As society continues to foster the growth and expansion of Internet and mobile-based communications, it has also started to provide new means for which people can express themselves. When comparing communication offline to its online counterparts, it becomes clear that due to the inherent “leanness” of the media, there may be a lack of emotional comprehension from the person who sends the message to the person who receives it. In this research study the relationship between perceived and intended emotional intensity, while in the scope of public messaging systems like Facebook’s wall posts and private messaging systems like text messaging, will be explained. After surveying 121 college students at Cornell University, results show that more negative emotions are both intended and perceived better while sending (online) private messages, and conversely, more positive emotions are intended and perceived more successfully through a more public messaging system.
Since the rapid expansion of the Internet and social networking, online computer mediated communication (CMC) has vastly changed the way that people interact. According to a Pew survey taken in 2013, 86% of all American adults use the Internet as compared to 14% back in 1995 (“Internet Use Over Time”, 2014). With this large non-trivial amount of growth in Internet usage in the past 19 years, it is significant to observe the extent to which people use it. Whereas 19 years ago the Internet was used primarily to read emails and share information, most users of the Internet did not stay on the computer for as many hours as they do today. It then becomes interesting to note the changes over the years as to how people interact and make connections in an Internet connected world.
The change that we are most interested in pursuing for the purposes of our research is in how emotion translates from face to face communications (FTF) to that of CMC. Whereas it is true that contextual analysis has often been conducted while focusing on word choice and syntax during online communication channels, the intent of our research was to stratify vastly diverse intense emotions in order to compare how the medium may also affect the interpretation of text based emotional cues.
Given the fact that the way that people interpret emotions through text is also not a new idea specific to an Internet age, one can simply open any book to be successfully transplanted from one mind set to another through the vivid storytelling that often occurs. No research has been done thus far to specify media choice as an influencer in emotional readability. All in all, the research that we accomplished, as will be described throughout the paper, will focus on media choice (public Facebook wall postings and private text messaging) and how emotions come through given those specific media choices.
In face-to-face conversations, there are various nonverbal cues expressed through facial movements and gestures within the voice. However, when communicating (CMC) facial movements and voice gestures are no longer present, which can be problematic when it comes to expressing various emotions. According to Hancock and Dunham (2001) “the reduction of nonverbal social and relational cues in CMC produces a de-personalized form of communication and decrease awareness of other, inhibiting interpersonal relation” (p. 326). Hence, there are limitations on how an individual can evoke certain body language to another individual through text-based CMC. This can lead receivers of a message to misinterpret the intended meaning of a message that can further pass along the miscommunication throughout a conversation. Therefore, as the progression of computer mediated communication continued users, especially frequent senders, use different approaches in order to compensate for those missing cues.
There has been research that looks at the importance of nonverbal cues. Several studies claim Social Information Processing Theory (SIP) employs text-based cues as a way to communicate nonverbal behaviors in text-based CMC. “The major thrust of social information processing theory is that CMC limits the way impression-relevant cues are exchanged during social interactions, rather than simply reducing or eliminating the amount of such information” (Hancock & Dunham, 2001, p. 328). Users achieve these exchanges by the degree of agreement presented in a reply, degree of negative affect terms, and the different usage of punctuation. Channel expansion theory takes into consideration that after a certain length of time using a specific medium, the communication produce a new sense of knowledge and skill in using the medium, therefore we can expect that there will be usage of emoticons, punctuations, and capitalizations in the sender’s messages to convey the different assigned emotions of regret, excitement, and contentment. According to the study by Shao-Kang Lo (2008), “that when Internet users are faced with pure text without emoticons, most people cannot perceive the correct emotion, attitude, and attention intents. However, when emoticons are added in the same context, the receiver’s perception of the messages starts to significantly change (p. 597).” Hence it is expected that the messages from the senders will contain some of these elements, in their endeavor to rightfully convey their message to the receiver. However, this also means that a skillful user of a specific CMC medium understands that, while they can take action to reduce the loss of emotional information within their message, there is still doubt whether the receiver of the user will be able to fully comprehend the emotion within the message.
