Vigilance behaviour of small cervids in South-West 
Finland: effects of group size, time of day, and  
predation risk in an agricultural landscape. 
 
Iina Kari 
 
 
 
Ecology and Evolutionary Biology  
Master’s thesis  
Credits: 30 ECTS 
 
Supervisors: 
Otso Huitu 
Toni Laaksonen 
 
31.10.2025 
Turku 
 
The originality of this thesis has been checked in accordance with the University of Turku quality as-
surance system using the Turnitin Originality Check service. 
 Master's thesis  
 
Subject: Ecology and Evolutionary Biology 
Author: Iina Kari 
Title: Vigilance of small cervids in South-West Finland: effects of group size, time of day and preda-
tion risk in an agricultural landscape 
Supervisors: Otso Huitu (LUKE) and Toni Laaksonen (UTU) 
Number of pages: 48 pages + Appendices 2 pages 
Date: 31 October 2025 
 
 
This master’s thesis examines the vigilance behaviour of white-tailed deer (Odocoileus virginianus) and 
roe deer (Capreolus capreolus) in agricultural landscapes in South-West Finland. The aim of the study 
was to investigate how potential predator (gray wolf, Canis lupus) presence, group size, time of day, 
and the distance from forest cover affect the vigilance of foraging deer. The main hypotheses were that 
1) deer in areas with higher predation risk (wolf presence) were expected to show increased vigilance 
as an adaptive response to the potential threat posed by predators; 2) vigilance was expected to decrease 
with increasing group size due to collective detection shared with group members; and 3) deer foraging 
farther from forest edge were expected to be less vigilant.  
The data were primarily collected by using video monitoring during May and June 2024. The influence 
of wolves was assessed by comparing vigilance-related variables in white-tailed deer observed within 
core areas of wolf territories (high predation risk/wolf present) and at the edges or outside the illustrated 
territory boundaries (low predation risk /wolf absent). Interspecific comparisons were conducted only 
with data from low-risk areas, as roe deer were too rarely observed within high-risk areas to be included 
in all analyses. During observation bouts, vigilance was measured as 1) the relative proportion of the 
time an animal was vigilant, 2) the number of vigilance bouts during a fixed observation period, and 3) 
the average duration of vigilance bouts. The data were analysed using generalized linear mixed models.  
The results showed that wolf presence did not increase vigilance in white-tailed deer, as hypothesised. 
No statistically significant effect was found in foraging distance to nearest forest edge. Instead, group 
size and time of day were the key predictors: vigilance decreased in larger groups and was highest during 
twilight hours compared to nighttime. Vigilance frequency was also significantly lower in May than 
June in white-tailed deer. In the interspecific comparisons, no statistically significant effects were found 
between the species in any vigilance-related variables. However, in low-risk areas the average vigilance 
bout duration was influenced by observation month across both species, with bout durations being 
shorter in June compared to May. These findings suggest that in human-modified landscapes, social and 
temporal factors may shape the vigilance behaviour of small cervids more than predator presence. Future 
research should also consider individual-level traits, interspecific competition, and the effects of human 
disturbance, such as hunting season, on deer vigilance. 
 
 
 
 
Key words: vigilance behaviour, indirect effects of predators, predation risk, ecology of fear
  
Pro gradu -tutkielma  
 
Pääaine: Ekologia ja evoluutiobiologia 
Tekijä: Iina Kari 
Otsikko: Vigilance of small cervids in South-West Finland: effects of group size, time of day and pre-
dation risk in an agricultural landscape 
Ohjaajat: Otso Huitu (LUKE) ja Toni Laaksonen (UTU) 
Sivumäärä: 48 sivua + liitteet 2 sivua 
Päivämäärä: 31.10.2025 
 
 
Tämä pro gradu -tutkielma tarkastelee valkohäntäkauriin (Odocoileus virginianus) ja metsäkauriin 
(Capreolus capreolus) valppauskäyttäytymistä Lounais-Suomen maatalousalueilla. Tutkimuksen 
tavoitteena oli selvittää, kuinka pedon (susi, Canis lupus) mahdollinen läsnäolo, ryhmäkoko, 
vuorokaudenaika sekä etäisyys metsän reunasta vaikuttavat pelloilla laiduntavien kauriseläinten 
valppauteen. Tutkimuksen keskeisten hypoteesien mukaan 1) kauriiden oletettiin olevan valppaampia 
alueilla, joilla saalistusriski on suurempi ja suden läsnäolo todennäköisempää, 2) valppauden oletettiin 
vähenevän ryhmäkoon kasvaessa, koska ryhmän jäsenet jakavat tarkkailuvastuun, ja 3) kauempana 
metsän reunasta laiduntavien yksilöiden oletettiin olevan vähemmän valppaita.  
Aineisto kerättiin pääosin videohavainnoinnin avulla touko-kesäkuussa 2024. Suden epäsuoraa 
vaikutusta ruokailukäyttäytymiseen arvioitiin vertaamalla valkohäntäkauriiden valppausmuuttujia 
susireviirien ydinalueilta (korkeampi saalistusriski/susi läsnä) sekä reviirien raja-alueilta tai niiden 
ulkopuolelta (matalampi saalistusriski/susi ei läsnä). Lajien välistä vertailua tehtiin vain matalan 
saalistusriskin alueiden havainnoista, sillä metsäkauriita ei havaittu riittävästi susireviirien sisäpuolella 
kerätyssä tutkimusaineistossa. Valppautta mitattiin tarkkailujaksojen aikana kolmella tavalla: 1) 
valppausajan suhteellisena osuutena tarkkailuajasta, 2) valppausjaksojen lukumääränä sekä 3) 
valppausjaksojen keskimääräisenä kestona. Aineisto analysoitiin yleistetyillä lineaarisilla sekamalleilla.  
Tulokset osoittivat, ettei suden läsnäolo lisännyt valkohäntäkauriiden valppautta hypoteesin mukaisesti. 
Myöskään etäisyydellä metsän reunasta ei havaittu tilastollisesti merkitsevää vaikutusta. Sen sijaan 
lauman ryhmäkoko ja vuorokaudenaika olivat keskeisiä selittäjiä: suuremmissa ryhmissä valppaus 
väheni ja oli suurimmillaan hämärän aikaan verrattuna yöhön. Valkohäntäkauriin valppausjaksojen 
määrä oli myös merkitsevästi pienempi toukokuussa kuin kesäkuussa. Lajien välisessä vertailussa 
matalan saalistusriskin alueilla ei havaittu tilastollisesti merkitseviä eroja lajien 
valppauskäyttäytymisessä. Havainnointikuukausi vaikutti valppausjaksojen keskimääräiseen kestoon 
molemmilla lajeilla siten, että jaksot olivat lyhyempiä kesäkuussa kuin toukokuussa. Tulokset viittaavat 
siihen, että ihmistoiminnan muovaamassa ympäristössä sosiaaliset ja ajalliset tekijät voivat vaikuttaa 
pienten hirvieläinten valppauskäyttäytymiseen enemmän kuin petojen läsnäolo. Tulevissa tutkimuksissa 
tulisi tarkastella laajemmin myös yksilötason muuttujia, lajien välistä kilpailua sekä ihmistoiminnan, 
kuten metsästyskauden, vaikutuksia eläinten valppauteen.  
 
 
 
 
Avainsanat: valppauskäyttäytyminen, petojen epäsuorat vaikutukset, saalistusriski, pelon ekologia   
  
Contents 
1 INTRODUCTION 1 
1.1 Indirect effects of large carnivores on prey species 1 
1.2 Escape behaviour and vigilance in prey species 2 
1.2.1 Presence of predators 3 
1.2.2 Group size effect and alarm signals 3 
1.2.3 Human influence and disturbance 4 
1.2.4 Other environmental and ecological factors 4 
1.3 Predator–prey context and the study species 5 
1.3.1 White-tailed deer 6 
1.3.2 Roe deer 7 
1.3.3 Wolves in Europe and in Finland 7 
1.3.4 Predator impacts and knowledge gaps 9 
1.4 Main objectives and research questions 10 
2 MATERIAL AND METHODS 11 
2.1 Study design 11 
2.2 Research equipment 13 
2.3 Data collection 14 
2.3.1 Identifying study species 16 
3 STATISTICAL ANALYSIS 18 
3.1 Dataset 18 
3.2 Data exploration and descriptive statistics 18 
3.2 Modelling 19 
4 RESULTS 21 
4.1. Vigilance behaviour of white-tailed deer in the presence of wolf 21 
  
4.1.1 Vigilance proportion 22 
4.1.2 Vigilance frequency 25 
4.1.3 Average vigilance bout duration 28 
4.2 Interspecific variation in vigilance behaviour under low predation risk 30 
4.2.1 Vigilance proportion 32 
4.2.2 Vigilance frequency 32 
4.2.3 Average vigilance bout duration 33 
5. DISCUSSION 35 
5.1 Overview of main findings 35 
5.2 Vigilance responses to predation risk and group-level effects in white-tailed deer 35 
5.3 Temporal variation in vigilance 36 
5.4 Interspecific comparison 37 
5.5 Data limitations and methodological considerations 38 
6. CONCLUSIONS 41 
ACKNOWLEDGMENTS 42 
REFERENCES 43 
APPENDICES 49 
Appendix 1: Study fields in high-risk areas 49 
Appendix 2: Study fields in low-risk areas 50 
 
 
  
 1 
 
1 Introduction 
1.1 Indirect effects of large carnivores on prey species 
Large carnivores are widely recognised for their role in limiting prey populations and influenc-
ing ecosystem functioning through both direct and indirect effects on their prey species 
(Preisser, 2008). Direct regulation occurs through predation, whereas indirect top-down effects 
influence prey behaviour, physiology, and ecology (Lima & Dill, 1990; Brown, 1999). Through a 
chain reaction, both direct and indirect effects of top predators on prey species can further affect 
species on lower trophic levels. This type of chain reaction is referred to as a cascading effect, 
and can potentially influence broader ecosystem dynamics (Preisser, 2008; Ripple et al., 2014). 
Such dynamics highlight the significance of predator-prey interactions in shaping and main-
taining ecosystem processes and biodiversity. 
In the presence of predators, prey species must navigate in environments with varying predation 
risk while allocating their energy to balance between costs and benefits of deciding when, 
where, and how to forage, rest, or reproduce. These indirect effects of predators on prey are 
known as the Ecology of Fear and found in diverse group of species (Brown et al., 1999). Ac-
cording to this concept, prey alter their behaviour to minimise predation risk, even without 
direct encounters with predators. The fear of predation can therefore create a Landscape of Fear, 
characterised by spatial variation in predation risk, which depends on the presence and density 
of predators (Laundré et al., 2001). These fear-driven responses can alter prey species’ foraging 
patterns, habitat use, and reproductive strategies (Lima & Dill, 1990; Zanette & Clinchy, 2025). 
If a prey has no evolutionary history with a specific predator, it may not have a response to 
avoid it either. Prey might also lose their adaptative anti-predator behaviour if a predator is 
absent for a long period of time, such as hundreds of years (Stankowich & Reimers, 2015). This 
kind of predation release might affect how prey are able to assess predation risk, and how they 
are able to escape from a recolonised predator. Small scale adaptation might also occur in cer-
tain low predation risk areas, where prey have not experienced any predator encounters in a 
while (Sirot & Pays, 2011).  
Evidence of indirect effects of predators, or predator-induced ecosystem level effects in terres-
trial ecosystems, have been primarily found in large conservation areas (Preisser, 2008; Ripple 
& Beschta, 2012; Gerber et al., 2024). In Europe conservation areas are relatively small and 
 2 
 
fragmented, forcing wildlife to adjust to a human modified landscape characterised by agricul-
tural land, deforestation, and habitat degradation. The limited size and isolation of protected 
areas may constrain the extent to which, for example, cascading effects can spread. This high-
lights the need for further research on species interactions and community structures under an-
thropogenic disturbances (Gerber et al., 2024; Kuijper et al., 2024). 
1.2 Escape behaviour and vigilance in prey species 
The escape behaviour of mammals consists of multiple stages, which vary depending on the 
presence of a predator and the immediacy of the predation risk (Stankowich & Reimers, 2015). 
Escaping is always costly, but benefits overcome the costs if the risk of being attacked is acute 
and the escape is successful. In escape, there is always the possibility of getting injured, which 
lowers the individual’s fitness and its ability to escape the next predator it encounters (Cooper, 
2015). However, if prey is killed by a predator, it loses all its fitness, which creates selection 
for an optimal escape behaviour that balances costs and benefits. 
Evolutionary, physiological, social, and environmental factors all influence how prey should 
evade predator attack, and when it is an optimal time to flee. Before the actual escape, vigilance 
behaviour is a fundamental strategy, in which prey individuals actively monitor their surround-
ings to detect potential threats, and prepare for predator encounters (Brown, 1999). Many stud-
ies indicate that vigilance behaviour can reduce unnecessary, costly escape attempts (Beau-
champ, 2015; Stankowich & Reimers, 2015). Strong evidence also suggests that the more an 
animal allocates its energy to vigilance, the less likely it is caught by a predator (Lima & Bed-
nekoff, 1999b). Despite the benefits, vigilance can be energetically costly, as the energy allo-
cated to detecting predators may be taken from other important tasks such as foraging or resting 
(Dimond & Lazarus, 1974; Lima, 1998; Lashley et al., 2014). On the other hand, reducing 
vigilance is sure to become costly if a predator approaches the prey without it noticing the threat 
early enough (Lima & Bednekoff, 1999b). 
Vigilance of vertebrates has traditionally been measured from the animal’s body position and 
especially from the head-up position (Pulliam, 1973; Beauchamp, 2015). This approach as-
sumes that the animal relies primarily on visual information for anti-predator behaviour. How-
ever, especially animals with laterally placed eyes might engage in vigilance even in head-down 
position, for example while foraging in terrain without visual obstacles (Lima & Bednekoff, 
 3 
 
