Twitter as a Vehicle of Political Conversation

My name is Grace Murray and I am a rising junior at the College of William & Mary. I am a Government and Psychology major looking to build a career as a political scientist, ideally specializing in the interplay of social media within American politics.

It is an honor to be researching this summer through the generosity of the Charles Center, and I look forward to the steep curve of learning ahead of me in the weeks to come. The study I will be conducting this summer was inspired by my work in William & Mary’s Social Networks and Political Psychology lab. Professor Jaime Settle has helped me carefully cultivate the variables and research paradigm and I am incredibly thankful for her continuing support and expertise.

The crux of this study is to evaluate the similarities of Twitter responses based on topics, and understand the opinion of the public on the issues as viewed in responses to tweets. A widely lauded typology based on congressional records from the Senate will be applied to the body of tweets for specific senators. This typology categorizes key words and stems frequently associated with 42 different categories of topics and ranks the frequency of mention in the Senate (Quinn 2010). In applying this model to senators’ tweets, the results will categorize the tweets by topic and allow for effective coding of the responses. Subsequently evaluating the responses to the tweets for normativity and positivity displays the vocalization of opinion of the public in regards to the topics discussed within the Senate.

I hypothesize that there will be significant variations in responses based on the topics mentioned and the party of the senators in question. I also hypothesize that social issues will warrant more nonnormative responses than economic issues and that Republican Senators will have lower rates of positivity. “Normativity” will describe whether the responses are in socially acceptable vernacular. Emotionality will measure whether the rhetoric used in responses is largely “positive” or “negative” based on key words and roots.

My independent variable will be the content of the tweets of senators. These are coded using the key topic typology. My dependent variable will be the normativity and emotionality of the responses to tweets. “Normativity” is found by coding for swear words as they are viewed as nonnormative social behavior on Twitter. “Emotionality” is found by comparing the degree of positive vs. negative semantics. Linguistic Inquiry Word Count software available to me through the Social Science Research Method Center (SSRMC) will be used to semantically evaluate for normativity and emotionality; this software has an extensive body of coding for swear words and positive/negative semantics available which I will apply and evaluate. This will give a quantified description of the findings of normativity and emotionality. I will then assess this dataset statistically in order to evaluate the strength of the relationships suggested by the hypothesis.


Quinn, Kevin, Burt Monroe, Michael Colaresi, Michael Crespin, and Radev Dragomir. 2010.

“How to Analyze Political Attention with Minimal Assumptions and Costs.” American Journal of Political Science 54 (1):209-28.