The influence of communication style on children’s perception : what influence does the communication style of a volunteer and child have on a child’s evaluation of the chat conversation?

(1)

Master Thesis

The Influence of Communication Style

on Children’s Perception

What influence does the communication style of a volunteer and child

have on a child’s evaluation of the chat conversation?

Tansy Jean Skidmore (10599592)

University of Amsterdam, Faculty of Economics and Business

MSc. in Business Studies – Marketing Track

Under supervision of: Prof. Dr. W.M. van Dolen

Second assessor:

J. Demmers MSc

(2)

ABSTRACT

The expanding use of the internet has given companies the opportunity to communicate and connect directly with customers in new ways, leading to an astonishing growth in e-services over the past decade. Much research has been conducted concerning language use in communication and this study explores and tests some of the research findings in the important new context of e-services. The influence communication style has on customers using e-services is investigated, based on chat data provided by the Kindertelefoon in combination with answers to a questionnaire filled out by children who had chatted with one of the volunteers of the Kindertelefoon. In total 679 questionnaires and chats are used in this study, offering a rich array of data in line with the aim of the research. In this case, the ‘customer’ is a child and the ‘employee’ is a volunteer. To analyze the data, the program LIWC, which is text analysis software that can extract a host of information from a text, is employed. The results indicate that the use of more positive emotive words in a chat will lead to higher satisfaction and to increased recommendation by the child. Likewise, the use of negative emotive words in a chat will lead to decreased recommendation. Moreover, a significant correlation is found between the positive emotive words used by child and volunteer, as well as between the negative emotive words used by child and volunteer. It does lead to the conclusion that the way a volunteer and child communicate certainly has an influence on the outcome of their conversation. These results may well be beneficial for companies, when translated into best business practices and applied in their e-services. Although no significant effect regarding emotive words leading to high or low arousal, LSM, or gender is found in this study, possible explanations are provided and future research opportunities are suggested that may offer more clarity on these topics.

(4)

CHAPTER I – INTRODUCTION

In the Netherlands, a staggering 92.9 percent of the population makes use of the internet (“Internet World Stats”, 2014). Especially children have grown up in the internet era. At the same time retailing has become increasingly competitive through online shopping sites such as Amazon (Ludwig et al., 2013). In response, many companies have started to embrace the internet (Rust & Kannan, 2003). It has become a standard business strategy for once-traditional retailers (such as department stores and specialty stores) to incorporate the internet into their marketing and sales systems (Kwon & Lennon, 2009), allowing customers to compare products and suppliers more easily. Specialized sites have even been developed with comparison as sole purpose.

It is clear that the internet has caused enormous changes in communication, which in turn has important implications for the success of businesses and marketers. This expanding use of the internet has given companies the opportunity to communicate with customers in new ways and to connect directly with their customers. Increasingly, companies are including live communication on their websites, providing customers with the opportunity to have online contact with a company, ask questions or post remarks, as well as rank the service and products of a company (e.g. Gefen & Straub, 2003). Developments in internet technology have led to serving online customers in better ways, for example through 24 hour shopping with more personalized support and reduced response times (Park, Chung, & Rutherford 2011), as well as to improving the efficiency and effectiveness of companies (Smith, 2011). Internet also allows for dialogue with customers, creating the option to individually negotiate product configuration (and price) or required services (Wilson, Daniel, & McDonald, 2002). Internet, combined with wireless communication, has led to an astonishing growth in services over the past decade, with more and more companies implementing customized e-services. E-services can be defined as ‘interactive software-based information systems received via the internet’ (Featherman & Pavlou, 2003, p. 451), which, generally speaking, means providing a service via electronic networks (Rust & Kannan, 2002). Research demonstrates that e-services serve as an effective marketing vehicle; to be used for establishing and maintaining desired relationships with customers (Park et al., 2011). One element of e-services is online chat. Another word for online chat is instant messaging. Instant messaging is defined as: ‘near-synchronous computer-based one-on-one communication’

(5)

(Nardi, Whittaker, & Bradner 2000, p. 80). Chat services are the latest form of computer-mediated communication to gain popularity in and outside the workplace (Garrett & Danziger, 2008).

There are two types of online chat implemented by companies. They either use robotic answering machines, where there is no interaction at all with a human service agent (Rowley, 2006), or (with larger organizations) interaction may involve (many) different human service agents. In this study the focus is completely on actual people chatting with customers. Online chat has various benefits, including the option to save an answer for rereading at any desired time. Online chat also provides the customer with control over when and with whom they interact (Baron, 2005). Moreover, both parties can think about what and how they are going to respond and customers and do not have to fear that others can overhear their conversation (Sindahl, 2011).

However, online chat services also provide challenges. Poorly defined technology can frustrate customers and encourage them to switch to competitors, as competing businesses are now only a ‘mouse click away’ (Anderson & Srinivasan, 2003). Moreover, the online chat advisor’s communication style might affect the success of the service. Advisors might behave with scripted friendliness towards customers, which may feel unnatural, or they might each use a different language style, confusing the customer. The characteristics of service encounters and the dynamics developing will each have impact on how customers respond to a service (Price, Arnould, & Deibler, 1995a). The communication style used by an advisor interacting with a customer may influence the customer’s evaluation. In general, for marketers this means that employee communication style can be used to achieve company objectives in terms of building desired relationships between the business and its customers (Dabholkar, Van Dolen, & De Ruyter, 2009). Therefore, to enhance a company’s objectives it is crucial to match advisor communication style with the target group and the purpose of the chat session (Van Dolen, Dabholkar, & De Ruyter, 2007). Matching the advisor’s communication style to the requirements of the customer will ultimately lead to more efficiency.

A lot of research has been conducted concerning language use in communication, and in this study some of the findings are explored and tested in a new context. As online communication via chat is a reasonably recent phenomenon, little research has been conducted to date regarding the most effective way for a company to communicate via chat. Marketers are interested in how the emergence of e-services influences customer satisfaction and customer

(6)

loyalty, with Park et al. (2011) arguing that e-services can be used to establish and maintain desired relationships with customers, by means of high interpersonal service quality. However, Park et al. (2011) do not elaborate on how e-services might best create high service quality or these desired relationships with customers. Some aspects of this will be investigated in this study.

Research question

The focus of this study is on online chat and customer satisfaction and recommendation. As this study looks at language use and types of words that may enhance customer satisfaction, the results may ultimately be relevant to companies employing human as well as those using robotic e-service chat functions. The study itself solely deals with person-to-person communication and the communication style of an employee and his/her customer during an online chat is linked to the subsequent evaluation of that conversation by the customer. The study is based on secondary data from the Kindertelefoon, where the employee is a volunteer and the customer a child. The research question has been formulated as: What influence does the communication style of a volunteer and child have on a child’s evaluation of the chat conversation?

