Does leadership show through behavior?
Ann-Christin Kock B.Sc. Thesis
June 2016
Faculty of Behavioral Management and Social Science
Supervisors:
Dr. ir. P.W. de Vries Dr. M. van Bommel
Psychology of Conflict, Risk and Safety Faculty of Behavioral management and Social Sciences University of Twente
Abstract
Because public concerts raise many public safety concerns with respect to the escalation of group excitement it seems important to monitor groups to prevent hazardous incidents.
Research suggests that it can be effective to engage with leaders of groups (Solera et al., 2015). The purpose of this paper is therefore to investigate whether it is possible to (validly) detect a leader within a group on the basis of behavioral data. It was supposed that (a)
(perceived) leaders show initiative behavior and have (b) less interpersonal distance compared with the followers. Groups where (c) leaders have closer physical proximity to the group are hypothesized to perform better. Furthermore it was asked whether the perceived leader can be recognized on the basis of walking distance. These effects are expected to be greater in groups with formal leaders than in groups with informal leaders. An experiment with groups of four was executed. The groups had to search locations while wearing GPS-trackers. Half of the groups had a directed (formal) leader and the other half had an informal leader. The perceived cohesion of the group and the representativeness of the chosen leader were also measured.
The results revealed that leaders tend to show initiative behavior. The representativeness also played a key role. It (significantly) predicted interpersonal distance, task performance and duration. The type of group had an influence on interpersonal distance and task performance.
From this it can be concluded that it is not the membership (being a leader or not) that has a crucial influence on behavioral patterns, but the representativeness of the leader.
Keywords: emergent leadership, formal and informal groups, perceived cohesion, leadership representativeness, Global Positioning System
Samenvatting
Het lijkt belangrijk groepen op openbare evenementen te monitoren om gevaarlijke incidenten te voorkomen. Uit onderzoek blijkt dat het hierbij effectief is de leiders van groepen te
benaderen (Solera et al., 2015). Daarom is het doel van dit paper te onderzoeken of het mogelijk is leiders in een groep op basis van gedragsmaten te detecteren. Er werd
verondersteld dat (a) (waargenomen) leiders initiatief tonen en (b) minder interpersoonlijke afstand ten opzichte van de volgers vertonen. Verder werd de verwachting opgesteld dat groepen waar de leiders minder fysieke afstand met de groep hebben beter presteren. De laatste hypothese die werd opgesteld houdt in dat de effecten in groepen met informele leiders sterker zijn dan in groepen met formele leiders. Ten slotte werd er gekeken of de
(waargenomen) leider op basis van de totale loopafstand gedetecteerd kan worden. Een experiment met groepen van vier participanten werd uitgevoerd. De groepen moesten diverse locaties opzoeken en daarbij GPS-trackers dragen. De helft van de groepen had een
toegewezen, formele leider en de andere helft had een informele, natuurlijke leider. Naast de gedragsmaten werden ook de waargenomen cohesie en de waargenomen representativiteit gemeten. De resultaten lieten zien dat (waargenomen) leiders de neiging hebben initiatief te tonen. Er werd ook duidelijk dat representativiteit een grote rol speelt wat betreft de leider.
Deze variabele voorspelt zowel de interpersoonlijke afstand en de prestatie alsook de benodigde tijd van een groep. De groepssoort (informele of formele leider) had een
significante invloed op de interpersoonlijke afstand en de prestatie van de groep. Hieruit kan worden geconcludeerd dat niet het lidmaatschap (het wel of niet zijn van een leider) een cruciale invloed op gedragsmaten heeft, maar de representativiteit van de (waargenomen) leider.
Sleutelwoorden: opkomende leiderschap, formele en informele groepen, waargenomen cohesie, leiderschap representativiteit, Global Positioning System
Although most events/occasions such as sporting events, concerts or festivals pass mostly peacefully and without problems, the occurrence of incidents during such events can have enormous consequences (Lee & Hughes, 2006). The consequences become apparent in headlines such as “1100 pilgrims killed during mass panic” (De Volkskrant, 2015) or in accidents such as the Love Parade in Duisburg in 2010 (Hitzler, Kirchner, & Betz, 2011).
Losing control over a crowd can result in a high number of deaths, injuries and causalties, and those affected often struggle for years with the psychological effects of trauma (Huber, 2012).
Lee and Hughes (2006) emphasized in their report the increasing number of reported crowd- related incidents during the last decade. In summary, controlling large groups can be difficult and losing control can have enormous consequences. To successfully monitor and control crowds and to prevent incidents, a well-founded knowledge of crowds is required. To be able to successfully intervene in the future, it is important to understand how crowds are
constituted, and what kinds of behavior they are likely to exhibit. Therefore, this paper investigates aspects of group behavior, focusing on the leader-follower relationship within groups. GPS-data gathered during a Scavenger Hunt are used to analyze the moving patterns of several groups and to identify the potential leaders.
Reviewing the literature, there is no single accepted definition of a crowd (Ge, Collins
& Ruback, 2009; Solera, Calderara & Cucchiara, 2015). However, it is important to make a distinction between a collection of people sharing a common location and a psychological crowd that shares a common social identity. The current paper refers to psychological crowds.
A crowd consists of several small groups that are defined as “a set of individuals who have direct interactions with each other over a period of time and share a common fate, identity, or set of goals” (Kassin, Fein & Markus, 2008, p. 297; Kenny et al., 2001).
Within groups, the members may take different roles, including the role of a leader (Solera et al., 2015). Leaders are those who often define the norms and values of a group and act as guides. They therefore offer orientation. Followers have confidence in the leader, who consequently has a great influence on the group. Leaders have the ability to transform,
whether intentionally or unintentionally, the behavior of followers from individual behavior into group behavior (Hogg et al., 2006).
