Content uploaded by Petar Gabrić
Author content
All content in this area was uploaded by Petar Gabrić on Jun 08, 2021
Content may be subject to copyright.
1
Petar GABRIĆ 1, 2
Clinical Linguistics, Institute for German Linguistics, Philipps University of Marburg,
Pilgrimstein 16, 35032 Marburg, Germany
Department of Psychiatry and Psychotherapy, Philipps University of Marburg, Rudolf-Bultmann-
Straße 8, 35039 Marburg
FAKE NEWS, MEDIA MANIPULATION, AND HEALTH EFFECTS OF 5G: A SMALL-
SAMPLE DISCOURSE-ANALYTIC CASE STUDY OF THE CROATIAN NEWS WEBSITE
INDEX.HR
2
SUMMARY
This study investigated whether the Croatian news website Index.hr manipulates information on
the health effects of 5G. We constructed one experimental corpus, containing all articles by
Index.hr on health effects of 5G, and two control corpora, one with articles about health effects of
5G published by reliable media, and one with articles about science published by Index.hr.
Compared to Index.hr science articles, Index.hr 5G articles were 288.14 times likelier to express
the author’s opinion, 16.95 times likelier to express a subjective opinion, 10.78 times likelier to
contain no references, 10.78 times likelier to contain misinformation, and 4.20 times likelier to
contain no scientific references. The simultaneous increase in misinformation and reduction in
referencing suggests that the misinformation doesn’t stem from other sources, but that it is
produced within Index.hr. An increase in opinion expression, and opinion subjectivity in the
context of misinformation suggests that Index.hr is manipulating the information on health effects
of 5G. Furthermore, all articles were written by different authors, indicating that this phenomenon
is systematic within Index.hr. Still, the small sample size warrants a degree of caution.
Key words: fake news ; misinformation in media ; electromagnetic fields ; 5G ; Croatia
SAŽETAK
Ova je studija istražila manipulira li hrvatski informativni portal Index.hr informacijama o
učincima 5G-a na zdravlje. Izradili smo eksperimentalni korpus koji je sadržavao sve članke o
učincima 5G-a na zdravlje objavljene na portalu Index.hr, te dva kontrolna korpusa, jedan s
člancima o učincima 5G-a na zdravlje objavljenima u pouzdanim medijima, i jedan s člancima o
znanosti objavljenih na portalu Index.hr. U usporedbi s znanstvenim člancima portala Index.hr, za
3
članke portala Index.hr o učincima 5G-a na zdravlje bilo je 288.14 puta vjerojatnije da će izraziti
autorovo mišljenje, 16.95 vjerojatnije da će izraziti subjektivno mišljenje, 10.78 puta vjerojatnije
da neće sadržavati referencije, 10.78 puta vjerojatnije da će sadržavati dezinformacije te 4.20 puta
vjerojatnije da neće sadržavati znanstvene referencije. Istovremeni porast u broju članaka s
dezinformacijama i pad broja članaka s referencijama sugerira da dezinformacije ne proizlaze iz
drugih izvora informacija, već da nastaju unutar portala Index.hr. Porast u broju članaka s
izraženim mišljenjem kao i subjektivnim mišljenjem sugerira da Index.hr manipulira
informacijama o učincima 5G-a na zdravlje. Nadalje, sve su članke napisali različiti autori što
ukazuje na sustavnost ovog fenomena unutar portala Index.hr. Ipak, potrebna je pažnja pri
interpretaciji rezultata zbog male veličine uzorka.
Ključne riječi: lažne vijesti ; dezinformacije u medijima ; elektromagnetska polja ; 5G ; Hrvatska
1. INTRODUCTION
The expansion of wireless information and communication technologies using man-made
electromagnetic fields has exploded in the last couple of decades. Once limited to particular social
and/or geographical groups of people, these technologies have become practically omnipresent
and they have been systematically incorporated into everyday human functioning. Currently, the
new generation of wireless information transfer – the 5G – is expected to be globally introduced.
It is predicted that 5G will provide faster and more extensive data transmission through the use of
additional higher frequency bands (Simkó & Mattsson, 2019, p. 1). Thus, 5G has been welcomed
by a multitude of interest groups, however, many have expressed their concerns about its possible
4
adverse effects on human health. While some of these concerns have been a priori assumed
unconvincing and/or categorized as “conspiracy theories” in the media, some of them have been
raised by scientists and published in reliable sources.
There is a far-reaching history of research on the health effects of wireless radiation (Belpomme
Hardell, Belyaev, Burgio, & Carpenter, 2018; Deruelle, 2020; Desai, Kesari, & Agarwal, 2009; Di
Ciaula, 2018; Doyon & Johansson, 2017; Havas, 2017; Kaplan et al., 2016; Kostoff & Lau, 2013,
2017; Lerchl et al., 2015; Levitt & Lai, 2010; Miller et al., 2019; Pall, 2016, 2018; Panagopoulos,
2019; Panagopoulos, Johansson, & Carlo, 2015; Russell, 2018; Sage & Burgio, 2018; van Rongen
et al., 2009; Yakymenko et al., 2016). Kostoff, Heroux, Aschner, & Tsatsakis (2020) summarize
these findings reporting that exposure to radio frequency radiation below the American Federal
Communications Commission guidelines can result in the genesis of several types of cancer, DNA
and chromatin damage and/or dysfunction, mutagenesis, teratogenesis, neurological and
psychiatric disorders (including Alzheimer’s disease and autism), reproductive problems,
excessive reactive oxygen species/oxidative stress, inflammation, apoptosis, blood-brain barrier
disruption, pineal gland/melatonin production dysfunction, sleep disturbance, headache,
irritability, fatigue, concentration difficulties, depression, dizziness, tinnitus, burning and flushed
skin, digestive disturbance, tremor, cardiac irregularities, and general dysfunction of the neural,
circulatory, immune, endocrine, and skeletal systems.