The Kindertelefoon

The Dutch Kindertelefoon is one of the many organizations turning to online chat services to connect with their customers. The Dutch Kindertelefoon is a non-profit organization, which has been offering anonymous telephone counseling and referral services to children since 1979. In 2003 an online ‘chat’ function was added to communicate with the children, which has proved a great success (Ott & Van Zant, 2013). The Kindertelefoon provides information, advice, and support to children aged between 8 and 18 years. The fact that the children remain anonymous reduces any discomfort regarding discussing controversial issues (Chen & Berger, 2013), making children chat in an honest and straightforward manner. In addition, the study of Fukkink and Hermanns (2009) suggests that the anonymous nature of online chat (‘faceless’ and ‘voiceless’) provides children with a non-threatening type of support, which may encourage children to discuss their personal problems, making online communication more open than offline communication. The Kindertelefoon now yearly receives more than 400,000 phone calls, while over 110,000 children connect via the chat, and 640,000 visits are made to the website. This study makes use of the database of the Kindertelefoon, where all chat conversations are stored, providing a host of information.

(7)

Academic and practical relevance

The central contribution of this study lies in the marketing field, however the study also draws on theories linked to psychology and communication. As e-services are a reasonably new phenomenon, little research has been undertaken concerning the impact of communicating through instant messaging or chat. In other words, little is known about e-services despite its impact and relevance.

The aim of this study is to build on previous literature and to contribute to scientific knowledge about the optimal use of e-services for companies, through the investigation of online chat at the Kindertelefoon, where children chat with volunteers every single day. This study provides insight for marketers, businesses and the society. If language use in e-services can be scientifically proven to be a successful marketing tool, the findings of this study could help lead to enhanced customer satisfaction and recommendation, aiding long-term relationship building, and ultimately resulting in increased revenues for a company. In particular, the Kindertelefoon might see the advantage of implementing the findings of this study. More generally, the findings could be translated into best business practices, illustrating for companies how to further optimize their use of e-services, thus ultimately resulting in higher company revenues or lower costs. Governmental organizations could also potentially implement such best practices, resulting in more efficient and satisfactory customer interaction. Moreover, in the long term, higher company revenues and efficient customer contact could have a positive impact on society by providing growth for companies and thereby stimulating the economy.

Research structure

The outline of this study is as follows. Chapter I introduces the subject matter, states the aim of this study and poses a research question. In Chapter II relevant literature and key conceptual issues are discussed. The focus is on emotion in communication, sharing through arousal, and linguistic style matching. Hypotheses are formulated. Chapter III is concerned with the methods used to test these hypotheses. A detailed account of results and analyses is provided in Chapter IV. Chapter V discusses the implications of the findings, addresses the limitations of the study and poses suggestions for further research.

(8)

CHAPTER II - LITERATURE REVIEW

Online chat

The strong rise in the use of internet has caused an enormous increase in computer-mediated communication. Computer-mediated communication includes a variety of electronic messaging systems. Increasingly interactions take place via e-mail and chat, privately as well as professionally (Derks, Fischer, & Bos, 2008). The chat function is one of the ways companies use e-services. Chat can be used within a company for employees to communicate internally amongst each other. On the other hand, chat can be used externally, so that a company can communicate with customers. The focus of this study is on the external use of chat, namely business-to-consumer, where employees chat with customers.

Opinions differ on whether online chat should be categorized as synchronous communication or as asynchronous communication. Berger and Iyengar (2013) state that online chat is usually somewhat asynchronous. Berger and Iyengar (2013) argue that people can take breaks between responding to chats and that the response can be revised and reviewed before sending, which grants the sender more control over the outgoing message, therefore characterizing it as asynchronous. On the other hand, online chat does allow people to interact in real time and accordingly Derks et al. (2008) assert that online chat is synchronous communication. In this study, online chat has been categorized as synchronous communication. Even though both parties do have some time to form their answer (a characteristic of asynchronous communication), there is no long time delay and communication happens in real time, which are both main characteristics of synchronous communication.

While in the setting of this study online chat is defined as synchronous communication, communication through chat might still be more difficult than face-to-face communication (Gefen & Straub, 2003). There is an absence of social context cues and physical human presence, such as body language and tone of voice, present in face-to-face communication, such as Skype video calls and, to a lesser extent, telephone calls (Byron & Baldridge, 2007). However, Greenfield and Subrahmanyam (2003) argue that chat participants are adapting to the online chat environment by using available cues, and that online chatters thus do not have fewer means at their disposable. Walther, Loh and Granka (2005) found similar results, stating that online chat or e-mail may be as constructive as meeting face-to-face or by videophone, since online chat or e-mail can be highly effective in conveying social

(9)

information through text-based cues. In chat conversations people use emotive words to communicate their feelings and these emotions can either be expressed directly, or substituted by emoticons (Byron & Baldridge, 2007).

Emoticons are defined as vehicles to express emotion (Wolf, 2000, p. 832), and are an abbreviation of ‘emotional icon’. Emoticons are graphic representations of facial expressions (Walther and D’Addario, 2001) and can be used in computer-mediated communication. Emoticons (e.g. smileys) or text-based symbols not only represent emotion, but at times also tone or affect (Byron & Baldridge, 2007). According to Derks et al. (2008) emoticons are mainly used to express emotions, moods, and to strengthen the verbal part of the message. Since emoticons may serve as nonverbal surrogates, suggestive of facial expression, they may add a paralinguistic component to a message (Derks, Bos & Grumbkow, 2007). Also, the use of emoticons reduces uncertainty on the part of the receiver by lowering ambiguity (Kruger, Epley, Parker, & Ng, 2005). Online employees, such as the volunteers at the Kindertelefoon, can be trained to use emoticons to convey a sense of social connection, leading to higher trust in the employee and more commitment (Dabholkar et al., 2009).

The goal of the Kindertelefoon is to help children and provide advice to children, and trust is an important aspect. All children who need someone to talk to or need support in the form of a single conversation may contact the Kindertelefoon (Fukkink & Hermanns, 2009). The Kinderelefoon does not attempt to provide professional counseling (Fukkink & Hermanns, 2009). Often, a child, after discussing his or her problem with an outsider, will see things in a different light and know what steps can be undertaken to solve the problem, ultimately leading to satisfaction and possibly telling a friend or peer about his or her experience with the Kindertelefoon. The focus of this study will be on child satisfaction and word of mouth, and in particular how language and emotion use during a chat session can influence satisfaction and word of mouth. In the following paragraphs these constructs will be discussed in more detail.