Successful interventions directed at preventing incidents where crowds are involved can therefore be focused on leaders of groups (Solera et al., 2015). Identifying leaders could possibly help to control a crowd. If the leader is an in-group member and naturally developed, the probability of the group accepting the leader’s guidance is high. Often, these leaders (are able to) represent and analyze the interests of the group’s members due to a shared common
identity. As a consequence they gain power (Haslam et al., 2011; Turner, 2005). An external figure – an out-group member – may want the group to do certain things that are not in their own interest. When an attempt is made to force a crowd into subordination, there is an
increased probability that the crowd will reject the guidance of the out-group authority figure.
This is another reason to approach the leaders of a crowd in order to control a crowd.
Identified leaders can be separated from the group to eliminate their possibly negative
influence on the group. On the other hand, it can be highly valuable to work together with the leaders, making use of the influence they have. Leaders may initialize, encourage, motivate and stimulate behavior and thereby take control over a crowd (King, Johnson, & Van Vugt, 2009; Solera et al., 2015). All these points underline the importance of leadership detection.
Therefore, the current paper investigates whether it is possible to detect leaders within a group. One focus of the current research will be to examine how leaders within a group can be detected through the analysis of GPS data.
Theoretical Background
For the detection of leaders, understanding the concepts of crowd, group, leader and follower is necessary. In the following sections, these four terms are explained in more detail.
As already noted, there is a difference between a group of people sharing the same location and a psychological crowd. A psychological crowd is characterized by a common social identity and is a synonym for group. In the literature, a group is often referred to as a unit consisting of individuals who share common goals, motives, norms and values (Sherif, Sherif & Murphy, 1956). These individuals are related to each other in a manner that
enhances the performance of the group. To act as a group, it is important that the members of the group share a common identity.
In a group different roles can be assumed, such as the role of leader. In the literature different definitions of a leader are used. Andersson, Gudmundsson, Laube and Wolle (2008), for example, define the leader as someone who does not follow anyone but is followed by several members of the group. This definition, however, only notes the relative distance between the leader and the followers. Social and psychological factors are here not taken into account. However, the current study is concerned with detecting the
(psychological) leader based on non-verbal behavior. Within the current paper, the following definition of a leader is to be used: a leader is a group member who influences other members and has a coordinating role. A leader is someone who is perceived by the majority of the group as the leader. He or she is characterized by showing an understanding of the group’s
needs (Solera et al., 2015). A process occurs in which not only individual but also collective performance is facilitated to reach common goals (Van Vugt, 2006). So the leader has not only the role of guiding the group and establishing stability, but also of stimulating and supporting the members of the group. A leader represents the group with its norms and values (Haslam et al., 2011). Several different types of leader can be enumerated, but such
distinctions are beyond the scope of this study.
Leaders and followers have a reciprocal relationship (Antonakis & Atwater, 2002; King et al., 2009; Solera et al., 2015). Trust is one of the most important factors that influences this relationship (King et al,. 2009). Due to a common identity with the other members of the group, a feeling of ‘we-ness’ arises which strengthens the cohesion of the group (Haslam et al., 2011). This feeling facilitates group behavior.
Now that the theoretical background concerning crowds, groups, and the different roles of those within them has been clarified, the different variables expected to influence the
recognition of leaders are to be considered.
Following on from this, hypotheses are put forward for investigation. The current literature of leadership detection is presented in more detail.
Initiative taking as predictor
As mentioned, leaders are often defined in the literature in terms of their physical location in relation to the other members of the group. For instance, they are referred to as those who walk in front of the group (Laube & Wolle, 2008). The current paper focuses on the psychological definition of a leader. Contrary to the former definition, the latter implies interpersonal factors, such as trust, perceived cohesion and interconnectedness, instead of purely examining the interpersonal distance between the members of a group. The extent to which these two different definitions cohere is examined. Can the (psychological) leader be detected by observing his initiative behavior? The recent literature gives some indications of answers to this question. Bullée (2015), for example, shows that taking the initiative in a group seems to be one indicator of perceived group leadership. In the experiment the participants were asked to state three attributes of the perceived leader. 20% of the
participants indicated that taking the initiative was an important characteristic of the leader.
The Evolutionary Game Model supports this conception of leadership. This model suggests that, within any group, there is high probability that the leader will exhibit initiative behavior (Van Vugt, 2006). This implies that those members of a group who tend to be active are most likely to fulfil the leadership roles within the group. Another study showed that executives
differ from other employees in terms of activity levels, willingness to make decisions and ambition (Bray & Howard, in 1983 as in Van Vugt, 2006). Therefore, in the current study it is expected that the perceived leader of a group will more often initiate behavior compared to other members of the group (hypothesis I).
Interpersonal Distance as Predictor
In this study, the (interpersonal) distance is determined by the proximity of followers to the group leader. In the literature physical distance is often related to social distance (Antonakis & Atwater, 2002). Social distance refers to “the psychological effects of actual and perceived […] differences between the supervisor and subordinate” (p.13, Antonakis &
Atwater, 2002). These differences involve values (beliefs, attitudes), power and
demographical factors (gender, race). It should be noted that similarities as well as differences are perceived by supervisors and subordinates to exist between them. Some authors have suggested that these two constructs, physical and social distance, act in a similar manner (e.g., Howell-Marenda & Hall, 1999). In the literature it is also suggested that social distance involves the physical distance of followers (e.g. Shamir, 1995). In contrast to that view, Antonakis and Atwater (2002) argue that social distance and physical distance should be sharply distinguished as concepts. Indeed, it is possible for a leader to be at a great physical distance from the followers and be socially and psychologically close to them. Therefore, the current paper treats these concepts as distinct and concentrates on physical distance.
