had to detect the interaction between delivery method
we predicted that participants in the video condition and warning was .73.
MEMORY 103

Fifty-nine percent of participants were undergraduate condition, followed by a warning that was based on
students recruited through a university, and 41% of partici- Blank and Launay’s (2014) meta-analysis on post-warning
pants were recruited through Prolific. There was no differ- studies; it was specifically adapted from Oeberst and
ence in age categories of participants recruited from the Blank’s (2012) study. Video participants were told:
student population or Prolific, χ2 (145, df = 3) = 3.59, p
Please read the following information carefully. This exper-
= .309. There was a difference in gender:1 a greater pro- iment is about the psychology of eyewitness memory and
portion of students identified as female (82.1%) than the the influence of incorrect post-event information. Imagine
proportion from Prolific (54.4%), χ2 (141, df = 1) = 12.69, you had witnessed a traffic accident. Later, you might
p < .001, Cramer’s V = .30. discuss it with other witnesses or read an article about it in
the newspaper. Some details of the accident may be missing
in such reports, and others may be wrong, without you necess-
arily noticing this at first. In this experiment, we are simulating
Design
such a situation: You watched a video about Eric the electri-
The design of the experiment was a 2 (post-event infor- cian, and—on the next page—you will see a TikTok video
about Eric’s actions, which was purportedly provided by a pre-
mation: misleading, control) × 2 (delivery method: video,
vious participant in this experiment. In fact, this TikTok video
text) × 3 (warning: none, minimal, detailed). Post-event was made by the researchers. In a later recognition test, you
information was a within-subjects factor; delivery will be asked questions about the electrician video. When
method and warning were between-subjects factors. answering the questions, please rely exclusively on your
Memory performance was examined through the memory of the electrician video, rather than any information
from the TikTok video.
number of errors that participants made in response to
the critical items on the recognition test. Confidence was Text participants received the same warning, except that
examined through participants’ mean confidence ratings the phrase “TikTok video” was replaced with “text
to these same critical items on the recognition test. description”.
Post-event information. Participants were randomly
allocated to receive post-event information in a 3-min
Materials
social media-style video or a 445-word text-based narra-
Stimulus video. Participants watched a 6 min 28 s video tive. The video was a “reaction” video in which a female
showing an electrician (named Eric) working in a client’s actor reacted while watching the video of Eric the electri-
house; he looked through their belongings and stole cian; it appeared to be hosted on the TikTok platform
some items (Takarangi et al., 2006). It has been used in because it had a frame consisting of the standard icons
many past misinformation experiments (e.g., Battista on the right-hand side and across the bottom of the
et al., 2020; Otgaar et al., 2014; Sharman & Powell, 2012). screen. Participants did not see any of the original electri-
The video contained eight critical details (magazine title, cian video again in the social media-style video. The social
picture on the wall, whether the bed was made, the media-style video, as well as the text-based narrative, are
colour of a cap, the name of the electrician’s company, both available through the OSF: https://osf.io/7c8v2/?
how Eric checked the time, the colour of a mug, and the view_only = f516ec2606d844cd97f906393f0a4237.
soft drink that Eric drank). Both the video and the text-based narrative contained
Warning. Participants were randomly allocated to the same information, including the eight critical items
receive one of three warnings: no warning, minimal from the electrician video. Half of the critical items were
warning, or detailed warning. In the no warning condition, described in a misleading way (misleading items), which
participants did not receive a warning. Video participants provided contradictory information to what participants
were told: “On the next page you will see an account of had seen in the electrician video. For example, in the
Eric’s actions, recorded by a previous participant in this video, Eric looked at a picture of the Eiffel Tower on the
experiment. Please watch it carefully”. Text participants wall. Participants who received misleading information
were told: “On the next page you will read an account of for this item were told that “Eric looked at a picture of
Eric’s actions, written by a previous participant in this the Leaning Tower”. The other four critical items (control
experiment. Please read it carefully”. items) were described using neutral language such that
Participants in the minimal warning condition received they did not contain any specific information, such as
the above information, along with a warning based on “Eric drank a soft drink”.
TikTok’s warning for videos whose content could not be Recognition test. The recognition test consisted of 20
verified by factcheckers. Video participants were told, questions in which participants chose between two
“Caution: the following video has been flagged for unver- options. Eight of the questions were about the critical
ified content”. Text participants were told, “Caution: the details; the other 12 questions were about non-critical
following text description has been flagged for unverified details and were included so that participants did not
content”. become suspicious about the other questions. For
Participants in the detailed warning condition received example, one of the questions about non-critical details
the same information as participants in the no warning asked participants to decide whether the tool that Eric