(10)

Emotions and satisfaction

Emotion plays an important role in the way people communicate on a daily basis. Not surprisingly, emotions also play a critical role in customer-employee interaction (Mattila & Enz, 2002; Wong, 2004) and in marketing. Here, emotion is conceptualized as: ‘the recognition, expression and sharing of emotions or moods between two or more individuals’ (Derks et al., 2008, p. 767). This study will only focus on emotions as ‘discrete feeling states, such as happiness, fear, and anger’ (Byron, 2008, p. 310), which can have a positive or negative loading and can differ in intensity. This study does not focus on moods, as moods are less likely to be caused by the other’s behavior (Byron, 2008).

Price et al. (1995a) state that it is reasonable to expect that, because the customer is more actively engaged in service encounters, customer emotions are enhanced more by service encounters than by advertisements or even many product purchases. Bitner (1990) is one of the first to investigate satisfaction in a service setting, such as the Kindertelefoon. The research of Bitner (1990) suggests that service employees can act in ways that contribute to customers’ perception of empathy and understanding, thereby enhancing customer satisfaction. Here satisfaction can be defined as: ‘The consumer’s fulfillment response. It is a judgment that a product/service feature, or the product or service itself, provided (or is providing) a pleasurable level of consumption-related fulfillment, including levels of under- or over-fulfillment’ (Oliver, 2010, p. 8). In other words, satisfaction occurs when the expected benefit is exceeded by the perceived benefit (Bauer, Grether, & Leach, 2002).

The findings of Bitner (1990) are in line with the later results of Price et al. (1995a) and of Price, Arnould and Tierney (1995b). Price et al. (1995b) state that as a service encounter is an interpersonal relationship, the affective component influences the outcome of the encounter. They also state that mutual understanding is something more than just a transaction between a service provider and a client, and that the service provider needs to be understanding and engage in emotion work. Authentic understanding and emotion work, in its turn, is strongly correlated to satisfaction by the customer (Price et al., 1995b). The results of the study by Price et al. (1995a) provide some insight into how service employees generate positive or negative customer emotions and therefore satisfaction. These findings may be beneficial for the Kindertelefoon, as the results show that when the service provider is more understanding and engages in emotion work, this will enhance satisfaction by the customer. When translating this to the Kindertelefoon, it implies that when a volunteer is more understanding

(11)

and uses emotion words, this will lead to higher satisfaction by the child.

Menon and Dubé (2000) state that, when the interaction between a salesperson and a customer is in tune with the customer’s emotions, this will improve satisfaction. Van Dolen, Lemmink, Mattsson and Rhoen (2001) found similar results when investigating affective customer responses in service encounters. In their study, they make use of the critical incident technique filled in by customers and by asking customers to evaluate their satisfaction on a ten-point scale. The results show that critical incidents indeed evoke emotional responses in a customer and that negative emotions have a negative impact on satisfaction, while positive emotions have a positive influence. Another finding was that the intensity of emotions plays a substantial role (Van Dolen et al., 2001).

In line with the findings of Menon and Dubé (2000) and Van Dolen et al. (2001), Oliver (2010) observed that emotions can be linked to a customer’s satisfaction response, based on a previous study by Oliver (1993). He found that positive and negative emotions influence satisfaction, in a positive and negative way, respectively: positive emotions increasing customer satisfaction and negative emotions decreasing customer satisfaction. Interesting results were found by Liljander and Strandvik (1997), who concluded that negative emotions have a stronger effect on satisfaction than positive emotions. They also found that situations where individuals experience strong negative emotions are most influential, thus having the highest effect on satisfaction.

Following on from these studies regarding emotion influencing satisfaction, this study specifically concentrates on how the use of emotive words may influence the satisfaction rating of a customer. The findings of Lench, Flores and Bench (2011) focus on word use and state that reading texts that contain emotions can elicit a reaction and influence thoughts and behavior. Thus, the use of positive emotive words may influence the customer, for instance by stimulating positive thoughts. Ludwig et al. (2013) also conducted a study concerning language and word use, and investigated the use of emotion in a study focusing on reviews on online sites. They found that positive affective content, namely ‘happiness words’, in online reviews increased conversion rate. Negative affective content had an even stronger effect than positive content, with negative affective content being more detrimental to a products site’s conversion rate (Ludwig et al., 2013). These findings could also be relevant for the Kindertelefoon. At the Kindertelefoon, the use of positive affective words by the volunteer (comparable to affective content in reviews) could lead to higher satisfaction by the child (as

(12)

in an increased conversion rate by the consumer) and similarly, the use of negative affective words by the volunteer could lead to lower satisfaction by the child.

However, this study does not only focus on the use of positive and negative emotive words by the volunteer at the Kindertelefoon, but also on the use of positive and negative emotive words by the child. Literature illustrates that when customers display positive emotions during an encounter, this will have a strong impact on the evaluation of the quality of the service (Mattila & Enz, 2002). The results of the study by Mattila and Enz (2002) demonstrate that customer emotional displays have important consequences for an organization, as they are associated with the customer’s satisfaction with the encounter. The content of such emotions may be expressed in language (Mattila & Enz, 2002), very similar to the way a child will express itself at the Kindertelefoon. Gallan, Jarvis, Brown and Bitner (2013) found similar results, revealing that as emotion levels of a customer become more positive, this improved the customer’s perception of the quality of the service provider and satisfaction with the service itself. This study was conducted in a health care setting, and in a way the Kindertelefoon can also be seen as a service setting providing mental health care to children.

This study based on Kindertelefoon data builds on this literature by investigating the use of emotion by both volunteer and child during chat conversations and the effect this emotion has on satisfaction. Following on from the results of the above mentioned literature, indicating that positive emotions of both employees and customers positively influence customer satisfaction, it is expected that the use of positive emotive words by the employee (or in the case of the Kindertelefoon: the volunteer) and the customer (or for the Kindertelefoon: the child) will ultimately lead to higher satisfaction by the customer (child). Extrapolating this line of thought to the chat conversations at the Kindertelefoon, this leads to:

Hypothesis 1a: The use of more positive emotive words in a chat will lead to a higher satisfaction rating by the child.