Most literature on physical distance stresses that the performance of a group is negatively correlated with the physical distance between leaders and followers. This implies that members of a group perform better if they are physically close to the leader. Napier and Ferris (1993), for example, argue that close physical proximity is associated with higher performance, greater follower satisfaction and lower probability of leader withdrawal.
From this the authors conclude that the follower-leader distance should be smaller in order to preserve a good atmosphere within the group; this is because physical abandonment has a negative effect on the quality of the leader-follower relationship and reduces the
influence exerted by the leader of the group (Bass, 1990). Bass (1998) explains this as due to the decrease of social interaction. Reduced physical distance gives leaders less opportunity to control, monitor and directly observe the followers. To summarize, the factor of physical distance between followers and leaders is often presented in the literature as a negative moderator for leadership results (Antonakis & Atwater, 2002). The papers stress that the
quality of the leader-follower relationship, as well as the performance of the group, is negatively influenced by a large physical distance
As mentioned in the definition of a leader, a leader wants to achieve the common goals of the group and wants to promote individual and collective performance. This implies that the leader (consciously or unconsciously) reduces physical distance in order to exert a positive influence on the group. Therefore, the assumption is made that the interpersonal distance of perceived leaders from their followers will be (significantly) less than the average interpersonal distance between the followers (hypothesis IIa). Furthermore it is hypothesized that groups with perceived leaders that have a smaller interpersonal distance from their followers will perform better during the experiment than groups where the perceived leader has a greater interpersonal distance from their followers (hypothesis IIb).
Formal and Informal Leader as Predictor
There are two types of groups - formal and informal groups (Sanchez-Cortes et al., 2010). To clarify these concepts, definitions of both terms are given. The concept of a formal group is often used in the context of organizations or institutions. Usually, this kind of group is deliberately composed by an ‘external’ person who has a higher position in the hierarchy.
However, these groups can also be created by voting process. Due to his position, the leader of a formal group usually has power or authority over the group. In contrast to the formal group, there is no formal hierarchy in an informal group. Here, all members are equal in principle. Furthermore, the members can decide individually whether to stay in the group, which means that they are members voluntarily. Members of an informal group are interconnected and have direct contact (Tonkens, Duyvendak, & Hurenkamp, 2006). In general, there is no allocation of roles, no system or formalization in informal groups (Van den Berg, Van Houwelingen, & De Hart, 2011). A spontaneous and often unconscious
process leads to the emergence of a leader. The member who seems the most competent to the other members usually takes the role of the leader. An example of an informal group is a group of colleagues who decide to go drinking after work, or to a concert.
The leaders of both kinds of group exert influence on their followers. Sanchez-Cortes et al. (2010) state that an informal leader has a stronger influence on the other members than the formal leader who is chosen by an authority. Members of an informal group perceive as the leader whichever member of the group seems to them to have the most skills. Therefore they trust and accept the leader. The chance that leaders who are selected by an authority are not accepted by the group is great because the group may, for example, distrust or not respect
the formal leader, or perceive the leader as someone without sufficient competences to execute the role of a leader. On the other hand, members of informal groups identify their informal leaders with the group, because these leaders arise, spontaneously and organically, from within the group. The norms and values represented by the informal leader are more quickly accepted because the leader is chosen by, and thereby represents, the (whole) group.
Because of shared identity which is more likely to develop with an informal leader, group behavior is mostly driven by the cognitive process in which individuals identify themselves in terms of group membership rather than separate individuals (Haslam, Reicher & Platow, 2011, Turner, 1982). Overall, an informal leader can therefore exert more influence on the group than a formal leader.
The current research will examine to what extent the influence of informal leaders and formal leaders on their groups will differ, as measured using behavioral data (GPS). It is expected that the observed effects which are predicted in hypotheses I, II, and III will be greater in groups with an informal perceived leader than in groups with a formal leader (hypothesis III).
Walking Distance as Predictor
Another measure by which group leaders could possibly be detected may be walking distance. For this measure no recent literature could be found that made assumptions about a relationship. However, other variables, for instance auditory factors such as talking time, volume of a person, or speaker turn duration, have been related to group leader recognition (Jayagopi, Hun Yeo & Gatica-Perez, 2009; Hung, Jayagopi, Ba, Odobez, & Gatica -Perez, 2008; Mast, 2002). All the studies conclude that the leader of a group exhibits active behavior (Jayagopi et al., 2009; Sachez-Cortez,Aran, Mast & Gatica-Perez, 2012). This active behavior becomes obvious from a longer speaking time, a higher voice volume, as well as from
proactive and dynamic actions. Not only have the behavioral measures in these studies showed that leaders are more active than followers, but also followers indicate that they perceive their perceived leader as active. The question is whether this active behavior of group leaders can also be observed in their walking behavior. To further analyze this, the current paper examines the following subsidiary question: can the perceived leader of a group be recognized based on the distance they walk relative to other members of the group?
Hypotheses
On the basis of the theoretical background an experiment was executed. Participants were divided into groups of four. After playing games to get to know each other, the groups had to find several locations using photos given to them by the researcher. They had to take photos of these locations, and the time they took to complete the Scavenger Hunt was also measured. GPS-trackers recorded the movement patterns of the participants during the task.
Perceived group cohesion and interconnectedness was measured before and after the Scavenger Hunt. Using a camera, it was also recorded which member of the group grasped the paper with the instructions for the Scavenger Hunt. In addition, the perceived leader was identified by asking the participants to complete a ranking of the members of their group. All hypotheses are presented in more detail below, also in figure 1.
