Gratifications and Seeding Behavior of Online Adolescents

Authors

  • C. Courtois,

    1. Research Group for Media & ICT, Department of Communication Sciences, Faculty of Political and Social Studies, Ghent University-Belgium. Korte Meer 7/9/11, BE-9000 Ghent-Belgium
    Search for more papers by this author
  • P. Mechant,

    1. Research Group for Media & ICT, Department of Communication Sciences, Faculty of Political and Social Studies, Ghent University-Belgium. Korte Meer 7/9/11, BE-9000 Ghent-Belgium
    Search for more papers by this author
  • L. De Marez,

    1. Research Group for Media & ICT, Department of Communication Sciences, Faculty of Political and Social Studies, Ghent University-Belgium. Korte Meer 7/9/11, BE-9000 Ghent-Belgium
    Search for more papers by this author
  • G. Verleye

    1. Research Group for Media & ICT, Department of Communication Sciences, Faculty of Political and Social Studies, Ghent University-Belgium. Korte Meer 7/9/11, BE-9000 Ghent-Belgium
    Search for more papers by this author

Abstract

Several studies indicate that only a small minority of Web 2.0 users actively participates, while the minority do not contribute at all. This article investigates whether a similar division applies for adolescents' Internet behavior. Using Szuprowicz’ (1995) typology of interactivity, we distinguish different types of user-generated content (UGC): media, narrative, and metadata UGC. Our results show a 20%–80% division between high- and low-frequency seeders. Furthermore, we utilize the uses-and-gratifications paradigm to investigate how these high- and low-frequency seeders differ in their overall gratifications obtained by WWW use. Although the gratifications' rank orders are identical for all groups, their magnitudes differ significantly. Finally, this article focuses on how these WWW gratifications can predict seeding, while controlling for socio-demographics and usage frequency.

Introduction

Online digital information has become ubiquitous and accounts for a major part of the economic and cultural activities in western society. Hence, our society calls itself an information or network society. Still, it may be worth remembering that the web is just a 14-year-old—if one considers the introduction of Windows 951 as its day of birth—and subject to all the angst, mood changes, and transformations typical for a young teenager. The start of the dotcom-crisis in 2000–2001 marked the advent of one of these transformations, resulting in new and more interactive and participative models of online communication.

In 2004 the O’Reilly Media group summarized these changes with the phrase Web 2.0. Analogous to the release numbers assigned to software packages, Web 2.0 refers to a newer, better version of the World Wide Web. This new generation of the WWW places emphasis on interactivity, cocreation, and the active role of website users, as user participation is considered to be one of the key notions of digital culture (Deuze, 2006).

Despite the optimistic discourse on participation and user-generated contributions, several studies (cfr. infra) demonstrate how reality draws another picture: only a small minority actively participates (referred to as seeders) while the majority does not contribute at all (free-riders or leechers). In this article we investigate online participation among adolescents. This group of digital natives (Prensky, 2001) has a unique status for they are and were raised with the Internet as omni-present.

This article describes Web 2.0, sketches the users' centrality in its structures, looks into the nature of users' contributions and offers an overview of contemporary findings regarding the inequality in contributing (referred to as seeding). The first research question investigates how contemporary adolescents, assumed to be technologically savvy, divide into low- and high-frequency seeders (RQ 1). The second research question looks into the gratifications that adolescents obtain by their WWW use and explores how these gratifications differ for the different seeding groups (RQ 2). The final research question investigates whether these obtained WWW gratifications' magnitudes predict seeding behavior (RQ 3).

Literature Review

Web 2.0: an architecture of participation?

The actual meaning of Web 2.0 is still subject to discussion. Moreover, various authors emphasize the hyped character of the phrase (see e.g. Graham, 2005; Stern & Wakabayashi, 2007; O’Reilly, 2007). Despite these difficulties Web 2.0 has become a central concept in contemporary discussions about the Internet. Web 2.0 helps the user to overcome the technical obstacles that were in his or her way earlier (Harrison & Barthel, 2009; Cormode & Krishnamurthy, 2008; Depauw, 2008), making the Internet an instrument for and by the mass user. One can easily consume (read, listen, watch, download, search, and buy), create (personalize, aggregate, and contribute), share (publish, upload), facilitate (tag, recommend, filter, subscribe to channels and items through RSS) and communicate (send messages, post comments, rate, and chat) online (Slot & Frissen, 2008).

Thus, websites are no longer merely static online pages but are evolving into platforms “(…) delivering software as a continually updated service that gets better the more people use it, consuming and remixing data from multiple sources, including individual users, while providing their own data and services in a form that allows remixing by others, creating network effects through an ‘architecture of participation’, and going beyond the page metaphor of Web 1.0 to deliver rich user experiences.” (O’Reilly, 2003, 2005).

One of the key aspects of Web 2.0 is the rise of user participation. The user network and the delivered contributions dominate (the discourse about) the Internet entirely, expressing itself in data streams from the user's side, the prominence that items obtain based on their popularity, the trust in the contribution of other users, the use of folksonomies2, etc. (O’Reilly, 2005; Andersen, 2007; Aguiton & Cardon, 2007). Not surprisingly, many definitions of Web 2.0 employ a user-centric approach to describe the concept (see e.g. Millard & Ross, 2006; Kolbitsch and Maurer, 2006; Maness, 2005; White, 2007; MacManus & Susan Mernit, both cited in Dybwad, 2005). Cooke and Buckley's definition serves as an example of this user-centric approach towards Web 2.0: “Web 2.0 is about making computing and media social.” (2008, p. 277). Web 2.0 enables and strongly encourages user-generated content (UGC). UGC is defined as content made publicly available on the Internet, reflecting a certain amount of creative effort. UGC is created outside of professional routines and practices (OECD, 2007) and exists in different shapes and sizes. Users of Web 2.0 platforms or services express themselves in a variety of ways: they upload photo, video, and audio files; tag, rate, or comment on online content; and converse in the blogosphere or on online forums. In the following paragraph we illustrate this by means of three popular and often-cited Web 2.0 websites; Flickr, YouTube and Delicious.