According to the literature cited above, the use of negative emotive words by the employee and the customer will lead to lower satisfaction by the customer. However, this literature does not consider that fact that the Kindertelefoon is a ‘negative’ service. A negative service is a service people in principle would prefer not to require, in which customers cope with unwanted or stressful situations (Miller, Kahn, & Luce, 2008). The phenomenon negative

(13)

service is understudied in the service marketing literature (Singh & Duque, 2012). Customers react differently to negative services than to neutral or positive services (Miller, Luce, Kahn & Conant, 2009). Miller et al. (2009) state that consumers are more likely to evaluate negative services based on feelings and emotions experienced during the service than on the outcome of the service, which would imply that the chat interaction with the Kindertelefoon has a greater influence on the child than the actual outcome of the chat. The study by Singh and Duque (2012) found that with negative services, preexisting stress moderates customers’ evaluation, influencing the satisfaction of the customer. Therefore, helping customers manage stress is important for satisfaction levels, while politeness and speediness of service are more important when customers are not stressed (Singh & Duque, 2012). However, as stress is not dealt with as part of the questionnaire used in this study it is not possible to test the stress level of customers/children contacting the Kindertelefoon. Consequently, this study relies on the general service marketing literature. Therefore is expected that:

Hypothesis 1b: The use of more negative emotive words in a chat will lead to a lower satisfaction rating by the child.

Emotions and word of mouth

Not only does the internet provide companies with new ways to communicate with customers, it also facilitates customer interconnectedness, which can enhance referrals and word of mouth between customers (De Bruyn & Lilien, 2008). Word of mouth may be conceptualized as: ‘Word of mouth is informal advice passed between consumers. It is usually interactive, swift, and lacking in commercial bias. Word of mouth is a powerful influence in consumer behavior.’ (East, Hammond, & Lomax, 2008, p. 215). Word of mouth can either be positive, neutral, or negative about a product or service (Anderson, 1998). Customers share word of mouth face-to-face, over social media, and through a host of other communication channels, with 59 percent of customers stating that they frequently forward online content to others (Allsop, Bassett, & Hoskins, 2007). Due to the internet, people can not only receive referrals from friends and family, but they can also read reviews and recommendations from anyone anywhere.

A key construct in marketing is the willingness of a customer to recommend a service or product to someone else. Therefore, word of mouth is not only frequent, but also important (Berger & Schwartz, 2011), and can affect awareness, sales and customer preference regarding products or services (Godes & Mayzlin, 2004; Berger & Iyengar, 2013). An

(14)

influential study is the one by Reichheld (2003), introducing the Net Promoter Score. This study shows that even if someone buys a product or uses a service multiple times that person does not necessarily needs to be loyal to that product or service. A customer recommending a product or service, e.g. by word of mouth, is what really makes a difference for a company (Reichheld, 2003). Godes and Mayzlin (2004) state that word of mouth may serve as an indicator of a firm’s success and research shows that word of mouth is one of the most influential channels of communication in the marketplace (Allsop, et al., 2007). For the Kindertelefoon, which has 95 percent brand awareness (Ott & Van Zant, 2013), word of mouth is likely to be an important medium in maintaining this high brand awareness. As the Kindertelefoon is a non-profit organization, presumably not much of their income is used for marketing, again implying the importance of word of mouth.

As stated, a significant amount of research has examined the consequences of word or mouth, such as the effect of word of mouth on brand awareness and brand choice. However, much less attention has been paid to its causes, or the reasons people talk about one thing versus another (Chen & Berger, 2013). As this aspect is equally important, a short overview is presented of the different motives for word of mouth by customers. Berger and Milkman (2012) state that people might share knowledge, which they deem to contain useful information. According to Sundaram, Mitra and Webster (1998) this is due to altruism regarding the receiver and, in other words, is helping others make the right decision. Word of mouth could also be used for self-enhancement purposes, thus for positive recognition by others (Wojnicki & Godes, 2008). Another reason for word of mouth is product or service involvement, resulting when customers are extremely happy or satisfied with a brand or service. Wojnicki and Godes (2008) investigated this effect of satisfaction on word of mouth and found that customer experts are generally inclined to generate more word of mouth regarding their satisfying experiences than regarding their unsatisfying experiences. Anderson (1998), however, found different results. He concluded that word of mouth increases when a customer is more satisfied as well as when the customer is more dissatisfied. The results also show that negative word of mouth has a more extreme effect than positive word of mouth, with extremely dissatisfied customers engaging in more word or mouth than extremely satisfied customers. Godes et al. (2005) found that there is some belief that customers are more likely to share negative word of mouth than positive word of mouth and that negative word of mouth can be more credible. Berger and Milkman (2012) also conducted research concerning the on-going debate whether people are more likely to share positive or negative

(15)

content. The findings of Berger and Milkman (2012) were that positive content is more likely to be shared than negative content.

Another factor influencing word of mouth, which needs to be considered, is emotion. Emotions have been shown to be determinants of post-consumption behavior such as word of mouth intentions (Nyer, 1997). Honing in on the causes of word of mouth, and in particular on how emotion influences word of mouth, Heath, Bell and Sternberg (2001) explored how memes can influence the willingness of people to pass on a story. They concluded that emotion plays a vital role in whether a story is shared or not and that, when a story has a particularly high level of disgust, it will easily be passed on. A reason why people communicate more about experiences evoking strong emotions might be to make sense of the situation (Heath et al., 2001).

Yu & Dean (2001) found that positive emotions are positively associated with positive word of mouth and negative emotions are negatively related to positive word of mouth. This means that companies that provoke customers in a negative way, creating negative emotions, will enhance negative word of mouth and companies that provoke customers in a positive way, creating positive emotions, will enhance positive word of mouth. Jones, Keynolds, Mothersbaugh and Beatty (2007) found similar results and state that negative emotions enhance negative word of mouth. White (2010) on the other hand, focused on results regarding positive emotions, and states that positive emotions influence positive word of mouth intentions. The study at hand acknowledges the reasoning that positive emotions are positively related to positive word of mouth and that negative emotions are negatively related to negative word of mouth. Therefore, it is expected that children will be more willing to recommend the Kindertelefoon if their own experience with the Kindertelefoon is positive and that children will be less willing to recommend the Kindertelefoon if their own experience with the Kindertelefoon is negative. Based on the use of positive and negative emotive words the following hypotheses are put to the test:

Hypothesis 2a: The use of positive emotive words in a chat will lead to increased recommendation of the Kindertelefoon to a friend or peer. Hypothesis 2b: The use of negative emotive words in a chat will lead to decreased recommendation of the Kindertelefoon to a friend or peer.