Specific research on the health effects of 5G has been scant. Nevertheless, the majority of the
existing literature demonstrates that exposure to 5G, i.e. higher frequency bands, has biological
effects on humans, but the relationship between these biological effects and human health is still
unclear, mainly due to methodological limitations of the studies (see Simkó & Mattsson, 2019 and
papers cited therein). As an illustration of the difficulties in relating these biological effects to
5
eventual implications for human health, the following citation is provided: “A study that was
conducted on bacteria and fungi showed an increase in cell growth … The other in vitro study was
performed on fibroblasts (25 GHz, 0.80 mW/cm2, 20 min), with genotoxic effects observed at high
SAR levels (20 W/kg) …” (Simkó & Mattsson, 2019, p. 5). However, in a number of these studies
the biological effects appear to be more indicative of possible adverse health effects on humans
than not (e.g. eye and skin damage in laboratory subjects). Kostoff et al. (2020) further predict that
compared to technologies using lower frequency bands exposure to high-band wireless radiation
as used in 5G would lead to disproportionate increases in skin and eye diseases, and, likely, effects
on the nervous system, heart, and the immune system (cf. Mehdizadeh & Mortazavi, 2019).
Interestingly, a recent review on the effects of 5G found no relationships between the effects of
exposure, and intensity, exposure time, and frequency (Simkó & Mattsson, 2019, p. 16). Be as it
may, there is no scientific consensus on the health effects of 5G (i.e. higher frequency bands),
mainly because this topic remains understudied, the existing studies employ heterogeneous
methodologies, and the methodology in this field is severely limited. As discussed by Kostoff et
al. (2020), most studies have been conducted in laboratory settings implying predominant use of
non-human compared to human subjects, frequent omission of extremely low frequencies which
are regularly present in all telecommunication, as well as frequent use of only one toxic stimulus
as a stressor, whereas in real-life settings humans are exposed to numerous stressors which can
exacerbate the existing adverse effects of radiation. Simkó & Mattsson (2019, p. 16) also criticize
the quality of some of the research: “[T]oo few studies fulfill the minimal quality criteria to allow
any further conclusions.” Kostoff et al. (2020) conclude: “[A]lmost all of the wireless radiation
laboratory experiments that have been performed to date are flawed/limited with respect to
showing the full adverse impact of the wireless radiation that would be expected under real-life
6
conditions.” What is more, “studies have shown that industry-funded research of wireless radiation
adverse health effects is far more likely to show no effects than funding from non- industry sources
[Carpenter, 2019; Huss et al., 2007].” (Kostoff et al., 2020). It should, however, also be emphasized
that technological innovations of 5G are expected to bring benefits for the public health as well,
particularly in the domains of telemedicine and extremity rehabilitation (Li, 2019; Li & Wang,
2019).
Despite the lack of consensus, the use of 5G technology clearly poses a possible public health issue
(cf. Bircher & Kuruvilla, 2014; Mason, Lindberg, Read, & Borman, 2018). In this regard it has
been recognized that media and journalists play a decisive role in how public health issues are
perceived in the public, how the public will respond to the issue, and what the public knows about
the issue (Leask, Hooker, & King, 2010; Wallington, Blake, Taylor-Clark, & Viswanath, 2010).
Furthermore, media can act as catalysts to promote particular health practices in the public
(Institute of Medicine 2003). However, there appears to be an increasing amount of misinformation
on health topics in the media (Dhoju, Rony, Kabir, & Hassan, 2019; Scheufele & Krause, 2019),
rendering the relationship between the media and public health complicated. We are not aware of
any studies specifically investigating media misinformation on the health effects of 5G. On the
other hand, a number of studies investigating media-related phenomena used articles from
Index.hr, among other sources, in their studies (Brautović & Miloslavić, 2009; Britvić, Đurić, &
Bužić, 2014; Capurso, 2019; Čižmešija, Sorić, & Lolić, 2017; Granić, Mitrović, & Maragunić,
2011; Hazdovac Bajić, 2013; Jakopović & Mikelić Preradović, 2016; Mijatov & Radenović, 2019;
Tanta, Barić-Šelmić, & Levak, 2017). Conversely, we are aware of only one (unpublished) study
which focused specifically on the news website Index.hr. Ivica Jeđud presented his study at the 7th
Stulikon Student Linguistic Conference in May 2018 in Zagreb, Croatia in which he found that
7
Index.hr articles do not adhere to the suggested people-first language prescriptions when
describing people with mental disorders, and that when the title does contain people-first language,
the news is far more likely to be of negative than positive emotional valence (people-first language
frequently didn’t appear in the main text, even when the title contained such constructions; for the
conference abstract see Jeđud, 2018).
2. METHODOLOGY
2.1. Materials (corpora)
News articles for the experimental corpus were extracted from the Croatian news website Index.hr
(https://www.index.hr/). We extracted all news articles marked with “#5G” or “#5G mreža (‘5G
network’)” from the website. We selected only those articles which addressed health effects of 5G.
News articles written by the Croatian News Agency and published on Index.hr were not included
in the experimental corpus. The search generated nine news articles altogether. Two of the nine
articles were identical in content. Thus, one was excluded with the experimental corpus finally
consisting of eight news articles. All articles were written by different authors. The oldest article
was published on April 8, 2019 and the newest article was published on April 7, 2020.
We constructed two control corpora with articles which were published in approximately the same
time frame as their experimental counterparts. The first control corpus consisted of eight randomly
chosen news articles about the health effects of 5G published by reliable mainstream media.