Flickr3 enables users to manage, organize, archive and share their photographs. It provides an open API4, pre-packaged licensing models, tagging and other community involvement mechanisms. YouTube5 has become the largest video sharing Web site online with over 100 million video accesses per day and 65,000 video uploads on a daily basis (Gill, Arlitt, Li, & Mahanti, 2007). YouTube is a free online video streaming service enabling users to view and share videos that have been uploaded by other users and providing mechanisms such as tagging, rating, and syndicating content. It offers anyone the means to communicate to a mass audience, a power once held exclusively by television networks. This strategy proved to be very successful; the total amount of content uploaded as of March 2008 is estimated on 78.3 million videos (Wesh, 2008). The site Delicious6 is a free social book marking website where users can store their bookmarks and access other users' favorite online resources. Tags (one-word descriptors) and bundles of tags are used to organize and store bookmarks. In this way, over 180 million unique URLs have been saved by Delicious users (Hood, 2008).

However, several authors have signaled the lack of participation and contribution on websites or Internet services (see e.g. Jones, Ravid, & Rafaeli, 2004; Joyce & Kraut, 2006; Kraut, Egido, & Galagher, 1990; Nonnecke & Preece, 1999; Nonnecke & Preece, 2000; Preece, Nonnecke & Andrews, 2004), even long before the advent of Web 2.0. For instance, Butler (1999, described in Cummings, Butler & Kraut, 2002) examined a random sample of listserv-based online groups and concluded that one-third of all lists saw no activity during a 3-month observation period. Whittaker, Terveen, Hill and Cherny (1998) studied Usenet and discovered a ratio of 580,000 low-frequency contributors versus 19,000 high-frequency contributors. Jones examined a sample of 578 Usenet newsgroups in 1999 and found that only 11.5% of people who posted in one month returned to post in a second month (Jones, Ravid & Rafaeli, 2004).

More recently, Bughin (2007) argued that only a fraction of Internet users is responsible for the majority of the content added to the Internet (seeders). For instance, figures show that about 2% of the Wikipedia users are responsible for 60% of the articles, on YouTube 6% of the users post 90% of the videos (Bughin, 2007). Prieur, Cardon, Beuscart, Pissard, & Pons (2008) report that 20% of the users of the website Flickr own more than 82% of the photos and usability expert Nielsen (2006) stated that in most online communities, 90% of users are lurkers who never contribute, 9% of users contribute a little, and 1% of users account for almost all the contributions. Several authors have created typologies describing users by means of their participation behavior or activity level, see for example: Barker & Schoggen (1978); Blanchard (2004); Haas, Trump, Gerhards, & Klingler (2007); and Charlene Li of the Forrester Group (2007).

These findings illustrate the persistent division between a large majority of WWW users who rarely contribute and a small minority that regularly engages in content seeding. As previously argued, Web 2.0 websites and services offer structures for the storage, management and the disclosure of user-generated content. Obviously, these sites refrain from rendering service when there is no content available. Without narrative and media content, web structures would lack substance, becoming completely irrelevant. The absence of metadata, accounting for the creation of folksonomies, recommendation systems, etc. would lead to chaos. By consequence user-driven contributions are a necessity for the experience, efficacy, usability, and searchability of Web 2.0 sites. A smoothly functioning Web 2.0 depends on user interactivity to nourish its participative architecture, collective intelligence, and dynamic structures.

Adolescent WWW use: A Net-Generation?

In this article we focus on adolescents, given their specific status of being digital natives (Prensky, 2001). Today's adolescents, regarded by Berk (2007) to be commonly aged somewhere between the age of 12 and 18 years, are growing up with the Internet as an absoluteness. They are part of the first generation ever to natively speak the digital language of the Internet and digital technology in general. Analogue to the WWW, they are experiencing a phase of transition, struggling out to reach maturity.

This transition into adulthood comprises physical as well as cognitive and emotional changes. It is a time in which the ability to think more effectively and complexly dramatically increases. This brings about critical thoughts on oneself, others, and the surrounding world (Lehalle, 2006; Byrnes, 2006). Teenagers reflect upon who they are, what they stand for, and where they are heading to, thus forming a concept of identity (Erikson, 1968; Marcia, 1980). At the same time, due to a amplified sense of autonomy, adolescents get more detached from their parents, increasingly make decisions of their own and become emotionally independent (Goossens, 2006). They de-idealize their parents while on the other hand peer relations gain importance (Scholte & Van Aken, 2006; Bradford Brown & Klute, 2006). Young adolescents are also keen on their privacy, on having a space of their own. Livingstone (2002) notices an emerging bedroom culture. Teenage bedrooms offer comfort and refuge, while at the same time these rooms become extensively mediatized. Technologies such as computers and the Internet are largely domesticated and are often at immediate and private disposal. No less than 42% of the Flemish youth have computers in their rooms, while 98% share one with the household (Graffiti jeugddienst & Jeugdwerknet, 2008). This is reflected in figures on Flemish teenage Internet usage: As early as the age of 10 to 12, 96% of the Flemish teenagers use the Internet, 63% even three times a week to daily (Valcke, Schellens, Van Keer & Gerarts, 2006). These dazzling figures remain stable or even increase with age (FOD Economie, 2006; OIVO, 2007).

Given the omni-presence and availability of computers and the Internet, several authors such as Tapscott (1998) and Leung (2003, 2004) propose the existence of a Net Generation: a technologically savvy generation that shares a global orientation, open-mindedness and a strong belief in equality. They are preoccupied with maturity and want to be treated with respect, based on their contributions, not on their age. However, Schulmeister (2008) strongly opposes the use of the term Net Generation, stressing the dissimilarities over the similarities in Internet usage and skills of young people. Furthermore, various authors (e.g. Livingstone & Bober, 2005; Jenkins, Purushotma, Clinton, Weigel & Robison, n.d.) point to the online participation gap, dividing adolescents for whom the Internet is of growing importance from those who use the Internet rarely and in a trivial, insufficient way.