(16)

Sharing (word of mouth) through arousal

Another factor influencing word of mouth is arousal. Ladhari (2007) found that arousal has a significant positive impact on the likelihood of generating word of mouth. The study by Ladhari (2007) shows that moviegoers are more willing to talk about the movie they watched if the movie induced intense emotions, and therefore caused arousal. Recent work on the social sharing of emotion also suggests that positive and negative emotion may increase transmission (word of mouth) when driven by physiological arousal (Berger, 2011). The study by Berger (2011) demonstrates that physiological arousal can plausibly explain transmission of news or information in a wide range of settings. This arousal is caused by emotion. Situations that heighten arousal will boost social transmission, regardless of whether they are positive (e.g. weddings) or negative (e.g. fights) in nature (Berger, 2011). This could mean that whenever a customer feels arousal during a transaction, word of mouth will increase. Other research by Berger and Milkman (2012) proves that online content that evokes high-arousal emotions is more viral, regardless of whether those emotions are positive or negative in nature. While marketers will always try to communicate positive things about their product, and hope customers will share this with others, the results of Berger and Milkman (2012) suggest that content will be more likely to be shared if it evokes emotions, which lead to arousal.

This study adopts the ‘sharing by arousal’ theory and investigates how word of mouth can create extra value for organizations and marketing campaigns. Based on the findings of Berger and Milkman (2012), it is expected that the use of more emotion, which in turn leads to more arousal, will positively influence word of mouth. However, emotions differ strongly and the level of arousal emotions induce may vary (Smith & Ellsworth, 1985). Therefore, the emotive words are separated in this study into emotive words that generate high arousal and emotive words that generate low arousal. Examples of high-arousal emotions are awe (positive emotion) and anger or anxiety (negative emotion). An example of a low-arousal or deactivating emotion is sadness (negative emotion) (Berger & Milkman, 2012). In the case of the Kindertelefoon, it is therefore expected that the use of emotive words in a chat that generate high arousal will lead to increased recommendation of the Kindertelefoon by the child to a friend or peer. The opposite effect is also expected to occur, namely that the use of emotive words in a chat that generate low arousal will lead to decreased recommendation of the Kindertelefoon by the child to a friend or peer. This leads to the following hypotheses:

(17)

Hypothesis 2c: The use of emotive words in a chat that generate high arousal will lead to increased recommendation of the Kindertelefoon to a friend or peer.

Hypothesis 2d: The use of emotive words in a chat that generate low arousal will lead to decreased recommendation of the Kindertelefoon to a friend or peer.

Mimicry

Mimicry refers to the copying of other people’s posture, behavior and facial expression during an interaction (Chartrand & Bargh, 1999). This happens intuitively. Thus, as we interact, we adapt to others by, for instance, copying their movements. Roberts and Aruguete (2000) assert that the reciprocity theory (mimicry) predicts that patients recognize different behavior of their physicians. Patients then respond to this behavior in thematically similar ways, duplicating the behavior of their physicians. Thus, the reciprocity theory assumes that patients recognize, understand, and evaluate their physician according to his/her behavior and also copy this behavior. There are many more examples where people react to others, mimicking behavior.

Emotions can also induce a “mimetic response”, meaning that others respond by expressing the same emotion or behavior (Menon and Dubé, 2000). In the study by Menon and Dubé (2000) this occurred between salesperson and customer. When the customer expressed anger and anxiety, they were faced by mimetic responses to their anger as the salesperson also expressed anger and anxiety. Söderlund and Rosengren (2007) found similar results and state that the sender’s overall emotional state influences the emotional state of the receiver. The sender’s emotions serve as a point of reference for emotional mimicry. Often emotional contagion occurs with the receiver mimicking the emotions of the sender (Söderlund & Rosengren, 2007). Mimetic responses to customers’ emotional displays can be particularly effective when the customer is expressing enjoyment or delight (Wong, 2004). This mimicking can occur across a wide range of behaviors and emotions, including in language use (Niederhoffer & Pennebaker, 2002).

Even though the above mentioned findings were all tested in an offline environment, it is, according to Preece and Maloney-Krichmar (2005), reasonable to expect that interaction styles and behavior of individuals is conveyed from offline to online environments. Following

(18)

the reasoning of Menon and Dubé (2000) and Söderlund and Rosengren (2007), and translating these findings to the Kindertelefoon, we predict that the volunteer and the child chatting at the Kindertelefoon will respond to each other by expressing the same emotion and that due to mimicry, the child and the volunteer will automatically match each other’s emotive words. It is expected that, if the volunteer uses a higher degree of emotive words, both positively and negatively, the child will consequently also use a higher degree of emotive words in response. This may be worded as follows:

Hypothesis 3a: The number of positive emotive words used by the volunteer is correlated to the number of positive emotive words used by the child.

Hypothesis 3b: The number of negative emotive words used by the volunteer is correlated to the number of negative emotive words used by the child.

Linguistic style matching

Language is used in many different settings and everyday people share their internal thoughts and emotions using words, and therefore in a form that others can understand (Tausczik & Pennebaker, 2010). People automatically change their language style as they communicate with different people, an example being talking on the telephone with a child or with a demanding boss (Ireland & Pennebaker, 2010).

Different employees, for instance, in a large company providing (non-automated) e-services, will each use a (more or less) different communication style when communicating with customers. Thus a customer trying to contact this company is faced each time with a different employee using a different language style. This could lead to confusion for the customer and possibly jeopardize the establishment of a long-term relationship. On the other hand, a ‘standard’ style, used by an automatic online assistant, may appear bland and culturally inappropriate to some customers.

The differing characteristics of all service encounters have an impact on how customers respond to a service (Price et al., 1995a). In general, for marketers this means that the communication style of an employee may make or break the relationship between a business and its customers (Dabholkar et al., 2009). Therefore, it is crucial to match an employee’s communication style with the target group and the purpose of the inquiry (Van Dolen et al.,

(19)

2007) in order to enhance the company’s objectives. Matching (an employee’s) communication style to the requirements of the customer may impart a positive impression on the customer (e.g. Ludwig et al., 2013).

In the language used between people, for example a retail advisor and customer, the synchronized use of words, i.e. the matching of words between these two people, is called ‘linguistic style matching’ or LSM (Niederhoffer & Pennebaker, 2002; Ireland & Pennebaker, 2010). LSM measures the degree to which two people use similar language patterns, by looking at nine function word categories. Niederhoffer and Pennebaker (2002) state that words, used by one person, prime the other person to respond in a certain and similar way. These findings are consistent across various settings, including at a conversational level, such as within groups, and on a turn-by-turn level, such as between two individuals. Ireland and Pennebaker (2010) also found that LSM occurs across different contexts, as individuals tend to match each other’s words. Interestingly, they found that individuals are not aware of the matching and are not able to control it.