Problematically, there exist no robust criteria for media reliability: “The matter of reliability is
subjective.” (Dhoju et al. 2019: 94). The reliability of the media in the present study was thus
8
determined by the author based on his experience with these media, and his expertise in linguistics
and communication. The media included news websites by BBC, Tagesschau, Spiegel Online,
Hrvatska radiotelevizija, Zeit Online, Frankfurter Allgemeine Zeitung, and Süddeutsche Zeitung.
All articles were published by different media. The oldest article was published on January 16,
2019 and the newest article on July 17, 2019. If there would be significant differences between the
Index.hr 5G, and reliable 5G corpora, this would suggest that Index.hr reports more unreliably on
the health effects of 5G than our corpus of reliable media articles, and would thus tentatively
indicate that Index.hr reports unreliably on the health effects of 5G.
The second control corpus consisted of eight randomly chosen news articles from Index.hr which
were about science, but not 5G. All articles in the second control corpus were written by different
authors and none of the authors was the author of any of the news articles from the experimental
corpus. The oldest article was published on June 4, 2019 and the newest on February 24, 2020. If
there would be significant differences between the Index.hr 5G, and Index.hr science corpora, this
would suggest that Index.hr has a different approach to 5G–health articles compared to other
articles on scientific topics, and would further possibly indicate that Index.hr manipulates
information on the health effects of 5G.
All news articles were assessed by the author on the presence of references, presence of scientific
references (clear information about and/or links to a scientific paper or a summary of a scientific
paper), presence of clear misinformation (in the present study: erroneous referencing, and denial
of the existence of scientific literature on a particular topic), expression of an opinion about the
issue, and expression of a subjective opinion about the issue. In the present study opinions were
identified as subjective if they were expressed with the use of subjective words, or if they were not
backed by any type of external information. Words used for opinion expression were categorized
9
as subjective if they were emotionally salient, or if they simultaneously had a weak ability in
delivering facts, and a high ability in expressing opinions (for the second criterion see Wang et al.,
2018).
We also assessed the number of hyperlinks in each article. Hyperlinks referencing to articles
published by the same medium were excluded from analyses. According to Dhoju et al. (2019:
94), articles published by reliable media contain more hyperlinks compared to articles published
by unreliable media. In their study reliable media had a median of eight hyperlinks per article,
while most articles published by unreliable media had less than two or no hyperlinks. Furthermore,
we assessed the number of visual media in each article. According to Dhoju et al. (2019: 95), no
differences in the number of visual media per article between reliable and unreliable news media
should be expected. In their study, articles from reliable media had a mean of 13.83 visual media,
while articles from unreliable media used 14.22 visual media on average.
2.2. Statistical analyses
Statistical analyses were conducted in JASP 0.11.1.0. The chi square test was used to compare
categorical variables between groups. We report the Likelihood ratio because of the small sample
size (< 30), along with Cramer’s V for effect sizes (ES). We also manually calculated the odds
ratio (note: not the relative risk, cf. Stare & Maucort-Boulch, 2016). If a contingency table
contained a count which was = 0, the Haldane-Anscombe correction was applied. We only report
odds ratios > 1, as proposed by Osborne (2006). The independent sample t-test, namely the Mann-
Whitney U and Welch’s tests were used for the analyses of group differences in the number of
hyperlinks and visual media. Rank-biserial correlation, and Cohen’s d were reported for ESs,
respectively. Normality of distribution was tested using the Shapiro-Wilk test of normality, and
the Levene’s test of equality of variances was used for the assumption of homogeneity of variance.
10
3. RESULTS
Group values of the categorical variables are shown in Table 1. References and scientific
references were detected in all corpora. Articles with the author’s opinion were found in the
Index.hr 5G and reliable 5G corpora, but not in the Index.hr Science corpus. Articles with
misinformation, and articles with a subjective opinion by the author were identified only in the
Index.hr 5G corpus. Regarding misinformation, two cases of erroneous referencing, and one case
of denial of the existence of scientific literature on the health effects of 5G were detected (Table
2). Subjective opinions are shown in Table 3. All subjective opinions favored the hypothesis that
5G had no adverse health effects. Subjective words in the expression of subjective opinions
included sulud ‘silly’, naravno ‘naturally, of course’, and lud ‘crazy’. Perhaps notably, the oldest
Index.hr 5G article was the only article in the Index.hr 5G corpus which contained scientific
references, and further contained no misinformation, nor a subjective opinion. The author’s
opinion was, nevertheless, expressed, conveying that science has thus far been unable to say
whether 5G has adverse effects on human health, and that it is necessary to continue exploring this
topic since 5G is putatively to unstoppably continue his expansion around the world. This oldest
article and the second oldest article were temporally divided by 103 days (April 8, 2019 compared
to July 20, 2019). The opinion in the second oldest article already denies without reservation that
5G has any possible adverse health effects, with the title reading: “Čuli ste za opasnost koja prijeti
od 5G mreža? To ne postoji” ‘Have you heard of the danger behind the 5G networks [sic]? That
[sic] doesn’t exist’. However, this article contains references, and contains neither misinformation
nor a subjective opinion. In all other Index.hr 5G articles an opinion is expressed which generally
11
states that 5G has no adverse effects on human health. One Index.hr 5G article which had no
references, no scientific references, and expressed a subjective opinion (but had no
misinformation) was written by a doctor of psychology. No other article displayed biographical
information on the author(s).
Table 1
Group values of the categorical variables
Group/Corpus
Articles with
references
Articles with
scientific
references
Articles with
misinformation
Articles with
the author’s
opinion
Articles with a
subjective
opinion by the
author
Index.hr 5G
5/8
1/8
3/8
8/8
4/8
Index.hr Science
8/8
3/8
0/8
0/8
0/8
Reliable 5G
7/8
6/8
0/8
3/8
0/8
Table 2
Cases of misinformation in the Index.hr 5G corpus
Subject/
Article no.