104 S. J. SHARMAN ET AL.

used in the kitchen was pliers or a screwdriver. Four of the calculated the total number of errors that participants
eight critical questions were about misleading items; the made. This value ranged from 0 (in which participants
other four critical questions were about control items. For did not make any errors; i.e., they always selected the
all eight questions, participants chose between the correct correct detail from the original electrician video) to 4 (in
detail from the electrician video and the incorrect detail which participants made errors for all four items; i.e.,
that was introduced in the post-event information. For they always selected the detail suggested in the misinfor-
example, participants were asked, “In the lounge, the mation). Similarly, for the four control items, we calculated
picture Eric looked at was the _______Tower” and they participants’ number of errors. This could range from 0 (in
were given the response options of (a) Eiffel and (b) which participants did not make any errors; i.e., they
Leaning. Participants’ error scores were calculated as always selected the correct detail from the original electri-
described below. For all questions, participants rated their cian video) to 4 (in which participants made errors for all
confidence in their answer from 1 (not at all confident) to 5 questions; i.e., they always selected the detail that was
(very confident). not presented in the original video). We only analysed
the number of errors that participants made – and not
their number of correct responses – because the number
Procedure
of correct responses can be inferred from the number of
Ethics approval for this study was granted from Deakin Uni- errors (i.e., the number of correct responses can always
versity’s ethics committee (approval number HEAG-H be determined by subtracting the number of errors from
159_2022). Participants were initially recruited through an a total of four). For example, if a participant made two
undergraduate cognitive psychology unit; they responded errors for misleading items and one error for control
to an advertisement about the experiment. Participants items, then the participant made two correct responses
were not compensated for their time. Another 60 partici- for misleading items and three correct responses for
pants were recruited through Prolific; they were compen- control items. Data are available from https://osf.io/
sated £3 for their participation. How participants were 7c8v2/?view_only = f516ec2606d844cd97f906393f0a4237.
recruited did not affect the number of errors they made
on the memory test, F(1, 143) = 0.80, p = .374, h2p = .006.
Results
Participants followed the same procedure. They com-
pleted the online experiment individually at a time and We present the results in two sections. First, we analysed
location of their choosing. First, they responded to an the errors that participants made for the critical items in
advertisement for a study “investigating the effects of the recognition test. Second, we analysed participants’
post-event information on people’s memories for wit- confidence ratings for these same items.
nessed events”. Participants were required to be aged
between 18 and 65 years and have fluent English to partici-
Errors
pate. Next, they provided informed consent. Participants
then watched the stimulus video before completing a 3- Overall, participants made more errors for misleading
min distractor task in which they matched number pairs. items (M = 1.48, SD = 1.03) than control items (M = 0.77,
This task was given to prevent participants from mentally SD = 0.72), t(144) = 7.45, p < .001, Cohen’s d = .619. In
rehearsing any details from the video. Next, participants other words, they showed the misinformation effect.
were randomly assigned to the no warning (n = 48), Before examining the effects of misleading information
minimal warning (n = 49) or detailed warning (n = 48) con- on participants’ memories, we first determined whether
dition. After reading the corresponding warning infor- there was any difference between conditions for control
mation, participants received the post-event information. items. We conducted a 2 (delivery method: video, text) ×
They were randomly assigned to receive the post-event 3 (warning: none, minimal, detailed) ANOVA on partici-
information in a video (n = 73) or in a text-based narrative pants’ errors for control items. The main effects for delivery
(n = 72). After watching the video or reading the narrative, method and warning were not significant, F(1, 139) = 0.06,
participants completed a 1-min distractor task in which p = .809, h2p < .001 and F(2, 139) = 2.49, p = .087, h2p = .035,
they put letters in alphabetical order. This task was given respectively. The interaction between delivery method
to prevent participants from mentally rehearsing any and warning was also not significant, F(2, 139) = 0.10, p
details from the post-event information. After the distrac- = .909, h2p = .001. Therefore, we did not consider partici-
tor task, participants completed the recognition test. pants’ errors for control items any further.
Finally, all participants were debriefed about the purpose To determine whether delivery method and warning
of the experiment. affected participants’ errors for misleading items, we con-
ducted a 2 (delivery method: video, text) × 3 (warning:
none, minimal, detailed) ANOVA. Neither the main effect
Data scoring
for delivery method nor the main effect for warning con-
Responses to the eight critical items on the recognition dition was significant, F(1, 139) = 1.66, p = .200, h2p = .012
test were examined. For the four misleading items, we and F(2, 139) = 0.63, p = .538, h2p = .009, respectively.
MEMORY 105

Figure 1. Mean errors for misleading items in the text and video conditions
displayed by warning type. Note that bars are standard error bars.
Figure 2. Mean confidence for misleading items in the text and video con-
ditions displayed by warning type. Note that bars are standard error bars.

However, there was a significant two-way interaction


between delivery method and warning, F(2, 139) = 3.93,
we conducted a 2 (delivery method: video, text) × 3
p = .022, h2p = .054 (see Figure 1). To follow-up the inter-
(warning: none, minimal, detailed) ANOVA. Neither the
action, three independent samples t-tests were conducted
main effect for delivery method nor the main effect for
comparing the number of errors that participants made
warning condition was significant, F(1, 139) = 0.01, p
when the post-event information was delivered via text
= .920, h2p < .001 and F(2, 139) = 2.28, p = .106, h2p = .032,
or video for each of level of warning. When there was no
respectively. However, the two-way interaction between
warning, participants in the text and video conditions
delivery method and warning was approaching signifi-
made a similar number of errors for misleading items, t
cance, F(2, 139) = 2.80, p = .064, h2p = .039 (see Figure 2).