1999a; Fernández-Juricic et al., 2011). In addition to visual monitoring, animals may also in-
crease vigilance in response to acoustic stimuli, underlining the importance of auditory infor-
mation in escape behaviour (Bhardwaj et al., 2022). Moreover, different patterns of head move-
ment have been linked to responses to varying types of threat (Jones et al., 2007). Vigilance has 
also been identified to occur either as a routine-like or stimulus-induced behaviour. For exam-
ple, routine behaviour has been studied by assessing whether a foraging animal continues chew-
ing during vigilance and whether its gaze direction changes during the vigilance bout 
(Blanchard & Fritz, 2007; McDougall & Ruckstuhl, 2018).  
1.2.1 Presence of predators 
According to the Landscape of Fear concept, the presence of predators can influence the vigi-
lance levels of prey, as animals must remain alert for potential threats while foraging (Brown, 
1999). Building on this, the Risk Allocation Hypothesis suggests that prey species are expected 
to show behavioural plasticity in response to the spatial and temporal variation in predation 
risk, increasing their vigilance when the perceived risk is higher (Brown et al., 1999; Lima & 
Bednekoff, 1999b). Predation risk can vary depending on the season, time of day or predator 
activity patterns. Moon phases may also influence this risk, as brighter nights during a full moon 
can enhance visual detection for both prey and predator (Lashley et al., 2014).  
1.2.2 Group size effect and alarm signals 
For prey species, gathering into groups while foraging can be beneficial for avoiding potential 
predator attack. In many taxa, studies have found evidence for a group size effect in vigilance 
behaviour, possibly through a so called Many Eyes Effect (Pulliam, 1973; Lima & Dill, 1990). 
In larger groups, individuals might rely on collective vigilance among their group members, 
and therefore reduce their own vigilance (Roberts, 1996; Lashley et al., 2014; Fattorini & Fer-
retti, 2019). Trusting in shared vigilance frees more time for other tasks, but might also increase 
the risk of not being able to detect the threat early enough. 
Alarm calls and signals play a significant role in vigilance and escape behaviour, both by alert-
ing conspecifics and by influencing predator–prey dynamics (Beauchamp, 2015). For example, 
McDougall & Ruckstuhl (2018) found that vigilance was more contagious among group mem-
bers in bighorn sheep (Ovis canadensis) if the gaze was fixed and chewing stopped during a 
vigilance bout. This behaviour was thought to indicate nearby individuals that the vigilance was 
 4 
 
triggered by an external stimulus rather than a routine scanning. Similarly, the tail flagging of 
white-tailed deer has been suggested to be social cue of potential threat to group members 
(LaGory, 1987).  
The evolutionary origins of alarm signals remain debated (Beauchamp, 2015). It is unclear 
whether such signals primarily evolved to warn conspecifics or to function as pursuit-deterrent 
signals directed at the predator, indicating that it has been detected and that an attack would 
therefore be less likely to succeed (Hasson, 1991; Beauchamp, 2015). In this way, prey can 
assess predation risk for a longer time without escaping immediately. For instance, roe deer 
(Capreolus capreolus) are thought to produce barking sounds (Reby et al., 1999) and white-
tailed deer (Odocoileus virginianus) use snorting signals (LaGory, 1987) in response to detect-
ing predators. Although these signals are generally low-cost to produce, they can become costly 
if they draw unwanted attention of a predator and potentially increase predation risk (Beau-
champ, 2015). However, when one individual signals, it can elevate vigilance in nearby group 
members, making collective vigilance more effective under acute threat.  
1.2.3 Human influence and disturbance 
As mentioned earlier, indirect effects and vigilance behaviour are much less studied under im-
mediate human influence than in more natural settings. Although some animals might benefit 
from foraging near humans due to artificial food sources and possibly reduced predation risk, 
studies have also shown humans being able to cause spatial and temporal fear for animals 
(Sönnichsen et al., 2013; Suraci et al., 2019; Zanette & Clinchy, 2025). Essentially, humans 
impact large carnivores’ habitat selection, hunting behaviour, and activity, which is reflected 
upon the behaviour of the prey (Oriol-Cotterill et al., 2015; Johnson-Bice et al., 2023). Evidence 
is also found for prey animals responding to human-posed risk differently compared to risk 
from natural predators. For example, roe deer have been shown to increase their vigilance levels 
during the hunting season, as well as in daylight hours, when human hunters were active 
(Sönnichsen et al., 2013). Also, red deer (Cervus elaphus) (Proudman et al., 2020) and elk 
(Cervus canadensis) (Ganz et al., 2024), have been observed to avoid humans during daytime. 
1.2.4 Other environmental and ecological factors 
Many environmental factors, such as habitat structure and vegetation composition, often inter-
act with predation risk by shaping the behaviour of prey species (Beauchamp, 2015). Also, food 
 5 
 
availability and quality play a role in the energy allocation of prey, and how much it can invest 
in anti-predator behaviour. If food is scarce, animals may have to prioritise foraging over vigi-
lance (Brown, 1999). The influence of foraging distance to cover on vigilance behaviour is 
debated and many conflicting results have been found across studies (Elgar, 1989). Different 
flight tactics can also influence habitat choice and therefore vigilance levels under higher pre-
dation risk (Lima, 1992; Bonnot et al., 2017; Dellinger et al., 2019). For example, human-made 
‘easy’ terrains, like roads or fields, can influence habitat choice of prey under the Landscape of 
Fear due to faster possibilities of escape (Johnson-Bice et al., 2023). On an open pasture, deer 
might be able to detect a predator earlier than in the forest with denser vegetation. Therefore, 
prey can be found less alert in open areas (LaGory, 1987). In dense vegetation prey species 
might have better chances of hiding, but also higher chance of losing sight of the predator. 
Similarly, structural features of habitat, such as edge density, can also shape perceived risk. 
Recent experimental work found that white-tailed deer showed significantly higher vigilance at 
high-density habitat edges compared to low-density ones, regardless of whether the patches 
were connected or isolated (Bartel et al., 2025). This suggests that not only vegetation structure, 
but also the spatial configuration of habitat boundaries can influence antipredator behaviour.  
Additionally, other social factors than group size and alarm signals have been discovered to 
influence vigilance behaviour. The presence of another species in a group may reduce individual 
vigilance through shared scanning and improved predator detection, especially when conspe-
cific numbers are low (Fitzgibbon, 1990b; Bartos et al., 2002). However, in larger conspecific 
groups, vigilance may be neutral or even increase vigilance depending on the specific species 
interactions (Fitzgibbon, 1990b). Vigilance may also vary among individuals within conspecific 
groups, depending on group composition. For example, female white-tailed deer have shown 
increased vigilance in the presence of fawns (Lashley et al., 2014), and when males are present 
in the same group (Cherry et al., 2015). 
1.3 Predator–prey context and the study species 
The large carnivore‒ungulate system in Southern Finland consists primarily of wolves (Canis 
lupus), Eurasian lynx (Lynx lynx), and several cervid species, including e.g. moose (Alces al-
ces), roe deer (Capreolus capreolus), and the introduced white-tailed deer (Odocoileus virgin-
ianus). Of these, roe deer and white-tailed deer are the smallest and most abundant cervids in 
the southern parts of the country, and are important prey for large carnivores. Brown bears 
 6 
 
(Ursus arctos) and wolverines (Gulo gulo) are also native large carnivores, but are rarely ob-
served in South-West Finland (Heikkinen et al., 2024; Mäntyniemi et al., 2025). The red fox 
(Vulpes vulpes) acts as a mesopredator in this system, and may pose a predation risk particularly 
to fawns of deer species (Jarnemo, 2004).  
Historically, wolves have been concentrated in Eastern Finland where they have coexisted pri-
marily with moose and wild forest reindeer (Rangifer tarandus fennicus). As a result of conser-
vation efforts, the wolf population has expanded westward into South-West Finland, a region 
which had previously lacked permanent wolf presence, but has long supported high densities 
of small cervids. In the 2024 population assessment, wolf territory densities were found to be 
highest in South-West Finland (Valtonen et al., 2024). In contrast, the Eurasian lynx has main-
tained relatively stable distribution across Southern Finland (Herrero et al., 2024) and is con-
sidered an established predator of small cervids. Although, the consumptive effects of lynx on 
the cervid populations have not been studied.  
1.3.1 White-tailed deer 
The white-tailed deer is medium sized (females 60–90 kg and males 90–150 kg) cervid species. 
It is a non-native species in Finland, but it has a status of game species in the Finnish legislation. 
It was intentionally introduced from the United States as a new game animal in 1934 and is 
therefore not classified as an invasive species (Kekkonen et al., 2012). During the first intro-
duction, only four females and one male individual survived the journey from Minnesota to 
Finland. The animals were initially managed as livestock, but after a tree fell on the enclosure 
fence and some individuals escaped, the rest were intentionally released into the wild in 1938. 
After that, new attempts were made to introduce more individuals to Finland to avoid inbreed-
ing among the first four animals, but these attempts were unsuccessful. However, the deer pop-
ulation began to increase gradually, and today the species is widely distributed across Finland, 
with the highest densities found in South-West Finland. In winter 2023, the population was 
estimated to be approximately 120 000 individuals (Aikio & Pusenius, 2023).  
The home range of white-tailed deer depends on population density, vegetation cover, and food 
sources, but generally males inhabit larger ranges than females (Poutanen et al., 2022). Females 
reach maturity around the age of one year and they can have one to three fawns, the earliest in 
late May and the latest by the first weeks of July. The dispersal of fawns has been studied in 
their native areas in North America, but limited data exists for Finland. Some studies have 
 7 
 
documented dispersal in spring of their first year (Poutanen et al., 2022). White-tailed deer 
mainly inhabits mixed forests near agricultural areas. It can utilize many types of vegetation in 
its diet and it frequently forages on a variety of crop species in fields. During winter, it shifts to 
spruce-dominated forests where snow cover is lighter and it is easier to forage dwarf shrubs, 
like bilberries. Supplementary feeding of deer by hunters is common in Finland during winter 
(Poutanen et al., 2023). However, the natural diet of white-tailed deer has not been fully studied 
in Finland. Increased white-tailed deer numbers have led to a significant rise in traffic accidents, 
particularly in South-West Finland. Additionally, damage to agriculture and forestry have been 
increasing rapidly, which has raised economic concerns of the negative impacts of this species 
(Matala et al., 2021). 
1.3.2 Roe deer 
The roe deer is a small cervid (20–30kg) native to Finland. It disappeared in the 16th century 
due overhunting, but was later reintroduced to Åland and Southern Finland in the 20th century. 
After that, its population started to grow, and finally became established. Nowadays, roe deer 
carry a legislative status as a game species like white-tailed deer, but there is not as much de-
tailed information available about its density. However, most roe deer sightings have been re-
ported in South-West Finland (Finnish Biodiversity Information Facility, 2025).  
The roe deer is a well-adapted species which originally inhabited deciduous forest, but nowa-
days it mostly occurs in agricultural areas where it effectively utilizes food sources of crop 
lands, and enjoys the predator release around human settlements (Polameri, 2007). Roe deer 
also successfully inhabit coniferous forests which can provide cover and more food during win-
ter. Males are highly territorial, especially during summer. The more food is available, the 
denser a population can be. Females reach maturity at the age of one year and the reproductive 
capacity remains throughout their life. Roe deer can have one to three, sometimes four, fawns 
every spring around May or June. Fawns can live with their mother until the next spring. 
1.3.3 Wolves in Europe and in Finland  
Like other large carnivores worldwide, wolves were driven to regional extinctions across Eu-
rope during the 19th and 20th centuries (Chapron et al., 2014; Ripple et al., 2014). Their recovery 
began when all Europe’s large carnivores received a conservation status of strictly protected 
species in EU's Habitat Directive (Council Directive 92/43/EEC), signed in 1992. Since then, 
 8 
 
wolves have been particularly successful in recolonising Europe and adapting into human-dom-
inated landscapes, although they continue to be widely persecuted (Chapron et al., 2014; Kui-
jper et al., 2024).  
Cimatti et al. (2021) identified several drivers facilitating the expansion of large carnivores in 
Europe, including changes in EU legislation, protected area coverage, reintroduction pro-
grammes, and rural land abandonment. Their findings suggest that suitable habitats for large 
carnivores seem to be areas with increased forest cover and decreased human population. How-
ever, they claim that areas with high human densities can also have stable large carnivore pop-
ulations, if landscape heterogeneity remains high. Due to the recolonisation of wolves, there is 
a lack of knowledge on the extent of the effects that wolves have on other species and even on 
entire ecosystems, especially in human dominated landscapes (Gerber et al., 2024; Kuijper et 
al., 2024). 
In Finland, the 19th century trend of declining wolf population was similar to what has been 
recorded elsewhere in Europe, and the population was nearly hunted to extinction. After the 
Habitat Directive was signed, the wolf population slowly re-established, but declined again 
between 2009‒2013. Since 2014, the population began to increase once more. In March 2024, 
the Finnish wolf population was estimated to consist of 277–321 individuals, inhabiting a total 
of 42 wolf pack territories (Valtonen et al., 2024). All these wolf territories located below the 
northern Finnish reindeer husbandry area, which covers nearly 36% of Finland’s surface area. 
The Finnish wolf population is divided into eastern and western subpopulation (Mäntyniemi et 
al., 2022; Valtonen et al., 2024). This has raised a concern of the viability of the western popu-
lation as it is considerably isolated from other populations, while the eastern population has 
gene flow through dispersing individuals from the Russian wolf populations.  
At the time of conducting this thesis in summer 2024, the hunting of wolves for the purpose of 
population management was not possible in Finland under either national or EU legislation. 
However, in March 2025, the legal status of the wolf changed (Council of the EU, 2025). Fol-
lowing a decision by the Standing Committee of the Bern Convention, the wolf was down-
graded from Appendix II (strictly protected) to Appendix III (protected). This change was trans-
posed into EU law through Directive 2025/1237, which removed the wolf from Annex IV (An-
imal and plant species of community interest in need of strict protection) and included it in 
Annex V (species of community interest whose taking in the wild and exploitation may be 
subject to management measures) under the Habitats Directive. This legal change allows EU 
 9 
 