I. The perceived leader of a group will more often initiate behavior compared to other members of the group. Perceived leaders will therefore more often be the ones who grasp the instructions and walk the greatest distance at the front of the group compared to (perceived) followers.
II. a The averaged interpersonal distance between perceived leaders and their followers will be (significantly) less than the average interpersonal distance between the followers.
b The groups with perceived leaders that have a smaller interpersonal distance from their followers will perform better during the experiment than groups in which the perceived leader has a greater interpersonal distance from their followers. Performing better is defined in this case as finding (significantly) more locations from photos given by the researcher (and take correct photos of them) and being (significantly) faster in executing the Scavenger Hunt.
III. The observed effects predicted by hypotheses I, II, and III will be greater in groups with a perceived informal leader than in groups with a formal leader. The examined effects are: initiative taking (grasping the paper, walking in front), interpersonal physical distance, and performance (number of correct photos, time required).
Figure 1. Conceptual model describing the relationships between variables measured in this study.
Method Participant
In total, 66 subjects participated in this study. Seven males (10.6%) and 59 females (89.4%) took part voluntarily and signed the informed consent form (Appendix A). 32 of the participants were German (48.5%) and 34 were Dutch (51.5%) The age of the participants varied from 18 to 24 (M= 20.76, SD= 1.55). Most of the participants had a VWO diploma (89.4%). Several participants already had a Bachelor’s degree (7.6%) or a HAVO diploma (3%). Only 28.8% of the participants stated that they lived on the campus, whereas 71.2%
stated that they lived in the city. 15 (22.7%) of the participants knew nobody else in the group, 19 (28.8%) knew one person, 13 (19.7%) knew two persons and 19 (28.8%) knew three persons, figure 2. In this study, the convenience sampling method was used. Several social media sites and the SONA system were used to get in touch with the participants. The SONA system is an online subject pool management system.
In this study, one selection criterion was necessary. Only respondents able to understand and speak Dutch were allowed to participate because the questionnaires were given in Dutch. This criterion ensured that all participants were able to comprehend the materials and to communicate accurately with each other.
Figure 2. Graphical overview of the number of participants knowing other participants of their group.
Design
In the study, a 2x2 factor design was used. The independent variables were
Membership (follower vs. leader) and Type of Group (informal vs. formal). The dependent variables were Cohesion, Leader Representativeness, Walking Distance, Interpersonal Distance, Velocity, Initiative Taking, Task Duration and Task Performance.
Procedure
Before the students entered the room, the researcher wrote down in which Type of Group the group was supposed to be (informal/formal) depending on the condition of the previous group. The study started with an informal group and then the groups were alternately assigned to the two types without any information about that. In the informal groups (N = 8) a leader could naturally emerge during the study. In formal groups (N = 9) a leader was
allocated by the order of the participants’ arrival. To avoid a systematic bias, it was decided that in the first experimental group the participant who arrived first would be designated as leader, in the second group the second participant, and so on. In the fifth group, the order was repeated so that the first person entering the room was designated as the leader of the group.
After welcoming the participants, it was emphasized that the participants were allowed to stop with the study without giving reasons and without dealing with any consequences.
Furthermore, it was stressed that the data recorded by the camera, the GPS tracker and questionnaires would be saved, treated anonymously and not given to a third party. The participants were asked whether they agreed with their data being recorded. After that, the participants filled in the informed consent form and the questionnaire with all demographic variables, see Appendix B.
0 5 10 15 20
0 1 2 3
Count
Number of Acquaintances
Overview of Number of Acquaintances
Then, the participants were asked to play several games to get to know the other (two or three) members of the group. After showing the participants where the instructions for the games were, the researcher turned on the camera and left the room. In the first game the participants had to briefly introduce themselves. In the second game, theyhad to throw one (or two) balls to another group member while saying his/ her name aloud. The goal of the game was to memorize each other’s names, and to establish the feeling of a group due to a common goal. In the last game, the participants had to (collectively) balance two folding rules on their fingers. Together, they had to squat once while balancing the rules. The goal of this game was to develop a strategy by communicating and through this a stronger “we”-feeling. For a more detailed description and the instructions for the games, see Appendix C. Subsequently, the participants filled in questionnaires about the (perceived) Cohesion of the group and the (perceived) Interconnectedness within the group, see Appendix D and Appendix E.
Next, the participants were asked to let the researcher into the room who then asked the participants to hang two GPS-trackers around their necks, on top of their clothes so that the trackers could accurately make contact with several satellites. With this system, spatio- temporal information about the location of a person could be registered weather
independently and reliably (Milner, 2016). For this space-based navigation I-gotU GT-600 GPS trackers were used which registered every second the latitude and longitude of the participant’s location.
Then, the researcher asked the participants to grasp the paper with the instructions about the Scavenger Hunt (Appendix F) and the camera (Panasonic Lumix DMC-FS10). Each group was supplied with a fully loaded camera with a cleared chip card to avoid participants getting information from previous groups. Then, the researcher wrote down who grasped the paper (Initiative Taking) and stopped the recording of the camera.
After that, the researcher and the participants went outside. The participants were supposed to find thirteen locations on the campus which were shown on thirteen photos they got with the instructions (Appendix G). They had to make photos of these locations with at least one group member in each photo within maximal 30 minutes. They were not permitted to move around by bike or to split up and individually search for the locations. The researcher also stressed that the participants should not go into or next to (large) buildings. The
participants were also asked to check the connections of the GPS-trackers to the satellites several times. The researcher emphasized that not only finding the photos was important, but also coming back as soon as possible. As motivation, the researcher explained that there was a secret prize the group could win, depending on their performance (correct photos and time).