(46) = 0.15, p = .883, d = .043. When there was a minimal
We explored this potential interaction with three indepen-
warning, participants in the text and video conditions
dent samples t-tests comparing participants’ confidence
made a similar number of errors for misleading items, t
when the post-event information was delivered via text
(47) = 0.59, p = .558, d = .169. However, when there was a
or video for each of level of warning. When there was no
detailed warning, there was a significant difference
warning, participants in the video condition were more
between the two conditions. Participants in the text con-
confident than participants in the text condition for mis-
dition made fewer errors for misleading items than partici-
leading items, t(46) = 2.09, p = .042, d = .603. When there
pants in the video condition, t(46) = 3.26, p = .002, d = .941.
was a minimal or detailed warning, there was no difference
To examine whether the warning that participants in
in confidence for misleading items between participants in
the text condition were given affected their reading
the text and video conditions, t(47) = 1.19, p = .241, d
speed, we examined how long they spent on the page
= .339 and t(46) = 0.79, p = .435, d = .227, respectively.
showing the post-event information (as a proxy). Overall,
Next, we examined participants’ confidence for mis-
participants spent an average of 103 s on this page (SD
leading items according to whether their answer was
= 73); it was not affected by type of warning, F(2, 71) =
correct or incorrect. A 2 (delivery method: video, text) × 3
0.67, p = .513, h2p = .019.
(warning: none, minimal, detailed) × 2 (answer: correct,
incorrect) repeated measures ANOVA2 revealed that only
the main effect of answer was significant, F(1, 107) =
Confidence
6.91, p = .010, h2p = .061. Participants were more confident
Overall, participants were more confident about mislead- about misleading items when their answer was correct
ing items (M = 3.62, SD = 0.75) than control items (M = (M = 3.64, SD = 0.96) than when it was incorrect (M = 3.27,
3.23, SD = 0.75), t(144) = 5.27, p < .001, d = .437. We first SD = 1.28), t(112) = 2.60, p = .011, d = .245. No other main
determined whether there was any difference in confi- effects or interactions were significant, all Fs < 1.91, all ps
dence ratings between conditions for control items. We > .152.
conducted a 2 (delivery method: video, text) × 3 To follow-up this finding, we calculated the accuracy-
(warning: none, minimal, detailed) ANOVA on participants’ confidence relationship for participants in the text and
confidence for control items. The main effects for delivery video conditions at each level of warning. As rec-
method and warning were not significant, F(1, 139) = 1.58, ommended by Nelson (1984), an ordinal measure of
p = .212, h2p = .011 and F(2, 139) = 1.00, p = .371, h2p = .014, association (i.e., the Goodman–Kruskal gamma correlation)
respectively. The interaction between delivery method was used. Table 1 shows that, for control items, there were
and warning was also not significant, F(2, 139) = 1.90, p significant negative correlations between number of errors
= .153, h2p = .027. and confidence in all conditions: as errors increased, confi-
To determine whether delivery method and warning dence decreased. For misleading items, the only significant
affected participants’ confidence for misleading items, negative correlation was for participants who received a
106 S. J. SHARMAN ET AL.

Table 1. Gamma correlations between confidence and number of errors for effective than the minimal warning and no warning is con-
control and misleading items in each condition. sistent with Blank and Launay’s (2014) review of post-
Delivery Control Misleading warnings and the elements that make them effective.
Warning mode N items items
Our detailed warning stated that misleading information
No warning Text 24 −.73*** −.23
Video 24 −.70*** −.20 was present in the post-event information, discredited
Minimal Text 24 −.66** .15 the source of the misleading information, and “enligh-
warning tened” participants about why misleading information
Video 25 −.58** −.17
Detailed Text 24 −.62** −.44* was included in our experiment. In contrast, our minimal
warning warning only mentioned the possibility of incorrect infor-
Video 24 −.65** −.30 mation being present in the post-event information; it
Note: ***p < .001, **p < .01, *p < .05. did not discredit the source of the misleading information
or enlighten participants. Our finding that the minimal
detailed warning in the text condition. For all other con- warning did not protect people against misleading infor-
ditions, the relationship was not significant. mation is consistent with other research (Bailey et al.,
2021). Bailey et al. (2021) showed participants a video
before asking them to read a text-based narrative contain-
ing control and misleading post-event information. Half
Discussion
the participants received a pre-warning before reading
We examined the effectiveness of pre-warnings about mis- the narrative. They were told that, “There may be differ-
leading post-event information when it was delivered in a ences between the [narrative] and the video”. Finally, par-
social media-style video and in a text-based narrative. We ticipants completed a recognition test about items from
found that delivery method of misinformation had no the original video. The pre-warning had no significant
main effect on the number of errors that participants effect on participants’ responses on the test.
