Chapter 5: Theory as a Lens

Listen in Dr. Leith’s voice

Chapter 4 ended on a finding. Across a body of survey research, livestream viewing looks more socially motivated than content motivated. People watch, the studies suggest, for the chat as much as for the game. Take that finding as settled for a moment. It still does not explain itself. Why would the social experience matter more than the content on the screen? What is actually happening when a viewer chooses a stream and stays?

Three researchers could look at the same finding and explain it three different ways. One, drawing on uses and gratifications theory (Katz, Blumler, & Gurevitch, 1973), might argue that viewers are active agents seeking to satisfy specific needs, and that a livestream happens to satisfy a social need that a finished, edited video cannot. A second, drawing on parasocial interaction (Horton & Wohl, 1956), might argue that viewers form one-sided emotional bonds with the streamer, and that the sense of social connection is the felt experience of that bond. A third, drawing on social identity theory (Tajfel & Turner, 1979), might argue that a stream community is a group, that viewers use it to construct part of their identity, and that the chat is a space of belonging.

Each explanation focuses attention on a different mechanism. Uses and gratifications emphasizes the individual’s active choice. Parasocial interaction emphasizes the relationship with the streamer. Social identity emphasizes the group. The same empirical pattern generates three different research programs depending on which lens you choose.

This is what theory does. It is not decorative, and it is not an abstract exercise you complete before getting to the real work of analysis. Theory is the architecture that transforms scattered observations into coherent understanding. It tells you what to look for, what to measure, and what would count as evidence for or against your explanation. It is also not specific to any one topic. The same theories that illuminate livestreaming also explain news consumption, political persuasion, brand loyalty, and health behavior. Choosing a theory is choosing a way of seeing.

Theory is not speculation

In casual conversation, “theory” often means “guess” or “hunch.” A theory about why the coffee shop is always crowded on Tuesdays. In research, the term means something more precise. A theory is a formal, systematic explanation of relationships between concepts or variables. It organizes knowledge, explains phenomena, and generates predictions. Good theories are logically coherent, meaning the parts fit together without contradiction. They are generalizable, meaning they apply across contexts rather than to one specific case. They are falsifiable, meaning they make predictions that could in principle be proven wrong. And they are generative, meaning they produce new hypotheses and research questions.

Consider social identity theory, developed by Henri Tajfel and John Turner (1979). It proposes that people derive part of their self-concept from the groups they belong to. Group membership creates in-group favoritism, a preference for “our” people, and out-group derogation, a tendency to distance from “them.” When a group’s status is threatened, members are motivated to protect it.

This is not speculation. It is a systematic framework built from decades of experimental research, and it generates testable predictions. Applied to livestreaming, it predicts that viewers will rate their preferred stream community more positively than rival communities, especially on subjective dimensions like authenticity. It predicts that when a streamer or community is criticized, regular viewers will respond defensively. Applied to a different domain entirely, it predicts that political partisans will rate news outlets aligned with their party as more credible than opposing outlets, even when the factual content is identical. The theory does not change. The domain does. Each prediction is testable, and that is what makes it theory rather than speculation.

The lens metaphor

Theory functions like a camera lens. It brings certain elements into sharp focus while leaving others blurred or outside the frame entirely.

If you study livestreaming using appraisal theory, which focuses on how individuals cognitively evaluate stimuli, you pay attention to individual psychological responses, the cognitive processes behind them, and personal experience. You design measures that capture individual reactions: surveys asking viewers to rate how a stream makes them feel, or experiments manipulating stream features and measuring responses.

If you study the same phenomenon using social identity theory, you focus on different elements: group-based meanings, collective identity, the cultural context in which a community formed. You design different measures, analyzing how community members talk about the stream as an identity marker, or how community boundaries are policed in chat.

Same phenomenon, different lenses, different research designs. Neither approach is wrong. They are answering different questions because they are guided by different theoretical frameworks. The key is choosing a lens that fits your research question and your data. You cannot use every lens at once, because that produces incoherent blur. You commit to a perspective, knowing it will illuminate certain aspects while leaving others in shadow.

Three paradigms

The way researchers use theory depends on their broader philosophical commitments, what is called a paradigm. Three major paradigms dominate communication research, and each treats theory differently.

The social scientific paradigm assumes there is an objective reality that can be observed, measured, and understood through systematic testing. Its approach to theory is deductive: start with theory, derive specific predictions, collect data to test those predictions. Its typical question is “does X cause Y?” A researcher might start with cultivation theory (Gerbner & Gross, 1976), which proposes that heavy media exposure cultivates perceptions consistent with media portrayals, hypothesize that heavy viewers of a particular kind of stream develop particular perceptions, and design a study to test that prediction. If the data support the hypothesis, confidence in the theory strengthens. If not, the theory or the hypothesis needs refinement. This cycle, from theory to hypothesis to data to conclusion to refinement, is how social science progresses.

The interpretive paradigm assumes reality is socially constructed and that meaning emerges through interaction and interpretation. Its approach to theory is inductive: start with detailed observations, identify patterns, build theory from the data. Its typical question is “what does X mean to the people who experience it?” A researcher might immerse themselves in a stream community, analyze how members talk, notice that they engage in shared interpretive labor, and propose a theory that the community builds identity through collaborative sense-making. The theory is the end point, grounded in the data, emerging from the bottom up. Methods common to this paradigm include grounded theory, thematic analysis (Braun & Clarke, 2006), and narrative analysis.