It is also important in which the way that the message is written because the subtle cues affects the general tone of the message. In our research we asked that people identify the ways that they signaled the various emotional states and the vast majority of respondents reported consistencies with these studies in that they used emotional markers such as smiley faces and punctuation. In Savicki and Kelly (2000) research, they stated that the “language that increase intimacy, social penetration, or message immediacy helps to combat the limitation of the medium” in text-based CMC (p. 824). Essentially, word choice, use of punctuation, and other literary elements help embellish a message meaning. However, while the sender may think that these cues are helping to communicate the meaning and emotion within their message, we are curious to what extend they are actually interpreted by a group of receivers. To a certain extent in our research we expect to see that intensity of the emotion in lost in translation to an audience because of the implications in previous research. In addition we seek to look closer at how the medium the message is exchanged across comes into play with the overall understanding and interpretation of the message’s emotional strength; combined with the influence of the medium on the way the sender of the message decides to convey the emotion and content to the receiver.
Using this information we developed three hypotheses for our study. The first is that Senders think the emotional intensity is higher for them than they think the receivers will perceive the intensity. As mention previous within the section, while a skillful user of a specific CMC medium knows how to reduce the loss of emotional intensity within the message. Therefore, when the user rates their emotional intensity they will rate it higher for themselves and expect a lower rating from another person who may read their message. Our second hypothesis is that Senders will rate the message emotional intensity is higher for them than they think receivers will perceive the intensity. When a sender sends a message they know the emotion that they are trying to convey, therefore, they will give themselves a higher rating of intensity in the perception of emotion. However, experienced CMC users know there is a chance that their emotional intensity will not be fully perceived by the receiver which will lead to the rate of the emotional intensity perceived by another person being lower than their own rating. Our third hypothesis is that regret will be better interpreted through text messaging, while Facebook will personify excitement better. When it comes to a negative emotion such as regret, we believe that not many people will be willing to be truly regretful on a Facebook Wall post where anyone can read the message. Therefore, it may be more comfortable for senders to express regret in a text message. Excitement on the other hand transcribes better through Facebook because it is associated with longer messages, which can be seen on a Facebook Wall post by the friends of each person, to truly express this emotion in a meaningful way.
For the purposes of our research question and the availability of our resources we decided to go with a survey based study. We understand the limitations of self-reporting that can occur in the gathering of our data, however despite such we believe the way that we structured our survey will lead to little self-reporting biases in this study. For this study, we used Cornell Qualtrics to create and deploy both of our surveys. The way that we approached the structure of our survey is that we first constructed the sender’s survey in order to use the messages that they generated for our second round receivers survey. The format of our senders and receivers survey begun with a brief statement of the procedure and what we were researching without giving detail that may bias our results. We carefully laid out that our data collection will remain private and all the information provided would be anonymous. After asking the participant to agree to our survey, we went forth with our questions.
For our sender’s survey we laid out a prompt asking the participants to imagine that they were dealing with a specific scenario. The prompt was the same for all participants, however, we changed the medium between Text messaging and Facebook Wall Post. When we specified the medium to pretend that they were to create the message pertaining to the prompt. In addition, we also changed between the emotions excitement, contentment, and regret when asking the participant to convey a specific emotion. Therefore, for the sender’s survey there were a total of six different prompts. To prevent any form of selection bias we randomized these six prompts, therefore, upon clicking on the survey a participant could have received any of these six prompts by chance rather than conducting six entirely different surveys. Before, the participants reached the prompt we ask them to think of things that made them feel the particular emotion similar to the emotion stated on the prompt. This was in effort to foster the imagination and visualization of the scenario to aid in retrieving the best messages we could for our research.
Depending on the medium stated within the prompt there was a word minimum to encourage the participants to write a lengthy message. For the text message prompt there was a 50-character minimum and for Facebook there was a 100-character minimum. The reason for such is that we believe that Facebook wall messages are typically lengthier than text messages and we wanted to encourage the Facebook wall post to be longer to be more authentic in the reciprocation of the message. After the prompt we asked the participants to rate on a seven point, Likert scale, how much they agreed that their message conveyed the given emotion. Where completely disagree was 1 and completely agree was 7. After that we asked the participants to rate on the same scale how much they agree that if someone one else was to read their message they would know it is trying to convey the given emotion. (See Appendix for all survey questions)
We approach the receiver’s survey using a different method. We compiled the messages that we received from the Sender’s survey and reviewed every message amongst each of the six categories. We then chose a total of 18 different message from the sender’s survey- three excitement and text message, three excitement and Facebook Wall post, three regret and text message, three regret and Facebook Wall post, three content and text message, and three content and Facebook Wall post.