Original text in Croatian
Translation into English (by the author)
2
Naravno, ne postoje nikakvi znanstveni
dokazi da peta generacija mobilne
komunikacije, poznatija kao 5G, izaziva ili
pogoršava novu bolest COVID-19 koju
izaziva koronavirus SARS-CoV-2. Tom se
temom, među ostalima, pozabavio i
There are, naturally, no scientific evidence
that the fifth generation of mobile
communication, better known as 5G, causes
or worsens the new disease COVID-19 [sic]
which is caused by the coronavirus SARS-
CoV-2. This topic has been discussed,
12
nezavisna britanska stranica za provjeru
činjenica (fact check) Full Fact.
among others, by the independent British
fact-checking website Full Fact.*
5
Hrvati se bune protiv 5G mreže. Ona je
opasna po zdravlje kao sušena šunka
Croatians are protesting against the 5G
network. It is as adverse to our health as is
smoked ham**
6
Jasno, kako to često biva kod takvih stvari,
priče o štetnosti 5G mreže su najobičnija
nagađanja i nemaju uporišta u znanosti.
Of course, as it often is with these things,
the stories about the adverse effects of 5G
network are pure guessing, and they have
no foundation in science.
Notes. The first two cases of misinformation were cases of erroneous referencing, while the third case was
one of denial of the existence of scientific literature on health effects of 5G. * A closer inspection of the
referenced hyperlink to the Full Fact article reveals that it doesn’t fact-check whether 5G causes or worsens
[the symptoms of] COVID-19, but whether it has adverse effects on the immune system which would make
one more susceptible to viral, including coronaviral, infections. Additionally, there are at least several
studies suggesting the possibility that exposure to man-made electromagnetic fields might lead to
immunosuppression and thus increase the risk of opportunistic infections (see Doyon & Johansson 2017
and papers cited therein). ** This claim is based on the fact that the International Agency for Research on
Cancer classified radio frequency exposure as possibly cancerogenic (International Agency for Research
on Cancer 2013), and the “pseudofact” that smoked ham is also classified as possibly cancerogenic.
Interestingly, this pseudofact is not expressed explicitly in the main text, but is implicitly expressed two
times, once in the title, and once in a subheading. Firstly, smoked ham is not specifically identified as a
(possibly) cancerogenic agent by the IARC. Secondly, processed, and red meat are not in the same category
as radio frequency radiation. Thirdly, carcinogenesis is not the only possible adverse effect of radio
frequency radiation. Fourthly, it is not clear how the label “possibly cancerogenic” should lead to the
implication that radio frequency exposure has no adverse health effects.
Table 3
Cases of a subjective opinion by the author
Subject/
Article no.
Original text in Croatian
Translation into English (by the author)
1
Iako se suluda teorija zavjere protiv 5G
mreže proširila cijelim svijetom […].
Although this silly conspiracy theory
against the 5G network has dispersed all
over the world […].
2
Naravno, ne postoje nikakvi znanstveni
dokazi da peta generacija mobilne
There are, naturally, no scientific evidence
that the fifth generation of mobile
13
komunikacije, poznatija kao 5G, izaziva ili
pogoršava novu bolest COVID-19 […].
communication, better known as 5G, causes
or worsens the new disease COVID-19 [sic]
[…].
3
Subjective opinion not backed by any type of external information.
4
Internetom kruže lude teorije da je 5G
tehnologija uzrokovala koronavirus
Crazy theories about the 5G technology
causing the coronavirus [sic] are circling
the internet
Table 4
Descriptive data on hyperlinks and visual media
Group/Corpus
Hyperlinks
Visual media
Index.hr 5G
1.750 (3.495)
2.750 (2.493)
Index.hr Science
1.000 (0.535)
1.125 (0.354)
Reliable 5G
2.375 (2.446)
2.500 (2.000)
Notes. The mean values are reported. Standard deviations appear in parentheses.
Group values of hyperlink and visual media frequency are shown in Table 4. Hyperlinks and visual
media were detected in all corpora. Articles without hyperlinks were found in all corpora. All
articles in all corpora displayed at least one visual medium.
3.1. Comparisons between Index.hr 5G, and reliable 5G articles
Categorical variables – Results of the chi square test are shown in Table 5. There were no
significant differences in the number of articles with references between the Index.hr 5G and
reliable 5G corpora, with a small-to-moderate ES. Still, the odds ratio analysis showed that
Index.hr 5G articles were 4.20 times likelier to contain no references than reliable 5G articles.
There were significantly less articles which contained scientific references in the Index.hr 5G
14
corpus compared to the reliable 5G corpus, with a large ES. Additionally, Index.hr 5G articles
were 21.02 times likelier to contain no scientific references compared to reliable 5G articles. The
Index.hr 5G corpus also displayed significantly more articles which contained misinformation
compared to the reliable 5G corpus, with a moderate-to-large ES. Index.hr 5G articles were 10.78
time likelier to contain misinformation compared to reliable 5G articles. The Index.hr 5G corpus
also contained significantly more articles in which the author gave his opinion on the issue, with a
large ES. Index.hr 5G articles were 26.73 times likelier to express the author’s opinion on the issue
compared to reliable 5G articles. Furthermore, the Index.hr 5G corpus contained significantly more
articles in which the author’s subjective opinion was expressed compared to the reliable 5G corpus,
with a large ES. Index.hr 5G articles were 16.95 more likely to include a subjective opinion on the
issue compared to reliable 5G articles.
