The concept was coined by Rheingold (1993, n.p.), who defined virtual communities as “social aggregations that emerge from the Net when enough people carry those public discussions long enough, with sufficient human feeling, to form webs of personal relationships in cyberspace.” Despite criticism of this definition as vague (Androutsopoulos, 2006; Zappavigna, 2011), the criteria established by Rheingold recur in Herring’s (2008, p. 921) overview of frequently-cited characteristics of online communities, namely:

The notion of spatiality has long been contested in sociolinguistics with the recognition that people belong not only to physically-located “speech communities” (Labov, 1966) but also to various discourse communities formed not through geographical proximity but through shared interests and goals (Barton & Tusting, 2005). However, the notion of “virtual space” can be useful in understanding the development of online affiliations in well-defined sites like TED.com. Gee’s (2004) concept of “affinity space” explains how people may be drawn to interact with others in an online location because of shared interests or activities and how this may lead informally to the formation of what users would recognize as community practices and norms. We see TED.com as comprising affinity spaces, in that the website provides the opportunity for active participation through the sharing of ideas and knowledge, which may in turn lead to shared practices, the emergence of specific roles, and a sense of shared history and group identity among some users (Herring, 2004).

Persistence of interactions (the second criterion) is likely not a feature of many contemporary online affinity spaces. In 1985, the WELL community described by Rheingold had 700 users, scattered over the United States, many of whom would log in daily and develop face-to-face relationships, despite its asynchronicity. By contrast, TED is a public, global, highly populated platform, whose users are likely to divide their attention among various online sites. One consequence is that users of modern social media have greater liberty to associate with, and move among, specific values or interests (boyd, 2012). The fact that shifting, multiple allegiances have also been described as characteristic of contemporary, urban offline communities (Blommaert & Varis, 2011) suggests that this phenomenon is not restricted to online interactions and may have wider significance in society. Despite this, in crowded and rapidly-updating spaces such as TED.com, attempts to affiliate with others nonetheless seem to shape much user behavior (Seargeant et al., 2012). Given the lack in many affinity spaces of prolonged interactions or binding exclusive membership, the signaling of allegiance to values or interests through language becomes an important way of affiliating with “like-minded” others. As discussed by Herring (2004, p. 361), such discourse behaviors as in-group references and the expression of solidarity can act as indicators of the development of virtual community.

Relevant here is the concept of imagined communities (Anderson, 1991), according to which criteria such as spatiality and sustained interaction become less relevant than online users’ perceptions of whom they are addressing and with whom they wish to affiliate, meaning that attempts to align practices with imagined others can be crucial in determining community. This process of affiliation may be more complex online than in typical face-to-face interactions, not only because the online audience is invisible and unknown but also because online audiences can potentially involve diverse groups of people which the user must address simultaneously (Marwick & boyd, 2014). The linguistic strategies used to address such complex audiences have been described across online sites (Frobenius, 2014; Tagg & Seargeant, 2014) and include pronouns, code-switching, and contextualized references.

As this suggests, particular uses of language – choices between languages, level of formality, use of respelt forms, and so on – help users affiliate with other users. The use of linguistic resources in community formation may be particularly important in online spaces where people do not have access to physical cues such as gesture and body language and where the burden of pragmatic work tends to fall on “code-centred” cues (Georgakopoulou, 1997, p. 58). Zappavigna’s (2011) linguistic study of Twitter, for example, suggests that Twitter conventions such as the hashtag allow users to cluster around “evolving topics of interest” (Zappavigna, 2011, p. 800), a phenomenon which Zappavigna calls “ambient affiliation.” According to Zappavigna, users who are not “acquainted” on Twitter – they do not know each other offline or have a lasting online relationship – can still form ephemeral alliances through adopting shared conventions and values. In online, largely text-mediated contexts like Twitter (and TED comment threads), where the exact composition of one’s audience is unknowable (boyd, 2012), affiliations can therefore be said to be interactively constructed through users’ use of language (Tagg & Seargeant, 2014). Thus users can potentially bring into being a community through their language choices, which are in turn shaped by their perceptions of the individuals they wish to affiliate with.

To conclude, linguistic analysis of how users imagine and address communities online can show how users employ language resources to signal and forge affiliations. The argument that online affiliations can be identified through people’s shared language practices underlies our study of TED.com, where we show how users’ expression of shared values and beliefs constitutes an important interpersonal resource in affiliating with like-minded individuals who share similar views.