The critical and cultural paradigm assumes that power structures shape what counts as knowledge, and that research should critique and challenge inequity. Its approach to theory is to use it as an explicit tool for revealing hidden power dynamics. Its typical question is “whose interests does X serve, and whose voices does it silence?” A researcher studying livestreaming might examine who gets to be a successful streamer, connect the visibility of certain creators to the platform’s design and economic structure, and argue that what looks like neutral popularity is shaped by forces that advantage some creators over others. The critical lens transforms description into argument. Other critical lenses include the political economy of media and discourse analysis.

Theory before data as an epistemological commitment. Beyond focusing your research question and guiding your choice of variables, theory serves a further function: it is what you use to derive testable predictions before you look at the data. This sequencing (theory → prediction → data → test) is not a formality. It is what distinguishes confirmatory from exploratory research. Nosek and colleagues locate pre-registration at exactly this seam, defining it as the act of fixing the analytic path before any outcome is visible:

“Preregistration of an analysis plan is committing to analytic steps without advance knowledge of the research outcomes.”

Nosek et al. (2018, p. 2601)

For your own study, draft the hypotheses section of an OSF pre-registration, and ask how precisely your theoretical framework allows you to state a directional prediction, and where it is too vague to pre-register.

Grand theories and middle-range theories

Not all theories operate at the same level of abstraction.

Grand theories explain fundamental aspects of human behavior across all contexts. Marxism holds that all social relations are shaped by economic structures and class conflict. Psychoanalysis holds that human behavior is driven by unconscious desires. Structuralism holds that meaning emerges from underlying universal structures. Grand theories are intellectually powerful, but they are difficult to test empirically. The scope is too broad and the predictions too general to design a study that could prove or disprove them.

Middle-range theories occupy a more useful zone: specific enough to generate testable hypotheses, broad enough to generalize beyond a single case. The sociologist Robert Merton coined the term to describe frameworks that sit between grand theoretical systems and the narrow descriptions of individual studies. Most communication research uses middle-range theories, and every theory in the catalog below is one.

A catalog of middle-range theories

The theories below are the workhorses of communication research. Each is presented with its foundational citation, its core idea, and an application to livestreaming, but each applies far beyond livestreaming, and the cross-domain examples make that point.

Parasocial interaction. Horton and Wohl (1956) observed that audiences develop one-sided emotional relationships with media figures that feel intimate but lack reciprocity. The theory predicts that people who consume more content from a creator report stronger feelings of connection, that parasocial bonds influence real behavior such as purchasing, and that disruptions like scandals produce feelings of betrayal. In livestreaming, parasocial interaction is the natural lens for studying why viewers donate to streamers they have never met, or why a streamer’s absence produces something like grief in a community. Beyond livestreaming, the theory is central to research on influencer marketing and political communication.

Social identity theory. Tajfel and Turner (1979) proposed that group membership shapes self-concept and motivates in-group favoritism and out-group derogation. The theory predicts that people attend preferentially to information that reflects well on their in-group, that threats to group status increase defensive behavior, and that group boundaries are maintained through symbolic markers. In livestreaming, stream communities are rich sites for this lens: viewers do not just watch a streamer, they identify with a community, and that identification shows up in how they talk, what emotes they use, and how they treat outsiders. Beyond livestreaming, social identity theory is foundational in political communication, sports fandom, and brand community research.

Agenda setting. McCombs and Shaw (1972) found that media do not tell people what to think, but they tell people what to think about. By emphasizing certain issues and ignoring others, media shape public priorities. The theory predicts that heavily covered topics are perceived as more important, that the effect is stronger for issues people have less direct experience with, and that elite media influence discourse more than tabloids. In livestreaming, the platform’s own structures, the front page, the recommendation system, the category rankings, function as agenda setters, shaping which streamers and games are perceived as relevant. Beyond livestreaming, agenda setting originated in election research and has been applied to health and environmental communication.

Framing theory. Goffman (1974) and later Entman (1993) established that how information is presented, the frame, influences how it is interpreted. The same facts, given different emphasis, produce different conclusions. The theory predicts that episodic framing leads audiences to attribute problems to personal responsibility, that thematic framing leads them to attribute problems to structural causes, and that frames aligning with existing beliefs are more persuasive. In livestreaming, framing is visible in how streamers title and categorize their own broadcasts, and in how media coverage frames streaming itself, as a career, as a waste of time, as a new form of celebrity. Beyond livestreaming, framing theory is central to political, health, and organizational communication.

Uses and gratifications. Katz, Blumler, and Gurevitch (1973) argued that audiences are active, not passive: people choose media to satisfy specific needs, including information, entertainment, social connection, and identity construction. The theory predicts that people select media that fulfill current needs, that different media satisfy different gratifications, and that media use changes when needs change. In livestreaming, uses and gratifications is the lens behind much of the viewer-motivation research from Chapter 4. Sjöblom and Hamari (2017) explicitly used it to study why people watch others play video games. Beyond livestreaming, the theory has been applied to every medium from radio to TikTok.

Cultivation theory. Gerbner and Gross (1976) proposed that long-term, cumulative exposure to media content shapes audiences’ perceptions of social reality, so that heavy consumers develop beliefs more consistent with media portrayals than with real-world conditions. The theory predicts that heavy viewers overestimate the prevalence of what they see, that the effect is strongest where audiences lack direct experience, and that cultivation is gradual rather than the result of single exposures. In livestreaming, a cultivation question might ask whether heavy viewers of a particular genre of stream develop skewed perceptions of how common certain behaviors are. Beyond livestreaming, cultivation originated in television research and remains most associated with it.