Adolescents and the creation of user-generated content

Research on the creation of online user-generated content is often conducted on American Internet users (e.g. Lenhart, Madden, Macgill, & Smith, 2007). Moreover, the scarce data on UGC in Europe and its member states is rather difficult to compare and interpret as measuring UGC is not straight forward (OECD, 2007). Below we describe three recent studies conducted on Flemish adolescent UGC activities. A study by OIVO (2008) encompassing (N=) 2662 teenagers showed that 32% publishes a personal blog and that 15% takes part on discussion forums. Another Flemish study surveying (N=) 919 secondary school students on their media use concluded that UGC activity is rather limited. For instance, only 12% of the teenagers adds posts to a forum, 8% posts on a personal blog, while 22% actively uses YouTube (for rating, commenting, and/or uploading video clips) (Graffiti jeugddienst & Jeugdwerknet, 2008). A study conducted on the perceived risks of Internet use by (N=) 1318 adolescents (12–18y) shows higher frequencies (TIRO, 2008). According to this study, conducted in 2008, 41% of the respondents has a blog account. The figures above show that it is difficult to make conclusive arguments on the creation of UGC by Flemish adolescents. As a result, the question remains as to what extent adolescents adopt the practice of contributing content to the Web. Do they heavily engage in seeding, given their desire to be noticed and to express themselves, or do they resemble the previously mentioned general Internet population that rarely contributes?

Motives for WWW use: uses-and-gratifications

It still remains unclear why an individual takes up contributing content, while he or she might as well free-ride upon the waves of collective contributions. What individual differences function as precursors for the engagement in content seeding? A plausible explanation lies within the activities' gratificational abilities. High- and low-frequency seeders undisputedly have one thing in common: they both use the WWW and obtain a certain amount of gratification from it. Could it not be possible that differences in WWW users' motives are related with the engagement in seeding activities? To investigate this matter we draw upon the uses-and-gratifications (U&G) paradigm.

The U&G paradigm embodies 40 years of functional media research. It assumes an active audience that is conscious of its needs while deliberately selecting media to gratify these needs. Media constantly compete with each other and other means of gratification (Katz, Blumler, & Gurevitch, 1974; de Boer & Brennecke, 1995; Rubin, 2002). At least three gratification sources are distinguished in the literature: medium content, the exposure to the medium and the social context of the situation in which the medium is used. Each medium offers a unique combination of these three sources. In this manner media differ in the way they offer satisfaction to a certain need (Katz et al., 1974).

U&G has been used to grasp the latent motives of numerous media. From traditional media, such as television, books, movies, and radio (McQuail, Blumler & Brown, 1972; Katz, Gurevitch, & Haas, 1973) to more recent, interactive media such as video games (Selnow, 1984), the video recorder (Cohen, Levy & Golden, 1988), and the mobile telephone (Leung & Wei, 2000). However, numerous attempts to theorize upon U&G largely failed. U&G has often been criticized for its vague conceptual framework, its lack of precision in its core concepts, its confusing declaring apparatus and its inability to consider perception as an active process (Swanson, 1977; de Boer & Brennecke, 1995). From a cultural angle its individualistic perception, seemingly oblivious to the broad social context, is pointed out. Furthermore it lacks attention for the construction of meaning and assumes that media have to be functional (Ang, 1990). Despite these imperfections, U&G as an approach offers a solid, reliable framework for the inquiry of mass media usage (Ruggiero, 2000). In this article we choose to pursue this path, drawing upon the methodological benefits of the U&G approach.

Recently, U&G has regained its prominence by focusing upon the active audience. Thus, it has become especially favorable in the research of the Internet. Table 1 offers an illustrative overview of existing research on U&G for the Internet.

Table 1.  “Overview of some Internet Uses-and-gratifications typologies.”
 Personal FunctionInformational FunctionEntertaining FunctionEscapist FunctionSocial Function
Grace-Farfaglia, (2006)Self-Improvement, Fame and AestheticsEconomic GainEntertainment,EscapeSocial Companionship
Lin (1999) Information-LearningEntertainmentEscapeInteraction
Parker & Plank (2001) SurveillanceExcitement and RelaxationEscapeSocial Relationships, Companionship
Papacharissi & Rubin (2000) Information Seeking, ConveniencePass time Entertainment Interpersonal Utility
Leung (2003)Social IdentitySurveillanceEntertainmentEscapeAffection, Social Bonding

These patterns of gratifications show remarkable parallels: Internet usage typologies resemble each other as well as those of other media. As such we recognize a personal function reflecting a representation of oneself, an informational function regarding the finding of adequate information, an entertainment function including enjoyment and relaxation, an escapist function enabling a flee from daily worries, and a social function that reflects the social interaction through the medium.

Inquiry

In the literature review we sketched out the prominence of user-generated content in Web 2.0 websites. However, numerous empirical findings report a sharp division between WWW users who abstain from seeding and those who do seed on a regular basis. Our first research aim is to investigate how WWW users divide into low- and high-frequency seeders (RQ 1). Furthermore we investigate the relations between the gratifications obtained by WWW use for both low- and high-frequency seeding. We explore whether gratifications obtained differ for the different seeding groups (RQ 2). Next, we examine the predictive value of obtained WWW gratifications concerning seeding group membership (RQ 3).

Sample

Our inquiry was conducted on a selective sample of (N=) 836 adolescents in Flanders, (45% male, 55% female) aged 12 to 18 years (M = 15, SD = 1.86). Flanders is the northern part of Belgium, a Western European country characterized by a high degree of Internet connectivity (60%) and computer penetration (67%) (FOD Economie, 2007). ICT use is strongly encouraged in Flanders; for example, the government imposes obligatory goals for formal education: e.g. “The pupils [have to] learn to (…) recognise the opportunities of the use of new technologies (such as ICT) and new media (…).” (DVO, 2001).

Our sample showed no significant age difference between both genders (t = 1.32, df = 834, p > .05). Nine secondary schools were carefully selected to obtain a balanced distribution of socioeconomic status (SES) based on a sensible mix of types of education7. The respondents were reached by distributing approximately 4500 flyers at the schools' gates. The flyers invited the adolescents to fill out an online survey that ran from October to November 2007. To promote participation, several incentives such as cinema tickets and a MP3-player were raffled among the respondents. No respondents were recruited online in order to avoid difficulties with the answers' validity.