The aim of our study is to determine whether and, if so, how evaluative language resources act as a solidarity-building resource for TED.com users. This aim rests on the assumption, explained above, that a sense of “community” can be forged through the expression of shared values. We address the following research questions:

TED videos are free to watch for anyone visiting the website, but only registered users (‘members’) can comment on them. Although the videos tend primarily to attract non-academic audiences (Sugimoto & Thelwall, 2013), the comments differ significantly from those left in response to YouTube videos, which tend to be more interactive, less focused on the topic of the video, and more negatively-orientated (Tsou et al., 2014).

We collected 22 comment threads, corresponding to the 22 TED talks categorized under the topic of “food” by May 2012. Given the emphasis placed by AF on Tenor, selecting videos with a common topic makes Field a less problematic variable. The topic of food was chosen for practical reasons: The corresponding comment threads were considered to contain sufficient data for building a small corpus that was manageable in terms of the researchers’ time limitations. Comments were arranged chronologically (from September 2006 until May 2012 inclusively), manually stripped of unwanted information (such as the number of negative/positive votes given by other users and the date of posting) and then compiled into a corpus (TEDCor) containing a total of 22 .txt files and 340,938 words. [1] For the AF study, we selected comment threads on three talks, to cover a breadth of food-related topics and different patterns of interaction. Below is a brief description of the selected talks. The fact that all speakers are male reflects the gender distribution across the site, in that only a quarter of talks are delivered by women (Sugimoto et al., 2013), although this may impact our results, given that female presenters tend to attract more “emotional” comments (Tsou et al., 2014).

Thread 1: Roger Doiron on home and community gardening (119 comments posted by 86 users between December 2011 and May 2012; characterized by a tendency for users to evoke their personal gardening experiences). [2]

Thread 2: Dr. Heribert Watzke on the brain and digestive tract (259 comments posted by 160 users between October 2010 and May 2012 (most posted October-November 2010); takes the form of a long, complex debate with numerous participants and various topics). [3]

Thread 3: Chefs Ben Roche and Homaro Cantu on innovative cooking methods (156 comments posted by 83 users from December 2011 to May 2012, 58 of which are the contribution of one of the speakers, H. Cantu). [4]

Each of these discussion threads was manually coded for instances of Appraisal by two coders, using the colour-coding function provided by CATMA 3.2., [5] a qualitative data analysis software tool. Although interrater reliability was not measured, disagreements were discussed and resolved. We also used CATMA to generate numerical data for each category and subcategory of Appraisal (see Table 1).

The second part of the analysis employed a more quantitative approach. To carry out our corpus analysis we used a corpus linguistics software tool, AntConc (Anthony, 2001). By using a collection of texts (a corpus) as input, AntConc can perform a series of quantitative queries, such as word-frequency lists or word-cluster lists (lists of words that most frequently occur with the searched word). AntConc can also generate a list of the most significantly frequent words in a corpus (keywords), by comparing the word frequencies with those in a larger, reference corpus. The reference corpus used was the 40-million-word sample of the enTenTen web corpus (a corpus of English language web pages collected in 2012, which can be accessed through Sketch Engine [Kilgarriff et al., 2004], [6] a corpus query system). Another functionality of AntConc is to generate concordance lines, which display all occurrences of a word and its immediate contexts. This type of manual, intuitive analysis of words in context allowed us to perform a more in-depth, qualitative type of analysis of those words that quantitative corpus analysis indicated as significant.

Ethical concerns arise when working with authentic data collected online, particularly around user privacy and confidentiality. TED.com comments are public, and quoting them does not breach privacy conditions (AoIR, 2002). However, to preserve confidentiality, we use only users’ initials in our examples.

Discussion

What many TED.com comments across the three threads have in common is the presence of largely positive evaluations, of judgment (moral evaluations of people), appreciation (evaluations of objects or ideas), and affect (emotional reactions), suggesting that users are in general agreement with each other and that they agree with the ideas expressed in the videos. This positivity is illustrated below:

Thread 3/C.M.: You guys are remarkable [judgment+]! [7] [referring to the two chefs]

Thread 1/T.R.: I loved [affect+] this talk! I’m trying to turn it into an ESL lesson for my students.