made for misleading items. This finding does not Our finding that the detailed warning was effective only
support our first hypothesis. There are several possible for participants in the text condition supported our third
explanations that should be examined in future research. hypothesis. Despite the detailed warning, participants in
First, it is possible that, like our video participants, text par- the video condition made more errors for misleading
ticipants relied on more heuristic processing of the post- items compared to participants in the text condition. It is
event information. Alternatively, our video participants possible that the detailed warning alerted participants to
may have relied on more systematic processing of the the possibility of encountering incorrect information,
post-event information, like our text participants. Second, which may have led to two possible strategies. First, par-
it is possible participants found the post-event information ticipants in the text condition may have slowed down
equally credible in the text and video conditions; that is, their reading speed to detect discrepancies between the
the video information was not perceived to be more cred- original event and the post-event narrative (Greene
ible (see Sundar et al., 2021). Many factors affect perceived et al., 1982; Tousignant et al., 1986). However, our data
source credibility (see Pornpitakpan, 2004, for a review). In showing how long participants spent on the page display-
the current experiment, the main difference between the ing the post-event information does not support this
video and text conditions was that participants saw the explanation: the detailed warning participants did not
actor in the video; there was no visual representation of spend longer on this page than the minimal or no
the “previous participant” in the text condition. It is poss- warning participants.
ible that participants in the video condition made credi- Second, the detailed warning may have encouraged
bility judgments based on the actor’s appearance (e.g., participants to make more careful source monitoring
Horai et al., 1974; Wang & Scheinbaum, 2018), which decisions at retrieval (Ecker et al., 2010). When participants
affected their acceptance of the misleading information. received a detailed warning, they may have more closely
As expected, the effectiveness of the pre-warning monitored the source of the remembered details to deter-
depended on its level as well as whether participants mine if they could be attributed to the original event or the
received the misleading information via video or text. post-event information. Text participants were better at
When there was no warning or a minimal warning, partici- making this attribution than video participants, possibly
pants made a similar number of errors for misleading items because the text-based narrative presentation of the
in both the text and video conditions. However, when par- post-event information encouraged participants to
ticipants received a detailed warning, they made signifi- process the post-event information more systematically
cantly more errors for misleading items in the video than compared to video participants (see also Sundar et al.,
the text condition. In other words, the detailed warning 2021). However, this processing-based explanation is not
was effective in protecting participants against misleading supported by our lack of main effect for delivery method
information, but only if this information was presented in a (video vs text) on number of errors for misleading items.
text-based narrative. The finding that – for participants in Instead, the detailed warning may have been effective
the text condition – the detailed warning was more only in the text condition because text participants had
MEMORY 107

additional source cues, which allowed them to capitalise presence of the misleading information, (2) given infor-
on the detailed warning. For text participants, they mation that discredited the source of the post-event infor-
needed to distinguish between details that appeared in mation (they were told that the researchers wrote it rather
a video (original electrician event) and details that than a previous participant) and (3) “enlightened” about
appeared in a text narrative (post-event information) at why misleading post-event information was included in
retrieval. In contrast, video participants needed to dis- the experiment. These three components correspond to
tinguish between two videos (original electrician event the three dimensions of post-warnings identified in Blank
and the post-event information). Research has shown and Launay’s (2014) meta-analysis. Their analysis showed
that the greater the similarity between two sources, the that the enlightenment component was the most
more difficult it is to discriminate between them (Ferguson effective in reducing the misinformation effect. However,
et al., 1992; Johnson et al., 1988; Lindsay et al., 1991). future research could examine these three dimensions in
Our confidence results showed that, overall, partici- pre-warning studies.
pants were more confident about their responses for mis- Second, we explained above that the minimal warning
leading items than control items, which is consistent with was not effective in reducing errors for misleading items
past research (Higham et al., 2017; Tomes & Katz, 2000; because it did not increase participants’ source monitoring
Weingardt et al., 1994). We also found that when they at retrieval. However, there are at least two reasons other
were not warned, participants in the video condition reasons why it was not effective. The first reason is that
were more confident about their answers for misleading participants did not find the minimal warning credible. In
items than participants in the text condition. When they the context of experiments, participants may make
received a minimal or detailed warning, there was no assumptions about the researchers based on the materials
difference in confidence between video and text partici- presented to them. In our experiment, it appeared that we
pants. It is possible that, when there was no warning, par- deliberately chose to expose participants to post-event
ticipants found the post-event information more credible information that had been “flagged for unverified
when presented in a video than in text (see also Sundar content”. As a result, participants may have been dismis-
et al., 2021). This increased credibility may have led to sive of the warning, which is why it did not effectively
increased confidence for their answers for the misleading reduce participants’ errors for misleading items. Future
items. It is also possible that post-event information pre- research could examine participants’ trust in the minimal
sented in the video was processed more heuristically warning.
