Professional Documents
Culture Documents
The Urban Security Image Database (USID)
The Urban Security Image Database (USID)
https://doi.org/10.1007/s11292-021-09490-x
Abstract
Objectives Researchers have been studying the most important environmental cues
that influence people’s fear and how to measure these emotions and perceptions in a
more valid way. In order to contribute to experimental studies, we develop and vali-
date the Urban Security Image Database (USID).
Method The construction and validation of the USID followed two stages: (a) the
obtainment by researchers of more than 300 naturalistic pictures in different urban
contexts of the city of Porto and (b) using a within-subject design, a large sample
(N = 1780) classified 49 selected pictures for fear of crime, risk perception of vic-
timization, arousal, and valence levels.
Results The validated Urban Security Image Database (USID) contains 49 pictures
that are grouped in three categories according to fear mean levels: low fear, neutral
fear, and high fear. Pictures of the low fear group depict residential areas, with high
prospect spaces and well-cared vegetation. Pictures in the high fear group represent
scenarios in night-time, with signs of incivilities and low prospect spaces. Fear of
crime was negatively correlated with valence and positively with arousal.
Conclusions USID is an important step to laboratorial experiments in the field of
fear of crime and its relationship with environmental features. Moreover, since fear
of crime is correlated with valence and arousal, it provides strength to the impor-
tance of considering fear a context-specific experience.
* Inês Guedes
iguedes@direito.up.pt
1
Present Address: School of Criminology, Faculty of Law, University of Porto, Rua dos Bragas
223, 4050‑123 Porto, Portugal
13
Vol.:(0123456789)
I. Guedes et al.
Introduction
One of the main efforts researchers have been trying to pursue in laboratorial
studies is the use of environmental-controlled stimuli for inducing specific emo-
tional states (Lang et al. 1999, 2008). To create those stimuli, a wide variety of
datasets have been developed (e.g. International Affective Picture System, IAPS;
Nencki Affective Picture System, NAPS; Geneva Affective Picture Database,
GADEP) using emotionally charged materials such as naturalistic or manipulated
pictures, static emotional faces, film clips, or auditory stimuli (Lang et al., 1997;
Wessa et al., 2010; Westermann et al., 1996). These databases are standardized
on the fact that the central dimensions characterizing the affective experiences
are the arousal (or activation), the valence (or pleasantness), and the dominance
(or control) (Lang et al., 1993). Arousal ranges from excited/aroused to relaxed/
unaroused and is associated to behavioural, autonomous, and cortical activity.
On the other hand, valence ranges from highly positive to highly negative, thus
indicating which motivational system is involved (a pleasant-approach one or an
unpleasant-avoidance one) (Lang et al., 1993; Prada & Garcia-Marques, 2006).
Finally, dominance represents the degree of perceived control over the affective
stimulus, ranging from a feeling of dominance, control, and self-assurance to
submission, passivity, and timidity. This dimension is often less homogeneous
than valence and arousal and explains only a small proportion of variance. Gen-
erally, to collect normalized ratings for these three dimensions, the Self-Assess-
ment Manikin (SAM) scale is used (Bradley & Lang, 1994). While the IAPS is
the most widely used database of natural pictures, certain issues related with this
database have been raised (e.g. Dan-Glauser & Scherer, 2011). One of the limita-
tions is the restricted number of pictures belonging to specific content categories.
Under the field of fear of crime, this is especially true.
Fear of crime might be defined as a more restricted phenomenon correspond-
ing to an emotional component (Ferraro & LaGrange, 1987; Warr, 2000). On the
other hand, fear of crime is also conceptualized as a wider phenomenon com-
prising affective, cognitive, and behavioural elements (e.g. Gabriel & Greve,
2003). While fear of crime is an emotion (Ferraro, 1995), arising from crime or
symbols associated with it, researchers have been trying to understand what are
the contextual cues that trigger higher levels of fear of crime. Moreover, to over-
come the limitations of surveys, there is a tendency to use innovative methods to
explore the relationship between fear of crime and environmental space, such as
eye tracking (e.g. Guedes et al., 2015), app-based measures (e.g. Solymosi et al.,
2020), and physiological measures (e.g. Castro-Toledo et al., 2017; Noon et al.,
2019a). Both eye tracking and physiological studies have historically used stand-
ardized and validated set of stimuli. Nevertheless, to the best of our knowledge,
there is not any image set with naturalistic urban pictures available and validated
to be used in laboratorial studies to explore the relationship between fear of crime
and environmental features. Therefore, the present study aims at validating the
Urban Security Image Database (USID) to be used in laboratorial studies in the
field of fear of crime.
13
The Urban Security Image Database (USID): development and…
Stimuli and databases
Among the different types of materials used for emotional stimulation, the vis-
ual (e.g. static pictures) is the most prevalent in experiments. When researchers
choose the stimuli for their laboratorial studies, it is important to define the types
of static pictures necessary for the setting, as well as the intrinsic value of these
stimuli. Pictures must be tested on various parameters prior to the experiments
since they will be used as controlled inducing material. Generally, that are two
possibilities for the researcher (Dan-Glauser & Scherer, 2011): one is to gather
the pictures for the specific study, and the other is to use databases from other
available laboratories/studies.