Member States to apply more flexible population management measures for wolves under Ar-
ticle 14 of the Directive. 
Despite the overall protection status and the recent changes to it, the EU’s Habitats Directive 
had already listed wolf populations in Finland’s reindeer management area in Annex V. This 
exception has loosened their protection status within that specific area, allowing the removal of 
individuals posing a risk to the reindeer industry (Council Directive 92/43/EEC). As a result, 
large carnivores found in Northern Finland are annually removed with derogation permits. 
1.3.4 Predator impacts and knowledge gaps 
Vigilance behaviour has been explored in various carnivore–ungulate systems, including wolf–
prey interactions, which fall within the scope of this thesis (Laundré et al., 2001; Dellinger et 
al., 2019; Gerber et al., 2024). However, while both roe deer and white-tailed deer serve as prey 
for large carnivores in Southern Finland, the quantitative impacts of large carnivores on small 
cervid populations remain to be investigated in the system, where populations live in a human-
dominated landscape and are primarily managed through hunting.  
The westward recolonisation of wolf has been linked to the growing white-tailed deer popula-
tion, which is suggested to be one of the drivers behind the return of wolves to the region 
(Matala et al., 2021). While wolf research in boreal forests has traditionally focused on larger 
herbivores, such as moose and wild forest reindeer, evidence from other parts of Europe indi-
cates that smaller cervids also play a role in wolf ecology. For example, research in the Western 
European Alps demonstrates that high densities of both red deer and roe deer are crucial deter-
minants for wolf habitat selection during their establishment phase (Roder et al., 2020). Alt-
hough red deer were identified as the most energetically profitable prey, roe deer density also 
showed a positive association with wolf occurrence. Similar results have been found in study 
of wolf populations in Sweden and Norway, where roe deer was recognised as the main prey 
for some wolf packs, and increasing prey density has been linked to wolves maintaining smaller 
home ranges (Mattisson et al., 2013). Furthermore, due to this recolonisation, wolves represent 
a relatively new predator for small cervids in South-West Finland, and how deer respond to 
their presence and predation pressure is still unclear. Understanding how small cervids respond 
behaviourally to this shift in predator pressure is central to this study. 
 10 
 
1.4 Main objectives and research questions  
The main aim of my thesis was to examine the vigilance behaviour of both white-tailed deer 
and roe deer in relation to the presence of grey wolves in South-West Finland, by comparing 
observations within and outside wolf territories. I hypothesised that deer inhabiting areas with 
higher predation risk would exhibit increased vigilance as an adaptive response to the potential 
threat posed by predators (Lima & Bednekoff, 1999b). Since it turned out in the data collection 
phase that very few roe deer were observed in the wolf territories, the main research question 
was eventually addressed with data on the white-tailed deer only, while the rest of the analyses 
were conducted on both species.  
In addition to predation risk, I examined how ecological and environmental factors, including 
group size, distance from the forest edge, and time of day influenced the vigilance behaviour in 
foraging deer in agricultural fields. I predicted that individuals in larger groups would exhibit 
lower vigilance frequency and proportion due to collective detection shared among group mem-
bers (Pulliam, 1973). Additionally, deer further from the forest edge were expected to be less 
vigilant, as open landscapes enhance visibility, allowing for more efficient threat detection in 
shorter and fewer vigilance bouts (Elgar, 1989; Lima, 1992; Dellinger et al., 2019).  
By addressing these research questions, I aim to enhance understanding on how white-tailed 
deer adjusts their vigilance behaviour in response to perceived predation risk of wolf and how 
other environmental factors affect the behaviour of small cervids in a human-dominated land-
scape. 
 11 
 
2 Material and Methods 
2.1 Study design 
Two types of study areas were selected based on locations of wolf territories, to represent nat-
ural variation of predation risk. Four study areas were selected from the core areas of different 
wolf territories (Figure 1) to represent wolf presence due to the higher predation risk on deer. 
Five areas were selected outside of the territories, or at the edge of the illustrated territory 
boundaries (Figure 1) to represent a lower risk of wolf being present in the area. Within each 
study area, there were one or two fields chosen as study sites, depending on area. All fields had 
a forest at least on one side of the field edge, and the minimum size of the field was one hectare. 
The crop on fields was cereals that varied between the sites: the species of cereal was not taken 
into consideration because no published research is available of what type of crop these cervids 
prefer the most in Finland.  
South-West Finland was selected for this study because this region has the highest white-tailed 
deer density in Finland (Aikio & Pusenius, 2023) as well as the densest wolf population in 
Southern Finland (Figure 1; Valtonen et al 2024). Also, roe deer is found very abundant there 
(Finnish Biodiversity Information Facility, 2025). 
 
Figure 1. Study site locations in South-West Finland and the illustrated territory boundaries of the wolf population 
estimated to occur in Finland in March 2024. Areas with high predatory risk of wolf on deer are labelled as H1‒
 12 
 
H4 and locations of the study fields within these high-risk areas are marked with red dots. Low-risk study areas 
are labelled as L1‒L5 and the study fields within these are marked with blue dots. High-risk sites are chosen 
particularly inside of the territories and low-risk sites outside of them or at the edge of the illustrated territory 
boundary, in order to estimate wolf presence in study area. CRS: ETRS-TM35FIN (EPSG:3067).  
Two of the high-risk areas were located in the municipality of Urjala, one in Pöytyä, and one in 
Laitila (H1, H2, H3, and H4, respectively; Figure 1). In Urjala, area H1 was located within the 
territory of the Humppila wolf family pack and area H2 inside of the territory of the Toijala 
family pack (Valtonen et al 2024). Study area H3 in Pöytyä was located within the family pack 
territory of Vaskijärvi and the territory around H4 belonged to the family pack of Ihode. There 
were two study fields within each of these territories which in total makes 8 study sites with 
high predation risk. 
The low-risk areas designated as L1, L2, L3, L4, and L5, were located in Turku, Kaarina, Salo, 
Lieto, and Masku, respectively (Figure 1). More specifically, area L1 in Turku was located on 
Ruissalo Island and area L2 in Kaarina on Kuusisto Island. A total of seven study fields were 
chosen across these low-risk areas, with L1 and L2 containing two fields each, and L3, L4, and 
L5 containing one field each.  
As shown in Figure 1, a significant number of wolf territories occur relatively near each other 
in South-West Finland. Due to this high wolf density, identifying control areas entirely free 
from wolf activity was challenging. Therefore, in low-risk areas, the possibility that some 
wolves may have passed through the area or that the observed ungulates might have had previ-
ous encounters with wolves at some point in their life, must be acknowledged. Overall, low-
risk areas represent areas without significant wolf presence and the high-risk areas assume con-
tinuous wolf presence in the territory. 
Other predators, like red fox and Eurasian lynx, were most likely present in both high-risk and 
low-risk areas during the study period. In this thesis the fear effects of other possible predators 
are assumed to be constant regardless of the presence of wolf, because there were no tools to 
exclude or measure their densities. However, according to the 2024 lynx population assessment 
(Herrero et al., 2024), lynx were distributed relatively evenly across South-West Finland, sup-
porting the assumption of consistent lynx presence in the study area.  
 13 
 
2.2 Research equipment 
Data was collected by i) detecting animals from a car with binoculars, ii) observing the animals 
on forest edges via telescope and by taking video recordings of the animals with a phone 
adapter, and iii) by using two types of surveillance cameras.  
All study areas with a camera setup had to have landowners’ permission which limited the lo-
cations of study areas. The camera equipment consisted of four Reolink wildlife cameras (Go 
Ranger PT) and nine surveillance cameras (SD49225XA-HNR-P Dahua PTZ Lite-AI IR 2 Mpx 
25x, 4.8–120 mm). Ruissalo island was the only study site where observations were collected 
only manually by using binoculars and telescope (Figure 2 a), but all the other study fields had 
one set of camera equipment each (Figure 2 b, c). The majority of the observations were col-
lected with the surveillance cameras. 
 
Figure 2. Alongside 10x42 binoculars, the research equipment consisted from a) a telescope with phone adapter, 
b) Reolink trail cameras with solar panel, and c) surveillance cameras attached to metal pole with external battery. 
Solar-powered, 355°-rotating Reolink cameras (Figure 2 b) were used exclusively on study ar-
eas H1 and H2 in Urjala. Reolinks were set to detect animal movement at nighttime from 8 p.m. 
to 7 a.m. From each movement detection the video turned on and kept recording for 30 seconds. 
This meant that some records lasted only 30 seconds and some several minutes. Reolink cam-
eras accessed internet via a SIM-card so it was possible to remotely turn the camera frame 
through the Reolink -application 24/7. Video recordings were saved on 128GB SD-cards.  
 14 
 
Other surveillance camera setups (Figure 2 c) were made for ecological research purpose by 
Airaksinen et al. (2024) when they conducted research of damage to fisheries caused by cor-
morants (Phalacrocorax carbo) and grey herons (Ardea cinerea). In the setup, cameras were 
placed on 2-meter-tall metal poles with three metal wires attached on concrete tiles which kept 
the camera pole steady even in harsh weather conditions. In this study one set of camera equip-
ment was placed on the side of each study field on areas H3–H4 and L2–L5. Cameras had an 
external battery (Tab HD 12V 100Ah LiFePO4 BT) which had to be changed about once a 
week. Each camera was equipped with an IP-protected industrial cellular router 
(RUT240/RUT241), which could access the internet via SIM-cards. Cameras rotated 365° and 
could be remotely controlled from Dahua Config Tool. These surveillance cameras recorded 
non-stop video material between a set time frame, from 8 p.m. to 7 a.m. Material was sent in 
60-minute video clips from the router to the FTP server every hour throughout the night. The 
recordings could be downloaded into an external data storage drive from the FTP server by 
using the FileZilla application. 
2.3 Data collection 
Data was collected by using the focal animal sampling method (Altmann, 1974) with a ca. five-
minute observation time. Only adult individuals were selected as samples and every observation 
bout that lasted a minimum of two minutes were included in the data set. 
At study sites, where data were collected by camera equipment, each animal visible in the video 
frame was selected as soon as it started to forage. If there were many deer present, individuals 
were chosen so that all group members were observed within the same time slot. However, in 
Ruissalo, where data were collected on-site, focal animals were selected randomly, either the 
first animal detected or the one most easily visible. When all group members were clearly vis-
ible, observations proceeded systematically (e.g., from left to right) to ensure each individual 
was recorded once. The wind direction and distance between observer and the animals was 
considered, to avoid disturbing the foraging deer. Additionally, ten minutes were waited on site 
before the start of the first observation bout, to make sure that the deer were not paying attention 
to the observer. 
In this study, vigilance behaviour was used as an indicator of possible indirect fear effects 
caused by wolf presence (Blanchard & Fritz, 2007; Sönnichsen et al., 2013; McDougall & 
Ruckstuhl, 2018). A detected animal was recorded as ‘vigilant’ if its head was above shoulder 
 15 
 
level (Figure 3 c), and ‘not vigilant’ if the head was at the shoulder level or below that (Figure 
a, b) (Fitzgibbon, 1990a). The total vigilance duration during observation bout was timed and 
the frequency of head-up positions were counted. 
When deer search food or move around while foraging, the head is usually at shoulder level 
(Figure 3 b). During these search walks, an animal was counted as ‘not vigilant’. Also, if the 
animal was feeding in long vegetation or if it scratched its back (head being above shoulder 
level), the behaviour was counted as ‘not vigilant’. If the animal laid down during the observa-
tion bout, it was counted as ‘not vigilant’ because most of the time the vegetation was too tall 
to tell whether the behaviour of laying deer was calm or vigilant.  
 
Figure 3 Most common head positions of foraging deer during observation bout. Head-down events can be seen 
in pictures a and b. Picture a) shows the foraging position where the head is clearly down below the shoulder level. 
When a deer searched a new food patch, its head was usually in a neutral, horizontal position (b), and this behaviour 
was counted as ‘not vigilant’, as well as the foraging position. Animal was counted as ‘vigilant’ if the head was 
above the shoulder level (c). These pictures of roe deer are taken with phone attached to the telescope.  
All observation bouts were timed with the Stopwatch+ application on an iPhone, with records 
exported directly via email to prevent typing errors. The observation bouts began when the deer 
was in the foraging position, so the first lap on stopwatch marked the start of the first vigilance 
bout. Timing ended after five minutes, unless interrupted by events such as the animal leaving 
view, video cuts, or human disturbance. In cases when the animal’s head was in an upward 
position at the five-minute timestamp, recording was continued until the head was returned 
down in the foraging position. All video material where humans were detected was removed, 
and no vehicles were detected in any recordings. 
The group size was defined as the number of deer visible in the beginning of the observation 
bout in the same field or video frame. It was recorded at the start because some group members 
 16 
 
might have been laying down when observations began or could have left the field or video 
frame during the observation bout. The distance of the focal animal to the nearest forest edge 
was also recorded in the beginning of each observation bout, since deer were usually moving 
while foraging. The distance from the nearest forest edge was a categorised estimate, based on 
the known diameter of the study field and the estimated standing location of the focal animal. 
Distance from the forest edge was divided into three categories (1 = <20 m, 2 = 20‒50 m, 3 = 
>50 m).  
Data were collected between 9th of May and 20th of June 2024. For analysis, the collection time 
of each observation bout was categorised as follows; sunset (8 p.m.‒12 a.m.), night (1 a.m.‒4 
a.m.), and sunset (5 a.m.‒9 a.m.). All cameras had a night vision feature, which worked well as 
the nights did not get too dark in midsummer due to the unique light conditions at Finland’s 
latitudes. Additionally, I requested and was granted data of weather conditions from Natural 
Resources Institute Finland; however these data were not included in the final models. 
2.3.1 Identifying study species  
Study species were identified from the video records by recognising their species-specific char-
acteristics. White-tailed deer are larger than roe deer and they have a longer tail which is white 
underneath and dark brown on top (Figure 4). In contrast, roe deer are smaller and their tail is 
almost invisible, but instead they have a white rump patch which is recognisable especially 
when they are fleeing (Figure 5). Roe deer also have rounder face than white-tailed deer. If the 
focal animal was far away, the body size and possible the tail flagging of white-tailed deer 
usually helped in identifying the species from each other.  
 17 
 
 
Figure 4 Two white-tailed deer captured with Reolink -trail camera.  
 