Research in the 1980s showed that synchrony and mimicking in interactions have a positive effect (Bernieri, Reznick, & Rosenthal, 1988). Findings by Chartrand and Bargh (1999) support this claim and state that synchrony will lead to smoother interaction and a higher ‘liking’. In 2002 Niederhoffer and Pennebaker tested the relation between liking and LSM. In contrast with the nonverbal mimicry literature, they found LSM to be unrelated to communication quality or interpersonal liking. Niederhoffer and Pennebaker (2002) state that this is due to the fact that individuals unintentionally and unconsciously match language, independent of liking.

The findings of Niederhoffer and Pennebaker (2002) are, however, not supported by a study conducted by Ludwig et al. (2013). They found that online reviews with a high degree of LSM with a particular interest group (target audience) have a positive impact on conversion rates, and that new reviews change the linguistic style such that there is a closer match with the interest group’s linguistic style. This positive impact of LSM on conversion rates influences consumer behavior.

Clearly, there are contradicting results concerning LSM and the response by a second party. In this study the LSM theory is also tested to learn whether marketing communication with high LSM by a company and its customers, leads to a higher satisfaction rating by the

(20)

customers. Following the line of reasoning by Ludwig et al. (2013), one would predict that at the Kindertelefoon, a volunteer and child will match each other’s language style. Therefore it is expected that the degree, to which volunteers at the Kindertelefoon accommodate the linguistic style of the child, may influence the satisfaction rating of the child. This leads to posing of the following hypothesis:

Hypothesis 3c: The higher the LSM rating the higher the satisfaction rating by the child.

A slightly separate issue, which might form an interesting topic for further research, is created by another finding, namely by Wolf (2000), that communication style differs between gender, with men and women adjusting their communication style to the gender of their interaction partner. Ireland and Pennebaker (2010) found similar results and state that women consistently score higher on LSM than men. This means that when two women converse together, they are more inclined to match each other’s words. In the setting of the Kindertelefoon, this will be put to an initial test. It is expected that when two females (volunteer and child) chat together this will lead to a higher LSM rating, compared to when one female and one male chat together or when two males chat together. This is formulated as follows:

Hypothesis 3d: Higher LSM will be found when both parties are female than when one or both parties are male.

Figure 1 provides an overview of the hypotheses tested in this study and figure 2 provides a conceptual model.

(21)

Figure 1: Overview of hypotheses

Figure 2: Conceptual model

(22)

CHAPTER III - METHODS

Participants – The Kindertelefoon

During a period of almost ten months, from July 1, 2010, to April 19, 2011, children who had used the Kindertelefoon chat service, were asked to fill in a questionnaire afterwards. These children had either initiated an online chat with the Kindertelefoon via the website or via Windows Live Messenger. To safeguard this important anonymity of the children, other more traditional methods of interviewing, such as face-to-face or by telephone, were not possible. In total, 702 questionnaires were filled in. Of the 702 questionnaires filled in by a child after chatting with a volunteer, 679 were suitable for use in this study. 23 questionnaires were omitted from the study, for reasons such as insincere answering of the questions or discussing a topic not regarded to be genuine. Of the valid sample of 679 participants, 101 were male, 569 were female, while 9 participants chatted in a so-called ‘group’, consisting of more than one person. A strength of the study lies in the total of 679 questionnaires, filled in by children, that have been matched with chats and can be used for the purpose of this study, providing valuable information for answering the research question.

Design

In this study, the independent variables employed are: emotive words used by both the volunteer and the child, and LSM (Niederhoffer & Pennebaker, 2002; Ireland & Pennebaker, 2010). The emotive words used are categorized into positive and negative emotive words, as well as emotive words that lead to high arousal and emotive words that lead to low arousal. LIWC is used to evaluate the emotive words used by the volunteer and the child, as well as to evaluate the LSM of the volunteer and the child. The satisfaction rating given by the child and the likelihood the child will recommend the Kindertelefoon to a friend or peer (word of mouth) are both dependent variables in this study. Also, the use of positive or negative emotive words by the child in response to the use of positive or negative emotive words by the volunteer is considered a dependent variable. The results are analyzed using SPSS.

Procedure

Matching questionnaire to chat

Firstly, the participants chatted with the Kindertelefoon, and thereafter were asked to fill in a questionnaire. The questionnaire included the following questions: the topic of the chat; the category of the topic; the gender of the child; the age of the child; whether the child had

(23)

chatted before with the Kindertelefoon and if so, whether the chat concerned a different subject or not; if the child ever called the Kindertelefoon before on the telephone; the feelings of the child; a satisfaction rating by the child of the Kindertelefoon; whether the child would recommend the service to a peer or friend; the time the child had to wait to be helped; and the duration of the chat. Appendix I provides the full version of the questionnaire.

Though both the chat sessions and the questionnaires were anonymous, almost all questionnaires could be linked to particular chat sessions, based on questions asked and the timing of both chat and completed questionnaire. Questions that aided in linking the questionnaires to a particular chat session were on the subject and the duration of the chat. After each chat session a volunteer would also record the subject of that chat. When linking the subject filled in by the child to the subject recorded by the volunteer, and the chat time estimated by the child to the true chat time, most questionnaires could be connected to the relevant chat. To make sure the correct chat was selected, a final verification was to confirm that the volunteer had indeed passed on the link to the questionnaire to the child at the end of that chat session.

LIWC

With recent advances in computer text analysis, new methods have been developed to explore basic social processes in new and rich ways. A decade ago, many of these analyses could not yet have been performed (Chung & Pennebaker, 2007). To evaluate the communication style of the volunteer, this study also makes use of these recent advances in computer text analysis methods. Specifically, this study makes use of Linguistic Inquiry and Word Count (LIWC). LIWC is text analysis software that can extract information from a text. Text analytics can be used for collecting, categorizing and monitoring customer sentiment as well as gather marketing intelligence. Based on previous studies, LIWC captures on average 86 percent of the words people use in writing and in speech. Such a program can help companies predict customer behavior and more effectively converse with customer target groups (Ludwig et al., 2013).

LIWC analyses one or more text files on a word-by-word basis comparing each word in a given file to the words in an internal dictionary. LIWC can analyze text in at least three ways: analyze all words of the entire conversation, analyze language use per individual during the conversation, and analyze language use per person for each separate turn in the conversation (Niederhoffer & Pennebaker, 2002). In this study, LIWC is used to analyze the chats between

(24)

advisors (in this study, the volunteers) and customers (in this study, the children). At the end of the content analysis, LIWC calculates the total number of times a dictionary word (for instance an emotive word) appears in a chat, divided by the total number of words in the chat, to determine the percentage of the text that falls into a particular linguistic category (Ludwig et al., 2013).