Thread 2/A.A.: BTW, I’m crazy [affect+] about steak tartare. Delicious [appreciation+] :P [affect+] [8]

Affect and judgment occur very infrequently compared to appreciation (241 and 325 times, respectively, in comparison to 1361 total occurrences of appreciation), for reasons discussed below. When it does occur, positive affect is mostly used to express enthusiasm for the talk or for stating personal preferences, as in the following examples:

Thread 1/Z.: YEAH! I am so happy [affect+] I saw this.

Thread 3/D.P.: I love [affect+] Moto, one of Chicago's finest [appreciation+] establishments.

As for judgment, although positive judgments are made by users to praise the speaker and other users, there are also a few examples in the corpus when negative judgments are used as insults. In this example, a commenter accuses another of giving false data.

Thread 2/T.K.I.G: Quack [judgment-] Doctor, Are you going to stick with your fabrications and false knowledge [appreciation-]?

A first glance at appreciation suggests that it often refers to an evaluation of the ideas in the video, as in these examples:

Thread 1/K.R.: This is a brilliant [appreciation+], inspiring [appreciation+] talk, and a well crafted [appreciation+] argument as a well told [appreciation+] story. Subversive plot, giving back power through gardening, gardens grow healthy [Judgement+] kids, and how do we grow more [force] gardens.

Thread 3/S.T.: That is awesome [appreciation+] talk... :) [affect+].

The fact that about one-fifth of the comments start with a positive appreciation suggests that users are trying to engage other users in discussion. These positive appreciations are frequently followed by either a personal story (18 examples, mainly from Thread 1) or a disclaim (66 examples), which we discuss in more detail in the following section. Commenters seem to use these initial positive appreciations as a means to align themselves favourably with their interlocutors and/or with certain ideas expressed in the video, in order to make relevant a personal story or hedge a subsequent disclaim:

Thread 1/A.C.: Wonderful [appreciation+]! so [FORCE] inspiring [appreciation+] and moving [appreciation+]. I am a community worker, the coordinator of 37 community gardens in the city of Jerusalem (and a TEDtranslator, too!) [personal story] Is there any way we can get TED to transcribe and allow to translate this talk so we can spread it around to the thousands of co-plotters in this part of the world?

Thread 2/L.L.: THis is all very [FORCE] good [appreciation+] however [disclaim] we are forgetting a very [FORCE] important [appreciation+] issue with the idea of eating just fruits and nuts.

Users also draw on appreciation to directly evaluate opinions and ideas present in the comments, as in these examples from Thread 2:

Thread2/A.S.: Brilliant [appreciation+] comment!

Thread2/P.B.: Excellent [appreciation+] observation if only more people were as observant [Judgment+].

Thus, users appear to use positive appreciation not only as a means for expressing their own appreciation of the videos, but also for engaging with each other’s ideas on the particular topic or in response to a video.

To conclude this section, positivity is an important feature of the TED comments. Users overwhelmingly choose to frame their comments as positive evaluations (or appreciations) of the ideas being expressed. A possible explanation for the relatively high use of appreciations can be found in the characteristics of the platform: TED videos and discussions are mainly about ideas, so the targets of evaluation are, in most cases, abstract concepts. The focus on ideas is less confrontational than evaluations of other users, and less egocentric than expressions of personal feelings; thus it may constitute a more positive community-building resource on the site. Positive appreciations in particular are used to stimulate dialogue and set a polite tone for the discussion, in combination with engagement (disclaims) or personal stories, two categories that we discuss in the following section.

ENGAGEMENT refers to how users acknowledge alternative viewpoints and how they relate to these “other voices” (Martin & White, 2005, p. 94). As evident in Table 1, heteroglossic discourse is a well-represented category across the three threads (with 392 occurrences in Thread 1, 940 in Thread 2, and 334 in Thread 3), showing that users are picking up on their interlocutors’ ideas and opinions. By contrast, monoglossic discourse has considerably fewer realisations (see, for example, Thread 2: 58 instances of monogloss compared to 940 instances of heterogloss). This reflects users’ predisposition for dialogism, as we might expect from an online discussion group.

The following sentence from Thread 2 illustrates all four sub-categories of heteroglossic engagement (proclaim, entertain, attribute, disclaim):

Thread2/J.W.: I reckon! [proclaim] Further, i can only guess [entertain] that he's talking about [attribute] cooking meat, and thus ignoring a raw food vegan or vegetarian diet, and yet [disclaim] he started by referencing this [attribute] in describing our teeth as being best suited to soft mushy food.