In the formulation of our receiver’s survey we placed one of each emotion from the same medium (i.e. one content, excitement, and regret via text message) on a block. Therefore, there were a total of six different possible message combinations that a participant could have received while taking the survey. We did not allowed participants that took our first survey to participate in our second survey. We ensured this by placing in a conditional question that asked participants if they had completed our sender’s survey, and if they answered yes, they were not allowed to continue with the survey. The first thing we asked the respondents was their current mood to make sure that this did not have an effect on the message interpretation and ratings, and also because there has been research on emotional contagion in messages, and after reading the message it would be interesting to see if they were affected by the content.
For each message we asked the participant how much they agreed that the message conveyed the emotion that the sender intended it to convey on the same seven point likert scale that the senders rated their messages on. After each message the participant was also asked to list any cues from the message that they believed help conveyed the given emotion within the message. After the evaluating all three messages we asked the participant to once again to describe their current emotion to once again screen for any type of biases that may occur in the evaluation of the messages.
For both the senders’ and receivers’ survey we asked the participants a list of basic demographics, which included, race, age, sex, and highest completion of education. This was in effort to see if there were any other correlations in our data that may help with guidance towards further research for our study.
Distributing the Survey
We used Facebook as a main way of gathering participants for our survey by posting a link to our wall, private messages and groups. For the sender’s survey we spent a week collecting data, and for the receiver’s survey we spent a week and a half. The receiver’s survey took longer because we needed to collect a new batch of participants who did not take the original test.
We conducted all our data analysis on excel and SPSS. From the senders’ survey the only information that was necessary for our analysis was the specific messages that we used for our receivers’ survey. All of our data was then coded and analyzed with basic statistics (i.e. mean, frequency, ratio, etc.), t-tests, and mixed model logistic regression.
After analyzing the data we collected it was evident that the majority of the participants were female. The first survey we conducted was brief- only asking the participants to construct one message for a friend they have not seen since high school graduation.
This is in order to look at the first hypothesis of our research in finding out if there is a significant difference between what people think the intensity of the emotion is, and how it relates to their expectations of others rating the intensity of the emotion. Because we are looking at how emotions transfer over the CMC process, it is important to start at the origin and how people are conceptualizing the interpretation of their messages. To see if certain patterns of text match to emotional triggers we also had the senders highlight what they perceived to be emotional markers in their text.
For the receivers’ survey we received a total of 51 participants, which is the remainder of those who dropped out by not completing the survey or had previously taken the senders’ survey. The majority of respondents were females (60 percent). Additionally, of the participants the ethnical group breakdown was: majority, white (76 percent), with a few blacks (10 percent) and about 14 percent of everything else (Latino, Asian, Middle Eastern, Other).
In addition to the collection of messages and other written responses, participants in our first round Sender’s survey were asked to provide values on a Likert Scale from 0-7 for the intensity of their randomly assigned emotion- Regret, Excitement, and Contentment. This resulted on average eight respondents for each of the three emotions under the conditions of text message and Facebook wall post, adding up to the total 49 completed surveys we collected. Across all conditions the mean for the self-reported intensities was greater than or equal to the intensity they thought others would attribute to their message. But there were multiple answers that did not apply to the mission of the research, or would not be appropriate so from these messages, we selected 18 responses to use in part two of our research-three for each of the six possible combinations of emotion and medium. And because these are the only messages that are present in both of the surveys we had to limit our analysis to these 18 surveys to maintain consistency. So by taking the self-reported intensity of the message and the intensity the sender thought others would give the message, we conducted a t-test analysis between the two values to examine the human construct and ideas on how people interpret intensity in messages versus how they perceive their own intensity.