Finally, the researcher demonstrated how to handle the GPS trackers. Every participant was given two trackers to avoid missing data. When no questions remained, the participants left and the researcher began recording the time.
At the arrival the researcher stopped the time for the participants (Task Duration).
Together, they turned off the GPS trackers and went to the research room. There, the participants filled in four questionnaires about perceived (perceived group) Cohesion, the (perceived) Interconnectedness, the (perceived) Leader and Membership (ranking), and Leadership Representativeness (see Appendix D, E, H & I). Subsequently, the participants were given candies for taking part which was the secret price. The aim of the study, to measure leadership on the basis of GPS data, the conditions and other relevant information was explained to the participants. Finally, the researcher answered remaining questions from the participants.
In the informal condition, the paper with the instructions about the game laid on the table whereas in the formal condition, the designated leader received the paper with the instructions of the game and the map with the pictures to control the situation when the researcher was outside. Controlling was defined as managing the time-component, having oversight and leading the group so that they completed the task in an orderly fashion while having essential information about the task. In the informal group everyone could grasp the map containing the photos and the instructions about the task. In both conditions, the instructions about the task had to be grasped (Initiative Behavior).
Measurements
Different measures were used in the course of this study. In more detail, several questionnaires and behavioral variables were measured.
Demographic variables. In this questionnaire the participants were asked to fill in several demographic variables: gender, age, highest education and nationality. Furthermore, they had to state whether they live on the campus and how many members of the group they already knew before taking part.
Interpersonal Connectedness. The questionnaire Inclusion of Other in the Self Scale (IOS) was used to measure the perceived interpersonal connectedness of the participants with the group. It is a single-item pictorial measurement where participants have to choose
between seven figures showing overlapping circles which represent the self and the group (Aron, Aron, & Smollan, 1992). The scale is not sensible to social desirable responses, is
psychometric suitable and shows to be “consistent with diverse theoretical orientations to closeness” (Aron et al., 1992).
Cohesion. The perceived cohesion of the group was measured by a subscale of the Revised Subsitute for Leadership Scale (Podsakoff, Niehoff, Machenzie, &Williams, 1993).
The chosen subscale “closely-knit, cohesive, interdependent work groups (COHES)”
(Podsakoff et al., 1993, p. 707) was translated from English to Dutch. The scale consists of six items with a five point Likert scale varying from 1 (“totally disagree”) to 5 (“totally degree”). An example item was “members of my group work together as a team”. In this study, the questionnaire was shown to have excellent reliability, Cronbach’s α = .902 (George, 2003).
Membership and Leader. To make it possible to detect the perceived leader of a group, each group member was asked to rank the members of the group with respect to their leadership of the group. In doing so, the participants made a ranking and distributed ten points across the group. This distribution had to clearly show one (perceived) leader of the group.
The summation of all scores given the group made it possible to detect the perceived leader.
The character of each participant’s membership of the group was determined through the use of this questionnaire. Two variables were created on the basis of this questionnaire, Leader (yes vs. no) and Membership. The variable Membership was a ranked variable. Here, “1”
indicated that the participants were, compared to the other members of the group, perceived as the ones who were the most leading. The rank “4” was assigned to members who were seen as the most following members.
Leader Representativeness. A subscale of the Identity Leadership Inventory (ILI) was used to measure the perceived representativeness of the group’s chosen leader. It was measured whether the chosen leader was perceived as being one of them. The subscale consists of four items with a five point Likert scale varying from 1 (“totally agree”) to 5 (“totally disagree”). An example of an item is “this leader is representative of members of the group” (Steffens et al., 2014). The reliability of this questionnaire has been shown to be acceptable in this study with a Cronbach’s α of .717 (George, 2003).
Behavioral Measures. The behavioral measures include variables gathered by the GPS-trackers. Furthermore, other measures recorded by other technical devices are taken into account.
Initiating Behavior. With the camera, it was recorded which member of the group grasped the instructions for the Scavenger Hunt (Initiative Taking). With the program which read the GPS files it was actually not possible to measure who walked most often in front of
the group. Therefore, only grabbing the paper was used as Initiative Behavior for analyzing hypothesis I.
Performance. For this variable, the number of locations each group correctly discovered was checked. This was done by counting all the correct pictures stored on the camera.
Task Duration. The time each group spent on executing the task of the Scavenger Hunt was taken into account for the overall performance. The time was measured by the researcher.
Walking Distance. This variable represents the total distance a participant walked during the Scavenger Hunt (in meter). The distances between all locations the GPS-trackers recorded (one location per second) were summed per participant.
Interpersonal Distance. This variable represents the (averaged) distance a participant had with the other group members. To be able to analyze the difference in interpersonal distance between leaders and followers, the interpersonal distance of the leader with the followers and the interpersonal distance of the followers among each other was computed and assembled to one variable, Interpersonal Distance.
Velocity. This variable represents how fast a participant was in average (in kmH). For this, it was calculated how much time passed between the locations the trackers recorded.
This time was averaged and computed to one variable.
Data Analysis
Before the data was analyzed, the final data set was compounded. Then, the dataset was extended (means were computed) and some analysis were done to get an overview over the data (correlation matrix). Finally, the hypotheses were tested.
For testing the relationships between the group, perceptual and behavioral variables (see figure 1), two-step hierarchical regression analyses were computed. At the second step, the interaction term between the two independent variables was added. A mean score of the independent variables was aggregated and a new variable was computed. The new variable was a subtraction of the participants’ score in the variable minus the mean score.