The development and systematization of the IAPS (Bradley & Lang, 2007;
Lang et al., 1997, 2008) marked a directive step in the use of stimuli when
studying emotions, psychological elements, or psychosocial processes. IAPS is
the most worldwide used naturalist pictures database, having been validated in
numerous studies (e.g. Déak et al. 2010; Lang et al., 1993; Weinberg & Hajcak,
2010). Its fundamental aim is “to develop a large set of standardized, emotion-
ally-evocative, internationally accessible, colour photographs that includes con-
tents across a wide range of semantic categories” (Lang et al., 1997: 1). There-
fore, datasets such as IAPS allow improved experimental control in the selection
of emotional stimuli, facilitate the comparison of results across various studies
conducted in the same or different laboratories, and, finally, encourage replica-
tions within and across laboratories (Lang et al., 1997, 2008). IAPS comprehends
a vast display of pictures representing a multiplicity of objects and situations such
as landscapes, plants, animals, children and adults, plants, food, natural disasters,
weapons, and erotica. Each picture is associated with a certain affective and emo-
tional response and includes normative ratings concerning valence, arousal, and
dominance measured through SAM (Bradley & Lang, 1994).
Several studies have found high stability of affective ratings worldwide (e.g.
Vila et al., 2001) suggesting that emotional reactions that stem from IAPS pic-
tures are quite stable cross-culturally. Also, other studies have shown the reliable
induction of emotional responses by these stimuli (Lang et al., 1993; Weinberg &
Hajcak, 2010). Nevertheless, authors such as Marchewka et al. (2014) and Dan-
Glauser and Scherer (2011) have pointed out several issues regarding IAPS data-
set, namely the existence of a limited set of pictures underlying specific catego-
ries, leading to situations in which individuals observe the same materials two or
more times, reducing the capacity of emotional induction. Moreover, some inter-
nal validity issues have been raised, as the quality of the images is not always sat-
isfactory, and studies have been shown that the physical properties of the image,
such as luminance, size, or complexity, might influence the affective processing
of visual stimuli (Marchewka et al., 2014).
Other datasets have been developed and validated in the last years. For
instance, Dan-Glauser and Scherer (2011) created the GADEP which comprises a
set of negative, positive, and neutral pictures and aims at increasing the availabil-
ity of emotional visual stimuli. Furthermore, Wessa et al. (2010) also developed
13
I. Guedes et al.
13
The Urban Security Image Database (USID): development and…
Present study
Taking into consideration the growing number of studies in the field of fear of crime
that have been using innovative methods to explore (a) the nature of fear of crime as
a context-specific multidimensional experience and (b) the environmental cues asso-
ciated with fear, there is a demand for additional pictorial databases providing infor-
mation about the properties of the stimuli (fear of crime rates and affective dimen-
sions, such as valence and arousal). Therefore, in the present work, we provide a set
of 49 naturalistic photographs of urban scenarios varying in features that research
has been suggesting to be related to fear of crime, such as presence/absence of inci-
vilities, night/daytime, well-cared vegetation/untended vegetation, low prospect/
high prospect, and others. In light of previous studies, we hypothesized that pictures
with higher levels of fear of crime are rated with higher levels of arousal and lower
levels of valence. Moreover, we expect that pictures rated with higher levels of fear,
comparing with pictures rated with low levels, will present a set of contextual fea-
tures that are generally associated with fear, such as signs of incivilities, poor light-
ning, and low prospect.
Considering the normative rating procedures of Lang and colleagues (2008),
for each picture, we collected ratings of valence and arousal, adding ratings of per-
ceived risk of victimization and fear of crime. One of the advantages of this database
is related to the diversity of pictures selected and the consideration of more than
two affective ratings. Also, this dataset of urban security enables the comparison
of results between distinct studies that use these stimuli and encourage the replica-
tion within and between research centres. Thus, the Urban Security Image Database
(USID) is available to the scientific community for non-commercial use.
Method
Participants
13
I. Guedes et al.
Table 1 Sample characterization
Total sample Female Male p
and staff from different Portuguese universities. Participants were between 16 and
74 years, and the mean age was approximately 25 years (M = 24.53, SD = 6.69).
About 67% were female and 33% male. Most of our sample was attending a degree
at the university (57.9%) and had already a degree (21.1%) or a post degree (15%).
Furthermore, more than 80% of the participants has never been victim of a crime
(Table 1).
Materials
13
The Urban Security Image Database (USID): development and…
1
Due to an error detected when the survey was ready to be launched (one of the pictures previously
selected was repeated), 24 instead of 25 pictures were evaluated by the external sample.