Measures

Seeding behavior

The measurement of seeding behavior draws upon interactivity as a multidimensional framework for user-generated content, conceptualizing interactivity by dividing information streams into sender and receiver. Employing the division between user-to-user, user-to-document, and user-to-system interactivity (McMillan, 2006), Szuprowicz (1995) distinguished three main categories of interactivity: (end- )user-to(-end- )user, (end- )user-to-documents and (end- )user-to-computer. The first interaction process concerns direct communication and cooperation between two or more users and addresses relational or interpersonal control. User-to-documents interaction refers to the consultation of ‘fixed’ content and the control that users exert over the document or content. The third category, user-to-computer interaction, can be seen as the interaction between the user and the computer or web server itself and implies control over the process or the interface. In our study we distinguish media, narrative, and metadata user-generated content (UGC):

Media seeding comprises UGC created through an intensive process of user-to-computer interaction. Three items are measured: uploading (a) a video file, (b) an audio file, and (c) a photo. All three actions are preceded by intensive interaction with the computer, the Internet (content has to be selected, compressed, uploaded, … ), and other types of new media (content has to be recorded with a digital (video)camera, mp3-recorder, smart phone, etc.). The items were preceded by the phrase “how often do you upload…” and were answered with (a) never, (b) monthly, (c) weekly, (d) every 2–3 days, and (e) daily.

Narrative seeding involves a user-to-user interaction. It can be viewed as a result of feedback, monologue, mutual discourse or responsive dialogue user-to-user interaction (McMillan, 2006). In the inquiry narrative seeding comprises two items: posting (a) a weblog entry and (b) a forum entry. The items were preceded by the phrase “how often do you post…” and were answered with (a) never, (b) monthly, (c) weekly, (d) every 2–3 days and (e) daily.

Metadata seeding involves UGC created in the process of user-document interaction. Metadata seeding is measured by three items: adding (a) a tag, (b) a rating or (c) a comment. Tags and ratings are examples of metadata that result in cocreated content (McMillan, 2006): tags create folksonomies while ratings enable collaborative filtering. Commenting is considered metadata because of its object-oriented nature: it only makes sense when it acts upon existing content. The items were preceded by the phrase “how often do you add…” and were answered with (a) never, (b) monthly, (c) weekly, (d) every 2–3 days, and (e) daily.

Gratifications obtained are measured by a scale of 20 Likert statements (5-point metric, see appendix) whereas every item is preceded by “By means of my WWW use…”. The items are distributed over the gratification types mentioned in the literature review: personal function (N = 3), informational function (N = 4), social function (N = 6), escapist function (N = 4), and entertaining function (N = 3). The items originate from sources mentioned in table 1, heavily drawing upon Leung (2003) for its focus on adolescents. The initial item pool's validity was checked by two preceding focus groups, tapping into motivations for using the WWW.

Quantity of Internet use is measured by 2 variables. Firstly: “How often do you use the Internet?” whereas answering categories include (a) monthly, (b) weekly, (c) every 2–3 days and (d) daily. Secondly: “How long do you use the Internet per session?” with (a) 0–1 hours, (b) 1–2 hours, (c) 2–3 hours, (d) 3–4 hours, and (e) 4+ hours as possible answering categories.

Sociodemographics

These measures comprise age, gender, and the education level one currently attends. Education level is regarded to be an ordinal variable: Vocational secondary school is the lowest, followed by technical secondary school. General secondary education is considered to be the highest in rank.

Results

RQ 1: How do users divide into high- and low-frequency seeders?

As mentioned in the measures section, user-generated content is divided into three types. To answer this research question, the k-means clustering algorithm was thus applied on three separate sets of variables. These sets of items implement media seeding (N = 3; Frequency of uploading photos, video clips, and audio files), narrative seeding (N = 2; Frequency of posting blogs and forum messages), and metadata seeding (N = 3; Frequency of adding rates, tags and comments). The default amount of clusters is set to 2, dividing all three groups into (a) low-frequency (low-f) seeders and (b) high-frequency (high-f) seeders. The choice for a k-means clustering approach over the use of cut offs is a deliberate and pragmatic one, offering an elegant and parsimonious solution. Figure 1 depicts the plots of the final cluster centers.

Figure 1.

“Plots of the k-means final clusters centers. 0 = never, 1 = monthly 2 = weekly, 3 = every 2–3 days and 4 = daily”

When we look at media seeding, the popularity of uploading photos immediately shows. Uploading movie clips shows a rather small differentiation between low- and high-f seeders. The same goes for posting on a forum, whereas blog posting diverges. As for metadata seeding, adding comments and rates exceeds the popularity of adding tags. Subsequently we investigate the properties of all three seeding groups. Table 2 offers a summary of descriptive statistics for all three groups.

Table 2.  “Overview of media, narrative and metadata seeding group properties.
1: K-means converged in 7 iterations, sig. differences for all thee clustering variables were found (Mann-Whitney, p < .001)
2: K-means converged in 5 iterations, sig. differences for all two clustering variables were found (Mann-Whitney, p < .001)
3: K-means converged in 6 iterations, sig. differences for all three clustering variables were found (Mann-Whitney, p < .001)”
Metadata seeding1Low-f seedersHigh-f seedersSig. 
(N = 836, 100%)(N = 697, 83.4 %)(N = 139, 16.6 %)diff.Test statistic
Gender46.2% male, 53.8% female38.8% male, 61.2 % femaleNoPearson χ2 = 2.53, df = 1, p > .05
AgeM = 15.03, SD = 1.88M = 14.84, SD = 1.79NoIndependent t = 1.09, df = 834, p > .05
Session frequencyMean rank = 410.32Mean rank = 459.53YesKruskal-Wallis, χ2 = 9.37, df = 1, p < .05
Session durationMean rank = 397.20Mean rank = 525.33YesKruskal-Wallis, χ2 = 35.23, df = 1, p < .05
Education levelMean rank = 429.35Mean rank = 364.09YesKruskal-Wallis, χ2 = 10.36, df = 1, p < .05
Narrative seeding2Low-f seedersHigh-f seedersSig. 
(N = 836, 100%)(N = 646, 77.3%)(N = 190, 22.7%)diff.Test statistic
Gender46.1% male, 53.9% female41.1% male, 58.9% femaleNoPearson χ2 = 1.53, df = 1, p > .05
AgeM = 15.11, SD = 1.89M = 14.63, SD = 1.73YesIndependent t = 3.15, df = 834, p < .05
Session frequencyMean rank = 404.64Mean rank = 465.62YesKruskal-Wallis, χ2 = 18.24 df = 1, p < .05
Session durationMean rank = 389.18Mean rank = 518.17YesKruskal-Wallis, χ2 = 45.23, df = 1, p < .05
Education levelMean rank = 421.03Mean rank = 409.88NoKruskal-Wallis, χ2 = .38, df = 1, p > .05
Metadata seeding3Low-f seedersHigh-f seedersSig. 
(N = 836, 100%)(N = 636, 76.1 %)(N = 200, 23.9 %)diff.Test statistic
Gender43.6% male, 56.4% female49.5% male, 50.5% femaleNoPearson χ2 = 2.17, df =, p > .05
AgeM = 15.02, SD = 1.90M = 14.93, SD = 1.73NoIndependent t = .68, df = 363, p > .05
Session frequencyMean rank = 402.88Mean rank = 468.18YesKruskal-Wallis, χ2 = 21.67 df = 1, p < .05
Session durationMean rank = 383.47Mean rank = 529.89YesKruskal-Wallis, χ2 = 60.39, df = 1, p < .05
Education levelMean rank = 426.06Mean rank = 394.46NoKruskal-Wallis, χ2 = 3.19, df = 1, p > .05