Results
First of all, the effectiveness of manipulation was assessed. It was analyzed whether assigned leaders were actually perceived as the leader of a group. A Fischer’s Exact test was performed because the sample of allocated leaders was too small for computing a Chi-Square
test of independence (N = 8). The test analyzed the relation between the assigned leaders and the perceived leaders. Therefore, the two variables Leader and Assigned Leader were used in the test. The test was not significant which shows that there is no significant association between Leader and Assigned Leader, p = .189. A total of 50% of the allocated leaders actually were perceived as the group’s leader. 22.4% of the members who were not allocated as leader finally served as a leader.
Descriptives
There are several correlations between the variables measured in this study. Table 1 gives an overview. Here, only results that probably could have systematically influence the results of the study are briefly illustrated. Especially the variable Number of Acquaintances had a great influence on different variables used in the statistical analyses. It correlated significantly with Walking Distance [r(64) = -.318, p = .010], Cohesion [r(64) = .688, p <
.001] and Task Performance [r(64) = -.267, p = .030].
The perceived Cohesion after playing the games was relatively high (M = 3.955, SD = 0.736). This shows that playing the games created a (temporal) connection between the members of the group. The goal of playing the games which was creating a feeling of togetherness and being a unit therefore succeeded. A Wilcoxon Signed-Rank Test indicated that the perceived cohesion after the Scavenger Hunt (M = 4.054, SD = 0.723) was
marginally higher than the perceived cohesion before the Scavenger Hunt, Z = -1.815, p = .069. By contrast, the Inclusion of Other in the Self Scale (IOS) showed another trend. A Wilcoxon Signed-Rank Test showed that the feeling of belonging to the group was before the Scavenger Hunt significant higher [M = 4.985, SD = 1.622)] than after the Scavenger Hunt [M = 4.545, SD = 1.980], Z = -3.473, p = .001.
1234567891011121314151617 1 Groupsize3.909 (0.230) 2 Gender -.0621.110 (0.310) 3 Age.156-.10620.76 (1.550) 4 Nationality-.201.334**-.429**1.520 (0.504) 5 Education-. 201.334**-.429**.863**4.3 80 (1.527) 6 Number of Acquaintances.246*.356**-.090.039.0831.55 (1.139) 7 Campus.032.002-.013.120.071.1001.71 (0.456) 8 Type of Group-.346**.081-.212.459**.334**-.144.0430.45 (0.502) 9 Membership.157-.021-.061-.142-.017.019.214-.0132.348 (1.045) 10 Walking Distance-.173-.223.021-.045-.037-.318**.173.009-.05811171.06( 2485.675) 11 Task Duration .356**.046.217.097.022-.198-.231.076.062-.410**1654.74 (201.987) 12 Speed-.202-.139-.006-.169-.131-.240.290*-.091.094.727**-.459**23.001 (8.562) 13 Interpersonal Distance
-.041.216.169.156.195-.009.061.359**.102-.145.353**-.14111.684 (4.703) 14 Cohesion.063.341**-.251*.205.202.688**.038.105-.079-.331**-.284*-.215-.1568.001(1.3 87) 15 Initiative Taking.055-.022-.025-.017.033.039.008-.019.398**.000.020.066-.120-.1171.740 (0.441) 16 Task Performance-.837**.036-.076.233.234-.267*.133.448**-.095.324**-.427**.290*.025-.039-.0463.11 (0.467) 17 Leader Allocation.206-.023-.028-.267*-.030.056.071-.407**.259*-.021-.002.216-.243-.052.206-.23212.030 (1.0667) 18 Leader Representative ness -.151-.055.030.040-.027.082.088.202-.048-.123-.339**-.014-.279*.320**-.111.244*-.1973 (0 Note. * p < .05, ** p < .001. At the end of each row the mean and standard deviation of the variable is given.
Analyzing relationships between variables
In the following, several hierarchical regression analysis are described for explaining the relationships presented in figure 1 (introduction). The relationship between the group variables (Type of Group & Membership), the perceptual variables (Cohesion & Leader Representativeness) and the behavioral variables (Task Performance, Task Duration, Walking Distance, Interpersonal Distance & Velocity) were assessed. For all analysis a significance level of .05 was used. Furthermore, it was checked for the necessary assumptions for assessing a regression (normality, collinearity, multicollinearity & homoscedasticity). Only the statistics of the significant results are presented in the text. A summary of statistical information including the non-significant regression can be found in the appendices. Also the predictive ability of the different models are presented in the appendices.
Assessing the initiative behavior of leaders and followers. Direct logistic regression was performed to assess the impact of Membership, Type of Group and the interaction of these variables on the likelihood that leaders would show initiative behavior. Going in front of the group could finally not be measured because of unforeseen difficulties. Therefore it was chosen to make only use of grabbing instructions as indicator for initiative behavior. The full model containing all predictors was significant, X² (3, N = 66) = 11.656, p = .009. This implies that the model was able to distinguish between participants grabbing the instructions and participants that did not. Only the variable Membership made a significant unique contribution to the model (p = .003). Therefore, hypothesis I is accepted. The more a group member is perceived as the leader of the group, the bigger is the probability of showing initiative behavior. Type of Group and the interaction were not significant predictors.
Figure 3. Initiative Behavior for different kind of Membership.