H1: Senders think emotional intensity is higher for them than they think others will perceive the emotional intensity in their message.
As can be seen in Figure 1 of the Appendix, for the 18 messages we used in the both parts of our research does not display an obvious trend between the two intensities. After running a t-test of unequal variances on the intensities, the self- reported mean was equal to 6, and the mean for what they predicted of others was 5.8333. Even though this does support hypothesis one, in that the sender’s mean is greater than what they expect of others, the p (T<=t) = 0.262108, is not significant on any confidence interval; so we fail to reject the null hypothesis and we can not conclude that the two intensities are significantly different from each other (Figure 2).
H2: Senders will rate the message’s emotional intensity higher than receivers will actually rate the intensity of the emotion.
Now we incorporate results from both rounds of our surveys. Before getting into the emotion and the medium effects on message intensity ratings, we hypothesize that there will be a difference between the Sender’s rated intensity and the Receiver’s rated intensity of a message across all messages giving evidence that some of the intentions are lost in the process of translation. The results are very interesting because as seen above, there was not a significant difference between what sender’s rated and what they thought others would rate, but after running a t-test of unequal variances across all 189 receiver responses in comparison to the original 18 sender’s ratings of their messages, it was discovered that the mean intensity that the receivers attributed to the messages was even lower than sender’s predicted in the first survey- this is seen with the mean above 5.8333 vs. the actual perception mean of receivers at 5.37037 (Figure 3). There was also a large variance attributed to this mean, of 1.819543. In the appendix Figure 4 provides a visual of the variance and differences between the self-reported intensity and the perceived intensity by receivers. The t-test shows that with a p (T<=t) = 0.0000000442, that there is a significant difference between what was self-reported and what was received, so we can reject the null hypothesis that the two are the same, and conclude that there is a loss in the strength of emotion in messages, in general across emotions and mediums.
H3: Of the three emotions- Regret will be better interpreted through a private medium (text
Message) and Excitement will be better detected in a public medium (Facebook wall post)
Looking into the various effects of the three different emotions and two different mediums have on the perceived intensity of the emotion, we had each of the respondents randomly receive three messages- one from each of the emotions from either medium because they were not aware of the medium the message originated from. Because the same group of subjects was used to answer multiple questions that were studied we ran a Mixed Effects Logistic Regression. A multi-factor ANOVA is used so we can factor see if the data is affected by having respondents answer three different messages. This test looks at how the independent variables of medium and emotion in conjunction with the respondent and the message number influence the dependent variable- the difference between the self-reported intensity and the perceived intensity. The within subjects factor of sampling the same subjects for various ratings, combined with the between subject factors of medium and emotion combine to influence the ratings of the emotions.
Firstly, the test revealed that Schwarz’s Bayesian Criterion (BIC) fit index, is 670.399 for our model, which signifies that the explanatory variables leave a considerable amount of unexplained variation. The smaller the BIC the better the model fits the data, so our results could find better variables to account for the variation in difference (DV). But even so, there are still significant relationships between the independent variables and the dependent variable of difference between intensities (Figure 5). As seen in the Type III Tests of Fixed Effects which tests for equal variances, the medium and medium*emotion have respective significances of 0.006 and 0.002. Revealing that at .05 significance level, these two factors have a statistically significant effect on the difference in intensities. This is important because the receiver’s were not told the medium so that fact that the messages were able to yield significant results based on medium associated with the message shows a difference between how the messages were originally crafted for a specific medium and the success of intensity translation is different between text message and Facebook wall post. Furthermore, the Estimates of Fixed Effects (Figure 6), shows that there is a significance present in Text message (medium=1.00) of .001, and of Regret (emotion=1.00) of .015, which is extremely important for our investigation of hypothesis three because it signals that the data under these two conditions yields a significant effect on the difference between self-reported and perceived intensity. This means that text messaging results in higher differences than Facebook, and regret causes higher differences in intensities than excitement and contentment. In addition to these important findings, this chart also shows us that the message number the respondent was randomly assigned did not have a significant effect on the dependent variable (p= .570 and .305) which is extremely important in making sure that our results have not be jeopardized by the use of three different messages for each emotion and medium condition. Similar to this the Receiver_ID (each one yielding three intensity ratings), also did not have a significant effect on the dependent variable, so we can continue to analyze our findings between emotion, medium, and difference without the results being clouded by these other random effects.