Apart from demonstrating the dialogic character of a comment or comment thread, engagement can also work as an indicator of group solidarity (Martin & White, 2005). The most relevant subcategory in this sense is the subcategory of entertainment, which consists of expressions that acknowledge one’s subjectivity and the existence of alternative views. Table 1 shows entertainment, followed by disclaim, to be the most frequent subcategory of engagement. In the following example, S.A. uses the modal “maybe” to entertain other possible explanations or points of view; furthermore, by using the verb “imagine” in the first person singular, S.A. underlines the fact that this represents his/her own, subjective position:

Thread 2/S.A: And maybe [entertain] there are some other categories. There are also some abrupt changes I can imagine [entertain] that take just a generation, maybe [entertain] induced because of radiation.

There are numerous such examples of entertainment across the three threads. Also, entertainment often co-occurs with the other frequent subcategory, that of disclaim, as illustrated by the following examples:

Thread 3/G.H.: I would say that [entertain] the "tuna" has a nutritional profile radically [FORCE] different than that of real [FOCUS] tuna, but [disclaim] it also is a striking [APPRECIATION+] example of the transformation that can [entertain] take place in food.

Thread 2/M.W.: I don't think [disclaim] he tried to justify eating junk food so much as [entertain] tried to make a joke about his own weight.

This reciprocal relation between entertainment and disclaims illustrates the point made by Martin and White (2005) that solidarity within a group is not necessarily expressed through total agreement, but rather through respect for alternative opinions. The detailed analysis of comments indicates that these two subcategories are working together, rather than excluding each other. A similar point was made in the previous section with respect to the combined use of appreciation and disclaim – the appreciative comments served to hedge and frame the divergent views expressed, as illustrated by the following example:

Thread 2/M.P.: Interesting [appreciation+], but [disclaim] I would like to know more about the other types of feelings we get in our stomachs.

These practices allow users to express divergent views while at the same time affiliating with certain topics or ideas and maintaining a polite tone. With a relatively high level of tolerance and the presence of alternative viewpoints, interactions between users of TED.com seem to differ from those on larger video platforms such as YouTube (Tsou et al., 2014); for example, in his study of YouTube dialogues between Christians and atheists, Pihlaja (2011) finds a lot of antagonism and disagreement.

A relation-building role is also played by the personal stories or experiences mentioned in the previous section, which occur in 116 occasions across the three threads, sometimes (18 cases) preceded by appreciation. The frequency of these personal stories by themselves or in combination with Appraisal resources – see the use of appreciation in the examples below – suggests that they represent another important interpersonal resource on TED.com, as also noted by Kouper (2010) in her study of online peer advice. The willingness to share personal experiences in a public space indicates an intention to belong, rather than a desire to be different. This bridge-building quality is visible in the following examples from Thread 1, in which two users from different countries share their common experiences with gardening:

Thread 1/Z.A.: simply inspirational [APPRECIATION+]!
I am from Karachi, Pakistan and here I am trying to make use of empty plots around the city. I am trying to inspire and educate people to use their rooftops and balconies to grow their own food [personal story]. Roger Dorion inspires me because its amazing [APPRECIATION+] how he has gathered thousands of gardeners from around the world for the same cause. […]

Thread 1/J.N.: im from india [personal story] and whatever your doing is really [FORCE] great [APPRECIATION+]..our countries are space constrained and we need ideas like this...ill spread this idea too.

In conclusion, the frequency of heteroglossic comments (as opposed to monoglossic) and particularly of entertain and disclaim, indicates that TED users do more than rating and commenting on a video: They imagine, and communicate to, a group of individuals who they perceive to have similar interests and values. Moreover, the presence of the personal stories in this public debate space not only suggests a certain degree of intimacy but also reinforces the idea that TED users seem inclined to bond through sharing.

graduation – the intensification or other specification of evaluation – is frequently used by TED commenters. Table 1 shows realisations of graduation to be even more numerous than those of attitude in Threads 1 and 2. As in the case of attitude, the distribution of the subcategories of graduation is uneven – in all three threads, there are considerably more instances of force (that is, intensification), than instances of focus (which expresses the typicality of a category). Generally, graduation works alongside other Appraisal categories to intensify or add nuance to evaluations:

Thread1/M.A.: Awesome [appreciation+] Talk. I really [force] enjoyed [affect+] watching. I ve been a TED Talks video watcher for years now [force]..... this is my first comment...