Trying two lenses on the same data

The fastest way to feel what a lens does is to apply two of them to the same fragment of data. Consider one short stretch of chat from a Twitch broadcast:

DragoBoi89: xQc actually playing well today wtf

emotelord: PogU PogU

realnamedfan: @xqcow do u remember when u said hi to me last week

streamfan22: chat is this real

Through the uses and gratifications lens, you ask what needs these viewers are meeting in the moment. DragoBoi89 is satisfying a need for evaluative commentary on performance. emotelord is satisfying a need for collective emotional expression. realnamedfan is satisfying a need for direct social acknowledgement from the streamer. streamfan22 is satisfying a need for collective sense-making with other viewers. The lens directs your attention to motivations and to the variety of needs a single chat satisfies. All four viewers are roughly equally interesting because each represents a different gratification.

Through the parasocial interaction lens, you ask what the messages reveal about the viewer-streamer bond. DragoBoi89’s comment treats xQc as a familiar figure whose performance can be assessed in casual register. realnamedfan’s @-direct address paired with a memory claim about a prior acknowledgement is a textbook parasocial move: the viewer is treating a one-sided interaction as a continuing relationship. emotelord and streamfan22 are not addressing the streamer at all, so the lens does not foreground them.

Same four messages, two different stories. The uses-and-gratifications lens treats all four viewers as equally interesting. The parasocial lens makes realnamedfan the central case and pushes emotelord and streamfan22 to the periphery. Neither reading is wrong. They are answering different questions, and the prospectus you write in Chapter 6 will commit to one of them.

Variables: the building blocks of hypotheses

Theories become testable through variables, measurable concepts that can take on different values.

An independent variable is what you manipulate, in an experiment, or measure as the predictor, in a correlational study. It is the proposed cause. In livestreaming research, an independent variable might be the stream’s category type (gaming or non-gaming), the streamer’s tenure on the platform, or a stream’s concurrent viewer count.

A dependent variable is the outcome you measure, the thing you think is influenced by the independent variable. It might be the number of chat messages per minute, viewer retention over the course of a broadcast, or the rate of subscriptions and donations.

Consider a simple hypothesis: in non-gaming streams, a higher proportion of chat messages are directed at the streamer than in gaming streams. The independent variable is category type, gaming or non-gaming. The dependent variable is whether a message is directed at the streamer or broadcast to the room. This is testable. You classify streams by category, code a sample of messages for whether each one addresses the streamer, and run a statistical test to see if the predicted relationship holds. The same structure works in any domain. News stories framed around individual human interest will generate more social media engagement than stories framed around systemic data: the independent variable is frame type, the dependent variable is engagement, and the logic is identical.

Two further concepts refine the picture. A mediator explains how or why an independent variable affects a dependent variable; it is the mechanism in between. If stream category affects the proportion of directed messages, the mediator might be the perceived pace of the stream: non-gaming streams may feel more conversational, and that conversational feel may be what actually invites viewers to address the streamer directly. A moderator changes the strength or direction of a relationship; it answers the question “under what conditions?” The relationship between category type and directed-message rate might itself be moderated by audience size, holding on small streams but weakening on large ones. Mediators and moderators are how research moves from “X relates to Y” toward “X relates to Y, through this mechanism, under these conditions.”

Applying theory to your research

Selecting a theory is not arbitrary. It should fit your research question, your data, and the kind of explanation you are seeking. The process has a consistent shape.

First, identify your research question, and be specific. “Livestreaming and community” is too broad. “Does the proportion of chat messages directed at the streamer differ between gaming and non-gaming streams?” is focused enough to act on. Second, map the question to a theoretical domain: psychological mechanisms point toward appraisal theory, social and group processes point toward social identity theory or uses and gratifications, questions about media influence on perception point toward cultivation, agenda setting, or framing. Third, review how others have studied similar questions, which is where the literature review from Chapter 4 pays off; convergence on a particular theory signals it is a productive framework for the domain. Fourth, derive specific hypotheses or research questions by letting the theory generate predictions. Fifth, design measures that operationalize the theoretical concepts, translating abstract ideas like identity or salience into variables you can actually measure. That last step, operationalization, is the work of Chapters 7 and 8.

HARKing and why it undermines inference. HARK stands for Hypothesizing After Results are Known. It occurs when a researcher examines data, notices a pattern, then writes up the report as if the pattern had been predicted in advance. The statistical test reports a p-value, but because the hypothesis was generated by the data rather than before it, that p-value does not carry its nominal meaning. Nosek and colleagues describe the inferential cost of blurring the line between predicting an outcome and explaining one after the fact:

“Failing to appreciate the difference can lead to overconfidence in post hoc explanations (postdictions) and inflate the likelihood of believing that there is evidence for a finding when there is not.”

Nosek et al. (2018, p. 2600)

Pre-registration is the structural solution: you submit your theory-derived hypotheses, variables, and analysis plan to a public registry (OSF) before data collection begins. The registry entry is time-stamped and permanent. For your own study, consider how Nosek et al. distinguish pre-registration from Registered Reports: what additional protection does a Registered Report provide, and which communication journals currently offer that format?

What a V2V pre-registration includes. (1) Research question and directional hypotheses; (2) unit of analysis, platform, and data collection window; (3) the codebook or extracted variables; (4) planned statistical tests and inferential criteria; (5) sample size justification (from Chapter 6). OSF’s standard template walks through each section.