Before entering the chat texts taken from the database of the Kindertelefoon into LIWC, the chat texts were all checked for spelling mistakes, so that LIWC could properly detect the words used. The children often made minor mistakes while chatting with the volunteers. In most cases it was clear what the child meant and the word could be adjusted accordingly. The children also often used abbreviations or ‘slang language’ while chatting. An example is ‘BRB’, meaning ‘be right back’. These were all translated into Dutch, as the dictionary of LIWC is also in Dutch. Next, words that were used frequently by the children but did not appear in the dictionary of LIWC had to be added to the dictionary. Even though LIWC on average captures 86 percent of the words spoken or written, not all words were known and in particular ‘new/ young’ words needed to be added to the LIWC dictionary manually and coded appropriately. Often a similar word or a synonym has been used for the coding, as then these new words could be coded in the same way. In total, 145 words were added to the original dictionary. These words are listed in Appendix II.

For the analysis, the whole chat text of both volunteer and child combined was split into separate parts. Firstly, the whole chat text was saved, and then the chat text of the volunteer and of the child were saved separately. Finally, all the chat texts were subjected to the LIWC program.

LSM

Linguistic style matching (LSM) metric is also calculated by means of LIWC. LSM measures the degree to which two people use similar language patterns by looking at nine function word categories. The nine function word categories used to calculate LSM are: personal pronouns (I, their), impersonal pronouns (it, those), articles (a, the), conjunctions (and, but, because), prepositions (about, in), auxiliary verbs (shall, be, was), high-frequency adverbs (very, rather, just), negations (never, no) and quantifiers (tons, fewer) (Gonzales, Hancock & Pennebaker, 2009). Unfortunately, not all these function word categories are included in the Dutch dictionary of LIWC used in this study. Conjunctions, high-frequency adverbs and quantifiers do not appear in the Dutch dictionary and personal pronouns and impersonal

(25)

pronouns are combined. Therefore, in this study, LSM is calculated based on five out of the nine function word categories.

Having established a usage intensity of a word by means of a percentage in LIWC, the LSM scores are calculated for each function word separately, according to the following formula (here for prepositions):

LSMpreps = 1 - ((│preps1 – preps2│) / (preps1 + preps2 + .0001))

where preps1 is the percentage of total words in text 1 (by the volunteer) that were

prepositions, and preps2 is the percentage of total words in text 2 (by the child) that were

prepositions (Gonzales et al., 2009). The outcome score of LSM is bounded by 0 and 1, and higher numbers represent more language similarity between two people (Ireland et al., 2010). Besides for prepositions, this same formula is also used for the other four function word categories. Thereafter LSM is calculated by averaging the LSM scores for the different function word categories to yield a composite measure of the degree of function word similarity between two texts (Gonzales et al., 2009). In addition to the traditional way of calculating LSM, in this study, LSM is calculated separately for positive and negative emotive words.

(26)

CHAPTER IV - RESULTS

Firstly, the word count per chat varied and the mean number of words per chat was 697 words (SD=424) in total, for volunteer and child combined. The LIWC program captured 80.35 percent of the words used in a chat (SD=2.97) in this study, which is lower than the 86 percent of words spoken or written the program captures on average.

One of the main focus areas of this study is emotive words. Striking about the data is that emotive words, both positive emotive words and negative emotive words, do not account for a large number of the words used by either child or volunteer. Namely, the mean of the positive emotive words used in a chat is 3.47 percent of the total number of words used (SD=1.29), while the mean of the positive emotive words used by a child is 3.73 percent and by a volunteer is 3.22 percent. The mean of the negative emotive words used in a chat is 1.91 percent of the total number of words used (SD=1.04), while the mean of the negative emotive words used by a child is 2.10 percent and by a volunteer is 1.77 percent. The use of affect words in a chat is also not extremely high (M=5.67 percent) (SD=1.47). These figures could be seen as counterintuitive as when children chat at the Kindertelefoon, it is often concerning their feelings.

As mentioned, during chat conversations much use is made of emoticons. Often these emoticons are used to convey emotions, or used complementary to emotive words, perhaps to provide cues to the reader. By counting the use of semicolons (the sign ‘;’) and colons (the sign ‘:’), which are both used to create an emoticon during chats, an estimation of the number of emoticons can be made. The number of colons (M=0.74 percent) (SD=0.83) and semicolons (M=0.25 percent) (SD=0.37) used is relatively high compared to the number of emotive words used, thus possibly providing part of the explanation why emotive words are used less than expected.

To test whether there is a correlation between the use of (semi)colons, emotive words and affect words, Spearman’s correlation (ρ) is used. This is a non-parametric test that can be used when the data have violated parametric assumptions. This test has been selected, because both semicolons and colons are non-normally distributed (Field, 2009). The use of colons and semicolons is positively and significantly correlated with the use of positive emotive words (ρ=0.166 and 0.123, respectively). The use of colons and semicolons is negatively and significantly correlated with the use negative emotive words (ρ=-0.148 and -0.083,

(27)

respectively). Interestingly, no correlation is found between the use of affect words and colons or semicolons. An overview is provided in Table 1.

Table 1: Correlation between the use of (semi)colons, emotive words and affect words using Spearman’s correlation coefficient (ρ)

Correlation (ρ) Colon Semi

Colon Positive emotive words Negative emotive words Affect words Colon 1.000 0.262** 0.166** -0.148** 0.059 Semi colon 1.000 0.123** -0.083* 0.038

Positive emotive words 1.000 -0.243** 0.671**

Negative emotive words 1.000 0.474**

Affect words 1.000

**. Correlation is significant at a 0.01 level (1-tailed) *. Correlation is significant at a 0.05 level (1-tailed)

Another main focus area of this study is satisfaction of the child after chatting with the Kindertelefoon. It is interesting to see that in general, children are satisfied with the services of the Kindertelefoon with a mean of 5.84 (SD=1.37). Satisfaction is tested on a 7-point Likert scale (1 = not satisfied at all, 7 = completely satisfied). The third focus area of this study is the willingness to recommend the Kindertelefoon to a friend or peer by the child after chatting with the Kindertelefoon. Here the mean is 8.82 (SD=2.43), tested on an 11-point Likert scale (0 = not at all likely to recommend the Kindertelefoon to a friend or peer, 10 = extremely likely to recommend the Kindertelefoon to a friend or peer). This shows that, on average, the child is likely to recommend the Kindertelefoon to a friend or peer.

Emotions and satisfaction

Hypothesis 1a: The aim is to test whether the use of more positive emotive words in a chat will lead to a higher satisfaction rating by the child. The use of positive emotive words and the satisfaction rating have both been plotted to test whether they are normally distributed. The graphs show that positive emotive words are normally distributed, but that satisfaction rating is not normally distributed and has negative values of skewness (Field, 2009).