Figure 5 Two roe deer recorded with surveillance camera.  
Individual’s sex on the other hand was usually hard to determine as the pixels on surveillance 
cameras faded the thin antlers in certain frame angles, especially on nighttime records. In unsure 
cases, whether the animal had antlers or not, the sex was not specified. This explains why sam-
ple size of confirmed sex differs from the total sample size. Therefore, sex was not included in 
any of the models.  
© Iina Kari 
© Iina Kari 
 18 
 
3 Statistical Analysis 
3.1 Dataset 
The final dataset contained 83 observations from low-risk areas and 107 observations from 
high-risk areas (Table 1). The low-risk dataset included 62 roe deer and 21 white-tailed deer 
observations, whereas in high-risk areas only three roe deer observations were recorded, com-
pared to 104 white-tailed deer observations. Due to this imbalance, the effects of wolf presence 
could not be analysed for roe deer across different predation risk contexts. Therefore, the first 
analysis focused solely on white-tailed deer, using data from both risk areas to assess the effects 
of wolf presence on vigilance behaviour. Additionally, to examine whether group size differed 
between the two risk areas of wolf presence, a Wilcoxon rank-sum test was conducted. 
Table 1 Number of species recorded from each study area during the study period. 
Study Area Roe Deer White-Tailed Deer 
Low-risk   
L1 40 0 
L2 17 13 
L3 0 3 
L4 0 5 
L5 5 0 
   
High-risk   
H1 0 32 
H2 0 14 
H3 2 15 
H4 1 43 
   
The second set of analyses examined interspecific differences in vigilance behaviour under 
shared ecological and temporal conditions. These comparisons included only observations from 
low-risk areas. Wilcoxon rank-sum test was run to find group size differences between the spe-
cies observations. 
3.2 Data exploration and descriptive statistics 
The distributions of all vigilance-related variables (proportion, frequency and duration) were 
checked for normality, both graphically and statistically. Since the original observation time 
varied, the counted vigilance frequency divided with observation time (s) and multiplied by 60, 
to calculate average vigilance frequency per minute for each observation. Levene’s -test was 
 19 
 
applied to test for homogeneity of variances prior to conducting parametric tests. All outlier 
individuals were cross-checked against field notes to ensure that no external disturbances (e.g., 
sudden human activity) had influenced the recorded behaviour.  
For both the species-specific analysis of white-tailed deer and the interspecific comparison with 
roe deer, the same set of descriptive statistics, including mean, variance (Var), standard devia-
tion (SD), and standard error (SE), were calculated for all vigilance-related variables. Variance 
and SD quantify the degree of individual behavioural variability within each group, while SE 
reflects the reliability of the sample mean estimates.  
For white-tailed deer, observation bout durations ranged from 2 minutes to 7 minutes and 25 
seconds, with a mean duration of 4 minutes and 68 seconds. Descriptive statistics were calcu-
lated separately for all vigilance-related variables in high-risk (wolf present) and low-risk (wolf 
absent) areas. Lower SE values suggest more precise estimates and are important for interpret-
ing differences between risk areas. As the analysis aims to compare vigilance behaviour in dif-
ferent predation risk contexts, these statistics provide a baseline for assessing whether observed 
differences in the model outputs are likely to reflect true biological patterns or sampling uncer-
tainty. For the interspecific comparison, roe deer and white-tailed deer were analysed in absence 
of wolf (low risk areas). Observation durations ranged from 2 to 8 minutes with average of 4 
minutes and 59 seconds in low-risk areas. The same descriptive statistics were computed for 
both species, allowing a direct comparison of vigilance behaviour under shared ecological con-
ditions. 
3.2 Modelling  
Generalized linear mixed models (GLMM) were fitted for vigilance proportion, frequency, and 
average vigilance bout duration using two datasets: one with white-tailed deer observations 
from both risk areas, and another including both species but limited to low-risk areas only. Each 
model included one vigilance-related response variable: 1) vigilance proportion (the proportion 
of time an animal was observed having head raised above shoulder level relative to the total 
duration of the observation bout), 2) vigilance frequency per minute (the number of vigilance 
bouts during the observation bout divided by observation time), or 3) average vigilance bout 
duration (calculated as vigilance proportion divided by total vigilance frequency during the 
observation bout). Vigilance frequency per minute was log-transformed for white-tailed deer 
models, using the natural logarithm (log(x)) to reduce right skewness in the data (Salinas Ruíz 
 20 
 
et al., 2023). This transformation improved the distributional properties of this response varia-
ble, allowing the use of a gaussian distribution in the model. For interspecific model on vigi-
lance frequency, the logarithmic transformation was not needed as data was normally distrib-
uted on low-risk areas. This analysis was only done in low-risk area due to lack of data in high-
risk area. 
Study field (nested within study area) and the group ID-number were entered as random effects, 
to account for the non-independent observations within specific areas, and group-level depend-
encies as individuals within the same group might share common influence between the group 
members. Wolf presence in the study area (low risk, high risk) was a fixed effect in the models 
for white-tailed deer. Group size, observation month (May, June), and categorial variables such 
as time of day (sunset, nigh, sunrise), and the distance from the forest edge (<20 m, 20‒50 m, 
>50 m) were included as fixed explanatory variables, depending on model. Data from low-risk 
areas included only two observations of furthest forest distance category (>50 m). Those obser-
vations were pooled with the intermediate distance category (20–50 m) in interspecific model-
ling, to avoid problems in model fitting. Similarly, roe deer observations included only a few 
sunrise observations, which were subsequently pooled with night observations. 
Modelling was conducted with the RStudio (4.1.2) glmmTMB-package (v. 1.1.7, Brooks et al., 
2017), in which hierarchical nesting of dependent variables was possible. A backwards selec-
tion approach was used in choosing the best model fits (Zuur et al., 2007). In this approach, all 
variables were included in the first model. Then, covariates were reduced one at time, until the 
lowest AIC-value was found. 
 
 21 
 
4 Results 
4.1. Vigilance behaviour of white-tailed deer in the presence of wolf 
Descriptive statistics of white-tailed deer show higher variation in all measured behavioural 
variables in high-risk areas (Table 2). However, the number of observations in high-risk areas 
were 79,81% higher compared to number of observations collected from low-risk areas. The 
total number of observations (N) of white-tailed deer was 125. However, some observations on 
low-risk areas lacked data on some variables, so sample size (n) in all of the following models 
were 120.  
Table 2 Descriptive statistics for white-tailed deer behaviour in low predation risk and high predation risk areas. 
N = Sample size, VAR = variance, SD = standard deviation, SE = standard error. 
Behaviour Variable Risk Area N Min Max Mean Var SD SE 
Vigilance Proportion (%)        
 Low-risk 21 7.8 47.9 25.7 154.0 12.4 2.7 
High-risk 104 <0.1 92.9 22.0 366.0 19.1 1.9 
         
Vigilance Frequency per minute       
 Low-risk 21 1.3 5.0 2.6 0.7 0.9 0.2 
 High-risk 104 0.2 6.5 2.1 1.7 1.3 0.1 
         
Average Vigilance Bout Duration (s)        
 Low-risk 21 3.0 16.0 6.8 12.1 3.5 0.8 
High-risk 104 1.0 68.0 8.7 128.0 11.3 1.1 
 
The Wilcoxon rank-sum test did not reveal a statistically significant difference (W = 864, p = 
0.111) in group sizes between risk areas, which indicates that the group size was relatively 
consistent across the study sites and unlikely to influence the vigilance models (Figure 6). 
 22 
 
 
Figure 6 The group sizes in which each white-tailed deer was observed in low-risk (n = 21, mean ± SD = 1.67 ± 
0.91) and high-risk areas (n = 104, mean ± SD = 2.48 ± 1.93). Solitary individuals were recorded as group size 
one. The boxplots illustrate the distribution of group sizes: the horizontal line marks the median, the boxes repre-
sent the interquartile range, and the whiskers extend to the extreme non-outlier values (no whisker is shown for 
the low-risk area due to the narrow data range). Individual observations are overlaid as dots, with points beyond 
the whiskers representing outliers.  
4.1.1 Vigilance proportion 
The proportion of time white-tailed deer spent vigilant while foraging, modelled using a beta 
distribution with a logit link (Salinas Ruíz et al., 2023), was significantly influenced by group 
size and time of day, but not by predation risk or distance from the forest edge (Table 3). Among 
the random effects, individual group-level differences contributed more to variation in vigilance 
proportion than spatial factors. The model included 120 observations after excluding incom-
plete records on predictor data. The sample size divided between risk areas so that 104 obser-
vations were from high-risk areas and 16 from low-risk areas. The predicted vigilance propor-
tion was 23.88% [95% CI: 15.03, 35.75] in low-risk areas and 21.40% [95% CI: 16.49, 27.26] 
in high-risk areas. 
Group size had a significant negative effect on vigilance proportion (Table 3). As shown in 
Figure 7, the proportion of time deer spent vigilant declined with increasing group size. Model 
predicted the group size declining approximately 14% with each group member. Time of day 
also significantly affected vigilance (p <0.05; Table 3). Vigilance was lowest during the night, 
compared to sunset and sunrise (Table 4 a). Pairwise comparisons showed that these differences 
 23 
 
were statistically significant (Table 4 b). Vigilance levels at sunset and sunrise did not differ 
significantly (p >0.05; Table 4 b), suggesting similar levels of alertness during twilight hours. 
Predicted values for vigilance proportion on different time categories are visualised in Figure 
8. 
Table 3 Fixed and random effects from beta regression GLM model explaining vigilance proportion in white-
tailed deer. Table includes chi-square test values, degrees of freedom, p-values, and coefficient estimate for wolf 
presence, group size, distance from forest edge, and time of day. Type III Wald chi-square tests were used to assess 
the significance of fixed effects which included wolf presence, group size, distance to forest and time of day. 
Group size and time of day appeared as statistically significant explanatory variables (p <0.05).  
Vigilance Proportion  Wald χ² Df p-value Estimate SE 
Fixed factors 
   
  
Intercept 7.87 1 0.005 -0.79 0.28 
Wolf Presence 0.29 1 0.587 0.14 0.26 
   (high risk, low risk)      
Group Size 4.21 1 0.040 -0.15 0.07 
Distance to Forest 0.38 2 0.826 - - 
   (<20 m, 20‒50 m, >50 m)      
Time of Day 7.46 2 0.024 - - 
   (sunset, night, sunrise)      
      
Random effects Variance Std. Dev. 
 
  
Study Area  <0.01 <0.01 
 
  
Field (nested within study areas) <0.01 <0.01    
Group ID  0.21 0.45     
Model fitted with beta distribution (logit link). Estimates are calculated with low risk as the reference category 
in the wolf presence. 
     
Table 4 Time of day effects on vigilance proportion with pairwise comparisons. (a) presents estimated marginal 
means, standard errors, 95% confidence intervals, and (b) Tukey-adjusted p-values comparing vigilance propor-
tions between sunset, night, and sunrise. 
Time of Day Estimated model mean SE 95% CI  
a)     
Sunset (8 p.m.–12 a.m.) -1.146 0.17 [-1.49, -0.81]  
Night (1 a.m.–4 a.m.) -1.689 0.30 [-2.21, -1.17]  
Sunrise (4 a.m.–9 a.m.) -0.857 0.33 [-1.50, -0.22]  
     
b)     
Comparison Estimate (log odds) SE z-ratio p-value 
Sunset–night 0.543 0.23 2.371 0.047 
Sunset–sunrise -0.290 0.31 -0.937 0.617 
Night–sunrise -0.832 0.35 -2.362 0.048 
Results averaged over Distance from Forest and Wolf Presence. Tukey-adjusted p-values. Estimates given on 
the logit scale. 
 
 24 
 
 
Figure 7 Effect of group size on the proportion of time white-tailed deer spent vigilant. Each point represents one 
observation bout of ca. 5 minutes. To improve visualization of overlapping data points, a small amount of random 
jitter was applied to the x-axis positions. The solid line shows modelled predictions with 95% confidence intervals, 
based on a beta regression model. All modelled values are backtransformed on the original (proportion) scale. 
 
Figure 8 Predicted proportion of time white-tailed deer spends vigilant at different times of day. Modelled means 
and 95% confidence intervals are shown on the backtransformed scale. Sample size (n) shows how data is distrib-
uted across categories. Vigilance was lowest during the night compared to twilight periods. Letters above the 
confidence intervals indicate statistically significant groupings, showing that sunset and sunrise (A) did not differ 
significantly from each other (p >0.05), while night (B) differed significantly from both (p <0.05). 
 25 
 
4.1.2 Vigilance frequency 
Vigilance frequency per minute (log-transformed) was analysed using gaussian distribution in 
GLM model (Salinas Ruíz et al., 2023). Vigilance frequency in white-tailed deer was signifi-
cantly affected by group size, time of day, and observation month (Table 5). A marginal differ-
ence indicated vigilance frequency being lower in high-risk areas compared to low-risk areas, 
but the difference did not reach statistical significance (p >0.05; Table 5). Predicted vigilance 
frequency was 2.5/minute [95% CI: 1.62, 3.74] in low-risk areas and 1.6/minute [95% CI: 1.21, 
2.04] in high-risk areas. Group size had a significant negative effect on vigilance frequency 
with fewer vigilance bouts observed in larger groups (Table 5). According to the model, each 
additional group member was associated with an approximately 8% decrease in vigilance fre-
quency. The modelled decline on vigilance frequency is presented in Figure 9.  
Time of day also significantly influenced vigilance frequency (p <0.05; Table 5). The frequency 
was lowest at nighttime and higher both during sunset and sunrise (Table 6 a). However, pair-
wise comparisons revealed that the differences were marginally non-significant between the 
time categories, in which sunset compared to night, and night compared to sunset differed more 
from each other, while vigilance frequency was quite similar between twilight hours (Table 6 
b). Results are shown in Figure 10 with original scale of the count data adjusted per minute. 
There was also a significant (p <0.05; Table 5) effect of month on vigilance frequency, in which 
the frequency was higher in June compared to May, as presented in Figure 11 with backtrans-
formed values. Study field explained a very small amount of variance among the random ef-
fects, while Study area and Group ID contributed none. Residual variance represents the amount 
of variation that is left unexplained by fixed and random effects. 
 
