Page 1 of 13
Advances in Social Sciences Research Journal – Vol. 8, No. 5
Publication Date: May 25, 2021
DOI:10.14738/assrj.85.10266.
Mulligan, B. P., & Koren, S. A. (2021). Geopsychology of instrumental aggression: daily concurrence of global terrorism and solar- geomagnetic activity (1970-2018). Advances in Social Sciences Research Journal, 8(5). 487-499.
Services for Science and Education – United Kingdom
Geopsychology of instrumental aggression: daily concurrence of
global terrorism and solar-geomagnetic activity (1970-2018)
Bryce P. Mulligan
Department of Psychology, The Ottawa Hospital, Ottawa, Canada;
Stanley A. Koren
Neuroscience Research Group & Psychology Department
Laurentian University, Sudbury, Canada
ABSTRACT
Formal scientific study of the geopsychology of human aggression dates back at
least a century and has consistently demonstrated a positive association between
solar-geomagnetic activity and aggressive behaviour. Advances in the theories,
methodologies, and practical applications of geopsychology could therefore
contribute to collective efforts to comprehend, to forecast, and to develop
interventions for aggressive behaviours such as those seen in terrorism. This
requires a rigorous and precise estimate of the magnitude of association between
solar-geomagnetic activity and aggression using a representative, contemporary
sample of strictly-operationalized behaviour. Here we show that days in recent
history (1970-2018) with the lowest levels of instrumental human aggression
(number of casualty-associated terrorism incidents) also had the lowest levels of
solar and geomagnetic activity, and that stepwise increases in human aggression
were mirrored by progressive increases in solar activity. We used Bayesian
methods robust to outliers and heterogeneity of variance to analyze the most
comprehensive and contemporary global database of terrorism incidents available,
which included more than 106,000 unique instances of instrumental aggression
spanning 48 years. We conclude that there is a small, nonzero promotional effect of
solar-geomagnetic activity on terrorism-related aggression. This may reflect the
fact that solar-geomagnetic activity serves as a zeitgeber that coordinates the
expression of instrumental aggression across an aggregation of susceptible
individuals. We propose that many behaviours – even instrumental acts such as
terrorism which are presumed to involve a degree of planning and intention – may
be subject to subtle geopsychological induction or suppression.
Keywords: biophysics; geophysics; behavioural ecology; aggregate human behaviour;
Bayesian data analysis.
A BRIEF HISTORY OF THE GEOPSYCHOLOGY OF AGGRESSION
Geopsychology, or geo-neuroscience, denotes the manifold impacts of the Earth's physical
environment on biology and behaviour [1,2]. Interest in the geopsychology of aggression
originates prior to recorded history, with indigenous peoples inhabiting northern regions of
the globe. Displays of aurora borealis – visible light phenomena reflecting transient
disturbances in the geomagnetic field [3] often linked to variations in solar activity [4] – were
Page 2 of 13
488
Advances in Social Sciences Research Journal (ASSRJ) Vol. 8, Issue 5, May-2021
Services for Science and Education – United Kingdom
in some cases likened to ancient warriors or considered omens foretelling of war, famine, and
disease [5].
In 1924, A.L. Chizhevski first published (in Russian) the seminal scientific work on the
geopsychology of aggression. His statistical analyses of historical data from 500 BCE to 1914
CE demonstrated that the most acute/severe human conflicts and violent revolutions occurred
around maxima of 11-year solar activity (sunspot) cycles and that cultural development tended
to flourish at solar minima [6]. In 1928, Chizhevski described a comparable association
between Russian terrorism and solar activity for the period 1902 to 1911 [7]. Later (1999), M.A.
Persinger demonstrated that higher yearly levels of solar-geomagnetic activity similarly
predicted higher levels of armed human conflict for the years 1904 to 1950 [8]. Then, in 2015,
Vares and Persinger showed that daily solar activity predicted 4% to 10% of the variance in the
daily number of force and confrontation events occurring in the years 2009 to 2013 [9].
While both innovative and informative, existing naturalistic geopsychology of aggression
studies are limited by sample size/representativeness, operational definitions of aggression,
and data analysis methodology. For one, some prior studies have tended to examine relatively
small samples of data and/or datasets with low time-resolution (e.g. yearly data). Considering
that solar and geomagnetic activity are known to show substantial variation at a period of 11
years (as well as shorter- and longer-period cycles of varying magnitudes), geopsychological
studies should at minimum include data spanning more than one 11-year cycle. Furthermore,
larger datasets and those with higher time-resolution provide the high precision necessary to
detect and generate meaningful statistical estimates of geopsychological effect sizes, which
tend to be small (i.e. ~0.1-0.2 SD group differences or ~5-15% shared variance) [2, 7]. Second,
existing studies have tended to employ relatively broad or non-specific operational definitions
of aggressive behaviour. However, reactive/impulsive and instrumental subtypes of aggressive
behaviour involve differences in the nature of the aggressive behaviour as well as its social
context and neuroanatomical underpinnings [10, 11]. Because geopsychological effects may
operate via multiple distinct and behaviour-specific geo-neurobiological mechanisms [1, 2, 12,
13], effects could be diluted by tallying multiple distinct subtypes of aggressive behaviour
under a common metric. Finally, previous studies of the geopsychology of aggression have
tended to rely on methods of data analysis that are not robust to outliers or heterogeneity of
variance across groups or factor levels, which could introduce a source of confound into effect
size estimates.