Different theories call for different methods

Choosing a theory is closely tied to a decision this book has so far left implicit: the choice of method. A theory tells you what to look for. A method is how you go and look. And different theories, because they ask different kinds of questions, tend to call for different methods. It is worth seeing the major social-science methods as a set before this book commits, deliberately, to one of them.

Surveys measure what people report about themselves: their attitudes, their motivations, their self-described behavior. A survey is the natural method for a uses-and-gratifications question, because uses and gratifications is a theory about what audiences need and seek, and the most direct way to learn what someone seeks is to ask them. The livestreaming motivation research from Chapter 4 is survey research. The limitation of surveys is the gap between what people say and what they do: a viewer can sincerely report a motivation that does not match their actual behavior.

Experiments isolate cause by manipulating one variable while holding others constant. An experiment is the natural method for a sharp causal question, the “does X cause Y?” of the social scientific paradigm. If you want to know whether a particular stream feature causes a change in viewer behavior, you manipulate that feature, hold the rest constant, and measure the result. The limitation of experiments is artificiality: the control that makes causal inference possible also makes the setting less like the world the behavior normally happens in.

Qualitative methods, including in-depth interviews, focus groups, and ethnography, capture meaning, process, and lived experience. They are the natural methods for the interpretive paradigm’s question, “what does X mean to the people who experience it?” If you want to understand how a stream community understands itself, you talk to its members and observe it at length. The limitation of qualitative methods is that they are not built to generalize in the statistical sense; they trade breadth for depth.

Content analysis systematically quantifies the features of communication content itself: the messages, the texts, the artifacts. It is the natural method for questions about what is in the content, asked at a scale too large to eyeball. It is also the method this course teaches end to end, for three reasons. The first is fit: the Twitch dataset is content, millions of chat messages and stream records, and content analysis is the method native to that kind of data. The second is pedagogical completeness: content analysis lets you practice the full research pipeline, operationalization, sampling, reliability, analysis, on a single dataset, without the separate apparatus of human-subjects recruitment. The third is transfer: the disciplines content analysis demands, a precise codebook, tested reliability, systematic sampling, are the same disciplines that make every other method work. Learning one method deeply is the fastest way to understand what all of them are trying to do.

Seen together, these four methods are not a random toolbox. They line up with the paradigms from earlier in this chapter. The social scientific paradigm, with its deductive testing of predictions, leans on surveys and experiments. The interpretive paradigm, building theory from the ground up, leans on qualitative methods. Content analysis is the flexible case: it can serve a social scientific study that counts and tests, or an interpretive study that reads for meaning, depending on how its codebook is built. A method is not just a procedure. It is the operational expression of a way of seeing, which is why the choice of method and the choice of theory are, in the end, the same decision approached from two directions.

This is not a claim that content analysis is the best method. It is a claim that it is the right method for this dataset and this course, and that committing to it is a decision worth making with the alternatives in view rather than by default.

Looking ahead

Chapter 6 turns the theory and the literature into a plan. A research prospectus is the document that states a research question, situates it in the literature, names a theoretical framework, and commits to a method, all before any data is collected. It is where the foundations of the first half of this book become a concrete proposal for the second half.

References

Braun, V., & Clarke, V. (2006). Using thematic analysis in psychology. Qualitative Research in Psychology, 3(2), 77-101. https://doi.org/10.1191/1478088706qp063oa

Entman, R. M. (1993). Framing: Toward clarification of a fractured paradigm. Journal of Communication, 43(4), 51-58. https://doi.org/10.1111/j.1460-2466.1993.tb01304.x

Gerbner, G., & Gross, L. (1976). Living with television: The violence profile. Journal of Communication, 26(2), 172-199. https://doi.org/10.1111/j.1460-2466.1976.tb01397.x

Goffman, E. (1974). Frame analysis: An essay on the organization of experience. Harvard University Press.

Horton, D., & Wohl, R. R. (1956). Mass communication and para-social interaction: Observations on intimacy at a distance. Psychiatry, 19(3), 215-229. https://doi.org/10.1080/00332747.1956.11023049

Katz, E., Blumler, J. G., & Gurevitch, M. (1973). Uses and gratifications research. Public Opinion Quarterly, 37(4), 509-523. https://doi.org/10.1086/268109

McCombs, M. E., & Shaw, D. L. (1972). The agenda-setting function of mass media. Public Opinion Quarterly, 36(2), 176-187. https://doi.org/10.1086/267990

Sjöblom, M., & Hamari, J. (2017). Why do people watch others play video games? An empirical study on the motivations of Twitch users. Computers in Human Behavior, 75, 985-996. https://doi.org/10.1016/j.chb.2016.10.019

Tajfel, H., & Turner, J. C. (1979). An integrative theory of intergroup conflict. In W. G. Austin & S. Worchel (Eds.), The social psychology of intergroup relations (pp. 33-47). Brooks/Cole.