In the Mixed Model analysis, the medium- Text message versus Facebook wall post was tested against the difference between self-reported and perceived intensities of the messages. In Figure 7, you can see the mean difference in intensities in a text (medium=1), is 0.920 with standard error 0.149, and in a Facebook post the mean is 0.323 with standard error 0.147. This shows us that text messages are associated with a greater loss in intensity when being translated, and that Facebook is more likely to preserve the intensity across all three of the emotions. To see if this is a statistically significant difference between the two mediums, we ran a Pairwise Comparison which yielded a mean difference of 0.597, SE= 0.209, with a significance of 0.006. So at the .05 significance level we can conclude that there is a significant difference between text message and Facebook and the resulting intensity differences. Also for this comparison, F= 8.151, which tests the effect of the medium based on the linearly independent pairwise comparisons.
For each emotion, the Mixed Model test provided an associated mean difference in intensities in order to see the variation in interpretation success as a function of the emotion without the medium factored in. The mean difference for Excitement was the highest of the three emotions with a mean of 0.879, SE=0.176. Regret was the next highest difference with a mean of 0.518, SE=0.176, and then Contentment with a mean of 0.467, SE= 0.176. This reveals that based on means, excitement was the worst at translating from sender to receiver. But in order to see if this information is significantly different from Regret and Contentment, a Pairwise Comparison was now done on emotion to reveal that there was not a significant difference between emotions in any combination (Figure 8). The resulting F stat was 1.692, which has a corresponding significance of 0.189, which supports that there is not significant difference between the three emotions.
Lastly, the combined effect of the two independent variables was tested to see the effects of intensity differences. Based on the estimates it was found that the mean difference for the emotions in text messages were Regret (Mean= 0.296, SE= 0.251), Excitement (Mean=1.425, SE= 0.251), Contentment (Mean= 1.038, SE= 0.251). This shows that in a text message regret has a very small mean difference between intensities**This is consistent with hypothesis three, stating that the emotion regret would be better interpreted in a private medium such as text messaging. Excitement has a relatively large difference in comparison. Whereas in a Facebook wall post the respective emotion mean differences are, Regret (Mean= 0.739, SE= 0.246), Excitement (Mean= 0.333, SE= 0.246), and Contentment (Mean= -0.104, SE= 0.246). This is consistent with earlier findings that we reported that overall Facebook is more successful at translating intensity, which is seen here in the smaller mean figures when compared with text messages. But it also shows that in a Facebook wall post, regret intensity is the least successful at being translated from sender to receiver. Excitement and Contentment are very close to zero, so they are very successfully communicated in the public medium of a Facebook wall post. In the Pairwise comparisons, found in Figure 9, it is evident that emotion 2= Excitement, and emotion 3= Contentment, there are statistically significant effects on the difference in intensity ratings coupled with the mediums. These significance figures are 0.002 for Excitement, and 0.001 for Contentment. But Regret is not significant with a figure of 0.209. This is a result of the relatively small mean difference in regret; text message because text messaging was attributed with higher discrepancies, but in regret there was only a 0.296 difference in intensities. For each of the three emotions, the F stat also indicates the goodness of fit. For Regret F= 1.592, Excitement F= 9.665, and Contentment F= 10.578 (Figure 10). From these F numbers we can see that there is something about Regret that makes it differ from Excitement and Contentment in message intensity interpretations. For the two mediums the Univariate Tests revealed that for text messages the F= 5.424, with a significance of 0.006, and Facebook wall post F= 3.029 and significance 0.052 (Figure 11). Also supporting that in text messages there is a significant difference in intensities at the 0.05 level, or 95% confidence interval, but not in Facebook wall posts.
Message Number Examination:
As mentioned earlier the Mixed Model Linear Regression also tested for the various message numbers that were randomly assigned to respondents, and the resulting influence that had on the intensity differences. But the message numbers each had significant levels greater than 0.05, so we concluded that in the 18 messages sent out, the three messages the person was given to evaluate did not have a significant impact on the resulting intensity ratings.