13
Table 2 Mean levels, SD, median, and confidence intervals of valence, arousal, fear of crime, and perceived risk of victimization
Picture Valence (1–9) Arousal (1–9) Fear of crime (1–5) Perceived risk (1–5)
n M ± SD Mdn 95% CI n M ± SD Mdn 95% CI n M ± SD Mdn 95% CI n M ± SD Mdn 95% CI
13
1 1478 3.92 ± 1.37 4 [3.88, 4.05] 1475 3.29 ± 1.59 3 [3.16, 3.37] 1477 3.05 ± .67 3 [2.99, 3.07] 1477 2.21 ± 1.11 3 [2.13, 2.27]
2 1425 3.59 ± 1.42 3 [3.47, 3.64] 1422 4.18 ± 1.79 4 [4.12, 4.37] 1426 3.79 ± .71 4 [3.79, 3.88] 1427 3.20 ± .1.07 3 [3.18, 3.31]
3 1366 4.44 ± 1.38 4 [4.38, 4.55] 1358 3.04 ± 1.46 3 [2.91, 3.10] 1364 2.65 ± .65 3 [2.59, 2.67] 1365 1.66 ± .99 1 [1.58, 1.70]
4 1340 3.18 ± 1.49 3 [3.03, 3.22] 1333 4.32 ± 1.83 4 [4.29, 4.53] 1341 3.65 ± .71 4 [3.64, 3.73] 1341 3.01 ± 1.17 3 [2.96, 3.10]
5 1314 3.01 ± 1.54 3 [2.91, 3.10] 1309 4.72 ± 2.07 5 [4.63, 4.90] 1317 4.02 ± .75 4 [3.96, 4.05] 1316 3.49 ± 1.15 4 [3.39, 3.51]
6 1301 4.46 ± 1.44 5 [4.38, 4.56] 1294 3.38 ± 1.53 3 [3.29, 3.48] 1300 2.80 ± .68 3 [2.56, 2.84] 1301 1.81 ± 1.07 1 [1.73, 1.86]
7 1265 4.18 ± 1.38 4 [4.10, 4.27] 1257 3.22 ± 1.49 3 [3.25, 3.45] 1265 2.82 ± .64 3 [2.76, 2.84] 1266 1.80 ± 1.06 1 [1.72, 1.85]
8 1247 5.06 ± 1.69 5 [4.98, 5.19] 1240 3.31 ± 1.66 3 [3.18, 3.40] 1250 2.38 ± .78 2 [2.32, 2.42] 1249 1.50 ± .93 1 [1.44, 1.55]
9 1241 2.17 ± 1.25 2 [2.04, 2.19] 1237 4.98 ± 2.26 5 [4.95,5.25] 1244 4.33 ± .71 4 [4.30, 4.39] 1244 4.09 ± .88 4 [4.04, 4.15]
10 1220 3.25 ± 1.37 3 [3.19, 3.36] 1213 4.28 ± 1.89 4 [4.20, 4.45] 1222 3.80 ± .69 4 [3.76, 3.84] 1222 3.21 ± 1.11 3 [3.13, 3.27]
11 1198 4.43 ± 1.30 4 [4.35, 4.52] 1188 3.25 ± 1.49 3 [3.11, 3.31] 1197 2.78 ± .64 3 [2.72, 2.80] 1199 1.76 ± 1.03 1 [1.66, 1.78]
12 1184 3.66 ± 1.83 3 [3.48, 3.71] 1177 3.85 ± 1.87 4 [3.77, 4.02] 1183 3.54 ± .83 4 [3.50, 3.60] 1186 2.90 ± 1.29 3 [2.83, 2.99]
13 1160 3.39 ± 1.30 3 [3.32, 3.48] 1156 3.89 ± 1.75 4 [3.77, 4.01] 1161 3.63 ± .67 4 [3.58, 3.67] 1163 2.97 ± 1.12 3 [2.88, 3.02]
14 1135 3.77 ± 1.45 4 [3.68, 3.86] 1127 3.66 ± 1.64 4 [3.59, 3.81] 1139 3.06 ± .71 3 [3.01, 3.09] 1140 2.16 ± 1.19 3 [2.07, 2.22]
15 1123 6.66 ± 1.52 7 [6.60, 6.78] 1119 3.62 ± 1.98 3 [3.50, 3.77] 1125 2.04 ± .72 2 [1.98, 2.07] 1127 1.27 ± .72 1 [1.21, 1.30]
16 1106 3.59 ± 1.53 3 [3.51, 3.71] 1102 4.12 ± 1.82 4 [4.03, 4.27] 1111 3.27 ± .73 3 [3.23, 3.32] 1111 2.50 ± 1.20 3 [2.43, 2.58]
17 1092 5.70 ± 1.63 6 [5.60, 5.81] 1088 3.34 ± 1.71 3 [3.23, 3.45] 1097 2.27 ± .67 2 [2.22, 2.31] 1099 1.34 ± .76 1 [1.28, 1.37]
18 1085 7.41 ± 1.46 8 [7.36, 7.54] 1082 3.84 ± 2.33 4 [3.68, 3.99] 1092 1.91 ± .84 2 [1.85, 1.96] 1091 1.26 ± .73 1 [1.20, 1.29]
19 1072 3.31 ± 1.25 3 [3.23, 3.39] 1059 3.69 ± 1.68 4 [3.63, 3.85] 1075 3.42 ± .68 3 [3.37, 3.45] 1075 2.72 ± 1.15 3 [2.65, 2.80]
20 1064 3.55 ± 1.53 3 [3.46, 3.65] 1059 4.07 ± 1.78 4 [4.01, 4.25] 1063 3.12 ± .72 3 [3.09, 3.18] 1065 2.33 ± 1.19 3 [2.26, 2.41]
21 1044 3.04 ± 1.29 3 [2.96, 3.13] 1033 4.09 ± 1.86 4 [4.00, 4.25] 1047 3.80 ± .65 4 [3.77, 3.85] 1047 3.23 ± 1.19 3 [3.18, 3.27]
22 1034 3.62 ± 1.45 4 3.53, 3.72] 1029 3.69 ± 1.69 4 [3.59, 3.81] 1038 3.45 ± .76 3 [3.40, 3.50] 1038 2.80 ± 1.24 3 [2.71, 2.87]
23 1022 6.49 ± 1.55 7 [6.40, 6.59] 1024 3.54 ± 1.92 3 [3.41, 3.67] 1028 2.15 ± .69 2 [2.10, 2.18] 1030 1.33 ± .81 1 [1.28, 1.38]
24 1017 3.44 ± 1.36 3 [3.35, 3.52] 1006 3.93 ± 1.76 4 [3.83, 4.06] 1019 3.42 ± .80 3 [3.37, 3.57] 1020 2.76 ± 1.23 3 [2.68, 2.84]