Hyperlinks – Because normality of distribution was violated in the Index.hr 5G group (p < .001),
and the assumption of homogeneity of variance was not violated (p = .787), we compared the two
values using the Mann-Whitney U test. Although visual examination of the descriptive data
suggested that reliable 5G articles contain slightly more hyperlinks than Index.hr 5G articles, the
test failed to reject the null hypothesis (U = 40.500, p = .369), showing a small ES (r = 0.266).
Visual media – Normality of distribution was violated in both the Index.hr (p = .012), and the
reliable 5G groups (p = .021). The assumption of homogeneity of variance was not violated (p =
.612). Mann-Whitney U test revealed no differences between groups in the number of visual media
per article (U = 31.500, p = 1), with a trivial ES (r = 0.016).
Table 5
Results of the chi square tests comparing Index.hr 5G and reliable 5G articles
15
Variable
Chi square
Cramer’s V
References
X2 = 1.381, p = .240
V = 0.289
Scientific references*
X2 = 6.904, p = .009
V = 0.630
Misinformation*
X2 = 4.857, p = .028
V = 0.480
Author’s opinion*
X2 = 9.290, p = .002
V = 0.674
Subjective opinion*
X2 = 6.904, p = .009
V = 0.577
Notes. df = 1, N = 16, * p ≤ .05.
3.2. Comparisons between Index.hr 5G, and Index.hr science articles
Categorical variables – Results of the chi square test are shown in Table 6. There were
significantly less articles with references in the Index.hr 5G corpus compared to the Index.hr
science corpus, with a moderate ES. The odds ratio analysis showed that Index.hr 5G articles were
10.78 times likelier to contain no references than Index.hr science articles. There were no
differences between the two corpora in the number of articles with scientific references, with a
small ES. Still, Index.hr 5G articles were 4.20 times likelier to contain no scientific references
compared to Index.hr science articles. The Index.hr 5G corpus also displayed significantly more
articles which contained misinformation compared to the Index.hr science corpus, with a moderate
ES. Index.hr 5G articles were 10.78 times likelier to contain misinformation compared to Index.hr
science articles. The Index.hr 5G corpus also contained significantly more articles in which the
author gave his opinion on the issue compared to the Index.hr science corpus, with a large ES.
Index.hr 5G articles were 288.14 times likelier to express the author’s opinion on the issue
compared to Index.hr science articles. Furthermore, the Index.hr 5G corpus contained significantly
more articles in which the author’s subjective opinion was expressed compared to the Index.hr
16
science corpus, with a large ES. Index.hr 5G articles were 16.95 more likely to include a subjective
opinion on the issue compared to Index.hr science articles.
Table 6
Results of the chi square tests comparing Index.hr 5G and Index.hr science articles
Variable
Chi square
Cramer’s V
References*
X2 = 4.857, p = .028
V = 0.480
Scientific references
X2 = 1.381, p = .240
V = 0.289
Misinformation*
X2 = 4.857, p = .028
V = 0.480
Author’s opinion***
X2 = 9.290, p < .001
V = 1.000
Subjective opinion**
X2 = 6.904, p = .009
V = 0.577
Notes. df = 1, N = 16, * p ≤ .05, ** p ≤ .01, *** p ≤ .001.
Hyperlinks – Normality of distribution was violated in both the Index.hr 5G (p < .001), and the
Index.hr science corpora (p = .005). Assumption of homogeneity of variance was violated as well
(p = .028). Thus, we compared the two values using the Welch’s test. Although visual examination
of the descriptive data suggested that Index.hr 5G articles contained slightly more hyperlinks, the
test demonstrated no apparent differences between the two corpora [t (7.327) = 0.600, p = .567],
with a small ES (d = 0.300).
Visual media – Normality of distribution was violated in both the Index.hr 5G (p = .012), and the
Index.hr science corpora (p < .001). The assumption of homogeneity of variance was also violated
(p = .004). Welch’s test revealed no significant differences between groups in the number of visual
media per article [t (7.281) = 1.825, p = .109), but with a large ES (d = 0.913). Furthermore, the t-
17
value was slightly higher when comparing the two corpora in the number of visual media per
article, than when comparing them in the number of hyperlinks per article.
4. DISCUSSION
Comparisons of news articles about health effects of 5G published by Index.hr, and news articles
about health effects of 5G published by reliable mainstream media revealed that the Index.hr 5G
articles compared to articles in the reliable 5G corpus were 26.73 times likelier to express the
author’s opinion, 21.02 times likelier to have no scientific references, 16.95 times likelier to
express a subjective opinion, 10.78 times likelier to contain misinformation, and 4.20 times likelier
to contain no references. Only the data concerning the presence of references yielded relatively
unclear results. The converging data nevertheless suggest that Index.hr is a less reliable source of
information on the health effects of 5G compared to the present corpus of reliable media articles.
If we would assume that our sample of reliable articles were a representative sample, we would be
in a position to say that Index.hr reports rather unreliably than reliably on the unreliable–reliable
continuum.
Comparisons of news articles about health effects of 5G published by Index.hr, and news articles
about science (but not 5G) published by Index.hr revealed that the Index.hr 5G articles compared
to Index.hr science articles were 288.14 times likelier to express the author’s opinion, 16.95 times
likelier to express a subjective opinion, 10.78 times likelier to contain no references, 10.78 times
likelier to contain misinformation, and 4.20 times likelier to contain no scientific references. Here,
only the data concerning the presence of scientific references yielded relatively unclear results.