 
 
 
 
 
 26 
 
Table 5 Fixed and random effects from a generalized linear mixed model (GLMM) explaining vigilance frequency 
per minute in white-tailed deer. The response variable was log-transformed to improve normality and meet model 
assumptions, as the original frequency data were positively skewed. Fixed effects included group size, wolf pres-
ence, forest distance, time of day, and observation month. Random effects included study area, nested study field, 
group size, and residuals representing variation which is not explained by fixed or random effects. Model was 
based on 120 observations (high-risk, 104; low-risk, 16). 
Vigilance Frequency (/min) Wald χ² Df p-value Estimate SE 
Fixed factors 
   
  
Intercept 5.68 1 0.017 0.53 0.22 
Wolf Presence 3.66 1 0.056 0.45 0.23 
   (high risk, low risk)      
Group Size 4.64 1 0.031 -0.08 0.04 
Distance from Forest 1.13 2 0.567 - - 
   (<20 m, 20‒50 m, >50 m)      
Time of Day 7.43 2 0.024 - - 
   (sunset, night, sunrise)      
Observation Month 4.86 1 0.027 0.41 0.19 
   (May, June)      
      
Random effects Variance Std. Dev. 
 
  
Study Area  <0.01 <0.01 
 
  
Field (nested within study areas) 0.04 0.20    
Group ID  <0.01 <0.01     
Residual 0.31 0.56    
Model fitted with gaussian distribution (identity link). Estimates are presented on the log scale, and are calculated 
with low risk as the reference category in the wolf presence and June in observation month. 
    
Table 6 Model means and pairwise comparisons of log-transformed vigilance frequency across time of day cate-
gories (sunset, night, sunrise). Vigilance was highest at sunrise and lowest at nighttime. None of the pairwise 
comparisons reached statistical significance (p >0.05), although two comparisons approached significance with 
marginally higher vigilance frequency at sunset compared to night, and marginally lower frequency at night com-
pared to sunrise. 
Time of Day  Estimated model mean SE 95% CI  
a)     
Sunset (8 p.m.–12 a.m.) 0.747 0.11 [0.52, 0.97]  
Night (1 a.m.–4 a.m.) 0.399 0.17 [0.06, 0.73]  
Sunrise (4 a.m.–9 a.m.) 0.879 0.21 [0.46, 1.30]  
     
b)     
Comparison Estimate (log odds) SE t-ratio p-value 
Sunset–night 0.348 0.15 2.337 0.055 
Sunset–sunrise -0.132 0.19 -0.690 0.770 
Night–sunrise -0.480 0.21 -2.318 0.058 
Results averaged over Wolf Presence, Distance from Forest, and Observation Month. Tukey-adjusted p-val-
ues. Estimates are given on log scale. 
 
 27 
 
 
Figure 9 Modelled relationship between group size and vigilance frequency (/min) in white-tailed deer. Dots rep-
resent frequency values adjusted to observation bouts (n = 120), with slight x-axis jitter applied to reduce overlap. 
The curve and shaded area represent modelled predictions and 95% confidence intervals from a gaussian model, 
presented on the backtransformed scale. According to the model, vigilance frequency decreases when group size 
increases. 
 
Figure 10 Predicted frequency of vigilance bouts per minute in white tailed deer on different times of day. Mod-
elled mean estimates and 95% confidence intervals are displayed on backtransformed scale. Vigilance frequency 
was lowest during night and higher during twilight hours. Sample size (n) was highest on sunset observations and 
lowest at sunrise. 
 28 
 
 
Figure 11 Predicted vigilance frequency per minute in May and June on the original count scale. Mean estimates 
show lower vigilance frequency in May compared to June. Means are represented with 95% confidence intervals. 
Sample size (n) in May was 44.5% smaller than in June.  
4.1.3 Average vigilance bout duration 
Average duration of vigilance bouts, modelled with a gamma distribution (log link) (Salinas 
Ruíz et al., 2023), was significantly influenced by time of day, while no significant effects were 
found in wolf presence, group size, distance from forest edge, or observation month (Table 7). 
Random variation was mainly explained by group identity, while spatial factors had no effects. 
This model was also based on the subset of 120 observations. Predicted vigilance duration was 
5.6 seconds [95% CI: 3.07, 10.20] in low-risk areas and 8.0 seconds [95% CI: 5.54, 11.46] in 
high-risk areas. 
Vigilance bouts were shortest during nighttime compared to both sunset and sunrise (Table 8 
a). According to the pairwise comparisons, the difference between night and sunrise was statis-
tically significant, showing vigilance bouts being longer during morning twilight (Table 8 b). 
The difference between time categories can be seen on Figure 12. The differences between 
sunset and other times were not significant (p >0.05; Table 8 b).  
 
 29 
 
Table 7 Results of gamma regression model for average vigilance bout duration. Displays fixed effects (wolf 
presence, group size, forest distance, time of day, month) and random effects (group ID, field study area) affecting 
vigilance bout duration. 
Vigilance Bout Duration (s) Wald χ² Df p-value Estimate SE 
Fixed factors 
   
  
Intercept 35.24 1 <0.001 2.18 0.37 
Wolf Presence 1.31 1 0.252 -0.35 0.30 
   (high risk, low risk)      
Group Size 0.15 1 0.697 -0.33 0.08 
Distance from Forest 0.46 2 0.794 - - 
   (<20 m, 20‒50 m, >50 m)      
Time of Day 7.46 2 0.024 - - 
   (sunset, night, sunrise)      
Observation Month 0.14 1 0.707 -0.10 0.26 
   (May, June)      
      
Random effects Variance Std. Dev. 
 
  
Study Area  0.01 0.01 
 
  
Field (nested within study areas) <0.01 <0.01    
Group ID  0.28 0.53     
Model fitted with gamma distribution (log link). Estimates are calculated with low risk as the reference cate-
gory in the wolf presence. 
    
    
Table 8 Vigilance bout duration by time of day with pairwise comparisons. (a) presents mean estimates, confi-
dence intervals, and (b) statistical contrasts showing significantly longer bouts at sunrise compared to night. 
Time of Day Estimated model mean SE 95% CI  
a)     
Sunset (8 p.m.–12 a.m.) 1.88 0.17 [1.55, 2.22]  
Night (1 a.m.–4 a.m.) 1.51 0.25 [1.02,2.00]  
Sunrise (4 a.m.–9 a.m.) 2.30 0.32 [1.67,2.94]  
     
b)     
Comparison Estimate (log odds) SE z-ratio p-value 
Sunset–night 0.37 0.21 1.737 0.192 
Sunset–sunrise -0.42 0.30 -1.414 0.333 
Night–sunrise -0.79 0.30 -2.608 0.025 
Results averaged over Wolf Presence, Distance from Forest and Observation Month. Tukey-adjusted p-val-
ues. Estimates given on the log scale. 
 
 30 
 
 
Figure 12 Average duration of vigilance bouts across times of day in white-tailed deer. Modelled values with 95% 
confidence intervals are shown on the original time scale (seconds). Sample size (n) varied between the time cat-
egories with fewest of observations at sunrise hours. Based on Tukey-adjusted pairwise comparisons, vigilance 
bout durations were significantly (p <0.05) longer at sunrise (A) than night (B), while sunset (AB) did not statis-
tically differ from either (p >0.05).  
 
4.2 Interspecific variation in vigilance behaviour under low predation risk 
The descriptive statistics including both species, white-tailed deer and roe deer, showed that 
roe deer has lower mean at vigilance proportion and frequency. Although, the variation of these 
variables were higher compared to white-tailed deer (Table 9). The average vigilance bout du-
ration was relatively similar in both species. Total sample size (N) was 83, including both spe-
cies from all of the low-risk areas. However, the following models are based on sample size (n) 
77 (white-tailed deer; 16, roe deer; 61), which had data on all included variables. 
 
 
 
 
 31 
 
Table 9 Descriptive statistic of data including both study species, roe deer (cc) and white-tailed deer (ov).  
N = Sample size, VAR = variance, SD = standard deviation, SE = standard error.  
Behaviour Variable Species* N Min Max Mean Var SD SE 
Vigilance Proportion (%)        
 cc 62 0.1 53.7 19.3 179.0 13.4 1.7 
ov 21 7.8 47.9 25.7 154.0 12.4 2.7 
         
Vigilance Frequency per minute       
 cc 62 0.4 4.6 2.0 1.0 1.0 0.1 
 ov 21 1.3 5.0 2.6 0.7 0.8 0.2 
         
Average Vigilance Bout Duration (s)        
 cc 62 2.0 17.0 6.4 11.5 3.4 0.4 
ov 21 3.0 16.0 6.8 12.1 3.5 0.8 
* cc: Capreolus capreolus, ov: Odocoileus virginianus 
 
According to Wilcoxon rank-sum test, group sizes did not differ significantly (W = 503.5, p = 
0.102) between white-tailed deer and roe deer in areas of low predation risk (Figure 13). 
 
Figure 13 The group sizes of white-tailed deer (n = 21, mean ± SD = 1.67 ± 0.91) and roe deer (n = 62, mean ± 
SD = 2.87 ± 2.47) observed in low-risk areas. Animals observed alone were counted as a group size of one. Box-
plots illustrate the spread of group sizes per species, with dots marking individual observations; the central line 
shows the median, boxes indicate the interquartile range, and whiskers extend to the furthest non-outlier points. 
 
 32 
 
4.2.1 Vigilance proportion 
Predicted vigilance proportion was 18.62% [95% CI: 8.63, 35.67] for white-tailed deer (n = 
16), and 23.64% [95% CI: 14.39, 36.33] for roe deer (n = 61). Group size, distance to forest 
edge, or time of day showed no statistically significant main effects or species-specific interac-
tion effects (Table 10). Therefore, results show no interspecific differences between the species. 
Study Field showed highest variance among the random effects, while variance in Group ID 
and Study Area was near zero.  
Table 10 Fixed and random effects from beta regression model of vigilance proportion between roe deer and 
white-tailed deer in low-risk areas. Includes main and interaction terms for species, group size, forest distance, and 
time of day. 
Vigilance Proportion Wald χ² Df p-value Estimate SE 
Fixed factors 
   
  
Intercept 6.13 1 0.013 -1.49 0.60 
Species 0.39 1 0.531 0.34 0.55 
   (roe deer, white-tailed deer)      
Group Size 0.09 1 0.764 -0.06 0.21 
Distance to Forest 2.71 1 0.100 0.61 0.34 
   (<20 m, 20–50 m)      
Time of Day 0.15 1 0.694 0.23 0.59 
   (sunset, night)      
      
Interactions     
Group Size:Species  0.01 1 0.912 0.02 0.21 
Distance to Forest:Species 1.93 1 0.164 -0.61 0.44 
Time of Day:Species 0.42 1 0.518 0.41 0.63 
      
Random effects Variance Std. Dev. 
 
  
Study Area  <0.01 0.73 
 
  
Field (nested within study areas) 0.37 0.61    
Group ID  0.04 0.21     
Model fitted with beta distribution (logit link). Estimates are calculated with following reference categories: 
Species; roe deer, Distance to Forest; 20–50 m, and Time of Day; night. 
     
4.2.2 Vigilance frequency 
A gaussian GLM model was conducted to investigate interspecific differences between white-
tailed deer and roe deer. According to this model, the average vigilance frequency for white-
tailed deer was 2.37 times per minute [95% CI: 1.48, 3.26; n = 16], and for roe deer 2.15 times 
per minute [95% CI: 1.85, 2.45; n = 61]. Model included fixed factors with species-specific 
interactions, and did not reveal any statistically significant predictors or differences in vigilance 
frequency between the species (Table 11). 
 33 
 
Table 11 Modelled interspecific comparisons on vigilance frequency per minute between white-tailed deer and 
roe deer. Fixed effects include species, group size, distance to forest, and time of day, along with the interactions 
with the species. Random effects account for study area, nested study field, and group identity. Residuals show 
variation of how much of the model is not explained by fixed or random effects. No significant effects (p <0.05) 
were found. 
Vigilance Frequency (/min) Wald χ² Df p-value Estimate SE 
Fixed factors 
   
  
Intercept 24.44 1 <0.001 2.79 0.56 
Species 2.17 1 0.141 -0.90 0.61 
   (white-tailed deer, roe deer)      
Group Size 0.12 1 0.730 0.08 0.24 
Distance to Forest 0.16 1 0.693 -0.20 0.50 
   (<20 m, 20–50 m)      
Time of Day 0.32 1 0.573 -0.46 0.82 
   (sunset, night)      
Observation Month 1.11 1 0.292 -0.61 0.57 
   (May, June)      
      
Interactions      
Group Size:Species 0.50 1 0.480 -0.17 0.24 
Distance to Forest:Species 0.18 1 0.674 0.23 0.56 
Time of Day:Species 1.01 1 0.314 0.88 0.87 
Observation Month:Species 3.39 1 0.066 1.14 0.62 
      
Random effects Variance Std. Dev.    
Study Area  <0.01 <0.01    
Field (nested within study areas) <0.01 <0.01    
Group ID  <0.01 <0.01    
Residual (σ²) 0.76 0.87    
Model fitted with gaussian distribution (identity link). Estimates are calculated with following reference catego-
ries: Species; roe deer, Distance to Forest; 20–50 m, Time of Day; night, and Observation Month; June. 
 
4.2.3 Average vigilance bout duration 
The GLM model, fitted with gamma distribution, predicted vigilance bout duration for white-
tailed deer being ~6 seconds [95% CI: 3.67, 9.77; n = 16], and ~7 seconds [95% CI: 5.64, 9.16; 
n = 61] for roe deer. Vigilance duration was significantly influenced by the observation month 
(p <0.05; Table 12) across both species (n = 77). The average bout duration in low-risk areas 
was ~20 seconds in May (95% CI: 8.06, 50.31; n = 42), and ~11 seconds in June (95% CI:  
2.99, 40.65; n = 35). Modelled interaction between species and observation month suggested 
some temporal variation in vigilance patterns between species, but this effect was not statisti-
cally significant (p >0.05). Other predictors did not show significant differences between spe-
cies or consistent effects on vigilance duration (Table 12). 
 34 
 
Table 12 Gamma regression GLM model of average vigilance bout duration comparing roe deer and white-tailed 
deer. Includes main effects and species interactions with ecological variables. Sample size (n) in the model was 
77. 
Vigilance Bout Duration Wald χ² Df p-value Estimate SE 
Fixed factors 
   
  
Intercept 34.56 1 <0.001 2.06 0.35 
Species 0.08 1 0.771 -0.10 0.35 
   (white-tailed deer, roe deer)      
Group Size 0.98 1 0.323  -0.13 0.13 
Distance to Forest 1.62 1 0.204 0.31 0.24 
   (<20 m, 20–50 m)      
Time of Day 2.59 1 0.108 0.73 0.45 
   (sunset, night)      
Observation Month 6.03 1 0.014 -0.88 0.36 
   (June, May)      
      
Interactions      
Group Size:Species 0.99 1 0.320 0.13 0.13 
Distance to Forest:Species 0.36 1 0.551 -0.16 0.27 
Time of Day:Species 1.98 1 0.159 -0.68 0.48 
Observation Month:Species 3.66 1 0.056 0.72 0.38 
      
Random effects Variance Std. Dev.    
Study Area  <0.01 <0.01    
Field (nested within study areas) 0.03 0.18    
Group ID  0.02 0.15    
Model fitted with gamma distribution (log link). Estimates are calculated with following reference categories: 
Species; roe deer, Distance to Forest; 20–50 m, Time of Day; night, and Observation Month; June. 
     