I. Guedes et al.
Table 2 (continued)
Picture Valence (1–9) Arousal (1–9) Fear of crime (1–5) Perceived risk (1–5)
n M ± SD Mdn 95% CI n M ± SD Mdn 95% CI n M ± SD Mdn 95% CI n M ± SD Mdn 95% CI
25 1006 2.80 ± 1.32 3 [2.72, 2.89] 996 4.36 ± 2.07 4 [4.25, 4.53] 1007 3.98 ± .72 4 [3.93, 4.02] 1010 3.48 ± 1.09 4 [3.41, 3.53]
26 297 2.80 ± 1.31 3 [2.68, 3.03] 294 4.62 ± 2.00 5 [4.59, 5.13] 296 3.99 ± .64 4 [3.85, 4.10] 296 3.43 ± .85 3 [3.32, 3.55]
27 292 6.25 ± 1.37 6 [6.04, 6.39] 289 4.06 ± 1.80 4 [3.81, 4.29] 291 2.28 ± .63 2 [2.16, 2.42] 290 2.16 ± .60 2 [2.08, 2.23]
28 282 4.16 ± 1.32 4 [3.99, 4.33] 281 4.15 ± 1.65 4 [3.95, 4.40] 282 3.44 ± .71 3 [3.33, 3.60] 281 3.11 ± .74 3 [2.97, 3.16]
29 278 5.17 ± 1.46 5 [4.98, 5.35] 277 4.11 ± 1.63 4 [3.90, 4.37] 278 2.82 ± .68 3 [2.63, 2.92] 277 2.53 ± .75 2 [2.31, 2.61]
30 274 5.81 ± 1.81 6 [5.59, 6.04] 272 4.56 ± 2.08 5 [4.25, 4.82] 275 2.19 ± .70 2 [2.09, 2.36] 274 1.99 ± .65 2 [1.90, 2.05]
31 274 3.34 ± 1.64 3 [3.13, 3.55] 271 4.74 ± 1.97 5 [4.62, 5.15] 273 3.54 ± 1.10 4 [3.32, 3.69] 271 3.86 ± 1.68 3 [3.28, 3.51]
32 271 5.25 ± 1.56 5 [5.01, 5.41] 270 3.86 ± 1.68 4 [3.59, 4.03] 270 2.69 ± .68 3 [2.54, 2.79] 269 2.42 ± .65 2 [2.30, 2.48]
33 269 2.29 ± 1.55 2 [2.10, 2.51] 266 5.91 ± 2.30 6 [5.70, 6.33] 268 4.43 ± .78 5 [4.10, 4.44] 266 4.27 ± .83 4 [4.16, 4.38]
34 268 7.76 ± 1.32 8 [7.76, 7.93] 266 5.33 ± 2.28 5 [4.99,5.61] 267 2.40 ± .76 2 [2.31, 2.49] 266 2.41 ± .76 2 [2.29, 2.48]
35 266 4.33 ± 1.57 3 [4.09, 4.48] 263 4.38 ± 1.65 5 [4.30, 4.74] 265 3.17 ± .66 3 [2.99, 3.26] 265 2.84 ± .75 5 [2.76, 2.95]
36 262 2.73 ± 1.55 3 [2.49, 2.88] 263 5.37 ± 2.18 5 [5.24, 5.82] 264 4.24 ± .67 4 [4.09, 4.37] 264 3.95 ± .84 4 [3.83, 4.05]
The Urban Security Image Database (USID): development and…
37 265 6.00 ± 1.66 6 [5.79, 6.20] 260 4.14 ± 1.93 4 [3.86, 4.36] 264 2.24 ± .72 2 [2.12, 2.39] 264 2.08 ± .66 2 [1.97, 2.14]
38 262 3.60 ± 1.40 4 [3.41, 3.77] 257 4.19 ± 1.78 4 [4.01, 4.50] 260 3.27 ± .66 3 [3.05, 3.33] 261 2.93 ± .74 3 2.83, 3.03]
39 261 7.09 ± 1.56 7 [6.92, 7.30] 260 5.19 ± 2.18 5 [4.92, 5.53] 261 2.57 ± .80 3 [2.40, 2.73] 261 2.39 ± .80 2 [2.27, 2.47]
40 259 6.54 ± 1.33 7 [6.33, 6.67] 256 4.65 ± 1.83 5 [4.37, 4.88] 257 2.47 ± .66 2 [2.31, 2.58] 258 2.26 ± .64 2 [2.17, 2.32]
41 256 2.78 ± 1.31 3 [2.57, 2.91] 251 4.95 ± 1.99 5 [4.71, 5.27] 257 3.77 ± .74 4 [3.54, 3.87] 256 3.41 ± .90 3 [3.28, 3.52]
42 255 2.31 ± 1.46 2 [2.15, 2.55] 252 5.46 ± 2.21 6 [5.17, 5.78] 255 4.32 ± .73 4 [4.07, 3.38] 254 4.19 ± .83 4 [4.09, 4.31]
43 252 7.12 ± 1.46 7 [6.90, 7.28] 251 4.92 ± 2.09 5 [4.63, 5.23] 253 2.05 ± .68 2 [1.87, 2.15] 253 1.91 ± .61 2 [1.83, 1.99]
44 244 6.42 ± 1.30 6 [6.21, 6.55] 239 4.47 ± 1.90 5 [4.16, 4.69] 248 2.44 ± .64 2 [2.28, 2.53] 243 2.20 ± .78 2 [2.12, 2.29]
45 244 3.69 ± 1.36 4 [3.51, 3.87] 239 4.14 ± 1.82 4 [3.87, 4.38] 244 3.54 ± .68 4 [3.38, 3.65] 243 3.24 ± .78 3 [3.15, 3.35]
46 241 5.51 ± 1.48 5 [5.31, 5.70] 237 4.12 ± 1.72 4 [3.83, 4.31] 240 2.53 ± .62 3 [2.47, 2.71 238 2.31 ± .64 2 [2.22, 2.38]
47 239 6.61 ± 1.50 7 [6.35, 6.75] 237 4.58 ± 1.94 5 [4.26, 4.81] 240 2.35 ± .70 2 [2.23, 2.51] 240 2.13 ± .60 2 [2.05, 2.21]
13
Table 2 (continued)
Picture Valence (1–9) Arousal (1–9) Fear of crime (1–5) Perceived risk (1–5)
n M ± SD Mdn 95% CI n M ± SD Mdn 95% CI n M ± SD Mdn 95% CI n M ± SD Mdn 95% CI
13
48 237 3.77 ± 1.37 4 [3.61, 3.97] 235 4.14 ± 1.70 4 [3.93, 4.41] 237 3.63 ± .69 4 [3.45, 3.75] 236 3.19 ± .79 3 [3.11, 3.31]
49 235 6.40 ± 1.62 7 [6.19, 6.62] 232 4.46 ± 2.07 5 [4.13, 4.71] 235 2.32 ± .72 2 [2.22, 2.52] 235 2.07 ± .58 2 [2.00, 2.15]