Graduate readings

Nosek, B. A., Ebersole, C. R., DeHaven, A. C., & Mellor, D. T. (2018). The preregistration revolution. Proceedings of the National Academy of Sciences, 115(11), 2600–2606. https://doi.org/10.1073/pnas.1708274114

--- title: "Chapter 5: Theory as a Lens" --- ::: {.chapter-audio} [Listen in Dr. Leith's voice]{.audio-label} ```{=html} <audio class="v2v-audio v2v-audio-ug" controls preload="none"><source src="../audio/ch05-ug.mp3" type="audio/mpeg">Your browser does not support audio.</audio> <audio class="v2v-audio v2v-audio-grad" controls preload="none"><source src="../audio/ch05-grad.mp3" type="audio/mpeg">Your browser does not support audio.</audio> ``` ::: Chapter 4 ended on a finding. Across a body of survey research, livestream viewing looks more socially motivated than content motivated. People watch, the studies suggest, for the chat as much as for the game. Take that finding as settled for a moment. It still does not explain itself. Why would the social experience matter more than the content on the screen? What is actually happening when a viewer chooses a stream and stays? Three researchers could look at the same finding and explain it three different ways. One, drawing on **uses and gratifications theory** (Katz, Blumler, & Gurevitch, 1973), might argue that viewers are active agents seeking to satisfy specific needs, and that a livestream happens to satisfy a social need that a finished, edited video cannot. A second, drawing on **parasocial interaction** (Horton & Wohl, 1956), might argue that viewers form one-sided emotional bonds with the streamer, and that the sense of social connection is the felt experience of that bond. A third, drawing on **social identity theory** (Tajfel & Turner, 1979), might argue that a stream community is a group, that viewers use it to construct part of their identity, and that the chat is a space of belonging. Each explanation focuses attention on a different mechanism. Uses and gratifications emphasizes the individual's active choice. Parasocial interaction emphasizes the relationship with the streamer. Social identity emphasizes the group. The same empirical pattern generates three different research programs depending on which lens you choose. This is what theory does. It is not decorative, and it is not an abstract exercise you complete before getting to the real work of analysis. Theory is the architecture that transforms scattered observations into coherent understanding. It tells you what to look for, what to measure, and what would count as evidence for or against your explanation. It is also not specific to any one topic. The same theories that illuminate livestreaming also explain news consumption, political persuasion, brand loyalty, and health behavior. Choosing a theory is choosing a way of seeing. ## Theory is not speculation In casual conversation, "theory" often means "guess" or "hunch." A theory about why the coffee shop is always crowded on Tuesdays. In research, the term means something more precise. A **theory** is a formal, systematic explanation of relationships between concepts or variables. It organizes knowledge, explains phenomena, and generates predictions. Good theories are logically coherent, meaning the parts fit together without contradiction. They are generalizable, meaning they apply across contexts rather than to one specific case. They are falsifiable, meaning they make predictions that could in principle be proven wrong. And they are generative, meaning they produce new hypotheses and research questions. Consider social identity theory, developed by Henri Tajfel and John Turner (1979). It proposes that people derive part of their self-concept from the groups they belong to. Group membership creates in-group favoritism, a preference for "our" people, and out-group derogation, a tendency to distance from "them." When a group's status is threatened, members are motivated to protect it. This is not speculation. It is a systematic framework built from decades of experimental research, and it generates testable predictions. Applied to livestreaming, it predicts that viewers will rate their preferred stream community more positively than rival communities, especially on subjective dimensions like authenticity. It predicts that when a streamer or community is criticized, regular viewers will respond defensively. Applied to a different domain entirely, it predicts that political partisans will rate news outlets aligned with their party as more credible than opposing outlets, even when the factual content is identical. The theory does not change. The domain does. Each prediction is testable, and that is what makes it theory rather than speculation. ## The lens metaphor Theory functions like a camera lens. It brings certain elements into sharp focus while leaving others blurred or outside the frame entirely. If you study livestreaming using **appraisal theory**, which focuses on how individuals cognitively evaluate stimuli, you pay attention to individual psychological responses, the cognitive processes behind them, and personal experience. You design measures that capture individual reactions: surveys asking viewers to rate how a stream makes them feel, or experiments manipulating stream features and measuring responses. If you study the same phenomenon using social identity theory, you focus on different elements: group-based meanings, collective identity, the cultural context in which a community formed. You design different measures, analyzing how community members talk about the stream as an identity marker, or how community boundaries are policed in chat. Same phenomenon, different lenses, different research designs. Neither approach is wrong. They are answering different questions because they are guided by different theoretical frameworks. The key is choosing a lens that fits your research question and your data. You cannot use every lens at once, because that produces incoherent blur. You commit to a perspective, knowing it will illuminate certain aspects while leaving others in shadow. ## Three paradigms The way researchers use theory depends on their broader philosophical commitments, what is called a **paradigm**. Three major paradigms dominate communication research, and each treats theory differently. The **social scientific paradigm** assumes there is an objective reality that can be observed, measured, and understood through systematic testing. Its approach to theory is deductive: start with theory, derive specific predictions, collect data to test those predictions. Its typical question is "does X cause Y?" A researcher might start with cultivation theory (Gerbner & Gross, 1976), which proposes that heavy media exposure cultivates perceptions consistent with media portrayals, hypothesize that heavy viewers of a particular kind of stream develop particular perceptions, and design a study to test that prediction. If the data support the hypothesis, confidence in the theory strengthens. If not, the theory or the hypothesis needs refinement. This cycle, from theory to hypothesis to data to conclusion to refinement, is how social science progresses. The **interpretive paradigm** assumes reality is socially constructed and that