than post-event information presented in the text (Lang, The second reason is desensitisation. It is possible that
2000; Sundar, 2008). This heuristic processing may have participants were already familiar with the TikTok-style
increased participants’ confidence ratings as people rely warning as they had seen it many times. With repeated
on different cues to make their ratings, including their exposure, the warning may have lost its effectiveness as
beliefs about memory (Koriat, 1997). people become desensitised. Indeed, research has shown
The above results reflect participants’ confidence that smokers become desensitised to the health warnings
regardless of whether their answers to the misleading on cigarette packets (Drovandi et al., 2018) and that critical
items on the recognition test were correct or incorrect. care works become desensitised to alerts in their environ-
Further analyses demonstrated that participants’ confi- ment (Ancker et al., 2017). It is possible that our minimal
dence ratings were related to accuracy, which suggests warning was not effective as participants were already
that participants had some level of insight. Gamma corre- desensitised to it and dismissed it without really applying
lations revealed that misinformation impaired resolution; it to the post-event information. One final limitation is that
that is, the ability of participants’ confidence ratings to dis- all our participants were aware that the study was investi-
criminate correct from incorrect answers (Higham et al., gating the effect of post-event information on memory for
2011). Our participants showed good resolution for witnessed events. This knowledge may have encouraged
control items – confidence ratings increased as number participants to pay more attention to the electrican
of errors decreased – which is consistent with past video and engage in rehearsal, which may be different
research (Bonham & González-Vallejo, 2009; Tomes & from real-world witness situations. However, a recent
Katz, 2000; Weingardt et al., 1994). They showed poor res- experiment examined the effects of pre-event instructions
olution for misleading items – there were no significant that informed participants they would watch a video of a
relationships between confidence and errors – except for crime and later need to identify the culprit in a line-up (Bal-
participants in the text condition who received a detailed dassari et al., 2023). Compared to participants who were
warning. Although these participants’ resolution was lower told only that they would watch a video, the “informed”
for misleading items than control items, their confidence participants did not perform better on the identification
ratings increased as their number of errors decreased. task. Thus, informing our participants that the experiment
The current experiment has several limitations. First, we was investigating witness memory may not have affected
do not know which component of the detailed warning their accuracy.
was responsible for its effectiveness for participants in Our study found that misinformation delivered via a
the text condition. Participants were (1) told about the social media video is particularly detrimental to witnesses’
108 S. J. SHARMAN ET AL.

memory reports, given that this misinformation was resist- Open Scholarship
ant to a detailed pre-warning. While our study focused on
adult participants, adolescents are accessing social media
at substantial rates (Martin et al., 2018), and misinforma-
This article has earned the Center for Open Science badge
tion effects have been shown to differ with age (e.g.,
for Open Data The data are openly accessible at https://os-
Sutherland & Hayne, 2001). Thus, future work should deter-
f.io/7c8v2/?view_only=e56dbb79036c4dc6bd9a53fd40dc
mine adolescent witnesses’ susceptibility to misinforma-
67ba
tion delivered via social media. Researchers could also
consider testing a range of social media delivery modes
(such as embedding misinformation in photographic Insta- Disclosure statement
gram posts and Facebook comment sections) because the
No potential conflict of interest was reported by the author(s).
method of delivering misinformation has been shown to
affect the size of the misinformation effect in past research
(e.g., Campbell et al., 2007; Greene et al., 2022; Ulatowska Funding
et al., 2016).
The author(s) reported there is no funding associated with the work
In conclusion, this study was the first to compare misin- featured in this article.
formation delivered via a social media style video with a
standard text-based narrative. We found misinformation
effects for both delivery methods. We additionally tested ORCID
the benefits of two pre-warnings to caution participants
Stefanie J. Sharman http://orcid.org/0000-0002-0635-047X
about upcoming misinformation. The minimal pre- Meaghan C. Danby http://orcid.org/0000-0001-7060-7121
warning did not reduce the misinformation effect for Atticus D. Gray http://orcid.org/0000-0002-0370-7488
either the social media video or the text condition. The
detailed pre-warning was successful in reducing the misin-
formation effect for the text condition, but not the social References
media video condition. Participants who received a Ancker, J. S., Edwards, A., Nosal, S., Hauser, D., Mauer, E., Kaishal, R.,
detailed warning in the text condition also had better res- with the HITEC Investigators. (2017). Effects of workload, work