18
Problematically, small sample sizes are associated with both false negatives and false positives
(Oakes, 2017; Vadillo, Konstantinidis, & Shanks, 2016), while also being associated with
overestimated odds ratios (Nemes, Miao Jonasson, Genell, & Steineck, 2009). However, we
extracted all articles on health effects of 5G from Index.hr, making this issue insurmountable at
the moment. Furthermore, the analyses were conducted without excluding the oldest article in the
Index.hr 5G corpus, which displayed notable qualitative differences (presence of scientific
references, and expression of an objective, non-polarized opinion) compared to other articles in
the corpus. Possibly, this oldest article might be an indicator that a phenomenon occurred within
Index.hr between April 8, 2019 and July 20, 2019, which was followed by a presumed switch in
the approach to reporting on health effects of 5G.
Nonetheless, the observed differences between the two Index.hr corpora demonstrate that there are
both quantitative and qualitative differences within Index.hr in the production of news articles on
health effects of 5G compared to the production of other scientific news articles. Compared to the
production of general news articles on science, the production of news articles on health effects of
5G is thus characterized by a reduction in referencing, and an increase in misinformation, opinion
expression, and opinion subjectivity. The simultaneous increase in misinformation and reduction
in referencing, as well as the quality of the observed pieces of misinformation suggest that
misinformation doesn’t stem from other unreliable sources of information, but that the
misinformation is produced within Index.hr. An increase in opinion expression, and opinion
subjectivity in the context of misinformation suggests that Index.hr is manipulating the information
on health effects of 5G. This is corroborated by the fact that the two types of misinformation
identified in the present study included erroneous referencing, and denial of the existence of
scientific literature on the topic. Furthermore, all articles on both 5G, and scientific topics were
19
written by different authors, indicating that this phenomenon is systematic within Index.hr.
Additionally, the fact that there were significant differences in the presence of scientific references
between the Index.hr 5G, and reliable 5G corpora, but not between the Index.hr 5G, and the
Index.hr science corpora, possibly suggests that Index.hr relies less on scientific references in
general. Although the present data point to manipulation of information on the health effects of
5G, caution should be warranted due to the small sample size.
We found no significant differences in the raw number of hyperlinks, and visual media between
both pairs of corpora. However, the ES for the difference in the number of visual media between
the Index.hr 5G, and Index.hr science corpora was large, suggesting that the sample size possibly
affected the results of the statistical analyses. It is difficult to interpret this result at the moment.
Our mean raw values of hyperlinks and visual media greatly differed from Dhoju et al. (2019).
They reported a median of eight hyperlinks for reliable media, while the reliable 5G corpus in the
present study had a mean of 2.375 hyperlinks (and a median of 2; not reported in the results
section). Furthermore, articles across corpora in our study contained on average between 1.125
and 2.750 visual media, while in the mentioned study articles from both reliable, and unreliable
media contained on average around 14 visual media. We suggest that the raw number of hyperlinks
may not be a valid measure of media reliability. Additionally, our results are in line with Dhoju et
al. (2019) who found no differences between reliable, and unreliable media in the number of visual
media per article.
5. LIMITATIONS
20
Firstly, the sample size is small, as discussed in different parts of the paper. Secondly, the reliability
of media for the reliable 5G corpus was determined introspectively by the author. Finally, this
study used a very narrow methodological frame. Nevertheless, we find it, in this case at least,
effective.
6. CONCLUSION
The present case study investigated whether the Croatian news website Index.hr manipulates
information on the health effects of 5G. We constructed one experimental corpus, containing all
articles by Index.hr on health effects of 5G, and two control corpora, one with articles about health
effects of 5G published by reliable media, and one with articles about science (but not 5G)
published by Index.hr. We assessed the presence of references, scientific references,
misinformation, opinion expression, and opinion subjectivity. Compared to Index.hr science
articles, Index.hr 5G articles were 10.78 times likelier to contain no references, 4.20 times likelier
to contain no scientific references, 10.78 times likelier to contain misinformation, 288.14 times
likelier to express the author’s opinion, and 16.95 times likelier to express a subjective opinion.
The simultaneous increase in misinformation and reduction in referencing, as well as the quality
of the observed pieces of misinformation suggest that misinformation doesn’t stem from other
unreliable sources of information, but that the misinformation is produced within Index.hr. An
increase in opinion expression, and opinion subjectivity in the context of misinformation suggests
that Index.hr is manipulating the information on health effects of 5G. This is corroborated by the
fact that the two types of misinformation identified in the present study included erroneous
referencing, and denial of the existence of scientific literature on the topic. Furthermore, all articles
on both 5G, and scientific topics were written by different authors, indicating that this phenomenon
21
is systematic within Index.hr. We conclude that our data point to manipulation of information on
health effects of 5G by Index.hr. Still, the small sample size warrants a degree of caution.
References
Belpomme, D., Hardell, L., Belyaev, I., Burgio, E., Carpenter, D. O. (2018). Thermal and non-
thermal health effects of low intensity non-ionizing radiation: an international perspective.
Environmental Pollution, 242, 643–58. doi:10.1016/j.envpol.2018.07.019
Bircher, J., Kuruvilla, S. (2014). Defining health by addressing individual, social, and
environmental determinants: new opportunities for health care and public health. Journal of Public
Health Policy, 35(3), 363–86. doi:10.1057/jphp.2014.19
Brautović, M., Miloslavić, I. (2009). Hrvatski online mediji kao sekundarni izvori u novinarstvu.
MediAnali: međunarodni znanstveni časopis za pitanja medija, novinarstva, masovnog
komuniciranja i odnosa s javnostima, 3(6), 89–102. Available at https://hrcak.srce.hr/44534
Britvić, I., Đurić, J., Bužić, D. (2014). Comparative analysis of Google and Bing SEO on leading
Croatian news portals. 37th International Convention on Information and Communication
Technology, Electronics and Microelectronics (MIPRO), Opatija, 474–478.
doi:10.1109/MIPRO.2014.6859614
Capurso, S. (2019). A content analysis of the most popular online news portals in Croatia regarding
child abuse. Psychotherapy in Achieving Health and Well-Being for Children and Young People,
2, 5–16.