Our research question and study design was all in collaboration to examine the successfulness of various emotions that interested us-regret, excitement, and contentment- and observe their various abilities to translate to receivers depending on the social medium that was used to construct the original message. This is a very relevant topic because of society’s dependence on computer-mediated communication for an increasing amount of conversations. Based on this increasing communicative interdependence, relationships of all intimacy levels are involving a higher ratio of CMC to FtF interactions than ever before. Text based exchanges are no longer limited to family or close friends, now people are using these mediums to exchange information with others in the professional environment as well. Because of this, it is tremendously important that the intentions of Sender’s be transferred to the receiver in efforts to limit miscommunications and the resulting consequences they create. According to our results, it was found in hypothesis one that people overestimate the intensity that the receivers will give their message, which shows that there is some discrepancy between what people think when they are constructing a message, and what people gather from reading the same message. This may be due to the fact of self-assessing human bias, which is the idea that what is obvious to oneself should be obvious to someone else. It was also found that there is a significant difference across all emotions and mediums between the self-reported intensity of the emotion and the perceived intensity. This is important because it attests to the fact that there is a loss in the translation regardless of the emotional content or the medium choice; messages as a whole are not interpreted the way senders would predict.
Our research offers a supplement to pre-existing research on topics such as the uses and gratification theory, because as that theory poses the question, why do people use certain media for specific needs; our results, specifically for regret, a negative emotion was more successfully transferred to receivers in a private sphere such as text messaging. This is consistent with the theory because the sincerity of the message could be magnified by the respectful consideration to express the emotion in a private message, where as posting the same message on the person’s Facebook wall may have come across as more of a selfish need by the Sender to get public attention, as opposed to actually expressing the genuine feelings intended for the receiver. But we see the opposite effect in excitement and contentment; in these emotions, a Facebook wall post was actually associated with lower mean differences in intensity ratings between senders and receivers. This may be a result of these emotions being more positive, so according to the UGT this happy expression could be exasperated by the willingness to display it to the entire social network of both the sender and the receiver. This also relates to the warranting theory because the sender may be signaling pride in the message with his or her willingness to make it a public message. Consequently increasing the cues that are “given off” by the text, in addition to the emotional cues that the sender is giving to the receiver.
According to previous research by Hancock et al (2008), “both positive and negative moods, as displayed by the confederate, induced a similar emotion in other participants,” and the data that they collected, “suggest that in communicative environments with non-verbal cues emotions can be contagious” (p. 296). In accordance to our findings receivers of our study should had been able to interpret excitement and regret better than interpreting contentment. However, our findings found that contentment has the smallest difference between the senders and receivers’ rating compared to excitement and regret. The reason for such could be due the prompt that we issued. It is more difficult for a participant to pretend that an event occurred than actually living that experience. Hence, the issue could have been that we were asking our sender participants to draw on false emotion and lead to the misinterpretation of their data. In the study The role of emotion in computer-mediated: A Review, the authors discuss that when it comes to emotion it is the distinct style and word choice that influences the conveyed emotion through computer mediated communication. (Derks, Fisher, & Bos, 2007). In our data collection, for the emotion regret, there were choice words such as sorry that help influence the presence of the emotion more clearly. Excitement on the other hand, while an expressive emotion, it is difficult to use particular choice words to convey this meaning. Instead, the usage of other indications is present such as punctuation and emoticons to help embellish the presence of the emotion. Therefore, when expressing such an emotion in a medium such as text messaging, which has a limited character set, choice words are more efficient in clearly articulating the meaning of the message over simple emoticons. This can explain why in text messaging regret was a more prevalent match between the senders and the receivers as opposed to excitement. To explain for the opposite in Facebook wall post messaging further research is necessary, but there is a plausible explanation. It seems that when it comes to emotions such as regret, is not necessary an emotion that people will associate with an open post on someone else wall where anyone can simply read it. Therefore, on both ends, the sender could be reluctant to add the necessary embellishments into the message to portray regret, in addition, to the receivers could had been reluctant to justify a message as full of regret because it would be odd to see such on a public wall post.