The frequent use of force (intensification), along with appreciation, suggests a high level of involvement – users seem very committed to what is being said. As shown in the following example, not only are emotion and personal opinion reinforced, but also factual information or argumentation:

Thread2/J.G.: Gorilla might not be a good example. It is a herbivore, but the dental structure is somewhat [focus] universal (hence - omnivorous), just as ours is. I'm not sure teeth is a good indicator - they change far [force] slower than the diet. To see what a pure [focus]-herbivore dental structure looks like, look at a deer skull (gorilla doesn't come anywhere near [force] that degree of adaptation to herbivorous diet), to see what a pure [focus] carnivore looks like, look at a wolf's skull - both are very [force] speciffic and practical designs, and don't look like ours much [force].[…]

Here, force and focus play a complex rhetorical role – on the one hand, intensifiers are used to impress on the reader the validity of the user’s point (“far slower,” “anywhere near”), while other forms often appear to help the user avoid coming across as overly authoritative in this informal forum (as in “somewhat universal”). The use of graduation and positive appreciation thus suggests that commenters are intending to persuade other commenters and that the Appraisal strategies indicate a high degree of commitment to the values or ideas expressed (as opposed to using entertainment or attribution, such as quotation or hedging, which might be considered typical of academic debate).

To summarise, analysis of Appraisal in the TED corpus reveals a high level of interactivity within the three selected comment threads. While contributing to the general discussion of ideas, users seem to foreground interpersonal, rather than ideational meanings; that is, they seem as much, if not more, preoccupied with aligning themselves with others than with stating privately-held views. In doing this, users employ graduation-force (which suggest a high level of commitment), as well as positive evaluation (taking a stance helps them align with certain ideas, and consequently affiliate with groups holding shared values that emerge from the comments). Despite their sometimes divergent views, users manage their affiliations by prefacing divergences with positive appreciation, and seem committed to bonding and sharing experiences (such as through personal stories). In fact, it appears that TED users in these threads share to some extent similar worldviews and values in relation to food. In the next section, we investigate this further with the help of corpus analysis.

Corpus analysis of the full data set of 22 comment threads (340,938 word tokens) suggests that the attitudes revealed through qualitative Appraisal analysis of three threads form part of a wider pattern across TED comments, at least those relating to food. By looking closely at the top 150 most frequent “keywords” (that is, words and phrases that occur significantly more frequently than in a general corpus of English-language texts published online), we noticed that a considerable number fit into one of the five ad hoc categories illustrated in Table 2. These manually-identified categories are: 1) food-related terms; 2) deictics, particularly pronouns; 3) debate-related terms (a group of words that suggest an ongoing debate or discussion); 4) positive evaluative terms; and 5) intensifiers and quantifiers (realized adjectivally and adverbially).

Table 2. Categories of keywords in TEDCor (with rank orderings)

With 2478 occurrences, the word food is the most significant lexical keyword in TEDCor. This and other food-related terms (eat, cooking) suggest the “aboutness” of the corpus (Scott, 1999) – food. Although unsurprising given the topics of the videos, the finding suggests that users generally do not go off-topic in their comments – a tendency noted elsewhere in the literature and in contrast to YouTube commenters (Tsou et al., 2014). As shown in the analysis of appreciations, these TED users come together around their evaluations of the ideas expressed in the videos.

We can then identify what commenters say about food by looking at the collocates of food – that is, the words that occur most frequently with food. In this case, the most common adjectives used with food include fast, healthy, junk, organic, good, and processed, as shown in Table 3. Many of these collocates of food are themselves keywords in the corpus, such as good (Table 2, column 4), healthy, organic, and local (column 1).

Table 3. Collocates of food (modifiers only) with frequencies (from Sketch Engine)

The frequency and clustering of these keywords in TEDCor suggest the presence of certain discourses or “conventional ways of talking” (Stubbs, 1996, p. 158) among these users about food (Hunston, 2011). These include a tendency to discuss food in terms of its source (e.g., organic, local), its naturalness (organic, processed), and its nutrition (junk, healthy). Although this is unsurprising, as these are widely-shared contemporary concerns, it is not inevitable – for example, the British National Corpus collected in the 1980s shows a different picture, with contributors discussing food in relation to industry (agriculture, fisheries, supply, production), safety (poisoning, hygiene, safety), and problems such as intolerance and shortages. However, the method shows the potential for uncovering less obvious worldviews.