A trend that immediately shows is the absence of significant differences among gender. Although high-f seeders tend to be slightly younger, all mean ages lay within close proximity of the sample's mean (which amounts to 15). The same goes for the standard deviations (Sample SD = 1.86). Only narrative seeders differ significantly in age. Another consistency is the lower education level of high-f seeders, although only significant for media seeding. High-f seeders share an elevated frequency of WWW use: they go online more often and spend more time per session.

When we look at the relations among media, narrative, and metadata seeders, a descriptive analysis reveals the majority of the respondents (N = 520; 62.2%) to be low-f seeders for all three groups, whereas only 67 (8%) are overall high-f seeders. 20.3% are high-f seeders in only one group (4.2% media seeders, 8.6% metadata seeders and 7.5% narrative seeders), while 9.4% combine membership of two high-f seeding groups. In order to fully explore the relations among seeding groups we performed a set of three binary logistic regressions using each seeder group as a dependent while the remaining two serve as covariates (low-f seeding = 0, high-f = 1). Note that all variables were standardized prior to the analysis, thus coefficients should be read as standardized ß's. Table 3 summarizes the regression results.

Table 3.  “Summary of the binary regressions. *: Wald statistics p < .001
1: Omnibus χ2 = 166.07 (df = 2, p < .001), Hosmer & Lemeshow χ2 = .604 (df = 2, p > .05), Nagelkerke R2 = .30
2: Omnibus χ2 = 187.42 (df = 2, p < .001), Hosmer & Lemeshow χ2 = .604 (df = 2, p > .05), Nagelkerke R2 = .31
3: Omnibus χ2 = 181.92 (df = 2, p < .001), Hosmer & Lemeshow χ2 = .604 (df = 2, p > .05), Nagelkerke R2 = .29”
DependentsCovariatesßSEWaldOdd Ratio
Media seeding1Narrative seeding.68.0952.62*1.97
 Metadata seeding.65.1047.21*1.92
 Constant−2.02.12266.54*.13
Narrative seeding2Metadata seeding.71.0969.57*2.04
 Media seeding.60.0852.62*1.83
 Constant−1.47.10214.05*.23
Metadata seeding3Media seeding.57.0847.21*1.77
 Narrative seeding.70.0869.57*2.01
 Constant−1.37.10200.16*.25

Table 3 immediately shows that all seeding groups offer significant predictions (p < .001). All of the odds ratios appear to be substantially larger than one. This indicates an increase in the likelihood of being a high-f seeder in the dependent group while being a high-f seeder in the covarying group. This increase equals the reported magnitude of the odd ratio, while controlling for the effect of the other seeding group. The regression results offer evidence for significant, though moderate relations among seeder groups. A very strong relation, or even a substantial overlap, would lead to a much higher amount of variance explained.

RQ 2: Do seeding groups differ in the gratifications obtained by WWW use?

To answer this question we first need to grasp the gratifications obtained by WWW use. Our measurement consists of 20 items, properly tested for internal consistency and item-total correlations. An exploratory factor analysis (EFA) using an orthogonal Varimax rotation reveals a 5-factor structure accounting for (cumulative R2= ) 62.44% of the variance.

These five factors and its contents strongly resemble those found in the studies summarized in table 1 of the literature review. The rotated factors are named Identity signaling (Cronbach α = .81, Eigenvalue = 2.10, R2 = 10.49; personal function), Surveillance (Cronbach α = .79, Eigenvalue = 2.52, R2 = 12.61; informational function), Social Relations (Cronbach α = .84, Eigenvalue = 3.48, R2 = 17.39; social function), Escapism (Cronbach α = .74, Eigenvalue = 2.33, R2 = 11.63; escapist function) and Entertainment (Cronbach α = .75, Eigenvalue = 2.07, R2 = 10.33; entertainment function).

Table 4 depicts the zero-order Pearson correlations between all items, showing that correlations among items of the same construct are consistently higher than those with unrelated items. As such, items measuring the same concept show convergence while discriminating themselves from other concepts. This occurrence offers supporting evidence for construct validity.

Table 4.  “I = Identity signaling, S = Surveillance, SR = Social Relations, E = Escapism, and EN = Entertainment. All item labels are printed in the appendix section.”
 I1I2I3S1S2S3S4SR1SR2SR3SR4SR5SR6E1E2E3E4EN1EN2EN3
I11,00                   
I20,711,00                  
I30,510,521,00                 
S10,120,080,091,00                
S20,140,160,080,671,00               
S30,100,110,130,520,491,00              
S40,090,100,060,420,400,431,00             
SR10,230,350,170,120,220,110,201,00            
SR20,280,390,250,100,120,090,180,551,00           
SR30,260,380,200,090,180,080,210,520,451,00          
SR40,230,280,200,070,170,100,020,510,460,451,00         
SR50,280,350,190,040,100,050,120,560,460,400,441,00        
SR60,240,350,190,120,190,130,280,470,420,570,370,401,00       
E10,230,260,210,040,010,010,130,170,170,270,100,140,241,00      
E20,290,310,250,060,080,010,140,160,220,210,170,140,200,581,00     
E30,200,190,170,060,050,010,060,120,110,120,090,120,120,410,371,00    
E40,230,230,23−0,06−0,060,020,010,090,140,120,060,100,120,420,380,341,00   
EN10,210,220,200,050,010,120,100,180,260,240,170,140,190,100,100,060,141,00  
EN20,200,240,210,020,030,130,150,250,310,290,230,200,250,130,090,090,190,671,00 
EN30,190,130,140,100,100,170,210,130,170,230,040,150,180,170,100,130,160,440,391,00

Subsequently a confirmatory factor analysis was performed, using structural equation modeling (SEM), proving the model to fit the data and yielding an acceptable relative goodness of fit (Tucker Lewis Index = .918, Comparative fit index = .931, Root mean square error of approximation = .055). Due to the large sample size, an absolute fit was not accomplished (χ2 = 565, df = 160, p < .001). Figure 2 depicts the model.