 35 
 
5. Discussion  
5.1 Overview of main findings 
This study investigated the vigilance behaviour of white-tailed deer and roe deer in agricultural 
landscapes of South-West Finland, with a focus on the influence of wolf presence, group size, 
time of day, and foraging distance from the nearest forest edge. The results indicated that wolf 
presence marginally, but non-significantly, influenced vigilance frequency in white-tailed deer 
to the other direction than predicted. Wolf presence did not significantly affect any other vigi-
lance-related variables. Similarly, distance to the nearest forest edge did not have a significant 
effect on vigilance. Instead, time of day emerged as consistent predictor across all vigilance 
metrics in white-tailed deer, with highest vigilance exhibited during twilight periods, especially 
at sunrise. Group size affected vigilance proportion and frequency, so that lower vigilance lev-
els were exhibited in larger groups. Additionally, vigilance frequency in white-tailed deer was 
lower in May compared to June. The interspecific comparisons revealed no significant differ-
ences in vigilance behaviour between the species, although roe deer tended to show slightly 
higher mean vigilance levels across variables. Within both species in low-risk areas the average 
vigilance bout duration was significantly affected by observation month, but it did not highlight 
any interspecific differences. 
5.2 Vigilance responses to predation risk and group-level effects in white-tailed deer 
Contrary to the primary hypothesis of deer increasing their vigilance in the presence of predator 
(Lima & Bednekoff, 1999b), white-tailed deer did not exhibit significantly higher vigilance 
levels in areas with wolf presence (high-risk areas). Although predator presence marginally 
affected vigilance frequency, the effect was in the opposite direction than expected: vigilance 
appeared being lower in high-risk areas. Overall, the lack of a significant difference in vigilance 
levels may reflect a reliance on habitat-based risk management or an insufficient stimulus to 
trigger a measurable behavioural response (Beauchamp, 2015). It should be noted that the sam-
ple sizes were unequal between the risk areas and particularly low in low-risk areas, which 
reduced statistical power. Nevertheless, the available data did not suggest the existence of dif-
ferences. However, as previously mentioned, wolves have only recently recolonised the study 
region and might not yet impose strong enough selection pressure to generate a Landscape of 
Fear that could be detected through altered vigilance levels (Zanette & Clinchy, 2025). Further-
more, the study areas were chosen based on mapped illustrations of wolf territories during the 
 36 
 
preceding winter (Valtonen et al., 2024) rather than confirmed wolf presence or activity during 
the study period, meaning that some observed individuals, particularly in high-risk areas, may 
not have encountered wolves or signs of their presence. This result may also reflect similar 
patterns with the central hypothesis of Dellinger et al. (2019), suggesting that white-tailed deer 
primarily manage predation risk through "fine-scale shifts" in habitat use, rather than broad-
scale avoidance or a generalized increase in vigilance. Given their flight strategy based on early 
detection and sprinting, white-tailed deer may select microhabitats that optimise their escape 
potential rather than investing energy in vigilance. The decision to conduct this study on open 
agricultural fields, which provide flat terrain and usually high visibility, supports this interpre-
tation.  
Group size on the other hand was a significant predictor of both vigilance proportion and fre-
quency in white-tailed deer, supporting the Many Eyes Hypothesis (Pulliam, 1973). Individuals 
in larger groups showed lower vigilance, which indicates of a shared detection benefit. This 
pattern aligns with prior findings across diverse ungulate groups (Fitzgibbon, 1990b; Lashley 
et al., 2014; Fattorini & Ferretti, 2019). It should be noted, that the average group size was 
slightly larger in high-risk areas, which could affect the detected marginal effect of vigilance 
frequency being lower on high-risk areas compared to low-risk areas. Also, results were not 
consistent with the hypothesis that vigilance would decrease in more open areas further from 
cover (Elgar, 1989), as none of the vigilance variables were influenced by foraging distance 
from the forest. Here, the interaction between group size and the predator presence, as well as 
foraging distance from forest and study area would have been interesting to investigate, but my 
data did not allow these additions in the model.   
5.3 Temporal variation in vigilance 
Time of day also significantly influenced vigilance behaviour in white-tailed deer. Vigilance 
was lowest during night compared to twilight periods, in all vigilance-related variables. In-
creased vigilance during sunset and sunrise may reflect risk-sensitive foraging strategies 
adapted to increased predator or human activity. Notably, vigilance bouts were also signifi-
cantly longer at sunrise than at night, suggesting a behavioural mechanism for enhanced risk 
assessment in the shift to a higher risk timeframe. Previous research has shown that wolves are 
most active at dawn and dusk, with increased movement and predation rates during these times 
(Theuerkauf et al., 2003; Theuerkauf, 2009; Ferreiro-Arias et al., 2024). Wolves may also shift 
their behaviour to increased nocturnal activity in response to human presence (Theuerkauf, 
 37 
 
2009; Ferreiro-Arias et al., 2024; Sunde et al., 2024). Similarly, cervids have been observed to 
modify their behaviour and diel patterns in response to the perceived risk posed by humans 
(Sönnichsen et al., 2013; Bonnot et al., 2020; Proudman et al., 2020). Interestingly, Theuerkauf 
et al. (2003) suggest that human influence on wolf activity may be indirect, driven by changes 
in prey behaviour, which in turn could affect wolf movement and hunting strategies.  
Observation month significantly affected vigilance frequency, but not vigilance bout duration 
in white-tailed deer. White-tailed deer tended to have more vigilance bouts in June compared 
to May. This could indicate the need to scan surroundings more often when the vegetation is 
taller and denser, both in the open field and forest edge. Another possible explanation regarding 
higher vigilance frequency in June could be the fawning season (Lashley et al., 2014; Costelloe 
& Rubenstein, 2018; Poutanen et al., 2022). Fawns at this time of year are small, and especially 
white-tailed deer hide their fawns well in the cover of tall vegetation. However, the presence of 
fawns could not be confirmed in this study. One addition could have been to repeat this study 
in autumn when fawns are a bit more independent, which may also affect vigilance dynamics 
within a group (Higdon et al., 2019). 
An important consideration not captured in this study is the potential seasonal variation in both 
predator activity and human disturbance. For instance, wolves are likely to exhibit seasonal 
variation in activity and territory use, which may influence the degree of predation risk per-
ceived by prey species (Lodberg-Holm et al., 2021). Similarly, human hunting pressure during 
the autumn may alter deer vigilance patterns, either by increasing overall alertness or shifts in 
habitat use across the landscape (Benhaiem et al., 2008; Proudman et al., 2020). Conducting 
this study during the hunting season would help reveal how temporal fluctuations in predator 
presence and human disturbance shape vigilance behaviour, and whether deer alter their strat-
egies of risk assessment in response. 
5.4 Interspecific comparison  
In low-risk areas, interspecific comparisons revealed no significant differences between roe 
deer and white-tailed deer across any vigilance-related variables. While roe deer appeared to 
show slightly higher mean vigilance proportion and longer vigilance bouts, these differences 
were not statistically significant. These findings suggest that both species may exhibit similar 
vigilance strategies in areas with lower wolf abundance and under similar ecological conditions 
(Bartos et al., 2002) 
 38 
 
Despite these observed similarities, species-specific differences in habitat preference and anti-
predator strategies could still influence behavioural responses. White-tailed deer have been 
shown to use open fields, where it can rely on early detection and rapid escape (LaGory, 1987; 
Dellinger et al., 2019), while roe deer is considered a woodland species, which can adapt to 
fragmented and human dominated landscapes by utilizing hedgerows if forest habitats are 
scarce (Morellet et al., 2011). Moreover, roe deer may show shorter risk assessment intervals 
and earlier flight responses in open habitats (Bonnot et al., 2017). These differences may par-
tially explain the slightly higher vigilance observed in roe deer. However, it remains to be stud-
ied to what degree these two ungulate species compete for habitats and resources. Previous 
research on roe deer and white-tailed deer living in same areas found very few aggressive en-
counters, suggesting minimal direct competition between these species, and instead, a tendency 
toward coexistence or even cooperative use of shared habitats (Bartos et al., 2002).  
Interestingly, the group size effects observed in the white-tailed deer-specific models were not 
apparent in the interspecific comparisons. This may reflect variation in group structure between 
the species or risk areas, or lack of statistical power due to smaller sample sizes in interspecific 
models. Similarly, while time of day was a significant predictor in white-tailed deer specific 
models, no such effects were detected when comparing the two species, likely due to merged 
time categories (e.g. sunrise pooled with night) and limited sample sizes. This result could also 
indicate that prey has no reason to adapt vigilance levels according to the time of day in areas 
where wolf is nearly absent (Theuerkauf, 2009; Ferreiro-Arias et al., 2024), although lynx might 
still be present.  
Finally, the relatively high amount of variance explained by group identity in interspecific 
model suggests that within-group dynamics may have a stronger influence on vigilance bout 
duration than species identity alone under low-risk conditions. This could indicate of contagious 
vigilance between group members (McDougall & Ruckstuhl, 2018) or other social hierarchies 
within groups, such as presence of males in female dominated group (Fitzgibbon, 1990b; Lash-
ley et al., 2014; Cherry et al., 2015). 
5.5 Data limitations and methodological considerations  
All findings presented in this thesis must be interpreted with caution, especially due to the 
asymmetric data distribution between species across predation risk context. Only 13.3% of all 
observed white-tailed deer individuals were observed in low-risk areas. In low-risk areas 20.8% 
 39 
 
of observations consisted of white-tailed deer while the rest of the observations were of roe 
deer. The relatively small sample size of white-tailed deer in low-risk areas may have limited 
the statistical power of the models.  
As roe deer were nearly absent from high-risk areas within wolf territories, full interspecific 
comparisons on the effects of wolf presence on vigilance behaviour could not be conducted. 
The reason for such a low number of observed roe deer can only be speculated. One possibility 
is that roe deer actively avoid open habitats within wolf territories, and the study’s restricted 
focus on agricultural fields may have limited the possibility to detect them  (Morellet et al., 
2011; Bonnot et al., 2017). Alternatively, roe deer may effectively avoid inhabiting areas where 
wolf territories have recently established. It is also possible that lynx, a key predator of roe deer, 
avoids these areas too. This would concentrate lynx presence in low wolf-risk areas, where it 
would make forest cover riskier for roe deer. If this were the case, it would explain why roe 
deer were observed more frequently in open fields in low-risk areas. However, there is no recent 
evidence to support such avoidance behaviour between these two predators (Schmidt et al., 
2008). Another explanation would be that wolves have preferentially established in areas with 
higher densities of white-tailed deer, which can serve as more profitable prey, similarly to the 
pattern observed with red deer in Germany (Roder et al., 2020).  
Furthermore, there are several methodological improvements that could have elevated the sci-
entific level of this thesis. Firstly, this study did not include direct measures of individual deer 
characteristics such as age, sex, or reproductive status, each of which could influence vigilance 
behaviour (Elgar, 1989; Lashley et al., 2014). All observed animals were adults, but it could 
not be specified whether they were young or old. Sex was recorded in field, but the problem in 
detecting antlers from video material left the data very imbalanced. Eventually, the variable 
contained many missing values, which was the reason it was not used in the analyses. Also, it 
could not be verified whether individuals had been observed earlier, which could create pseu-
doreplication by unintentional repeated measures of vigilance behaviour of the same individuals 
on different days or within different groups on same study fields. In addition, social variables 
beyond group size, such as the presence of fawns, alarm signals or mixed-species groups, were 
not recorded, but may have influenced the within-group variation of vigilance behaviour 
(LaGory, 1987; Reby et al., 1999; Bartos et al., 2002; Costelloe & Rubenstein, 2018;). Moreo-
ver, predator presence data lacked precision. Wolf presence was based on broad territory maps 
(Valtonen et al., 2024) rather than confirmed activity during the study period. Lynx presence 
 40 
 
was assumed constant due to the absence of spatial data, even though national assessments 
provide visual estimates of the distribution of female lynx home ranges derived from cub ob-
servations (Herrero et al., 2024). 
With respect to video monitoring, Reolink cameras did not serve the study purposes as well as 
the other surveillance cameras. It would thus be recommendable to employ only the larger cam-
era setup in future. The telescope with a phone adapter worked very well, as the observer could 
make audio notes (with a lowered voice) on the videos of factors happening outside of the 
camera frame, such as people walking by, or how many individuals were on the field at the time 
of observation bout. However, working with the telescope takes time and effort, and requires a 
careful approach, so that animals would not detect the observer.  
Future studies should incorporate more fine-scale metrics of habitat characteristics and both 
predator and human disturbance, by potentially including distance to the nearest roads and 
buildings, or even use GPS tracking to collect more information of the movement patterns and 
habitat use of both predator and prey (Vander Vennen et al., 2016; Kröschel et al., 2017; Sunde 
et al., 2024). Also, social composition and hierarchy within animal groups should be considered 
as an explanatory variable in studying vigilance patterns (Elgar, 1989). A knowledge gap about 
the distribution and habitat choices of roe deer in Finland should be studied, as well as, the 
aspects of competition between roe deer and white-tailed deer. Also, the percentage of these 
two cervid species should be investigated in wolf diet, especially in this part of Finland.  
 