0 5 10 15 20
1 2 3 4
Count
Membership
Initiative Behavior
Yes No
Assessing relationship between Interpersonal Distance, Task Performance and Duration. It was hypothesized that (perceived) leaders with a smaller distance to the
followers will perform better compared to (perceived) leaders with a bigger distance. For this, solely the data of the leaders was used. For analyzing this hypothesis (IIb) two linear
regressions were computed. The first regression model tested whether the Interpersonal Distance predicts the Task Performance of groups. Furthermore, no evidence was found that there is a predictive relationship between the Interpersonal Distance and Task performance, t(15) = -0.559, p = .584, β = -.143. There was no evidence found that the smaller the distance between the leaders and their group is, the more photos a group makes. A second linear regression was calculated with Interpersonal Distance as independent and Task Duration as dependent variable. It was tested whether groups with leaders that have a smaller
Interpersonal Distance are faster in executing the Scavenger Hunt. Interpersonal Distance was not able to predict the independent variable, indicating that there is no significant relationship between the interpersonal distance of the leaders with the followers and the required time for executing the task, t(15) = 1.707, p = .109, β = .403. As a result of these two regression analysis, hypothesis IIb cannot be accepted. It could not be proved that groups were the perceived leaders have a smaller distance to the group will perform better than leaders with a bigger distance.
Assessing relationship between the group variables and perceptual variables. The independent variables of the analysis were Type of Group (formal vs. informal) and
Membership (ranking of leadership) and the interaction between the two variables. Appendix J gives an overview of all statistical information with respect to these regressions.
In the first hierarchical regression it was analyzed to what extent the independent variables are able to predict the (perceived) Cohesion of a group (dependent variable). The independent variables were not able to (significantly) predict the (perceived) Cohesion of a group. Also the interaction between the two independent variables was not able to predict the Cohesion.
The second hierarchical regression with the independent variables Type of Group and Membership examined the ability for predicting the (perceived) Leader Representativeness (dependent variable). The Type of Group, Membership, and the interaction between Type of Group and Membership were not able to predict the perceived representativeness of the chosen leader. No main-or interaction effect was found.
Assessing relationship between group variables and behavioral variables. The independent variables for assessing the relationships between the group variables and the
behavioral variables were Type of Group, Membership and the interaction between the two variables. The dependent variables were: Task Performance, Task Duration, Walking Distance, Interpersonal Distance and Velocity. Appendix K shows an overview of all statistical information about the executed regressions.
Firstly, a hierarchical regression analysis with the mentioned independent variables (Type of Group, Membership and interaction between Type of Group and Membership) was assessed to test whether the independent variables were able to (significantly) predict Task Performance. The regression revealed that Type of Group significantly predicted the Task Performance, t(62) = 3.983, p < .001, β = .466. The positive beta reveals a positive
relationship between the variables indicating that groups with a formal leader performed better during the Scavenger Hunt. Better performing implies finding more locations (making more correct photos).
For assessing predictive ability of Type of Group, Membership and the interaction of Type of Group and Membership on the Task Duration, a hierarchical regression was assessed.
No main effect or interaction effect was found. The independent variables could not predict the dependent variable.
In addition to that hierarchical regression was assessed to analyze whether the
independent variable Type of Group, Membership and the interaction between Type of Group and Membership could predict the Walking Distance of a group. The independent variables did not have a (significant) predictive ability for the Walking Distance of a group. Therefore, the sub question ‘can the perceived leader of a group can be recognized based on the walking distance’ with which was dealt in the introduction, can be negated. Membership does not predict the distance a group walked during the task.
Furthermore, a hierarchical analysis was assessed to predict the (dependent variable) Interpersonal Distance. In the (significant) model the independent variable Type of Group was able to predict the Interpersonal Distance, t(62) = 3.064, p = .003, β = .361. The positive beta indicates a positive relation between Type of Group and Interpersonal Distance. This implies that in formal groups, the Interpersonal Distance is significant larger than in informal groups. Membership did not have a significant influence on Interpersonal Distance. The fact that Membership did not have any predictive ability implies that being a follower or leader does not lead to a certain trend in interpersonal distance. As a result, hypothesis IIa which stated that the interpersonal distance between the perceived leaders would be smaller than the interpersonal distance between the followers could not be accepted.
Finally, it was assessed whether it is possible to predict the Velocity of a group based on Type of Group, Membership and the interaction. The independent variables did not show any predictive ability for the dependent variable.
Because none of the interactions showed predictive value, hypothesis III was rejected.
Informal groups did not significantly higher on the behavioral measures.
Assessing relationship between perceptual variables and behavioral variables. Several hierarchical regression analysis were computed to analyze whether Cohesion and Leader Representativeness are able to predict several independent variables (Task Performance, Task Duration, Walking Distance,Velocity & Interpersonal Distance). The interaction effect was added in the second block to test whether the interaction effect adds predictive value to the model. In Appendix L the most important results of the regression analysis are presented.
First of all, a hierarchical multiple regression with the independent variables and the dependent variable group Task Performance was assessed. The interaction between the two independent variables (Cohesion and Leader Representativeness) was entered at stage two in the model. Although the regression equation was not significant, Leader Representativeness showed to significantly predict the Task Performance of groups, t(62) = 2.239, p = .029, β = .286. This implies that the more a leader is perceived as representative for the group, the better a group performs. The other independent variables did not significantly predict Task Performance.
Another hierarchical regression was calculated to analyze a possible prediction of the duration of the task (dependent variable) based on Cohesion and Leader Representativeness.
The interaction between Cohesion and Leader Representativeness was added at the second stage to analyze whether it adds any predictive ability to the model. The model was
significant in predicting the required time for the groups for executing the task, F(2, 62) = 5.534, p = .006. The independent variable Leader Representativeness significantly predicted Task Duration, t(62) = -2.257, p = .027, β = -.277. The negative beta shows a negative relation between the two variables. This implies that the more representative a perceived leader of a group is, the less time a group needs for the execution of the task. The two other variables had no significant effect on Task Duration.