Figure 2.

“Confirmatory factor analysis of the gratifications obtained by WWW use.”

To fully answer the second research question, we executed a series of t-tests comparing the mean score per seeding group for the five gratifications obtained (Table 5).

Table 5.  “Overview of the independent t-test results. * = p < .05, ** = p < .001. + = low-frequency seeders, ° = high-frequency seeders”
 Media seedersNarrative seedersMetadata seeders
MeantdfMeantdfMeantdf
Lf+Hf°Lf+Hf°Lf+Hf°
Identity signaling2.412.71−3.66*8342.372.76−5.33**8342.412.61−2.77*834
Surveillance3.183.19−.068343.163.25−1.218343.123.39−3.67**834
Entertainment4.214.41−3.23*8344.204.39−3.45*8344.174.48−6.12**383
Escapism2.642.85−2.52*8342.602.92−4.29**8342.602.92−4.57**834
Social relations3.313.77−6.03**8343.283.75−7.89**3833.283.71−6.35**834

A striking consistency reveals itself: All three seeding groups (low-f as well as high-f) share the same relative ranking of obtained gratifications. This implies that all seeding groups share a similar pattern of WWW gratification. The contemporary WWW appears to be especially satisfying the need for entertainment, followed at distance by social relations and surveillance. Escapism and identity signaling appear to be of lesser importance. Nonetheless, the absolute scores for these gratifications are consistently higher for high-f seeding groups. All but two mean differences (surveillance by media and narrative seeders) appear to be significant.

RQ 3: Do obtained WWW gratifications predict seeding group membership?

In answering this question three logistic regression functions are computed, each time using a different seeding group as dependent (table 6). The covariates include all five gratifications obtained by WWW use, while session duration, session frequency, gender (0 = male , 1 = female ), age, and education serve as control variables. As literature on U&G points out, the mere usage of a medium is a source of gratification (Katz et al., 1974; Stafford, Stafford & Schkade, 2004). As such, the exclusion of quantity of Internet use, being much higher for high-f seeders, would inflate the gratifications' predictions of seeding behavior. The sociodemographic variables are also entered to further isolate the obtained gratifications and to examine their unique relations with seeding behavior. Again, all variables were standardized prior to the analysis.

Table 6.  “Summary of the binary regressions. * Wald statistics p < .05, ** p < .001
1: Omnibus χ2 = 3.15 (df = 10, p < .001), Hosmer & Lemeshow χ2 = 80 (df = 8, p > .05), Nagelkerke R2 = .15
2: Omnibus χ2 = 112.43 (df = 10, p < .001), Hosmer & Lemeshow χ2 = 14.27 (df = 8, p > .05), Nagelkerke R2 = .19
3: Omnibus χ2 = 115.59 (df = 10, p < .001), Hosmer & Lemeshow χ2 = 12.87 (df = 8, p > .05), Nagelkerke R2 = .19”
DependentsCovariatesßSEWaldOdd Ratio
Media seeding1Gender.23.114.38*1.26
 Age−.09.11.78.91
 Education level−.24.105.86*.79
 Session duration.40.1014.67**1.50
 Session frequency.35.146.11*1.42
 Identity Signaling.16.112.081.18
 Surveillance−.07.11.42.93
 Entertainment.10.12.741.11
 Escapism−.03.11.12.96
 Social relations.38.138.96*1.47
 Constant−1.87.11268.71**.16
Narrative seeding2Gender.22.105.12*1.25
 Age−.34.1012.48**.72
 Education level.05.09.281.05
 Session duration.43.1020.53**1.54
 Session frequency.41.1310.33*1.51
 Identity Signaling.24.105.58*1.28
 Surveillance.05.10.241.05
 Entertainment−.03.11.07.97
 Escapism.09.10.891.10
 Social relations.37.129.86*1.44
 Constant−1.47.10212.08**.23
Metadata seeding3Gender−.01.10.01.99
 Age−.12.091.75.89
 Education level−.11.091.49.90
 Session duration.44.0922.67**1.55
 Session frequency.41.139.67**1.50
 Identity Signaling−.14.101.88.87
 Surveillance.24.096.51*1.27
 Entertainment.28.116.77**1.32
 Escapism.15.092.481.16
 Social relations.32.118.20*1.38
 Constant−1.39.10198.74**.25

All three models yield significance and account for 15 to 19% of the variance in seeding group membership. Session frequency and session duration positively predict high-f seeding in all three cases. Girls tend to contribute media more frequently. Education level serves as a negative predictor for media seeding while the social relations gratification does the opposite. Younger adolescents are more likely to be high-f narrative seeders while girls are more likely to contribute content. A higher score for the identity signaling and social relations gratifications indicate a greater likelihood for high-f narrative seeding. Metadata seeding on the other hand is positively predicted by the surveillance, entertainment and social relations gratifications.

Discussion

We divided our sample in two groups, based on the frequency of their seeding behavior for each seeding type (media, narrative, and metadata seeding). Throughout the sample low-f seeding groups account for roughly 80% while the high-f seeders amount to 20% of the sample. It is difficult to compare this ratio with earlier studies on this topic due to (a) the restriction of our sample to adolescents, (b) the time frame of our survey, and (c) the methodology for measuring (frequency, not quantity) and analyzing (k-means on bundled variants of UGC).

Nonetheless, our results show a seeder-leecher ratio that is similar to previous research. Moreover, activities requiring substantial effort and skills, such as media seeding, are more sharply divided than rather less effort related activities such as metadata seeding. Still, our results illustrate a remarkable discrepancy: A majority (62.2%) of Web users rarely engage in seeding behavior and only a minority (8%) of Web users are overall high-f seeders. Contrary to their extensive Web usage, adolescents do not adopt the practice of content seeding on a massive scale. This finding conforms with Schulmeister (2008), Jenkins (et al., n.d.) and Livingstone & Bober (2005) who stress that there is no such thing as a homogenous Net-Generation.