 41 
 
6. Conclusions  
This thesis demonstrates that in agricultural and human-dominated landscapes, vigilance in 
small cervids is shaped more strongly by social and temporal factors than by predator presence. 
Group size, time of day, and observation month influences vigilance in white-tailed deer, 
whereas wolf presence had no significant effect. In areas of low wolf density, vigilance patterns 
were similar for both roe deer and white-tailed deer, suggesting that under similar ecological 
conditions, the two species display broadly similar vigilance strategies. These findings highlight 
the importance of ecological and anthropogenic factors in shaping anti-predator behaviour. Fu-
ture research should focus on the role of human disturbance, predator-prey dynamics, and in-
terspecific competition, to improve understanding of behavioural adaptations in landscapes 
with recolonising predators. 
 
 42 
 
Acknowledgments 
I would like to thank the Natural Resources Institute Finland for the opportunity to collect data 
for this thesis as part of my internship. I am grateful to everyone involved in the PeVaKa project 
for warmly welcoming me as part of the team. I also wish to thank Pyry Toivonen for his help 
in setting up the camera equipment, and both Pyry and Tuukka Ståhlberg for assisting with 
transporting the equipment to the study sites. Special thanks go to the kind landowners who 
allowed me to place cameras on their fields for data collection, thereby making this research 
possible.  
 43 
 
References 
Aikio, S. & Pusenius, J. (2023). Valkohäntäpeurakanta talvella 2022–2023: Arvio Suomen 
valkohäntäpeurakannan koosta ja rakenteesta sekä kuvaus kanta-arvion laskentamenetelmistä (2. edition). 
Luonnonvara- ja biotalouden tutkimus 38/2023. Luonnonvarakeskus. Helsinki. 22 s.  
Airaksinen, S., Ekblad, C., Jormailainen, V., Kankainen, M., Laaksonen, T., Lindberg P., Lindqvist, M., Niukko, 
J., Olin, M., Ovaskainen, A., Pohjankukka, J., Tiittinen, M., & Westerbom, M. (2024) Merimetson ja 
harmaahaikaran aiheuttamien taloudellisten vaikutusten arviointi merialueen kalankasvatuslaitoksilla. 
Westerbom, M., Airaksinen, S., Arkko, V., Byholm, P., Ekblad, C., Hopkins, J., Jormalainen V., Kankainen, 
M., Laaksonen, T., Lehmonen, R. Lehmuskallio, J., Lindberg P., Lindén, A., Lindqvist, M., Moilanen, N., 
Niukko, J., Olin, M., Ovaskainen, A., Piha, M., Pohjankukka, J., Seimola, T., Svels, K. & Tiitinen, M. 
(2024). Merimetson ja harmaahaikaran suorat kalatalousvahingot: Määrän ja laadun arviointi Suomen 
merialueilla. Luonnonvara- ja biotalouden tutkimus 56/2024. Luonnonvarakeskus. Helsinki. 177 s. 
Altmann, J. (1974). Observational study of behavior: sampling methods. Behaviour, 49(3–4), 227–265. 
DOI:10.1163/156853974X00534 
Bartel, S. L., Hakkila, L., & Orrock, J. L. (2025). Deer vigilance and movement behavior are affected by edge 
density and connectivity. Ethology, 131(8), 22–30. DOI:10.1111/ETH.13574 
Bartos, L., Vankova, D., Miller, K. V., & Siler, J. (2002). Interspecific competition between white-tailed, fallow, 
red, and roe deer. The Journal of Wildlife Management, 66(2), 522. DOI:10.2307/3803185 
Beauchamp, G. (2015) Vigilance, alarm calling, pursuit deterrence, and predation inspection. In: Escaping from 
predators: An integrative view of escape decisions, ed. Cooper, W. E., & Blumstein, D. T. United Kingdom, 
Cambridge University Press 2015.  265–286. DOI:10.1017/CBO9781107447189 
Benhaiem, S., Delon, M., Lourtet, B., Cargnelutti, B., Aulagnier, S., Hewison, A. J. M., Morellet, N., & Verhey-
den, H. (2008). Hunting increases vigilance levels in roe deer and modifies feeding site selection. Animal 
Behaviour, 76(3), 611–618. DOI:10.1016/J.ANBEHAV.2008.03.012 
Blanchard, P., & Fritz, H. (2007). Induced or routine vigilance while foraging. Oikos, 116(10), 1603–1608. 
DOI:10.1111/j.0030-1299.2007.15799.x 
Bonnot, N. C., Couriot, O., Berger, A., Cagnacci, F., Ciuti, S., De Groeve, J. E., Gehr, B., Heurich, M., Kjellander, 
P., Kröschel, M., Morellet, N., Sönnichsen, L., & Hewison, A. J. M. (2020). Fear of the dark? Contrasting 
impacts of humans versus lynx on diel activity of roe deer across Europe. Journal of Animal Ecology, 89(1), 
132–145. DOI:10.1111/1365-2656.13161 
Bonnot, N. C., Hewison, A. J. M., Morellet, N., Gaillard, J. M., Debeffe, L., Couriot, O., Cargnelutti, B., Chaval, 
Y., Lourtet, B., Kjellander, P., & Vanpé, C. (2017). Stick or twist: roe deer adjust their flight behaviour to 
the perceived trade-off between risk and reward. Animal Behaviour, 124, 35–46. 
DOI:10.1016/J.ANBEHAV.2016.11.031 
Brooks, M.E., Kristensen, K., van Benthem, K.J., Magnusson, A., Berg, C.W., Nielsen, A., Skaug, H.J., Maechler, 
M., Bolker, B.M. (2017). glmmTMB Balances speed and flexibility among packages for zero-inflated gen-
eralized linear mixed modeling. The R Journal, 9(2), 378–400. DOI:10.32614/RJ-2017-066. 
Brown, J. S., Laundré, J. W., & Gurung, M. (1999). The ecology of fear: optimal foraging, game theory, and 
trophic interactions. Journal of Mammalogy, 80(2), 385–399. DOI:10.2307/1383287 
Chapron, G., Kaczensky, P., Linnell, J., Von Arx, M., Huber, D., Andrén, H., López-Bao, J., Adamec, M., Álvares, 
F., Anders, O., Balečiauskas, L., Balys, V., Bedõ, P., Bego, F., Blanco, J.C., Breitenmoser, U., Brøseth, H., 
 44 
 
Bufka, L., Bunikyte, R., Ciucci, P., Dutsov, A., Engleder, T., Fuxjäger, C., Groff, C., Holmala, K., Hoxha, 
B., Iliopou-los, Y., Ionescu, O., Jeremić, J., Jerina, K., Kluth, G.,  Knauer, F., Kojola, I., Kos, I., Krofel, M., 
Kubala, J., Kunovac, S., Kusak, J., Kutal, M., Liberg, O., Majić, A., Männil, P., Manz, R., Marboutin, E., 
Marucco, F., Me-lovski, D., Mersini, K., Mertzanis, Y., Mysłajek, R.W., Nowak, S., Odden, J., Ozolins, J., 
Palomero, G., Paunović, M., Persson, J., Potoečnik, H., Quenette, P.Y., Rauer, G., Reinhardt, I., Rigg, R., 
Ryser, A., Salvatori, V., Skrbinšek, T., Stojanov, A., Swenson, J.E., Szemethy, L., Trajçe, A., Tsingarska-
Sedefcheva, E., Váňa, M., Veeroja, R., Wabakken, P., Wölfl, M., Wölfl, S., Zimmermann, F., Zlatanova, D., 
and Boitani, L., (2014) Recovery of large carnivores in Europe’s modern human-dominated landscapes. 
Science (346)6216, 1517–1519. DOI: 10.1126/science.1257553. 
Cherry, M. J., Conner, L. M., & Warren, R. J. (2015). Effects of predation risk and group dynamics on white-tailed 
deer foraging behavior in a longleaf pine savanna. Behavioral Ecology, 26(4), 1091–1099. 
DOI:10.1093/beheco/arv054 
Cimatti, M., Ranc, N., Benítez-López, A., Maiorano, L., Boitani, L., Cagnacci, F., Čengić, M., Ciucci, P., 
Huijbregts, M., Krofel, M., López-Bao, J.V., Selva, N., Andren, H., Bautista, C., Ćirović, D., Hemming-
moore, H., Rein-hardt, I., Marenče, M., Mertzanis, Y., Pedrotti, L., Trbojević, I., Zetterberg, A., Zwijacz-
Kozica, T., and Santini, L. (2021) Large carnivore expansion in Europe is associated with human population 
density and land cover changes. Diversity & distributions (27)4, 602–617. DOI: 10.1111/ddi.13219. 
Cooper W. E. Jr. (2015). Theory: models of escape behavior and refuge use. In: Escaping from predators: An 
integrative view of escape decisions, ed. Cooper, W. E., & Blumstein, D. T. United Kingdom, Cambridge 
University Press 2015. 17–60. DOI:10.1017/CBO9781107447189 
Costelloe, B. R., & Rubenstein, D. I. (2018). Temporal structuring of vigilance behaviour by female Thomson’s 
gazelles with hidden fawns. Animal Behaviour, 145, 87–97. DOI:10.1016/J.ANBEHAV.2018.09.007 
Council of the EU (2025). Habitats directive: council agrees on the targeted proposal to change the protection 
status of the wolf. Press release, 16.4.2025. <https://www.consilium.europa.eu/en/press/press-re-
leases/2025/04/16/habitats-directive-council-agrees-on-the-targeted-proposal-to-change-the-protection-sta-
tus-of-the-wolf/> [cited: 1.8.2025]. 
Dellinger, J. A., Shores, C. R., Craig, A., Heithaus, M. R., Ripple, W. J., & Wirsing, A. J. (2019). Habitat use of 
sympatric prey suggests divergent anti-predator responses to recolonizing gray wolves. Oecologia, 189(2), 
487–500. DOI:10.1007/S00442-018-4323-Z 
Dimond, S., & Lazarus, J. (1974). The problem of vigilance in animal life. Brain, Behavior and Evolution, 9(1), 
60–79. DOI:10.1159/000123655 
Eccard, J. A., Meißner, J. K., & Heurich, M. (2017). European roe deer increase vigilance when faced with imme-
diate predation risk by Eurasian lynx. Ethology, 123(1), 30–40. DOI:10.1111/eth.12420 
Elgar, M. A. (1989). Predator vigilance and group size in mammals and birds: a critical review of the empirical 
evidence. Biological Reviews, 64(1), 13–33. DOI:10.1111/J.1469-185X.1989.TB00636.X 
Fattorini, N., & Ferretti, F. (2019). To scan or not to scan? Occurrence of the group-size effect in a seasonally 
nongregarious forager. Ethology, 125(5), 263–275. DOI:10.1111/eth.12844 
Ferreiro-Arias, I., García, E. J., Palacios, V., Sazatornil, V., Rodríguez, A., López-Bao, J. V., & Llaneza, L. (2024). 
Drivers of wolf activity in a human-dominated landscape and its individual variability toward anthropogenic 
disturbance. Ecology and Evolution, 14(10): e70397. DOI:10.1002/ECE3.70397 
Finnish Biodiversity Information Facility (2025) Roe Deer – Capreolus capreolus. Laji.fi. [cited: 9.5.2025] 
<https://laji.fi/en/taxon/MX.47507/occurrence>  
 45 
 
Fitzgibbon, C. D. (1990a). Anti-predator strategies of immature Thomson’s gazelles: hiding and the prone re-
sponse. Animal Behaviour, 40(5), 846–855. DOI:10.1016/S0003-3472(05)80985-6 
Fitzgibbon, C. D. (1990b). Mixed-species grouping in Thomson’s and Grant’s gazelles: the antipredator benefits. 
Animal Behaviour, 39(6), 1116–1126. DOI:10.1016/S0003-3472(05)80784-5 
Ganz, T. R., DeVivo, M. T., Wirsing, A. J., Bassing, S. B., Kertson, B. N., Walker, S. L., & Prugh, L. R. (2024). 
Cougars, wolves, and humans drive a dynamic landscape of fear for elk. Ecology, 105(4): e4255. 
DOI:10.1002/ECY.4255 
Gerber, N., Riesch, F., Bojarska, K., Zetsche, M., Rohwer, N., Signer, J., Isselstein, J., Herzog, S., Okarma, H., 
Kuijper, D. P. J., & Balkenhol, N. (2024). Do recolonising wolves trigger non‐consumptive effects in Euro-
pean ecosystems? A review of evidence. Wildlife Biology, 2024(6): e01229. DOI:10.1002/wlb3.01229 
Hasson, O. (1991). Pursuit-deterrent signals: communication between prey and predator. Trends in Ecology & 
Evolution, 6(10), 325–329. DOI:10.1016/0169-5347(91)90040-5 
Heikkinen, S., Mäntyniemi, S., Valtonen, M., & Kojola, I. (2024). Ahmakanta Suomessa 2024. Luonnonvara- ja 
biotalouden tutkimus 91/2024. Luonnonvarakeskus. Helsinki 12 s. 
Herrero, A., Mäntyniemi, S., Helle, I., Holmala, K., & Valtonen, M. (2024). Ilveskanta Suomessa 2024. 
Luonnonvara- ja biotalouden tutkimus 45/2024. Luonnonvarakeskus. Helsinki 22 s.  
Higdon, S. D., Diggins, C. A., Cherry, M. J., & Ford, W. M. (2019). Activity patterns and temporal predator 
avoidance of white-tailed deer (Odocoileus virginianus) during the fawning season. Journal of Ethology, 37, 
283–290. DOI: 10.1007/S10164-019-00599-1 
Jarnemo, A. (2004). Predation processes: behavioural interactions between red fox and roe deer during the fawning 
season. Journal of Ethology, 22(2), 167–173. DOI:10.1007/S10164-004-0118-2 
Johnson-Bice, S. M., Gable, T. D., Homkes, A. T., Windels, S. K., Bump, J. K., & Bruggink, J. G. (2023). Logging, 
linear features, and human infrastructure shape the spatial dynamics of wolf predation on an ungulate neo-
nate. Ecological Applications, 33(7): e2911. DOI:10.1002/eap.2911 
Jones, K. A., Krebs, J. R., & Whittingham, M. J. (2007). Vigilance in the third dimension: head movement not 
scan duration varies in response to different predator models. Animal Behaviour, 74(5), 1181–1187. 
DOI:10.1016/J.ANBEHAV.2006.09.029 
Kekkonen, J., Wikström, M., & Brommer, J. E. (2012). Heterozygosity in an isolated population of a large mam-
mal founded by four individuals is predicted by an individual-based genetic model. PLoS one, 7(9): e43482. 
DOI:10.1371/JOURNAL.PONE.0043482 
Kröschel, M., Reineking, B., Werwie, F., Wildi, F., & Storch, I. (2017). Remote monitoring of vigilance behavior 
in large herbivores using acceleration data. Animal Biotelemetry, 5(1), 1–15. DOI:10.1186/s40317-017-
0125-z 
Kuijper, D. P. J., Diserens, T. A., Say-Sallaz, E., Kasper, K., Szafrańska, P. A., Szewczyk, M., Stępniak, K. M., 
& Churski, M. (2024). Wolves recolonize novel ecosystems leading to novel interactions. Journal of Applied 
Ecology, 61(5), 906–921. DOI:10.1111/1365-2664.14602 
LaGory, K. E. (1987). The influence of habitat and group characteristics on the alarm and flight response of white-
tailed deer. Animal Behaviour, 35(1), 20–25. DOI:10.1016/S0003-3472(87)80206-3 
 46 
 