olution: their confidence ratings were related to their accu- complexity, and repeated alerts on alert fatigue in a clinical
racy for misleading items. No other participants decision support system. BMC Medical Informatics Decision
Making, 17(1), Article 36. https://doi.org/10.1186/s12911-017-
demonstrated this resolution for misleading items. These 0430-8
results have important implications for eyewitnesses. Anderson, M. (2015). 5 facts about online video, for YouTube’s 10th
People who witness an event and then share details birthday. Pew Research Center http://www.pewresearch.org/fact-
(including incorrect information) about it online in a tank/2015/02/12/ 5-facts-about-online-video-for-youtubes-10th-
social media video may detrimentally influence other wit- birthday/
Bailey, N. A., Olaguez, A. P., Klemfuss, J. Z., & Loftus, E. F. (2021). Tactics
nesses’ memories. Even warning potential viewers that the for increasing resistance to misinformation. Applied Cognitive
video that they are about to watch contains incorrect infor- Psychology, 35(4), 863–872. https://doi.org/10.1002/acp.3812
mation does not protect them against it. Given the speed Baldassari, M.J., Moore, K.N., Hyman, I.E., Hope, L., Mah, E. Y., Lindsay,
and scope with which misleading information can be D. S., Mansour, J., Saraiva, R., Horry, R., Rath, H., Kelly, L., Jones, R.,
spread, this is particularly concerning. Vale, S., Lawson, B., Pedretti, J., Palma, T. A., Cruz, F., Quarenta, J.,
Van der Cruyssen, … Wiechert, S. (2023). The effect of pre-event
instructions on eyewitness identification. Cognitive Research:
Principles and Implications, 8(16). https://doi.org/10.1186/s41235-
023-00471-4
Funding Battista, F., Mangiulli, I., Herter, J., Curci, A., & Otgaar, H. (2020). The
effects of repeated denials and fabrication on memory. Journal
The author(s) reported there is no funding associated with the work of Cognitive Psychology, 32(4), 369–381. https://doi.org/10.1080/
featured in this article. 20445911.2020.1767626
Blank, H., & Launay, C. (2014). How to protect eyewitness memory
against the misinformation effect: A meta-analysis of post-
warning studies. Journal of Applied Research in Memory and
Cognition, 3(2), 77–88. https://doi.org/10.1037/h0101798
Notes
Bonham, A. J., & González-Vallejo, C. (2009). Assessment of calibration
1. Note that participants who identified as non-binary were not for reconstructed eye-witness memories. Acta psychologica, 131(1),
included in this analysis as they had an expected count less 34–52. https://doi.org/10.1016/j.actpsy.2009.02.008
than 5. Campbell, J. M., Edwards, M. S., Horswill, M. S., & Helman, S.
2. This analysis was conducted on the 113 participants whose (2007). Effects of contextual cues in recall and recognition
answers for the misleading items were a mixture of correct memory: The misinformation effect reconsidered. British Journal
and incorrect, as accuracy was the within-subjects variable in of Psychology, 98(3), 485–498. https://doi.org/10.1348/000712606
this repeated measures ANOVA. It excluded participants who X160768
only got the misleading items correct (n = 29) or incorrect (n Chaiken, S. (1980). Heuristic versus systematic information processing
= 3). and the use of source versus message cues in persuasion. Journal
MEMORY 109

of Personality and Social Psychology, 39(5), 752–766. https://doi. Horai, J., Naccari, N., & Fatoullah, E. (1974). The effects of expertise and
org/10.1037/0022-3514.39.5.752 physical attractiveness upon opinion agreement and liking.
Chambers, K. L., & Zaragoza, M. S. (2001). Intended and unintended Sociometry, 37(4), 601–606. https://doi.org/10.2307/2786431
effects of explicit warnings on eyewitness suggestibility: Johnson, M. K., Foley, M. A., Suengas, A. G., & Raye, C. L. (1988).
Evidence from source identification tests. Memory & Cognition, 29 Phenomenal characteristics of memories for perceived and ima-
(8), 1120–1129. https://doi.org/10.3758/BF03206381 gined autobiographical events. Journal of Experimental
Chen, Q., Min, C., Zhang, W., Ma, X., & Evans, R. (2021). Factors driving Psychology: General, 117(4), 371–376. https://doi.org/10.1037/
citizen engagement with government TikTok accounts during the 0096-3445.117.4.371
COVID-19 pandemic: Model development and analysis. Journal of Johnson, M. K., Hashtroudi, S., & Lindsay, D. S. (1993). Source monitor-
Medical Internet Research, 23(2), e21463. https://doi.org/10.2196/ ing. Psychological Bulletin, 114(1), 3–28. https://doi.org/10.1037/
21463 0033-2909.114.1.3
Cullen, H. J., Paterson, H. M., & van Golde, C. (2021). Stopping crime? Karanian, J. M., Rabb, N., Wulff, A. N., Torrance, M. G., Thomas, A. K., &
The effect of crime re-enactments on eyewitness memory. Race, E. (2020). Protecting memory from misinformation: Warnings
Psychiatry, Psychology and Law, 28(2), 286–309. https://doi.org/10. modulate cortical reinstatement during memory retrieval.
1080/13218719.2020.1775151 Proceedings of the National Academy of Sciences, 117(37), 22771–
Drovandi, A., Teague, P.-A., Glass, B., & Malau-Aduli, B. (2018). 22779. https://doi.org/10.1073/pnas.2008595117
Australian school student perceptions of effective anti-tobacco Koriat, A. (1997). Monitoring one’s own knowledge during study: A
health warnings. Frontiers in Public Health, 6, Article 297. https:// cue-utilization approach to judgments of learning. Journal of
doi.org/10.3389/fpubh.2018.00297 Experimental Psychology: General, 126(4), 349–370. https://doi.