AIMS AND PURPOSES OF THE PRESENT STUDY
To address these limitations and contribute to advances in the geopsychology of aggression, we
used contemporary datasets and analysis methods to produce representative, rigorous, and
precise geopsychological effect size estimates (e.g. to inform policy development and the design
and interpretation of experimental studies). As such, we used the most complete and
representative terrorism dataset available (Global Terrorism Database, GTD) [14], reflecting
the global daily occurrence of a narrowly-defined subtype of instrumental aggressive behaviour
and spanning 48 years of recent history (1970-2018). We also employed Bayesian methods of
data analysis that provided informative detail about the range of credible parameter values (i.e.
geopsychological effect sizes) and that accommodated for outliers and for heterogeneity of
variance across groups/levels [15].
Page 3 of 13
489
Mulligan, B. P., & Koren, S. A. (2021). Geopsychology of instrumental aggression: daily concurrence of global terrorism and solar-geomagnetic
activity (1970-2018). Advances in Social Sciences Research Journal, 8(5). 487-499.
URL: http://dx.doi.org/10.14738/assrj.85.10266
In addition to the theoretical and methodological aims just outlined, we chose to study the
geopsychology of terrorism for humanistic reasons. Terrorism has far-reaching and long- lasting detrimental physical, social, psychological, and economical impacts [16-18]. We thus
aspired to re-examine and to enhance the current understanding and conceptualization of the
phenomenology of terrorism-related aggression in order to better anticipate and disrupt
intended acts of violent coercion before they occur; we aimed to facilitate development of just
economic and social policy to redress modifiable biopsychosocial risk factors underlying
terrorism [19, 20].
STUDY METHODS
Instrumental Aggression (Terrorism) Data
The Global Terrorism Database (GTD) [14] is in ongoing, active development; is freely available
for non-commercial use; currently includes information on more than 190,000 terrorism
incidents; and presently spans complete years 1970 to 2018 (all data from 1993 were
irretrievably lost). There are 4 data collection epochs in the GTD wherein distinct
methodologies were employed: 1 January 1970 to 31 December 1997; 1 January 1998 to 31
March 2008; 1 April 2008 to 31 December 2011; and 1 January 2012 to present. As such, the
GTD project leaders recommend treating each epoch distinctly when conducting statistical
analysis [21].
After retrieving the GTD, any terrorism incident with an unknown/missing value for date,
month, year, or casualties (number of wounded or killed persons) was omitted. For the
purposes of the present study, all GTD terrorism incidents associated with at least one casualty
(≥1 wounded or killed persons) were considered casualty-incidents. The number of such
casualty-incidents was then counted within each day, where days with no terrorism casualty- incidents were assigned a score of zero. The total number of distinct terrorism casualty- incidents occurring on each day from 1970 to 2018 served as our daily metric of global
instrumental human aggression.
There was a disproportionate number of days with zero terrorism casualty-incidents, giving
the distribution of daily casualty-incident counts a pronounced positive skew. Casualty- incident counts were hence indexed to quintiles within each study epoch using the quantcut
function [22]. In other words, each day was assigned to a subgroup (nominal factor level)
corresponding to its epoch-wise instrumental aggression (terrorism casualty-incident)
quintile.
Solar-Geophysical Data
Solar-geophysical data for years 1970 to 2018 (spanning solar cycles 20 through 24) were
obtained from GFZ German Research Centre for Geosciences (www.gfz-potsdam.de/en/kp- index/) and then matched by date and merged with the terrorism casualty-incident dataset.
Daily (sum) KP index (a measure of global geomagnetic activity) [23] and daily 10.7 cm radio
flux (a measure of solar activity) [24] were each separately regressed on case (day) number
and the standardized residuals were saved for use as predicted variables in our analyses. (This
was done in order to unit-normalize and linearly detrend the solar and geomagnetic data series
prior to analysis.)