The three different types of seeding do not show a substantial overlap, thus supporting the division in UGC based on literature on Szuprowicz’ (1995) conception of interaction between sender and receiver. Despite the moderate, significant relations among seeding groups, the variance accounted for does not exceed 31%. As such it does not empirically negate the a priori, theoretically-based segmentation in three seeder groups. The relation between narrative and metadata seeding is the strongest. This is not surprising given the moderate degree of interconnectedness between narratives. For instance, blogs often refer to other blogs and users comment each other's entries (Herring, Kouper, Paolillio, Scheidt, Tyworht, Welsch, et al. 2005).

A similar, yet weaker, association is apparent between media and metadata seeding. Analogue to narrative material, media content seeding requires involvement with the uploaded content. For example: Video uploaders are encouraged to create metadata for their content and website visitors are enabled to rate content (Gill et al., 2007). The slight difference in association strength between narrative and metadata seeding and between media and metadata seeding might be located in the characteristics of narrative and media content. Producers of narrative content (such as bloggers and forum users) are continuously confronted with new, interrelated content, whereas the production of media content (Web users posting music, video, text or pictures) is far more unidirectional.

Interestingly, the seeder groups in our sample share some characteristics on gender, education, age and online attendance. For example, females are more likely to seed narrative and media content: the most intensive forms of user-generated content8. This feels a bit contra-intuitive as the related stereotype assumes Internet-engaged adolescents to be male nerds. Especially as previous research on gender differences in computer and WWW use revealed that young males share a higher degree of computer control and self-efficacy (Broos & Roe, 2006).

Other characteristics include the lower level in education (significant for media seeding) and the lower age (significant for narrative seeders) for high-f seeders. High-f seeders also share higher rates of online attendance: Both session frequency and duration positively predict all three seeding groups. A lower level of education predicts high-f media seeding whereas age negatively predicts narrative seeding. This might appear strange at first, given the plausibility for younger, less formally educated adolescents to be confronted with linguistic (our sample is Dutch-speaking) and operational barriers. However, popular Web 2.0 websites often come in local versions and the design of Web 2.0 websites is focused on intuition and usability, thus eliminating operational obstacles. Further, previous research by Peter and Valkenburg (2006) points out that a higher age and educational level are associated with the use of the WWW as an informational medium. The Web's social function appears to be unrelated with age, whereas a formal education offers a slightly negative prediction. However, age and education are negatively associated with the use of the WWW as an entertainment medium.

When we look at the sample's gratification pattern (entertainment, social relations, surveillance, identity signaling, and escapism), we notice that our sample uses the Web especially for entertainment purposes. As such, the previously mentioned findings on education and age carry no inconsistency. The prominence of entertainment in our findings may be a result of Web 2.0 websites' focus on the enhancement of the attractiveness and pleasance of its content and user experience. The social relations gratification, comprising a social, interpersonal communicative Web use, received the second highest score. Such a finding does not surprise given the ongoing shift towards a Social Web that focuses on user-involvement. The third most important gratification is surveillance, embodying an information-seeking WWW use. Obviously, the Internet remains a Walhalla for finding up-to-date information. In contrast, identity signaling, comprising a status-oriented Web use tends to score much less. The gratification escapism, which reflects an avoidance of daily worries, scores the lowest. However, with the Internet being social and omni-present, one could ask to what extend it may still offer the opportunity to bring about such avoidance. In summary, we state that our typology of WWW U&G shows little surprises given its consistency with previous research (see table 1 for comparison). We notice that the three seeding groups share identical relative rank orders on the gratifications obtained. Despite their robust consistency, the differences in absolute magnitudes between low- and high-f seeders are significant in all but two cases whereas high-f seeders score consistently higher.

A study of the obtained gratifications' predictive abilities shows that identity signaling significantly predicts narrative seeding. This is likely to be related with the process of compiling content. Units of narrative content, such as blog posts, are usually the result of a substantial effort and creative activity. The author takes pride in this process and attaches his or her (nick) name to the created content. As such, the created content might be considered as a physical and/or ideological representation of oneself, as a taste performance expressing prestige, differentiation, authenticity, or theatrical persona (Liu, 2007). Weblogs are personal entities where people share information about themselves and expose their feelings and opinions with whoever is (believed to be) out there. As a consequence, Huffaker and Calvert (2005) stress out the importance of blogs as an environment where adolescent identity exploration and construction play an important role, suggesting that adolescents create a consistent public face through their online representations. Moreover, other Web users are able to comment, rate and tag blog posts, thus creating a loop of reinforcement (Nardi, Shiano, Gumbrecht & Swartz, 2004). The same goes for forum posts which Internet users use to share information, creating a sense of self-esteem. Forum posters are positively reinforced by interaction with and the appraisal of their fellow forum members.

The social relations gratification, reflecting a socially oriented WWW use, positively predicts membership for all three seeding groups. Unlike the early days of the WWW, Web 2.0 is now largely driven by social engaged activities, especially linked with social network sites (SNS) such as Myspace, Netlog, and Facebook. These sites afford for the three types of seeding behavior: furnishing a profile page with music, photographs, and videos; tagging, rating, and commenting on pictures of friends; and sharing daily experiences and feelings by means of blog entries (Boyd, 2007). As Livingstone (2008) points out, SNS are used by young adolescents to play and display, while older ones express a notion of identity through authentic relationships with others.

The other WWW gratifications that we detected in our sample (entertainment and surveillance) only predict metadata seeding behavior. Ratings, tags and comments are part of a wide variety of websites. These metadata types are used to canalize content. They can be used to fuel recommender systems based on ratings (Rashid, Albert, Cosley, Lam, McNee, Konstan, et al. 2002), to build folksonomies using tags (Mathes, 2004) and to enhance social interaction by encouraging comments. Websites ranging from media consumption (e.g. YouTube - associated with entertainment) to news websites (e.g. CBSnews.com - associated with the concept of surveillance) adopt these practices on a massive scale. Given the commonness of these activities, our findings hardly surprise. However, the question why these specific gratifications are unrelated to narrative and media seeding remains. Further (qualitative) research into this matter is needed.