meaning emerges through interaction and interpretation. Its approach to theory is inductive: start with detailed observations, identify patterns, build theory from the data. Its typical question is "what does X mean to the people who experience it?" A researcher might immerse themselves in a stream community, analyze how members talk, notice that they engage in shared interpretive labor, and propose a theory that the community builds identity through collaborative sense-making. The theory is the end point, grounded in the data, emerging from the bottom up. Methods common to this paradigm include grounded theory, thematic analysis (Braun & Clarke, 2006), and narrative analysis. The **critical and cultural paradigm** assumes that power structures shape what counts as knowledge, and that research should critique and challenge inequity. Its approach to theory is to use it as an explicit tool for revealing hidden power dynamics. Its typical question is "whose interests does X serve, and whose voices does it silence?" A researcher studying livestreaming might examine who gets to be a successful streamer, connect the visibility of certain creators to the platform's design and economic structure, and argue that what looks like neutral popularity is shaped by forces that advantage some creators over others. The critical lens transforms description into argument. Other critical lenses include the political economy of media and discourse analysis. ::: {.graduate-extension} **Theory before data as an epistemological commitment.** Beyond focusing your research question and guiding your choice of variables, theory serves a further function: it is what you use to derive testable predictions *before you look at the data*. This sequencing (theory → prediction → data → test) is not a formality. It is what distinguishes confirmatory from exploratory research. Nosek and colleagues locate pre-registration at exactly this seam, defining it as the act of fixing the analytic path before any outcome is visible: > "Preregistration of an analysis plan is committing to analytic steps without advance knowledge of the research outcomes." > > Nosek et al. (2018, p. 2601) For your own study, draft the hypotheses section of an OSF pre-registration, and ask how precisely your theoretical framework allows you to state a directional prediction, and where it is too vague to pre-register. ::: ## Grand theories and middle-range theories Not all theories operate at the same level of abstraction. **Grand theories** explain fundamental aspects of human behavior across all contexts. Marxism holds that all social relations are shaped by economic structures and class conflict. Psychoanalysis holds that human behavior is driven by unconscious desires. Structuralism holds that meaning emerges from underlying universal structures. Grand theories are intellectually powerful, but they are difficult to test empirically. The scope is too broad and the predictions too general to design a study that could prove or disprove them. **Middle-range theories** occupy a more useful zone: specific enough to generate testable hypotheses, broad enough to generalize beyond a single case. The sociologist Robert Merton coined the term to describe frameworks that sit between grand theoretical systems and the narrow descriptions of individual studies. Most communication research uses middle-range theories, and every theory in the catalog below is one. ## A catalog of middle-range theories The theories below are the workhorses of communication research. Each is presented with its foundational citation, its core idea, and an application to livestreaming, but each applies far beyond livestreaming, and the cross-domain examples make that point. **Parasocial interaction.** Horton and Wohl (1956) observed that audiences develop one-sided emotional relationships with media figures that feel intimate but lack reciprocity. The theory predicts that people who consume more content from a creator report stronger feelings of connection, that parasocial bonds influence real behavior such as purchasing, and that disruptions like scandals produce feelings of betrayal. In livestreaming, parasocial interaction is the natural lens for studying why viewers donate to streamers they have never met, or why a streamer's absence produces something like grief in a community. Beyond livestreaming, the theory is central to research on influencer marketing and political communication. **Social identity theory.** Tajfel and Turner (1979) proposed that group membership shapes self-concept and motivates in-group favoritism and out-group derogation. The theory predicts that people attend preferentially to information that reflects well on their in-group, that threats to group status increase defensive behavior, and that group boundaries are maintained through symbolic markers. In livestreaming, stream communities are rich sites for this lens: viewers do not just watch a streamer, they identify with a community, and that identification shows up in how they talk, what emotes they use, and how they treat outsiders. Beyond livestreaming, social identity theory is foundational in political communication, sports fandom, and brand community research. **Agenda setting.** McCombs and Shaw (1972) found that media do not tell people what to think, but they tell people what to think about. By emphasizing certain issues and ignoring others, media shape public priorities. The theory predicts that heavily covered topics are perceived as more important, that the effect is stronger for issues people have less direct experience with, and that elite media influence discourse more than tabloids. In livestreaming, the platform's own structures, the front page, the recommendation system, the category rankings, function as agenda setters, shaping which streamers and games are perceived as relevant. Beyond livestreaming, agenda setting originated in election research and has been applied to health and environmental communication. **Framing theory.** Goffman (1974) and later Entman (1993) established that how information is presented, the frame, influences how it is interpreted. The same facts, given different emphasis, produce different conclusions. The theory predicts that episodic framing leads audiences to attribute problems to personal responsibility, that thematic framing leads them to attribute problems to structural causes, and that frames aligning with existing beliefs are more persuasive. In livestreaming, framing is visible in how streamers title and categorize their own broadcasts, and in how media coverage frames streaming itself, as a career, as a waste of time, as a new form of celebrity. Beyond livestreaming, framing theory is central to political, health, and organizational communication. **Uses and gratifications.** Katz, Blumler, and Gurevitch (1973) argued that audiences are active, not passive: people choose media to satisfy specific needs, including information, entertainment, social connection, and identity construction. The theory predicts that people select media that fulfill current needs, that different media satisfy different gratifications, and that media use changes when needs change. In livestreaming, uses and gratifications is the lens behind much of the viewer-motivation research from Chapter 4. Sjöblom and Hamari (2017) explicitly used it to study why people watch others play video games. Beyond livestreaming, the theory has been applied to every medium from radio to TikTok. **Cultivation theory.