Eakin, D. K., Schreiber, T. A., & Sergent-Marshall, S. (2003). org/10.1037/0096-3445.126.4.349
Misinformation effects in eyewitness memory: The presence and Lang, A. (2000). The limited capacity model of mediated message pro-
absence of memory impairment as a function of warning and mis- cessing. Journal of Communication, 50(1), 46–70. https://doi.org/10.
information accessibility. Journal of Experimental Psychology: 1111/j.1460-2466.2000.tb02833.x
Learning, Memory, and Cognition, 29(5), 813–825. https://doi.org/ Lee, J., & Shin, S. Y. (2022). Something that they never said:
10.1037/0278-7393.29.5.813 Multimodal disinformation and source vividness in understanding
Echterhoff, G., Hirst, W., & Hussy, W. (2005). How eyewitnesses resist the power of AI-enabled deepfake news. Media Psychology, 25(4),
misinformation: Social postwarnings and the monitoring of 531–546. https://doi.org/10.1080/15213269.2021.2007489
memory characteristics. Memory & Cognition, 33(5), 770–782. Lindsay, D. S., & Johnson, M. K. (1989). The eyewitness suggestibility
https://doi.org/10.3758/BF03193073 effect and memory for source. Memory & Cognition, 17(3), 349–
Ecker, U. K., Lewandowsky, S., & Tang, D. T. (2010). Explicit warnings 358. https://doi.org/10.3758/BF03198473
reduce but do not eliminate the continued influence of misinfor- Lindsay, D. S., Johnson, M. K., & Kwon, P. (1991). Developmental
mation. Memory & Cognition, 38(8), 1087–1100. https://doi.org/ changes in memory source monitoring. Journal of Experimental
10.3758/MC.38.8.1087 Child Psychology, 52(3), 297–318. https://doi.org/10.1016/0022-
Eisen, M. M., Frenda, S. F., Jones, J. M., & Williams, T. Q. (2020, January/ 0965(91)90065-Z
February). How exposure to social media affects eyewitness Loftus, E. F. (2005). Planting misinformation in the human
memory. The Champion, 18–24. www.nacdl.org/Article/ mind: A 30-year investigation of the malleability of memory.
JanFeb2020-HowExposuretoSocialMediaAffectsEyewitne Learning & Memory, 12(4), 361–366. https://doi.org/10.1101/lm.
Ferguson, S. A., Hashtroudi, S., & Johnson, M. K. (1992). Age differences 94705
in using source-relevant cues. Psychology and Aging, 7(3), 443–452. Loftus, E. F., Miller, D. G., & Burns, H. J. (1978). Semantic integration of
https://doi.org/10.1037/0882-7974.7.3.443 verbal information into a visual memory. Journal of Experimental
Gabbert, F., Hope, L., Fisher, R. P., & Jamieson, K. (2012). Protecting Psychology: Human Learning and Memory, 4(1), 19–31. https://doi.
against misleading post-event information with a self-adminis- org/10.1037/0278-7393.4.1.19
tered interview. Applied Cognitive Psychology, 26(4), 568–575. Martin, F., Wang, C., Petty, T., Wang, W., & Wilkins, P. (2018). Middle
https://doi.org/10.1002/acp.2828 school students’ social media use. Journal of Educational
Goodwin, K. A., Hannah, P. J., Nicholl, M. C., & Ferri, J. M. (2017). The Technology & Society, 21(1), 213–224.
confident co-witness: The effects of misinformation on memory Mitchell, K. J., & Johnson, M. K. (2000). Source monitoring:
after collaborative discussion. Applied Cognitive Psychology, 31(2), Attributing mental experiences. In E. Tulving, & F. I. M. Craik
225–235. https://doi.org/10.1002/acp.3320 (Eds.), The Oxford handbook of memory (pp. 179–195). Oxford
Greene, C. M., Bradshaw, R., Huston, C., & Murphy, G. (2022). The University Press.
medium and the message: Comparing the effectiveness of six Nelson, T. O. (1984). A comparison of current measures of the accu-
methods of misinformation delivery in an eyewitness memory racy of feeling-of-knowing predictions. Psychological Bulletin, 95
paradigm. Journal of Experimental Psychology: Applied, 28(4), 677– (1), 109–133. https://doi.org/10.1037/0033-2909.95.1.109
693. https://doi.org/10.1037/xap0000364 Oeberst, A., & Blank, H. (2012). Undoing suggestive influence on
Greene, E., Flynn, M. S., & Loftus, E. F. (1982). Inducing resistance to mis- memory: The reversibility of the eyewitness misinformation
leading information. Journal of Verbal Learning and Verbal Behavior, effect. Cognition, 125(2), 141–159. https://doi.org/10.1016/j.
21(2), 207–219. https://doi.org/10.1016/S0022-5371(82)90571-0 cognition.2012.07.009
Hekkanen, S. T., & McEvoy, C. (2002). False memories and source- Otgaar, H., Howe, M. L., Memon, A., & Wang, J. (2014). The develop-
monitoring problems: criterion differences. Applied Cognitive ment of differential mnemonic effects of false denials and forced
Psychology, 16(1), 73–85. https://doi.org/10.1002/acp.753 confabulations. Behavioral Sciences & Law, 32(6), 718–731.