Furthermore, a hierarchical regression was assessed with the independent variables Cohesion and Leader Representativeness, and the dependent variable Walking Distance. At the second stage, the interaction between Cohesion and Leader Representativeness was added.
The model revealed to be significant, F(2, 62) = 3.836, p = .027, adj. R² = .081. The
(perceived) Cohesion showed to be significantly predicting the Walking Distance of a group,
t(62) = -2.572, p = .012, β = -.323. The negative beta indicates a negative relation between the two variables. In this case, it means that the more a groups perceives cohesion between the members, the less a group totally walked during the Scavenger Hunt. Leader
Representativeness did not have any predictive value on the Walking Distance of a group. An interaction effect was also not found.
In addition, a hierarchical regression with Cohesion, Leader Representativeness and the interaction between the two variables as independent variables and Interpersonal Distance as dependent variable was assessed. The perceived representativeness of the chosen leader showed to be significantly predict the Interpersonal Distance, t(62) = -2.002, p = .050, β = - .255. The negative beta value indicates a negative relation between the variables. It suggests that the more group members perceive their leader as representing the group, the less distance there is between the members among each other and the leader.
Finally, a hierarchical regression with Cohesion and Leader Representativeness as independent variables and Velocity as dependent variable was assessed. Cohesion and Leader Representativeness did not significantly contributed to the regression model. Adding the interaction did not add explain any addition variation in Velocity significantly. The independent variables were not able to predict Velocity.
Additional Analysis
The analysis were computed to check whether Number of Acquaintances influenced the predictive ability of variables with which it significantly correlated. Therefore the hypotheses containing the variables Task Performance (hypothesis III & IIb) and Walking Distance (subquestion) were tested with Number of Acquaintances as covariate. All results can be seen in appendix M.
First of all, an additional hierarchical regression was executed to test whether Number of Acquaintances had an influence on the predictive ability of Interpersonal Distance of leaders on the Task Performance of a group. The Number of Acquaintances was added as covariate and Interpersonal Distance was added in the second block. As in the regression executed without the covariate, the additional regression showed to be not significant, t(15) = -0.540, p = .598, β = -.143. This indicates that Interpersonal Distance does not predict the Performance of a group also when controlling for the covariate Number of Acquaintances.
Secondly, an additional analysis was done for measuring whether the regression for hypothesis III was influenced by Number of Acquaintances. Therefore Number of
Acquaintances was added as covariate in block one, Membership and Type of Group in block two and the interaction in block three. Only the model with the covariate showed to be
significant. Membership, Type of Group and the interaction does not have any predictive ability on Task Performance. Therefore, as also in earlier analysis without the covariate, hypothesis III could not be accepted.
Furthermore, it was also checked whether number of Acquaintances changes the results with respect to the sub question about the Walking Distance. The hierarchical regression revealed that only Number of Acquaintances significantly predict the Walking Distance, t(63) = -2.662, p = .010, β = -.318. Like in earlier analysis, the other independent variables had no predictive value on Walking Distance.
Figure 2. Significant relations between the variables.
Conclusion and Discussion
With respect to monitoring groups during public events to maintain public safety, it seems to be an effective option to address the leaders of groups (Solera et al., 2015). But how is it possible to (validly) detect a leader within a group on the basis of behavioral data? This paper analyzed to what extent it is possible to detect leaders within a group on the basis of GPS-data. For this, an experiment was conducted where GPS trackers recorded a group’s moving pattern during a task. The perceived leadership was measured using a ranking made by the participants.
Several (unexpected) relationships were found, as figure 2 reveals. First of all, being perceived as the leader of a group increases the probability of initiative taking behavior.
Therefore, hypothesis I can be accepted. Furthermore, it was found that groups which perceived the cohesion within the group as high walked less than groups with low cohesion.
The more a perceived leader is regarded as representative, the more photos a group found and the less time they needed for the Scavenger Hunt. Also the interpersonal distance of the group was smaller the more a leader was perceived as representative of the group. On top of that, groups where the leader could emerge naturally had a smaller interpersonal distance and performed better.
Discussion
It was hypothesized that (perceived) leaders of groups will more often initiate behavior compared to the followers (hypothesis I). Therefore it was expected that perceived leaders would more often grab the instructions than followers. The results show that actually displaying this form of initiative behavior predicts being perceived as the leader or not. This study is consistent about the literature about initiative behavior and therefore supports the findings. Previous work showed that the frequency of initiating behavior in an experimental setting correlates with being nominated as the leader of a group (Kremer & Mack, 1983).
Emergent leaders were shown to be more often the ones who actively take part in group activities (Riggio, Riggio, Salinas & Cole, 2003). Van Vugt (2006) explains that groups often have difficulties with making decisions. An (emergent) leader therefore facilitates this process by coordinating and initiating behavior. Due to this behavior a group is able to act like a unit.
So, leaders persuade other group members to follow them which facilitates the group decision-making process.
Furthermore, it was expected that the interpersonal distance between (perceived) leaders and their followers would be smaller compared to the mutual interpersonal distance of the followers. This hypothesis (IIa) was rejected in the course of the study. In the theoretical background it was argued that the interpersonal distance between leaders and followers would be small because leaders want to communicate more with the members, create a good
atmosphere and therefore perform better with the group (Napier & Ferris, 1993). In the literature a larger physical distance is often referred to as a negative moderator for leadership results. When taking the potential initiative behavior of leaders into account one could assume that (theoretically) leaders cannot be often in the middle of the group because they are
walking at the front, initiating the behavior of the whole group. This would be a possible explanation for the different findings, as to have the smallest interpersonal distance, the leader would have to be in the middle of the group. Further research could measure this explanation