** Gerbner and Gross (1976) proposed that long-term, cumulative exposure to media content shapes audiences' perceptions of social reality, so that heavy consumers develop beliefs more consistent with media portrayals than with real-world conditions. The theory predicts that heavy viewers overestimate the prevalence of what they see, that the effect is strongest where audiences lack direct experience, and that cultivation is gradual rather than the result of single exposures. In livestreaming, a cultivation question might ask whether heavy viewers of a particular genre of stream develop skewed perceptions of how common certain behaviors are. Beyond livestreaming, cultivation originated in television research and remains most associated with it. ## Trying two lenses on the same data The fastest way to feel what a lens does is to apply two of them to the same fragment of data. Consider one short stretch of chat from a Twitch broadcast: > **DragoBoi89:** xQc actually playing well today wtf > > **emotelord:** PogU PogU > > **realnamedfan:** @xqcow do u remember when u said hi to me last week > > **streamfan22:** chat is this real Through the **uses and gratifications** lens, you ask what needs these viewers are meeting in the moment. DragoBoi89 is satisfying a need for evaluative commentary on performance. emotelord is satisfying a need for collective emotional expression. realnamedfan is satisfying a need for direct social acknowledgement from the streamer. streamfan22 is satisfying a need for collective sense-making with other viewers. The lens directs your attention to motivations and to the variety of needs a single chat satisfies. All four viewers are roughly equally interesting because each represents a different gratification. Through the **parasocial interaction** lens, you ask what the messages reveal about the viewer-streamer bond. DragoBoi89's comment treats xQc as a familiar figure whose performance can be assessed in casual register. realnamedfan's @-direct address paired with a memory claim about a prior acknowledgement is a textbook parasocial move: the viewer is treating a one-sided interaction as a continuing relationship. emotelord and streamfan22 are not addressing the streamer at all, so the lens does not foreground them. Same four messages, two different stories. The uses-and-gratifications lens treats all four viewers as equally interesting. The parasocial lens makes realnamedfan the central case and pushes emotelord and streamfan22 to the periphery. Neither reading is wrong. They are answering different questions, and the prospectus you write in Chapter 6 will commit to one of them. ## Variables: the building blocks of hypotheses Theories become testable through **variables**, measurable concepts that can take on different values. An **independent variable** is what you manipulate, in an experiment, or measure as the predictor, in a correlational study. It is the proposed cause. In livestreaming research, an independent variable might be the stream's category type (gaming or non-gaming), the streamer's tenure on the platform, or a stream's concurrent viewer count. A **dependent variable** is the outcome you measure, the thing you think is influenced by the independent variable. It might be the number of chat messages per minute, viewer retention over the course of a broadcast, or the rate of subscriptions and donations. Consider a simple hypothesis: in non-gaming streams, a higher proportion of chat messages are directed at the streamer than in gaming streams. The independent variable is category type, gaming or non-gaming. The dependent variable is whether a message is directed at the streamer or broadcast to the room. This is testable. You classify streams by category, code a sample of messages for whether each one addresses the streamer, and run a statistical test to see if the predicted relationship holds. The same structure works in any domain. News stories framed around individual human interest will generate more social media engagement than stories framed around systemic data: the independent variable is frame type, the dependent variable is engagement, and the logic is identical. Two further concepts refine the picture. A **mediator** explains how or why an independent variable affects a dependent variable; it is the mechanism in between. If stream category affects the proportion of directed messages, the mediator might be the perceived pace of the stream: non-gaming streams may feel more conversational, and that conversational feel may be what actually invites viewers to address the streamer directly. A **moderator** changes the strength or direction of a relationship; it answers the question "under what conditions?" The relationship between category type and directed-message rate might itself be moderated by audience size, holding on small streams but weakening on large ones. Mediators and moderators are how research moves from "X relates to Y" toward "X relates to Y, through this mechanism, under these conditions." ## Applying theory to your research Selecting a theory is not arbitrary. It should fit your research question, your data, and the kind of explanation you are seeking. The process has a consistent shape. First, identify your research question, and be specific. "Livestreaming and community" is too broad. "Does the proportion of chat messages directed at the streamer differ between gaming and non-gaming streams?" is focused enough to act on. Second, map the question to a theoretical domain: psychological mechanisms point toward appraisal theory, social and group processes point toward social identity theory or uses and gratifications, questions about media influence on perception point toward cultivation, agenda setting, or framing. Third, review how others have studied similar questions, which is where the literature review from Chapter 4 pays off; convergence on a particular theory signals it is a productive framework for the domain. Fourth, derive specific hypotheses or research questions by letting the theory generate predictions. Fifth, design measures that operationalize the theoretical concepts, translating abstract ideas like identity or salience into variables you can actually measure. That last step, operationalization, is the work of Chapters 7 and 8. ::: {.graduate-extension} **HARKing and why it undermines inference.