22
Carpenter, D. O. (2019). Extremely low frequency electromagnetic fields and cancer: how source
of funding affects results. Environmental Research, 178, 108688.
doi:10.1016/j.envres.2019.108688.
Čižmešija, M., Sorić, P., Lolić, I. (2017). The interrelationship between media reports and the
recession in Croatia. Croatian Review of Economic, Business and Social Statistics, 3(1), 16–34.
doi:10.1515/crebss-2017-0002
Deruelle, F. (2020). The different sources of electromagnetic fields: dangers are not limited to
physical health. Electromagnetic Biology and Medicine, Latest Articles.
doi:10.1080/15368378.2020.1737811
Desai, N. R., Kesari, K. K., Agarwal, A. (2009). Pathophysiology of cell phone radiation: oxidative
stress and carcinogenesis with focus on male reproductive system. Reproductive Biology and
Endocrinology, 7, 114. doi:10.1186/1477-7827-7-114
Dhoju, S., Rony, M. M. U., Kabir, M. A., Hassan, N. (2019). A large-scale analysis of health
journalism by reliable and unreliable media. Studies in Health Technology and Informatics, 264,
93–7. doi:10.3233/SHTI190190
Di Ciaula, A. (2018). Towards 5G communication systems: are there health implications?.
International Journal of Hygiene and Environmental Health, 221(3), 367–75.
doi:10.1016/j.ijheh.2018.01.011
Doyon, P. R., Johansson O. (2017). Electromagnetic fields may act via calcineurin inhibition to
suppress immunity, thereby increasing risk for opportunistic infection: conceivable mechanisms
of action. Medical Hypotheses, 106, 71–87. doi:10.1016/j.mehy.2017.06.028
23
Granić, A., Mitrović, I., Maragunić, N. (2011). Exploring the usability of web portals: a Croatian
case study. International Journal of Information Management, 31, 339–349.
doi:10.1016/j.ijinfomgt.2010.11.001
Havas, M. (2017). When theory and observation collide: can non-ionizing radiation cause cancer?.
Environmental Pollution, 221, 501–5. doi:10.1016/j.envpol.2016.10.018
Hazdovac Bajić, N. (2013). Prezentacija religijskoga sadržaja na najposjećenijim hrvatskim
mrežnim portalima. Obnovljeni život: časopis za filozofiju i religijske znanosti, 68(4), 473–482.
Available at https://hrcak.srce.hr/113701
Huss, A., Egger, M., Hug, K., Huwiler-Müntener, K., Röösli, M. (2007). Source of funding and
results of studies of health effects of mobile phone use: systematic review of experimental studies.
Environmental Health Perspectives, 115(1), 1–4. doi:10.1289/ehp.9149
Institute of Medicine (2003). The Future of the Public's Health in the 21st Century. Washington,
DC: The National Academies Press.
International Agency for Research on Cancer (2013). Non-Ionizing Radiation. Part. 2:
Radiofrequency Electromagnetic Fields. Lyon, France: International Agency for Research on
Cancer.
Jakopović, H., Mikelić Preradović, N. (2016). Identifikacija online imidža organizacija temeljem
analize sentimenta korisnički generiranog sadržaja na hrvatskim portalima. Medijska istraživanja,
22(2), 63–82. doi:10.22572/mi.22.2.4
24
Jeđud, I. (2018, May). People-first language in use: case of Index.hr and mental health reporting.
In: Stulikon – studentska lingvistička konferencija. 10.–12. svibnja 2018., Filozofski fakultet u
Zagrebu. Paper presented at 7th Stulikon Student Linguistic Conference, Zagreb (14–14).
Kaplan, S., Deniz, O. G., Onger, M. E., Turkmen, A. P., Yurt, K. K., Aydin, I., … Davis, D. (2017).
Electromagnetic field and brain development. Journal of Chemical Neuroanatomy, 75, 52–61.
doi:10.1016/j.jchemneu.2015.11.005
Kostoff, R. N., Heroux, P., Aschner, M., Tsatsakis, A. (2020). Adverse health effects of 5G mobile
networking technology under real-life conditions. Toxicology Letters, 323, 35–40.
doi:10.1016/j.toxlet.2020.01.020
Kostoff R. N., Lau C. G. Y. (2013). Combined biological and health effects of electromagnetic
fields and other agents in the published literature. Technological Forecasting & Social Change,
80(7), 1331–49. doi:10.1016/j.techfore.2012.12.006
Kostoff, R. N., Lau, C. G. Y. (2017). Modified health effects of non-ionizing electromagnetic
radiation combined with other agents reported in the biomedical literature. In: Geddes, C. D (ed.).
Microwave Effects on DNA and Proteins. Cham: Springer International Publishing AG, 97–157.
doi:10.1007/978-3-319-50289-2_4
Leask, J., Hooker, C., King, C. (2010). Media coverage of health issues and how to work more
effectively with journalists: a qualitative study. BMC Public Health, 10, 535. doi:10.1186/1471-
2458-10-535
Lerchl, A., Klose, M., Grote, K., Wilhelm, A. F. X., Spathmann, O., Fiedler, T., ... Clemens, M.