I. Guedes et al.
The Urban Security Image Database (USID): development and…
Measures
To evaluate the pictures for the database, a survey was developed. The first part of
the survey consisted of four variables measuring valence, arousal, fear of crime, and
perceived risk of victimization. Regarding the valence scale, subjects were asked
to evaluate the degree of pleasantness of the image presented in a scale that ranged
between 1 (very unpleasant) and 9 (very pleasant). To measure arousal, subjects
were asked to evaluate the degree of arousal of the image presented from 1 (not
intense) to 9 (very intense). Both measures (valence and arousal) followed Brad-
ley and Lang (1994)’s work concerning the use of SAM. In addition, two ques-
tions to measure fear of crime and perceived risk of victimization were included.
For the former, subjects were asked “Imagine that you are in the place represented
by the image. Please evaluate the degree of insecurity you would feel” in a scale
that ranged between 1 (very secure) to 5 (very insecure). Subjects were also asked
to evaluate their perception of victimization risk through the following question:
“Imagine that you are in the place represented by the image. What is the likelihood
of being a crime victim from 1 (unlikely) to 5 (very likely)?”. The second group of
questions was related to the personal characterization of participants. Sex, age, edu-
cation, and victimization in the last 12 months were included.
Procedure
An online survey built in Survey Monkey Software was used in the present study.
Before starting answering to the survey, participants were informed about the goals
of the study, the expected duration of filling the survey, and that a gift card would
be drawn in order to reward participants for their time. We also informed that the
participation was voluntary and that the data would be only used for the purposes of
scientific research. After subjects provided informed consent, they would be directed
to the questionnaire starting with the phrase “watch the next image for a few sec-
onds” appeared. Since it was not possible to control the viewing time of each image,
due to the fact of subjects answered the questionnaire on their personal computer,
each picture remained on the screen until the participant intentionally proceed to
the next page. To control for potential order effects, we randomized the order of the
images presented. After rating the images, participants completed the personal char-
acterization measures.
Data analysis
Data was analysed through IBM Statistics 26. Mean and standard deviation levels
were obtained for each of the four main dimensions: arousal, valence, fear of crime,
and perceived risk of victimization. In order to constitute three groups based on the
levels of fear of crime, the percentiles 33.3 and 66.6 of fear of crime mean levels
were calculated. Then, taken those means levels into consideration, the variable fear
of crime was recoded into a new variable that ranged between 1 (low fear) and 3
13
I. Guedes et al.
(high fear). These three groups were compared for fear, perceived risk, arousal, and
valence means, using one-way repeated measures ANOVA to analyse if participant’s
ratings differenced significantly across the image categories. To explore the relation-
ship between the four measures in each group, correlations tests were made using
Pearson coefficient. Lastly, to build the affective space, the quadratic relationship
between arousal and valence was considered.