** HARK stands for Hypothesizing After Results are Known. It occurs when a researcher examines data, notices a pattern, then writes up the report as if the pattern had been predicted in advance. The statistical test reports a *p*-value, but because the hypothesis was generated by the data rather than before it, that *p*-value does not carry its nominal meaning. Nosek and colleagues describe the inferential cost of blurring the line between predicting an outcome and explaining one after the fact: > "Failing to appreciate the difference can lead to overconfidence in post hoc explanations (postdictions) and inflate the likelihood of believing that there is evidence for a finding when there is not." > > Nosek et al. (2018, p. 2600) Pre-registration is the structural solution: you submit your theory-derived hypotheses, variables, and analysis plan to a public registry (OSF) before data collection begins. The registry entry is time-stamped and permanent. For your own study, consider how Nosek et al. distinguish pre-registration from Registered Reports: what additional protection does a Registered Report provide, and which communication journals currently offer that format? ::: ::: {.graduate-extension} **What a V2V pre-registration includes.** (1) Research question and directional hypotheses; (2) unit of analysis, platform, and data collection window; (3) the codebook or extracted variables; (4) planned statistical tests and inferential criteria; (5) sample size justification (from Chapter 6). OSF's standard template walks through each section. ::: ## Different theories call for different methods Choosing a theory is closely tied to a decision this book has so far left implicit: the choice of **method**. A theory tells you what to look for. A method is how you go and look. And different theories, because they ask different kinds of questions, tend to call for different methods. It is worth seeing the major social-science methods as a set before this book commits, deliberately, to one of them. **Surveys** measure what people report about themselves: their attitudes, their motivations, their self-described behavior. A survey is the natural method for a uses-and-gratifications question, because uses and gratifications is a theory about what audiences need and seek, and the most direct way to learn what someone seeks is to ask them. The livestreaming motivation research from Chapter 4 is survey research. The limitation of surveys is the gap between what people say and what they do: a viewer can sincerely report a motivation that does not match their actual behavior. **Experiments** isolate cause by manipulating one variable while holding others constant. An experiment is the natural method for a sharp causal question, the "does X cause Y?" of the social scientific paradigm. If you want to know whether a particular stream feature causes a change in viewer behavior, you manipulate that feature, hold the rest constant, and measure the result. The limitation of experiments is artificiality: the control that makes causal inference possible also makes the setting less like the world the behavior normally happens in. **Qualitative methods**, including in-depth interviews, focus groups, and ethnography, capture meaning, process, and lived experience. They are the natural methods for the interpretive paradigm's question, "what does X mean to the people who experience it?" If you want to understand how a stream community understands itself, you talk to its members and observe it at length. The limitation of qualitative methods is that they are not built to generalize in the statistical sense; they trade breadth for depth. **Content analysis** systematically quantifies the features of communication content itself: the messages, the texts, the artifacts. It is the natural method for questions about what is in the content, asked at a scale too large to eyeball. It is also the method this course teaches end to end, for three reasons. The first is fit: the Twitch dataset is content, millions of chat messages and stream records, and content analysis is the method native to that kind of data. The second is pedagogical completeness: content analysis lets you practice the full research pipeline, operationalization, sampling, reliability, analysis, on a single dataset, without the separate apparatus of human-subjects recruitment. The third is transfer: the disciplines content analysis demands, a precise codebook, tested reliability, systematic sampling, are the same disciplines that make every other method work. Learning one method deeply is the fastest way to understand what all of them are trying to do. Seen together, these four methods are not a random toolbox. They line up with the paradigms from earlier in this chapter. The social scientific paradigm, with its deductive testing of predictions, leans on surveys and experiments. The interpretive paradigm, building theory from the ground up, leans on qualitative methods. Content analysis is the flexible case: it can serve a social scientific study that counts and tests, or an interpretive study that reads for meaning, depending on how its codebook is built. A method is not just a procedure. It is the operational expression of a way of seeing, which is why the choice of method and the choice of theory are, in the end, the same decision approached from two directions. This is not a claim that content analysis is the best method. It is a claim that it is the right method for this dataset and this course, and that committing to it is a decision worth making with the alternatives in view rather than by default. ## Looking ahead Chapter 6 turns the theory and the literature into a plan. A research prospectus is the document that states a research question, situates it in the literature, names a theoretical framework, and commits to a method, all before any data is collected. It is where the foundations of the first half of this book become a concrete proposal for the second half. ## References Braun, V., & Clarke, V. (2006). Using thematic analysis in psychology. *Qualitative Research in Psychology*, *3*(2), 77-101. https://doi.org/10.1191/1478088706qp063oa Entman, R. M. (1993). Framing: Toward clarification of a fractured paradigm. *Journal of Communication*, *43*(4), 51-58. https://doi.org/10.1111/j.1460-2466.1993.tb01304.x Gerbner, G., & Gross, L. (1976). Living with television: The violence profile. *Journal of Communication*, *26*(2), 172-199. https://doi.org/10.1111/j.1460-2466.1976.tb01397.x Goffman, E. (1974). *Frame analysis: An essay on the organization of experience*. Harvard University Press. Horton, D., & Wohl, R. R. (1956). Mass communication and para-social interaction: Observations on intimacy at a distance. *Psychiatry*, *19*(3), 215-229. https://doi.org/10.1080/00332747.1956.11023049 Katz, E., Blumler, J. G., & Gurevitch, M. (1973). Uses and gratifications research. *Public Opinion Quarterly*, *37*(4), 509-523. https://doi.org/10.1086/268109 McCombs, M. E., & Shaw, D. L. (1972). The agenda-setting function of mass media. *Public Opinion Quarterly*, *36*(2), 176-187. https://doi.org/10.1086/267990 Sjöblom, M., & Hamari, J. (2017). Why do people watch others play video games? An empirical study on the motivations of Twitch users. *Computers in Human Behavior*, *75*, 985-996. https://doi.org/10.1016/j.chb.2016.10.019 Tajfel, H., & Turner, J. C. (1979). An integrative theory of intergroup conflict. In W. G. Austin & S. Worchel (Eds.), *The social psychology of intergroup relations* (pp. 33-47). Brooks/Cole. ::: {.graduate-extension} ### Graduate readings Nosek, B. A., Ebersole, C. R., DeHaven, A. C., & Mellor, D. T. (2018). The preregistration revolution. *Proceedings of the National Academy of Sciences*, *115*(11), 2600–2606. https://doi.org/10.1073/pnas.1708274114 :::