As the satisfaction rating violates the parametric assumption of normal distribution, the Kruskal–Wallis test has been used to test the influence of positive emotive words on satisfaction. This is the non-parametric counterpart of the one-way independent ANOVA. As the hypothesis is directional, a one-tailed test has been selected. The null hypothesis (no

(28)

correlation) is rejected (p=0.000), thus there is a significant positive correlation between the use of positive emotive words in a chat and the satisfaction rating by the child. A regression has been conducted to validate these results. A multinomial logistic regression has been chosen, as the dependent variable is categorical. The results are similar, showing that the model is significant (p=0.000), adding further support to hypothesis 1a.

Hypothesis 1b: The aim is to test whether the use of more negative emotive words in a chat will lead to a lower satisfaction rating by the child. The graph of negative emotion words shows that this variable is normally distributed. Again, as the satisfaction rating violates the parametric assumption of normal distribution, the Kruskal–Wallis test has been used to test the influence of negative emotive words on satisfaction. The same steps as described above have been implemented. The result of the Kruskal–Wallis test is not significant (p=0.139), showing that there is no significant influence of the use of negative emotive words in a chat on the satisfaction rating by the child. Again, a regression has also been conducted to validate the results. A multinomial logistic regression has been chosen, as the dependent variable is categorical. The results are similar, showing no significant result (p=0.097), meaning hypothesis 1b is not supported.

Emotions and word of mouth

Hypothesis 2a: To test whether the use of positive emotive words in a chat will lead to increased recommendation of the Kindertelefoon to a friend or peer, two different tests have been employed. As the likelihood of recommendation is not normally distributed, the Kruskal-Wallis test has been chosen as an alternative for the ANOVA. The result of the test is significant, (p=0.026), thus providing support for hypothesis 2a. As the dependent variable is categorical, a multinomial logistic regression has been used to validate the result. The outcome again being significant (p=0.018), thus also supporting hypothesis 2a.

Hypothesis 2b: To test whether the use of negative emotive words in a chat will lead to decreased recommendation of the Kindertelefoon to a friend or peer, the same tests have been used as for hypothesis 2a. The Kruskal–Wallis is significant (p=0.028), as is the multinomial logistic regression (p=0.038), thus confirming hypothesis 2b.

Sharing (word of mouth) through arousal

Hypothesis 2c: To test whether the use of emotive words in a chat that generate high arousal will lead to increased recommendation of the Kindertelefoon to a friend or peer by the child,

(29)

some pre-processing steps need to be undertaken. Based on the article by Berger and Milkman (2012), the emotive words have been split into words leading to high-arousal and words leading to low-arousal. LIWC makes it possible to split negative emotive words, as the program automatically detects anger, anxiety, and sadness words. Of these, anger and anxiety are categorized as high-arousal emotions and sadness is categorized as a low-arousal emotion. As in more than 50 percent of the chat conversations no anxiety words are used, these data are excluded from the analysis, leaving only anger words as leading to high arousal.

Neither the likelihood of recommendation, nor the use of anger words, is normally distributed, and therefore the Kruskal-Wallis test has been used to test hypothesis 2c. From this test can be concluded that there is no significant effect of the use of anger (high arousal) words on recommendation of the Kindertelefoon by the child (p=0.229) and therefore hypothesis 2c is not supported. Multinomial regression confirms this with a similar non-significant result (p=0.188).

Hypothesis 2d: The aim is to test whether the use of emotive words in a chat that generate low arousal will lead to decreased recommendation of the Kindertelefoon to a friend or peer. The sadness words (leading to low arousal) are linked to recommendation and tested by means of the Kruskal-Wallis test. The Kruskal-Wallis test is chosen because the data are not normally distributed. The conclusion is that there is no significant effect of the use of sadness words on the recommendation of the Kindertelefoon by the child (p=0.152), thus rejecting hypothesis 2d. Again, multinomial regression confirms this result.

Mimicry and linguistic style matching

Hypothesis 3a: To test whether the number of positive emotive words used by the volunteer is correlated to the number of positive emotive words used by the child, a graph has first been plotted of the positive emotive words used by the volunteer and by the child. As the positive emotive words by the volunteer and the positive emotive words by the child are both normally distributed, the Pearson correlation coefficient is applied. This leads to the conclusion that there is a very weak but significant correlation (0.124) between the positive emotive words used by the volunteer and those used by the child, as shown in Table 2. Thus, hypothesis 3a is supported.

(30)

Table 2: Correlation between positive emotive words used by volunteer and child based on Pearson correlation coefficient (r)

Correlation (r) Positive emotive words volunteer

Positive emotive words child

Positive emotive words volunteer 1.000 0.124**

Positive emotive words child 1.000

**. Correlation is significant at a 0.01 level (1-tailed)

Hypothesis 3b: To test whether the number of negative emotive words used by the volunteer is correlated to the number of negative emotive words used by the child, the same test is conducted as for hypothesis 3a. From the Pearson correlation coefficient can be concluded that there is significant, medium correlation (0.426) between the use of negative emotive words by the volunteer and by the child, supporting hypothesis 3b. The results are depicted in Table 3.

Table 3: Correlation between negative emotive words used by volunteer and child based on Pearson correlation coefficient (r)

Correlation (r) Negative emotive words volunteer

Negative emotive words child

Negative emotive words volunteer 1.000 0.426**

Negative emotive words child 1.000

**. Correlation is significant at a 0.01 level (1-tailed)

Hypothesis 3c: The aim is to see whether a higher LSM rating (based on function words) leads to a higher satisfaction rating by the child. How LSM is calculated is explained in the methods section. A multinomial logistic regression has been used to see if LSM has an effect on the satisfaction rating by the child. This test is chosen as the outcome variable (satisfaction) is a categorical variable, and the predictor variable (LSM) is continuous. The results show that LSM does not have a significant effect on satisfaction (p=0.973), and therefore hypothesis 3c is not supported. Additional multinomial logistic regressions are then conducted with regard to LSM, in particular to see whether a higher LSM rating leads to a higher service and accessibility rating by the child. However, the outcome of both these tests is also non-significant.

The LSM rating specifically of positive emotive words is then tested with regard to the satisfaction, the accessibility as well as the service rating by the child. Again, multinomial logistic regressions are used. The tests regarding the satisfaction and accessibility ratings by

The influence of communication style on children’s perception : what influence does the communication style of a volunteer and child have on a child’s evaluation of the chat conversation?

Master Thesis