Page 4 of 13
490
Advances in Social Sciences Research Journal (ASSRJ) Vol. 8, Issue 5, May-2021
Services for Science and Education – United Kingdom
Statistical Analysis
Bayesian MCMC (Markov Chain Monte Carlo) simulation was used to estimate a credible range
of solar and geomagnetic activity values (metric-scale predicted variables) for days grouped
according to their terrorism casualty-incident quintile (a nominal factor with 5 levels). To
account for differing methodologies and sample sizes across the 4 GTD study epochs, epoch was
included as a second nominal factor (with 4 levels) in the models. Distinct means and standard
deviations were thus estimated for each casualty-incident quintile within each epoch (i.e. there
was no assumption of homogeneity of variance in solar/geomagnetic data across epochs or
aggression factor levels). Noise in the data was represented with a t-distribution that included
a normality parameter to prevent bias of geopsychological effect estimates as a result of
extreme values [15].
Prior distributions for all parameters were selected to be broad on the scale of the data and to
exert minimal influence over the posterior distributions. Broad normal priors were used for
baseline solar/geomagnetic estimates and for estimates of deflections from baseline associated
with each casualty-incident factor level. Standard deviation parameters were given broad
gamma prior distributions and the prior for the normality parameter was a broad exponential
distribution. Our main interest lie in examining and reporting the magnitude of differences in
solar/geomagnetic activity between instrumental aggression factor levels. This was estimated
by examining factor-level contrasts (relative differences) in the posterior distribution of
estimated solar-geomagnetic activity parameter values (i.e. using a conservative approach that
compared, in turn, values from each individual casualty-incident quintile to those from all other
quintiles combined).
Data analyses were conducted in R [25] with JAGS [26] and runjags [27] for Bayesian MCMC
simulation. We employed functions and a modified script for analysis of metric predicted
variables with two nominal factors made available and extensively described by Kruschke [15].
All of the MCMC chains were checked for convergence. Effective sample size (ESS) was
examined and reported as a measure of estimate stability/accuracy for parameters of interest.
Further detail regarding data processing and analysis can be obtained from the corresponding
author.
RESULTS
The GTD (downloaded 6 November 2020), spanning the 17,532 days from 1970 to 2018 (data
from 1993 missing), contained 191,464 unique terrorism incidents. After omitting those with
missing date or casualty information, 172,579 incidents remained for analysis. Of those,
106,998 (~62%) incidents were associated with ≥1 wounded or killed persons and so were
considered terrorism casualty-incidents. Daily tallies of terrorism casualty-incidents varied
from 0 to 56, with a mean (SD) of 6.10 (7.55) terrorism casualty-incidents per day. Upper and
lower limits for casualty-incident quintiles (i.e. casualty-incident factor levels), computed
separately for each of the 4 GTD study epochs, revealed qualitative across-epoch differences
(Table 1), reflecting in part the appreciable growth in terrorism behaviour from the late 20th to
the early 21st Century.
Page 5 of 13
491
Mulligan, B. P., & Koren, S. A. (2021). Geopsychology of instrumental aggression: daily concurrence of global terrorism and solar-geomagnetic
activity (1970-2018). Advances in Social Sciences Research Journal, 8(5). 487-499.
URL: http://dx.doi.org/10.14738/assrj.85.10266
Table 1. Epoch-wise terrorism casualty-incident quintile ranges.
Global Terrorism Dataset (GTD) data-collection epoch
1 January 1970-
31 December 1997
1 January 1998-
31 March 2008
1 April 2008-
31 December 2011
1 January 2012-
31 December 2018
Casualty
quintile
Lower Upper Lower Upper Lower Upper Lower Upper
1st 0 0 0 1 0 5 0 14
2nd 1 1 2 2 6 7 15 18
3rd 2 3 3 3 8 9 19 22
4th 4 5 4 5 10 12 23 27
5th 6 30 6 36 13 34 28 56
Each day from 1970 to 2018 was assigned a quintile rank based on the number of terrorism
casualty-incidents occurring on that day. Upper and lower limits of casualty-incident quintiles
were computed separately for each of the 4 data collection epochs (reflecting methodological
differences) included in the Global Terrorism Database (GTD).
Examination of terrorism casualty-incident factor-level differences in the posterior
distributions showed that solar activity levels were credibly lower (i.e. the 95% HDI for the
estimated differences excluded zero) on days falling in the 1st and 2nd terrorism casualty- incident quintiles, and credibly higher for days in the 3rd, 4th, and 5th quintiles. In fact, point
estimates of solar activity showed a rank-ordered stepwise increase across casualty-incident
quintiles. And while geomagnetic activity was likewise credibly lower on days in the 1st
casualty-incident quintile and credibly elevated on days falling in the 4th quintile, estimates of
differences in geomagnetic activity for days in the 2nd, 3rd, and 5th casualty-incident quintiles
had 95% HDIs that included zero (i.e. null contrasts). These results are summarized in Table 2
and in Figures 1 and 2.