Lashley, M. A., Chitwood, M. C., Biggerstaff, M. T., Morina, D. L., Moorman, C. E., & DePerno, C. S. (2014). 
White-tailed deer vigilance: the influence of social and environmental factors. PLoS one, 9(3): e90652. 
DOI:10.1371/journal.pone.0090652 
Laundré, J. W., Hernández, L., & Altendorf, K. B. (2001). Wolves, elk, and bison: reestablishing the ‘landscape 
of fear’ in Yellowstone National Park, U.S.A. Canadian Journal of Zoology, 79(8), 1401–1409. 
DOI:10.1139/cjz-79-8-1401 
Lima, S. L. (1992). Strong preferences for apparently dangerous habitats? A consequence of differential escape 
from predators. Oikos, 64(3), 597–600. DOI:10.2307/3545181 
Lima, S. L. (1998). Stress and decision making under the risk of predation: recent developments from behavioral, 
reproductive, and ecological perspectives. Advances in the Study of Behavior, 27, 215–290. 
DOI:10.1016/S0065-3454(08)60366-6 
Lima, S. L., & Bednekoff, P. A. (1999a). Back to the basics of antipredatory vigilance: can nonvigilant animals 
detect attack? Animal Behaviour, 58(3), 537–543. DOI:10.1006/anbe.1999.1182 
Lima, S. L., & Bednekoff, P. A. (1999b). Temporal variation in danger drives antipredator behavior: the predation 
risk allocation aypothesis. The American Naturalist, 153(6), 649–659. DOI:10.1086/303202 
Lima, S. L., & Dill, L. M. (1990). Behavioral decisions made under the risk of predation: a review and prospectus. 
Canadian Journal of Zoology, 68(4), 619–640. DOI:10.1139/z90-092 
Lodberg-Holm, H. K., Teglas, B. S., Tyers, D. B., Jimenez, M. D., & Smith, D. W. (2021) Spatial and temporal 
variability in summer diet of gray wolves (Canis lupus) in the Greater Yellowstone Ecosystem. Journal of 
Mammalogy, 102(4),1030–1041. DOI:10.1093/jmammal/gyab060 
Mäntyniemi, S., Heikkinen, S., Helle, I., Valtonen, M. & Kojola, I.  (2025). Karhukanta Suomessa 2024. 
Luonnonvara- ja biotalouden tutkimus 33/2025. Luonnonvarakeskus. Helsinki. 24 s.  
Mäntyniemi, S., Valtonen, M., Helle, I., Johansson, H., Ponnikas, S., Nivala, V., Harmoinen, J., Herrero, A., 
Heikkinen, S., Kvist, L., Aspi, J., Kojola, I., & Holmala, K. (2022). Suomen susikannan suotuisan 
suojelutason viitearvojen määrittäminen: Loppuraportti 2022. Luonnonvara- ja biotalouden tutkimus 
80/2022. Luonnonvarakeskus. Helsinki 147 s. 
Matala, J., Nikula, A., Pellikka, J., Aikio, S., Forsman, J., Henttonen, H., Holmala, K., Huitu, O., Jauni, M., Kojola, 
I., Melin, M., Paasivaara, A., & Pusenius, J. (2021). Hirvieläinten vaikutuksia yhteiskuntaan, elinkeinoihin 
ja ekosysteemiin. Luonnonvara- ja biotalouden tutkimus 38/2021. Luonnonvarakeskus, Helsinki 142 s. 
Mattisson, J., Sand, H., Wabakken, P., Gervasi, V., Liberg, O., Linnell, J. D. C., Rauset, G. R., & Pedersen, H. C. 
(2013). Home range size variation in a recovering wolf population: evaluating the effect of environmental, 
demographic, and social factors. Oecologia, 173(3), 813–825. DOI:10.1007/S00442-013-2668-X 
McDougall, P. L., & Ruckstuhl, K. E. (2018). Vigilance behaviour is more contagious when chewing stops: ex-
amining the characteristics of contagious vigilance in bighorn sheep. Behavioral Ecology and Sociobiology, 
72(9), Article 143, 1–10. DOI:10.1007/s00265-018-2536-0 
Morellet, N., van Moorter, B., Cargnelutti, B., Angibault, J. M., Lourtet, B., Merlet, J., Ladet, S., & Hewison, A. 
J. M. (2011). Landscape composition influences roe deer habitat selection at both home range and landscape 
scales. Landscape Ecology, 26(7), 999–1010. DOI:10.1007/S10980-011-9624-0 
Oriol-Cotterill, A., Valeix, M., Frank, L. G., Riginos, C., & Macdonald, D. W. (2015). Landscapes of coexistence 
for terrestrial carnivores: the ecological consequences of being downgraded from ultimate to penultimate 
predator by humans. Oikos, 124(10), 1263–1273. DOI:10.1111/OIK.02224 
 47 
 
Polameri, S. (2007). Metsäkauris. (8. edition). Metsästäjäin keskusjärjestö. Kopijyvä Oy. 
Poutanen, J., Fuller, A. K., Pusenius, J., Royle, J. A., Wikström, M., & Brommer, J. E. (2023). Density‐habitat 
relationships of white‐tailed deer (Odocoileus virginianus) in Finland. Ecology and Evolution, 13(1): e9711. 
DOI:10.1002/ECE3.9711 
Poutanen, J., Wikström, M., & Brommer, J. E. (2022). White-tailed deer Odocoileus virginianus (Zimmermann, 
1780). In: Terrestrial Cetartiodactyla. Handbook of the mammals of Europe, ed. Corlatti, L., Zachos, F.E.   
Springer, Cham. 277–288. DOI:10.1007/978-3-030-24475-0_27 
Pöysä, H. (1994). Group foraging, distance to cover and vigilance in the teal, Anas crecca. Animal Behaviour, 
48(4), 921–928. DOI:10.1006/ANBE.1994.1317 
Preisser, E. (2008). Trophic Structure. In: Encyclopedia of Ecology, ed. Jørgensen S.E & Fath B. D. Elsevier B.V. 
3608–3616. DOI:10.1016/B978-008045405-4.00558-9 
Proudman, N. J., Churski, M., Bubnicki, J. W., Nilsson, J. Å., & Kuijper, D. P. J. (2020). Red deer allocate vigi-
lance differently in response to spatio-temporal patterns of risk from human hunters and wolves. Wildlife 
Research, 48(2), 163–174. DOI:10.1071/WR20059 
Pulliam, H. R. (1973). On the advantages of flocking. Journal of Theoretical Biology, 38(2), 419–422. 
DOI:10.1016/0022-5193(73)90184-7 
Reby, D., Cargnelutti, B., & Hewison, A. J. M. (1999). Contexts and possible functions of barking in roe deer. 
Animal Behaviour, 57(5), 1121–1128. DOI:10.1006/ANBE.1998.1056 
Ripple, W. J., & Beschta, R. L. (2012). Trophic cascades in Yellowstone: the first 15years after wolf reintroduc-
tion. Biological Conservation, 145(1), 205–213. DOI:10.1016/j.biocon.2011.11.005 
Ripple, W. J., Estes, J. A., Beschta, R. L., Wilmers, C. C., Ritchie, E. G., Hebblewhite, M., Berger, J., Elmhagen, 
B., Letnic, M., Nelson, M. P., Schmitz, O. J., Smith, D. W., Wallach, A. D., & Wirsing, A. J. (2014). Status 
and ecological effects of the world’s largest carnivores. Science 343(6167): 1241484. 
DOI:10.1126/science.1241484 
Roberts, G. (1996). Why individual vigilance declines as group size increases. Animal Behaviour, 51(5), 1077–
1086. DOI:10.1006/ANBE.1996.0109 
Roder, S., Biollaz, F., Mettaz, S., Zimmermann, F., Manz, R., Kéry, M., Vignali, S., Fumagalli, L., Arlettaz, R., 
& Braunisch, V. (2020). Deer density drives habitat use of establishing wolves in the Western European 
Alps. Journal of Applied Ecology, 57(5), 995–1008. DOI:10.1111/1365-2664.13609 
Salinas Ruíz, J., Montesinos López, O. A., Hernández Ramírez, G., & Crossa Hiriart, J. (2023). Generalized linear 
mixed models with applications in agriculture and biology. Cham, Switzerland. Springer Nature Switzerland 
AG. 59–97, 129–319.  DOI:10.1007/978-3-031-32800-8 
Samelius, G., Andrén, H., Kjellander, P., & Liberg, O. (2013). Habitat selection and risk of predation: re-coloni-
zation by lynx had limited impact on habitat selection by roe deer. PLoS one, 8(9): e75469. 
DOI:10.1371/journal.pone.0075469 
Schmidt, K., Jędrzejewski, W., Okarma, H., & Kowalczyk, R. (2008). Spatial interactions between grey wolves 
and Eurasian lynx in Białowieża Primeval Forest, Poland. Ecological Research, 24(1), 207–214. 
DOI:10.1007/S11284-008-0496-Y 
Sirot, E., & Pays, O. (2011). On the dynamics of predation risk perception for a vigilant forager. Journal of Theo-
retical Biology, 276(1), 1–7. DOI:10.1016/J.JTBI.2011.01.045 
 48 
 
Stankowich, T., & Reimers, E. (2015) Mammals. In: Escaping from predators: An integrative view of escape 
decisions, ed. Cooper, W. E., & Blumstein, D. T. United Kingdom, Cambridge University Press 2015.  63–
87. DOI:10.1017/CBO9781107447189 
Sunde, P., Kjeldgaard, S. A., Mortensen, R. M., & Olsen, K. (2024). Human avoidance, selection for darkness and 
prey activity explain wolf diel activity in a highly cultivated landscape. Wildlife Biology 2024(6): e01251. 
DOI:10.1002/WLB3.01251 
Suraci, J. P., Clinchy, M., Zanette, L. Y., & Wilmers, C. C. (2019). Fear of humans as apex predators has land-
scape-scale impacts from mountain lions to mice. Ecology Letters, 22(10), 1578–1586. 
DOI:10.1111/ELE.13344 
Sönnichsen, L., Bokje, M., Marchal, J., Hofer, H., Jedrzejewska, B., Kramer-Schadt, S., & Ortmann, S. (2013). 
Behavioural responses of European roe deer to temporal variation in predation risk. Ethology, 119(3), 233–
243. DOI:10.1111/eth.12057 
Theuerkauf, J. (2009). What drives wolves: fear or hunger? Humans, diet, climate and wolf activity patterns. 
Ethology, 115(7), 649–657. DOI:10.1111/J.1439-0310.2009.01653.X 
Theuerkauf, J., Jȩdrzejewski, W., Schmidt, K., Okarma, H., Ruczyński, I., Śniezko, S., & Gula, R. (2003). Daily 
patterns and duration of wolf activity in the Białowieza Forest, Poland. Journal of Mammalogy, 84(1), 243–
253. DOI:10.1644/1545-1542(2003)084<0243:DPADOW>2.0.CO;2 
Valtonen, M., Heikkinen, S., Johansson, H., Härkälä, A., Helle, I., Mäntyniemi, S., & Kojola, I. (2024). Susikanta 
Suomessa maaliskuussa 2024. Luonnonvara- ja biotalouden tutkimus 54/2024. Luonnonvarakeskus. 
Helsinki 41 s. 
Vander Vennen, L. M., Patterson, B. R., Rodgers, A. R., Moffatt, S., Anderson, M. L., & Fryxell, J. M. (2016). 
Diel movement patterns influence daily variation in wolf kill rates on moose. Functional Ecology, 30(9), 
1568–1573. DOI:10.1111/1365-2435.12642 
Zanette, L. Y., & Clinchy, M. (2025). Ecology and neurobiology of fear in free-living wildlife. Annual Review of 
Ecology, Evolution, and Systematics, 51(1), 297–318. DOI:10.1146/annurev-ecolsys-011720-124613 
Zuur, A. F., Ieno, E. N., & Smith, G. M. (2007). Analysing Ecological Data. NY: New York, Springer-Verlag. p. 
134, 472. DOI:10.1007/978-0-387-45972-1  
 
Open Data:  
National Land Survey of Finland, Regions (vector) 1:10 000. 
National Land Survey of Finland, Orthophoto. 
Heikkinen, S. (2024). “Wolf territory boundaries” – Susireviirien tietovarannot. Natural Resourses Institute Fin-
land. DOI:10.23729/6ab1400f-7b8d-408e-b65d-7667e652c492 
OpenAI's ChatGPT model (GPT-4, 2025) was utilised in language management, R-code optimisation in statistical 
analysis and in visualising the results.  
 49 
 
Appendices 
Appendix 1: Study fields in high-risk areas 
Study fields in high-risk areas marked with orange polygons. The polygon outlines represent 
the approximate area from which observations were collected using the camera system, and 
therefore do not always cover the full field area. Dot marks the placement of the camera, which 
could be remotely rotated 355–360°. 
   
 50 
 
Appendix 2: Study fields in low-risk areas 
Study fields in low-risk areas marked with orange polygons. The polygon outlines represent the 
approximate area from which observations were collected, and therefore do not always cover 
the full field area. In Ruissalo (L1), the observations were conducted with telescope or binocu-
lars from covered place under the current wind direction. In study areas L2–L5, dot marks the 
placement of the surveillance cameras which could be remotely rotated 360°.