Limitations, Implications, and Further Research

Despite its sample size and the balanced distribution of age and gender, our study holds some limitations. Firstly, our sample was not a-select. Secondly, we used an online survey. Although web-based survey research has reached a level of maturity (Witte, 2009) it encompasses some limitations and disadvantages. Several technical glitches can occur while a respondent is filling out a survey (e.g. the browser freezes or the Web server crashes). Ganassali (2008) stresses the influence of the design of web survey questionnaires on the quality of responses. On a methodological level issues such as self-selection, multiple submissions, nonserious responses, and dropouts are pointed out to be problematic (Gosling, Vazire, Srivastava, & John, 2004). However, due to the subject and the population we wanted to study, a web-based survey approach seemed to be the most suited.

Thirdly, the technique we applied to divide into low- and high-f seeders (k-means on a predefined amount of two groups) can be criticized. However, our approach grants an easily interpretable and parsimonious solution that unmistakably leads to several interesting findings.

This study demonstrates how a multidimensional interactivity-based approach to user-generated content can provide new insights in seeding behavior. Moreover, it reveals how none of the dimensions of UGC are widely incorporated in adolescents' Internet use. We therefore approach the optimistic interactive discourse on Web 2.0 with a certain caution: the general Internet population rarely seeds and adolescents do not form an exception to the rule, despite being nurtured in the constant presence of technology.

When we look at the gratification patterns of seeding groups, we notice no differences in rank order: All adolescents, low- or high-f seeder, gain exactly the same set of gratification from their Internet use. However, we cannot ignore the differences in magnitude, albeit the patterns are consistently parallel. We showed that specific gratifications obtained by WWW use can predict separate forms of content seeding behavior, even when controlling for quantity of use and sociodemographics.

Nonetheless, we want to emphasize the explorative nature of our research. Up until today, scientific literature on Web 2.0 is very scarce. As such, it remains a difficult task to frame our findings. By consequence, we wish to stress the need for further research. This study might guide academics to perform in-depth (qualitative) research into the different types of seeding behavior and how these types of UGC relate to gratifications obtained by the medium. Concurrently, this study reminds stakeholders from different fields (e.g. educationalists, marketeers, etc.) that online adolescents are not a homogenous mass who massively and fully adopt whatever kind of technological affordances, thus emphasizing a critical attitude towards the popular, optimistic discourse on Web 2.0.

Aknowledgements

The authors would like to thank Pieter Verdegem, Dimitri Schuurman, Geertrui Berghmans, Cedric Praet and Sarah Van Poecke for their practical and moral support.

Notes

  • 1

    Unlike earlier computer operating systems, Microsoft Windows 95 contained a build-in TCP/IP-stack (a software implementation of two communication protocols which enable a computer to make external connections) and dial-up software that made connecting to the Internet relatively simple. Later, a new Windows 95 service release also included a build-in Internet browser (Internet Explorer 2.0). The introduction of Windows 95 onto the consumer market can thus be considered as a significant moment in time because all the tools that a computer user needs to browse the Internet became available in one user friendly package.

  • 2

    The term folksonomy, generally attributed to Thomas Vander Wal (Smith, 2004), refers to online tagging systems intended to make information increasingly easy to search and navigate over time. A combination of the words folk and taxonomy, it literally means ‘people’s classification management'. The process of ‘tagging’ (assigning particular keyword(s)) helps storing and filing digital content. Folksonomies move a user from a binary in-or-out classification system to an analogue one, only requiring a conceptual association with a resource (Shirky, 2005).

  • 3

    http://www.flickr.com

  • 4

    An application programming interface (API) is a source code interface that a website provides in order to support requests for (data) services by other websites.

  • 5

    http://www.youtube.com

  • 6

    http://www.delicious.com

  • 7

    16% attends vocational education, 53% technical education, and 31% general education. These levels are indicative of job prospects and access to higher learning whereas vocational education is the lowest and general education the highest. It is commonly accepted that this measure indirectly reflects SES.

  • 8

    This confirms data from the Pew Internet American Life Project on teens and social media (Lenhart, Madden, Macgill, & Smith, 2007). Lenhart et al. (2007) reported that 35% of online girls were bloggers compared to 20% of online boys.

About the Authors

Cédric Courtois has a master's degree in Communication Sciences and works as a junior researcher at the research group for Media and ICT (MICT), Ghent University. This article is derived from his master thesis. His research currently focuses on motives of digital media use.

Peter Mechant has a master's degree in Communication Sciences and works as a researcher at the research group for Media and ICT (MICT), Ghent University. He focuses his research on topics such as social software, folksonomies, on line communities and Web 2.0-websites.

Prof. Dr. Lieven De Marez has a master's degree in Communication Sciences and wrote a PhD on innovation segmentation forecasting and introduction strategies for ICT innovations. He teaches ‘new communication technologies' and ‘market- & consumer research’ at Ghent University.

Prof. Dr. Gino Verleye has a master's degree in economic psychology and wrote a PhD on the performance of missing data solutions for structural equations models. He teaches research methodology and statistics at Ghent University.

Contact: Research Group for Media & ICT (MICT, www.mict.be), Department of Communication Sciences, Faculty of Political and Social Studies, Ghent University-Belgium. Korte Meer 7/9/11, BE-9000 Ghent-Belgium. MICT is a part of the Interdisciplinary Institute for Broadband Technology (IBBT, www.ibbt.be). E-mail: Cedric.Courtois@Ugent.be, Peter.Mechant@Ugent.be, Lieven.DeMarez@Ugent.be, Gino.Verleye@Ugent.be

Appendix

This section offers an overview of the U&G Likert-statements. Every item is preceded by ‘Through my WWW use. ’ The answer categories range from ‘strongly disagree’ to ‘strongly agree.’

Factor 1: Identity signaling

I1: I feel important

I2: I impress others

I3: I pretend to be popular

Factor 2: Surveillance

S1: I keep track of the international news

S2: I keep track of the local news

S3: I keep track of worldwide events

S4: I stay informed of occasions and events

Factor 3: Social Relations

SR1: I let people know I care about their feelings

SR2: I stay in touch with people who understand me

SR3: I encourage other people

SR4: I comfort others

SR5: I talk about my problems

SR6: I feel involved with what happens with others

Factor 4: Escapism

E1: I flee form what I am doing

E2: I escape form my responabilities

E3: I postpone tasks that I should complete first

E4: I forget about my daily occupations

Factor 5: Entertainment

EN1: I amuse myself

EN2: I have a good time

EN3: I relax

Ancillary