Results
For each picture, the four dimensions were evaluated: valence, arousal, fear of crime,
and perceived risk of victimization. Table 2 presents the mean levels (M), standard
deviation (SD), median (Mdn.), and confidence intervals (CI, 95%) for the 49 pic-
tures of the database. As it can be observed in Table 2, the confidence intervals are
small thorough all the ratings of each image, showing an agreement in ratings.
The following procedure was to group the 49 images according to fear of crime
levels as explained in the “Data analysis” section. Therefore, we obtained three sets
of images with low, neutral, and high levels of fear of crime.
Table 3 presents mean levels of valence, arousal, fear of crime, and perceived
risk of victimization and the differences between groups. The results of one-way
ANOVA revealed that across the different groups, there were significant differ-
ences on classifications of each variable. First, valence mean levels are higher in
the low fear group (6.43) compared to neutral group (4.05) and high fear group
(3.00) (F = 139.15, p < − 001). The results from the post hoc test showed mean dif-
ferences between the three groups. Concerning the arousal dimension, it is possible
to observe that both the low fear (4.28) and the high fear (4.62) present higher mean
levels compared to neutral fear group (3.77) (F = 9.96, p < 0.001). Nevertheless,
low fear and high fear groups do not differ in arousal means (p = 0.199). Regard-
ing the fear of crime variable, one-way ANOVA and post hoc test showed differ-
ences between three groups. The mean levels of fear of crime are higher in high fear
(3.90) group compared to neutral fear (3.08) and low fear groups (2.29) (F = 153.81,
p < 0.001). Lastly, perceived risk of victimization is also different between three
groups (F = 46.82, p < 0.001), and it can be observed that the mean levels of this var-
iable are higher in the high fear group (3.48) compared to neutral fear group (2.47)
and with the low fear group (1.91).
Relationships between ratings
Table 4 presents the correlations between variables both for all the pictures together
and for each group based on fear of crime levels.
Considering all the 49 pictures, and as expected, we observed a negative corre-
lation (r = -0.926, p < 0.001) between fear of crime and valence. Therefore, images
with higher levels of fear of crime are considered less positive or pleasant for
13
Table 3 Mean comparisons of Fear of crime, arousal, valence, and risk perception of victimization between the different groups
Groups by fear levels Cronbach alpha Image number Valence M ± SD Arousal M ± SD Fear of crime M ± SD Perceived risk M ± SD
Low fear 8, 15, 17, 18, 23, 27, 30, 34, 6.43 ± .71a 4.28 ± .63a 2.29 ± .18a 1.91 ± .42a
37, 39, 40, 43, 44, 46, 47, 49
Neutral fear 1, 3, 6, 7, 11, 14, 16, 19, 20, 4.05 ± .57b 3.77 ± .40b 3.08 ± .29b 2.47 ± .54b
22, 24, 28, 29, 32, 35, 38, 48
High fear 2, 4, 5, 9, 10, 12, 13, 21, 25, 3.00 ± .48c 4.62 ± .59a 3.90 ± .30c 3.48 ± .42c
26, 31, 33, 36, 41, 42, 45
Comparisons between groups: 139.15 (p = < .001) 9.96 (p = < .001) 153.81 (p = < .001) 46.82 (p = < .001)
The Urban Security Image Database (USID): development and…
Fa
a
One-way ANOVA of variance test
b, c,d
Denotes the differences between groups using Tukey’s HSD post hoc test
Note: Cut-off vales were 2.622 for the percentile 33.3 and 3.477 for the percentile 66.6. Low fear images vary between the lowest rating and 2.62; neutral fear varies
between 2.63 and 3.48; high fear varies between 3.49 and the highest rating
13
I. Guedes et al.
individuals. Moreover, higher levels of fear are also correlated with higher arousal
(r = 0.382, p < 0.001). Lastly, and as expected, fear of crime is strongly correlated
with perceived risk victimization (r = 0.914, p < 0.001). Regarding the relationship
of these variables taking into consideration the three groups of pictures, it is possible
to observe relevant results. First, the relationship between fear of crime and valence
is only observed in groups 2 and 3 (neutral and high fear, respectively) but not in
the low fear group. Moreover, the negative relationship between valence and fear is
stronger in group 3 compared to group 2. The same can be visualized when observ-
ing the relationship between fear of crime and arousal. In fact, in group 3, where
mean levels of fear are higher, a stronger correlation between fear and arousal can be
observed when compared to group 2. Finally, in the three groups, a strong relation-
ship between fear of crime and perceived risk of victimization can be observed.
Figure 1 shows the distribution of the ratings of the 49 images in the valence and
arousal affective space. It can be observed that the distribution of the 49 images
in the affective space adopts the U shape when the valence is positioned on the
X-axis and the arousal on the Y-axis. The value of R2 quadratic is 0.546, considered
a high quadratic correlation between arousal and valence. Therefore, arousal ratings
are highest at the most pleasant (positive valence) and most unpleasant (negative
valence).