Whereas our research thus far has been comprehensive, a lot of work still needs to be done to further the overall assessment on how media affects emotional cues, and how emotions are understood and interpreted through various online communication outlets. In an ideal world we would have had significantly fewer limitations than we ended up perceiving after conducting and analyzing the data that we collected, and so it is worth it to note all of them. Chiefly, we only selected three emotions, regret, content and excited, mainly due to the fact that we wanted to use more intense emotions that the commonly studied happy and sad, but also because asking the participants of the study to rate each message on a multi-generated emotional scale would have skewed the sampling information as a whole. We also did not account for the cultural differences between how different groups of people tend to assess certain text-based patterns online. We would have focused on this aspect as well as other culture based limitations like non-native language translation (English into Chinese) but given our sampling of about 150 Cornell University students who generally were in one demographic bracket, that task had to be pushed aside for future studies. We also specified the prompt and generalized the relationship between the sender and the receiver of the messages in our study, which could have also limited our results as a whole. It could also be a point in emotional interpretations that family members who text or Facebook each other will pick up on more subtle emotions through texting than an acquaintance would. Accounting or the relational intensity of the participants would be a worthwhile venture for another study.
Another very large limiting factor that we were initially confronted with was the fact that we had to deploy the survey at a rapid pace and did not have time to create a simulated experience of actually using Facebook or actually sending or reading a text message. Our results could be partially biased because the people who were taking the survey had to pretend that they were using the services that we were asking them to create examples for, as compared to actually being on those services or actually sending text messages. The atmosphere of survey questions and responses was contrary to the experiences one goes through when using Facebook, so having a study that is a diary study through the mediums would likely be a suggestion for future works. And finally we did not account for how frequently a person uses each or one of the mediums that we asked. It could also be hypothesized that individuals who use a certain medium over another may grow more accustomed it and understand how to get their point across in a more efficient manner. For example, take a person of about 60 through 65 years, doesn’t really know anything about cell phone usage as well as Facebook and then promptly tell them to convey a certain message generalization; their answers would be very dissimilar to a tech-savvy college student. So in order for us to have a more well-rounded research conclusion, we would need to factor in and test how frequently a person uses the medium, what their cultural background is in order to see if there is a influencer in cultural norms in text based conversations, try to get the participants to use the actual mediums that we are testing under and have a more diverse emotional range so that we can thoroughly and universally prove our hypotheses. Overall however, the research that we conducted did prove that at least for certain emotions, the media that one chooses to use does have a large impact on how the message will be interpreted.
Derks, Daaantje. & Fisher, Agneta H. & Bos, Arjan E.R. The Role of Emotion in Computer- Mediated Communication : A review. Vol 24. no 3. 766-785. http://dx.doi.org/10.1016/j.chb.2007.04.004
Hancock, Jeffery T., & Dunham, J. Phillips. (2001). Impression Formation in Computer Mediated Communication Revisited: An analysis of the Breadth and Intensity of Impressions. Communication Research. vol. 28 no. 3 325-347. doi: 10.1177/009365001028003004.
Hancock, Jeffery T. & Gee, Kailyn, & Ciaccio, Kevin & Mae-Hwah Lin, Jennifer. I’m Sad, You’re sad: Emotional contagion in CMC. Social and Behavior Systems: Psychology. http://sml.comm.cornell.edu/wordpress/wp- content/uploads/2013/12/Im-sad-youre-sad.pdf
“Internet Use Over Time”. Pew Research Center, Washington, D.C. (2014) http://www.pewinternet.org/data-trend/internet-use/internet-use-over-time/, accessed on 5/1/2014
Lo, Shao-Kang. (2008). A Pragmatic Multimodal analysis of emoticons and gender in social networks. http://www.academia.edu/6669015/A_pragmatic_and_multimodal_analysis_of_emoticons_a nd_gender_in_social_networks. Accessed on 5/1/2010
Savicki, Victor, & Kelly, Merie. (2000). Cyber Psychology & Behavior. Vol. 3 no. 5. 817-
826. Retrieved from http://online.liebertpub.com/doi/pdf/10.1089/10949310050191791.