(2015). Tumor promotion by exposure to radiofrequency electromagnetic fields below exposure
25
limits for humans. Biochemical and Biophysical Research Communications, 459(4), 585–90.
doi:10.1016/j.bbrc.2015.02.151
Levitt, B. B., Lai, H. (2010). Biological effects from exposure to electromagnetic radiation emitted
by cell tower base stations and other antenna arrays. Environmental Reviews, 18, 369–95.
doi:10.1139/A10-018
Li, D. (2019). 5G and intelligence medicine – how the next generation of wireless technology will
reconstruct healthcare?. Precision Clinical Medicine, 2(4), 205–208. doi:10.1093/pcmedi/pbz020
Li, E.-L., Wang, W.-J. (2019). 5G will drive the development of health care. Chinese Medical
Journal, 132(23), 2895–6. doi:10.1097/CM9.0000000000000534
Mason, K., Lindberg, K., Read, D., Borman, B. (2018). The importance of using public health
impact criteria to develop environmental health indicators: the example of the indoor environment
in New Zealand. International Journal of Environmental Research and Public Health, 15, 1786.
doi:10.3390/ijerph15081786
Mehdizadeh, A. R., Mortazavi, S. M. J. (2019). 5G technology: why should we expect a shift from
RF-induced brain cancers to skin cancers?. Journal of Biomedical Physics and Engineering, 9(5),
505–6. doi:10.31661/jbpe.v0i0.1225
Mijatov, N., Radenović, S. (2019). (Mis)use of football: analysis of media reports about matches
between Serbia and Croatia in 2013. SPORT: Science & Practice, 9(2), 23–32.
Miller, A. B., Sears, M. E., Morgan, L. L., Davis, D. L., Hardell, L., Oremus, M., Soskolne, C. L.
(2019). Risks to health and wellbeing from radio-frequency radiation emitted by cell phones and
other wireless devices. Frontiers in Public Health, 7, 223. doi:10.3389/fpubh.2019.00223
26
Nemes, S., Miao Jonasson, J., Genell, A., Steineck, G. (2009). Bias in odds ratios by logistic
regression modelling and sample size. BMC Medical Research Methodology, 9, 56.
doi:10.1186/1471-2288-9-56
Oakes, L. M. (2017). Sample size, statistical power, and false conclusions in infant looking-time
research. Infancy, 22(4), 436–469. doi:10.1111/infa.12186
Osborne, J. W. (2006). Bringing balance and technical accuracy to reporting odds ratios and the
results of logistic regression analyses. Practical Assessment, Research & Evaluation, 11, 7.
doi:10.7275/z7we-dg46
Pall, M. L. (2016). Microwave frequency electromagnetic fields (EMFs) produce widespread
neuropsychiatric effects including depression. Journal of Chemical Neuroanatomy, 75, 43–51.
doi:10.1016/j.jchemneu.2015.08.001
Pall, M. L. (2018). Wi-Fi is an important threat to human health. Environmental Research, 164,
405–16. doi:10.1016/j.envres.2018.01.035
Panagopoulos, D. J. (2019). Comparing DNA damage induced by mobile telephony and other
types of man-made electromagnetic fields. Mutation Research: Reviews in Mutation Research.
781, 53–62. doi:10.1016/j.mrrev.2019.03.003
Panagopoulos, D. J., Johansson, O., Carlo, G. L. (2015). Real versus simulated mobile phone
exposures in experimental studies. Biomed Research International, 2015, 607053.
doi:10.1155/2015/607053
27
Russell, C. L. (2018). 5 G wireless telecommunications expansion: public health and
environmental implications. Environmental Research, 165, 484–95.
doi:10.1016/j.envres.2018.01.016
Sage, C., Burgio, E. (2018). Electromagnetic fields, pulsed radiofrequency radiation, and
epigenetics: how wireless technologies may affect childhood development. Child Development,
89(1), 129–36. doi:10.1111/cdev.12824
Scheufele, D. A., Krause, N. M. (2019). Science audiences, misinformation, and fake news.
Proceedings of the National Academy of Sciences of the United States of America, 116(16), 7662–
7. doi:10.1073/pnas.1805871115
Simkó, M., Mattsson, M.-O. (2019). 5G wireless communication and health effects – a pragmatic
review based on available studies regarding 6 to 100 GHz. International Journal of Environmental
Research and Public Health, 16, 3406. doi:10.3390/ijerph16183406
Stare, J., Maucort-Boulch, D. (2016). Odds ratio, hazard ratio and relative risk. Metodološki zvezki,
13(1), 59–67.
Tanta, I., Barić-Šelmić, S., Levak, T. (2017). The level of credibility of traditional and new media
in reporting on human rights issues. Collegium Antropologicum, 3, 215–229. Available at
https://hrcak.srce.hr/200344
Vadillo, M. A., Konstantinidis, E., Shanks, D. R. (2016). Underpowered samples, false negatives,
and unconscious learning. Psychonomic Bulletin Review, 23, 87–102.
van Rongen, E., Croft, R., Juutilainen, J., Lagroye, I., Miyakoshi, J., Saunders, R., … Xu, Z.
(2009). Effects of radiofrequency electromagnetic fields on the human nervous system. Journal of
28
Toxicology and Environmental Health. Part B: Critical Reviews, 12(8), 572–97.
doi:10.1080/10937400903458940
Wallington, S. F., Blake, K. D., Taylor-Clark, K., Viswanath, K. (2010). Challenges in covering
health disparities in local news media: an exploratory analysis assessing views of journalists.
Journal of Community Health, 35, 487–494. doi:10.1007/s10900-009-9217-x
Wang, H., Wu, F., Lu, W., Yang, Y., Li, X., Li, X., & Zhuang, Y. (2018). Identifying objective
and subjective words via topic modeling. IEEE Transactions on Neural Networks and Learning
Systems, 29(3), 718–730. doi:10.1109/TNNLS.2016.2626379
Yakymenko, I., Tsybulin, O., Sidorik, E., Henshel, D., Kyrylenko, O., Kyrylenko, S. (2016).
Oxidative mechanisms of biological activity of low-intensity radiofrequency radiation.
Electromagnetic Biology and Medicine, 35(2), 186–202. doi:10.3109/15368378.2015.1043557