After the constitution of the three groups differing according fear level, the environ-
mental cues of the pictures composing the groups were analysed (Fig. 2). A set of
contextual cues were chosen based on literature review, namely time of the day (day/
night), incivilities (presence/absence), vegetation (well-cared/untended), residential
area (yes/no), and prospect (open/close).
Group 1 (low fear, 16 pictures) is fully composed of pictures obtained in day-
time (100%), and most of the pictures depict a residential area (81.3%). Further-
more, 93.3% of the pictures do not present any signs of incivilities. Concerning the
state of vegetation, it is possible to observe that all of the pictures present well-cared
vegetation and have high prospect or openness. The second group (neutral fear)
is composed of pictures with mixed features. Of the 17 images that constitute this
13
The Urban Security Image Database (USID): development and…
Fig. 1 Distribution of 49 images as a function of the averages for the total sample subjects in the affec-
tive space defined by the dimensions of the valence and the arousal. The respective regression curve
(quadratic) is also observed
13
I. Guedes et al.
group, only 2 pictures (11%) depict night-time instead of daytime. Moreover, while
10 images represent residential areas (58.8%), 7 images (41.2%) depict other places
such as different streets of Oporto downtown. The presence of incivilities can be
observed in 9 of 17 pictures, such as some graffiti on the wall and buildings under
construction which are mostly the features of non-residential areas presented on the
images. Furthermore, analysing the pictures of this group, it is possible to notice
that only 5 images present vegetation, and in three of them the vegetation is well-
cared. Lastly, prospect can be understood as an important feature to distinguish the
previous and this group of images. In fact, in the total of 17 pictures, 61.1% depict
places with low prospect or openness. The last group (“high fear”) is composed of
16 pictures. Half of the pictures (50%) depict night-time scenarios. Of these 8 pic-
tures, 7 are from residential areas. The remaining pictures are from daytime sce-
narios, and only one is from a residential area. The other 7 pictures present features
such as tunnels, alleys or corners, isolated streets under construction, or with signs
of incivilities. In fact, of the 16 pictures, 12 present scenarios with physical disor-
der—highly visible signs of graffiti, abandoned buildings, and untended vegetation.
Lastly, the prospect seems to be a very relevant feature of this group since in a total
of 16 pictures, 14 of them (87.5%) depict scenarios with low prospect or openness
(e.g. corners or poor lightning places).
Discussion
The purpose of this research was to develop and validate a set of urban naturalistic
pictures varying in valence, arousal, fear of crime, and perceived risk of victimi-
zation to be used in experimental studies in the field of fear of crime. Therefore,
departing from a large set of images obtained by researchers both in situ and through
Google Street View, we selected and validated 49 of these images which are part of
the Urban Security Image Database (USID).
Based on fear of crime levels, we aggregated the images into three groups: low
fear (16 pictures), neutral fear (17 pictures), and high fear (16 pictures). The results
showed that the mean levels of valence, arousal, fear of crime, and perceived risk of
victimization were different between groups. The analysis of temporal and physical
environmental features of each group showed a set of relevant distinctions. Low fear
group was constituted of pictures at daylight, without signs of disorder, with well-
cared vegetation, and with open prospect. Pictures in the high fear group, in turn,
mainly depicted scenarios with signs of incivilities or symbols associated with crime
(vandalism, graffiti, bars on the windows, untended vegetation), with low prospect,
and at night-time. These results are consistent with prior research on fear of crime.
A large body of research suggests that people feel safer during daylight com-
pared to night-time (Nasar & Jones, 1997; Painter, 1994; Warr, 1990; Welsh &
Farrington, 2007). For instance, in an experimental study, Castro-Toledo and
colleagues (2017) found that the lack of luminosity in public spaces triggered
experiences of fear of crime. Specifically, physiological reactions of arousal as
an indicator of fear were higher in the experimental group where poor lighting
was introduced. In an extensive review, Lorenc (2013) concluded that lighting
13
The Urban Security Image Database (USID): development and…
13
I. Guedes et al.
Limitations
13
The Urban Security Image Database (USID): development and…
Although several image datasets already exist, USID offers a novel way to explore
the relationship between fear of crime and environmental cues. Researchers will be
able to select, according to their design, the most suitable pictures from the data-
base, rated for fear of crime, arousal, valence, and perceived risk of victimization.
In a laboratorial context, researchers can measure physiological reactions to the dif-
ferent environments depicted by the pictures or record eye movements directed to
the pictures through eye tracking techniques. Moreover, as previously mentioned,
replication is highly recommended both to generalize and compare results across
different populations and contexts.
Overall, USID provides important material for innovative studies dealing with the
relationship between fear of crime and the characteristics of urban spaces.
Acknowledgements We would like to thank Mariana Machado for the valuable insights during the revi-
sion process.
Funding The first author was funded by the Portuguese Foundation for Science and Technology through
a doctoral grant, financed by national funds of the Portuguese Ministry of Science, Technology and
Higher Education and the European Social Fund through the Human Capital Operational Programme.