Higham, P. A., Blank, H., & Luna, K. (2017). Effects of postwarning https://doi.org/10.1002/bsl.2148
specificity on memory performance and confidence in the Parker, S., Garry, M., Engle, R. W., Harper, D. N., & Clifasefi, S. L. (2008).
eyewitness misinformation paradigm. Journal of Experimental Psychotropic placebos reduce the misinformation effect by
Psychology: Applied, 23(4), 417–432. https://doi.org/10.1037/ increasing monitoring at test. Memory (Hove, England), 16(4),
xap0000140 410–419. https://doi.org/10.1080/09658210801956922
Higham, P. A., Luna, K., & Bloomfield, J. (2011). Trace-strength and Paterson, H. M., & Kemp, R. I. (2006). Comparing methods of encoun-
source-monitoring accounts of accuracy and metacognitive resol- tering post-event information: The power of co-witness sugges-
ution in the misinformation paradigm. Applied Cognitive tion. Applied Cognitive Psychology, 20(8), 1083–1099. https://doi.
Psychology, 25(2), 324–335. https://doi.org/10.1002/acp.1694 org/10.1002/acp.1261
110 S. J. SHARMAN ET AL.

Pornpitakpan, C. (2004). The persuasiveness of source credibility: A Tousignant, J. P., Hall, D., & Loftus, E. F. (1986). Discrepancy
critical review of five decades’ evidence. Journal of Applied Social detection and vulnerability to misleading postevent information.
Psychology, 34(2), 243–281. https://doi.org/10.1111/j.1559-1816. Memory & Cognition, 14(4), 329–338. https://doi.org/10.3758/
2004.tb02547.x BF03202511
Rantzen, A., & Markham, R. (1992). The reversed eyewitness testimony Tseng, Y.-C., Bi, N., Chang, Y.-J., & Yuan, C. (2022). Investigating correc-
design: More evidence for source monitoring. Journal of General tion effects of different modalities for misinformation about COVID-
Psychology, 119(1), 37–43. https://doi.org/10.1080/00221309.1992. 19. In Companion Computer Supported Cooperative Work and Social
9921156 Computing (CSCW’22 Companion), November 08–22, 2022, Virtual
R v Alexander. (2013). 1 CrAppR 26. Event, Taiwan. https://doi.org/10.1145/3500868.3559455
Sharman, S. J., & Powell, M. B. (2012). A comparison of adult witnesses’ Ulatowska, J., Olszewska, J., & Hanson, M. D. (2016). Do format differ-
suggestibility across various types of leading questions. Applied ences in the presentation of information affect susceptibility to
Cognitive Psychology, 26(1), 48–53. https://doi.org/10.1002/acp.1793 memory distortions? The three-stage misinformation procedure
Strass v Police. (2013). SASC 3. reconsidered. American Journal of Psychology, 129(4), 407–417.
Sundar, S. S. (2008). The MAIN model: A heuristic approach to under- https://doi.org/10.5406/amerjpsyc.129.4.0407
standing technology effects on credibility. In M. J. Metzger, & A. J. Vosoughi, S., Roy, D., & Aral, S. (2018). The spread of true and false
Flanagin (Eds.), Digital media, youth, and credibility (pp. 72–100). news online. Science, 359(6380), 1146–1151. https://doi.org/10.
The MIT Press. 1126/science.aap9559
Sundar, S. S., Molina, M. D., & Cho, E. (2021). Seeing is believing: Is video Wang, S. W., & Scheinbaum, A. C. (2018). Enhancing brand credibility
modality more powerful in spreading fake news via online messa- via celebrity endorsement. Journal of Advertising Research, 58(1),
ging apps? Journal of Computer-Mediated Communication, 26(6), 16–32. https://doi.org/10.2501/JAR-2017-042
301–319. https://doi.org/10.1093/jcmc/zmab010 Weingardt, K. R., Leonesio, R. J., & Loftus, E. F. (1994). Viewing eyewit-
Sutherland, R., & Hayne, H. (2001). Age-related changes in the misin- ness research from a metacognitive perspective. In J. Metcalfe, & A.
formation effect. Journal of Experimental Child Psychology, 79(4), P. Shimamura (Eds.), Metacognition: Knowing about knowing (pp.
388–404. https://doi.org/10.1006/jecp.2000.2610 157–184). The MIT Press.
Takarangi, M. K., Parker, S., & Garry, M. (2006). Modernising the misinfor- Williams, K. D., Loftus, E. F., & Deffenbacher, K. A. (1992). Eyewitness
mation effect: The development of a new stimulus set. Applied evidence and testimony. In D. K. Kagehiro et al. (Ed.), Handbook
Cognitive Psychology, 20(5), 583–590. https://doi.org/10.1002/acp. of Psychology and Law (pp. 141–166). Springer.
1209 Zaragoza, M. S., & Lane, S. M. (1994). Source misattributions and the
Tomes, J. L., & Katz, A. N. (2000). Confidence–accuracy relations for real suggestibility of eyewitness memory. Journal of Experimental
and suggested events. Memory (Hove, England), 8(5), 273–283. Psychology: Learning, Memory, and Cognition, 20(4), 934. https://
https://doi.org/10.1080/09658210050117708 doi.org/10.1037/0278-7393.20.4.934

You might also like