References
Andrews, M., & Gatersleben, B. (2010). Variations in perception of danger, fear and preference in a simu-
lated natural environment. Journal of Environmental Psychology, 30, 473–481.
Bishop, I. D., & Rohrmann, B. (2003). Subjective responses to simulated and real environments: A com-
parison. Landscape and Urban Planning, 65(4), 261–277. https://doi.org/10.1016/S0169-2046(03)
00070-7
Bradley, M. M., & Lang, P. J. (1994). Measuring emotion: The Self- Assessment Manikin and the seman-
tic differential. Journal of Behavioral Therapy and Experimental Psychiatry, 25, 49–59.
Bradley, M. M., & Lang, P. J. (2007). Emotion and motivation. In J. T. Cacioppo, L. G. Tassinary, & G.
G. Berntson (Eds.), Handbook of psychophysiology (pp. 581–607). Cambridge University Press.
Castro-Toledo, F. J., Perea-García, J. O., Bautista-Ortuño, R., & Mitkidis, P. (2017). Influence of envi-
ronmental variables on fear of crime: Comparing self-report data with physiological measures in an
experimental design. Journal of Experimental Criminology, 13(4), 537–545.
Chiang, Y., & Nasar, J. (2014). Influence of visibility and situational concern on appraisals of forest trails.
Landscape & Urban Planning, 125, 166–173.
Crosby, F., & Hermens, F. (2019). Does it look safe? An eye tracking study into the visual aspects of fear
of crime. Quarterly Journal of Experimental Psychology, 72(3), 599–615.
Dan-Glauser, E. S., & Scherer, K. R. (2011). The Geneva Affective PicturE Database (GAPED): A new
730-picture database focusing on valence and normative significance. Behavioural Research, 43,
468–477.
Deák, A., Csenki, L., & Révész, G. (2010). Hungarian ratings for the International Affective Picture Sys-
tem (IAPS): A cross-cultural comparison. Empirical Text and Culture Research, 4, 90–101.
Doran, B., & e Lees, B. . (2005). Investigating the spatiotemporal links between disorder, crime and the
fear of crime. The Professional Geographer, 57(1), 1–12.
Farrall, S. D., Jackson, J., & Gray, E. (2009). Social order and the fear of crime in contemporary times.
Oxford: Oxford University Press.
Ferraro, K., & LaGrange, R. (1987). The measurement of fear of crime. Sociological Inquiry, 57(1),
70–97.
13
I. Guedes et al.
13
The Urban Security Image Database (USID): development and…
Solymosi, R., Buil-Gil, D., Vozmediano, L., & Guedes, I. (2020). Towards a place-based measure of
fear of crime: A systematic review of app-based and crowdsourcing approaches. Environment and
Behavior, 53(9), 1013–1044.
Stamps, A. (1990). Use of photographs to simulate environments: A meta-analysis. Perceptual and Motor
Skills, 71(3), 907–913.
Talbot, J., & Kaplan, R. (1984). Needs and fears: The response to trees and nature in the inner city. Jour-
nal of Arboriculture, 10, 222–228.
Vila, S., Sánchez, M., Ramírez, I., Fernández, M. C., Cobos, P., Rodríguez, S., & Moltó, J. (2001). El
Sistema Internacional de Imágenes Afectivas (IAPS): Adaptación espanõla. Segunda parte. Revista
De Psicología General y Aplicada, 54, 635–657.
Vrana, S. R., Cuthbert, B. N., & Lang, P. J. (1986). Fear imagery and text processing. Psychophysiology,
23, 247–253.
Warr, M. (1990). Dangerous situations: Social context and fear of victimization. Social Forces, 68(3),
891–907.
Warr, M. (2000). Fear of crime in the United States: Avenues for research and policy. In D. Duffee (Ed.),
Criminal justice 2000: Vol. 4. Measurement and analysis of crime and justice (pp. 451–490). Wash-
ington, DC: National Institute of Justice.
Weinberg, A., & Hajcak, G. (2010). Beyond good and evil: The time-course of neural activity elicited by
specific picture content. Emotion, 10, 767–782.
Welsh, B. .C. ., & Farrington, D. .P. . (2007). Improved street lighting and crime prevention: A systematic
review. Stockholm: Swedish: National Council for Crime Prevention.
Westermann, R., Spies, K., Stahl, G., & Hesse, F. W. (1996). Relative effectiveness and validity of mood
induction procedures: A meta-analysis. European Journal of Social Psychology, 26, 557–580.
Wessa, M., Kanske, P., Neumeister, P., Bode, K., Heissler, J., & Schönfelder, S. (2010). EmoPics: Subjek-
tive und psychophysiologische Evaluationen neuen Bildmaterials für die klinisch-biopsychologische
Forschung. Zeitschrift für Klinische Psychologie und Psychotherapie, 39(Suppl 1/11), 77.
Wilson, J. Q., & Kelling, G. (1982). Broken windows: The police and neighborhood safety. Atlantic
Monthly, 249(3), 29–38.
Wong, K., & Domroes, M. (2005). The visual quality of urban park scenes of Kowloon Park, Hong Kong:
Likeability, affective appraisal, and cross-cultural perspectives. Environment and Planning b: Plan-
ning and Design, 32, 617–632.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published
maps and institutional affiliations.
13