The frequency of deictics and debate-related words (columns 2 and 3, Table 2) seem to confirm what we could only speculate based on the qualitative analysis of the three conversation threads: that, despite the lack of community-building functions, there is a considerable amount of interaction among TED commenters.

The presence of deictics – grammatical words whose referent is not fixed but contextual – can suggest a shared understanding among interlocutors. For example, a random sample of 150 concordance lines containing the first-person plural pronoun we revealed that it is mostly (in 85% of cases) used inclusively (meaning that the addressee is also included), which is a typical means of engagement in academic discourse (Hyland, 2005) and which, in the context of user comments on TED talks about food, seems to serve the same purpose: to position commenters and their imagined audiences along the same lines. The function of we in these comments is either to persuade, as in “I think we can all agree that: Humans have nails instead of claws, used for grasping, not for tearing at things like a tiger” or to invoke solidarity, as in “We as a society need to assume responsibility for ALL our actions, purchases and habits.” Also used to invoke solidarity is the possessive our, another important keyword, whose most frequent collocates are food, body, child, health, and diet, suggesting that users are invoking a common concern for these issues: “Getting back to nature in our food, organic nature, IS the key!”

Turning to debate-related words, the fact that point, fact, idea, agree, and example are significantly frequent in TED conversation threads about food (column 3, Table 2) underlines the dialogic and deliberative nature of these threads, as previously suggested by analysis of GRADUATION. This is illustrated by two keywords, talk and idea, in Tables 4 and 5 below. Concordance analysis shows that talk (737 occurrences) when used as a noun (90% of cases) is the target of evaluations in more than half of its occurrences, with a ratio of positive to negative evaluations of 6:1. Also, as shown in Table 4, the verbs love and enjoy are frequent collocates of the noun talk, as are modifiers such as great, interesting, good, and wonderful. Similarly, idea (288 occurrences) is the target of evaluation in half of the cases, and it mostly collocates with verbs such as like or love and positive adjectives such as great, good, or nice (Table 4), with an overall positive to negative evaluations ratio of 4:1. Using the “Concordance plots” function in AntConc, we looked at the distribution of the keywords talk and idea and their most frequent collocations – great talk, good talk, great idea, good idea. We found that these words and collocations are evenly distributed across the corpus, occurring once or twice in several comment threads. Moreover, a key phrase analysis using the Sketch Engine indicates great talk to be the most significant key phrase in the corpus. It can thus be concluded that TED users – in the interactions included in the corpus – tend to rate positively the ideas and talks presented on the website.

Tables 4 & 5. Most frequent collocates for idea and talk (from the Sketch Engine)

Finally, the evaluative terms, intensifiers, and quantifiers (columns 4 and 5 in Table 2) indicate that two other patterns identified in the analysis of Appraisal can also be attributed to the TED corpus as a whole. The frequent use of intensifiers and quantifiers (which, in AF, are typical realisations of GRADUATION), their status as keywords, and their relatively even distribution across the 22 files of the TED corpus suggest that the frequent use of GRADUATION is characteristic of the entire corpus. In a similar manner, the unusual frequency and the balanced distribution of positive evaluative terms such as good, great, love (verb), and like (verb) in TEDCor suggests that positive ATTITUDES are often expressed by TED users. The trend also emerges from the list of the 30 most frequent adjectives in the corpus. As evident in Table 6, positive qualifiers – such as good, great, healthy, important, and interesting, as well as adjectives such as real, local, sustainable, and organic (in bold in Table 6), which have positive meanings in the context of TED food-related comments –outnumber negative adjectives such as bad and wrong (in italics).

Table 6. The 30 most frequent adjectives in the TED corpus

By showing how particular words are used across the corpus, the corpus analysis confirmed and extended the Appraisal analysis. Firstly, it confirmed that the main topic of conversation in the comments is food, showing that users did not go off-topic; to the extent that they engaged with each other, they did so mainly through their views on food. The analysis of food also confirmed that many of the users who commented on these talks share the same widely-held views about good food (local, organic, fresh) and bad food (processed, junk), as well as a general tendency to align around positive comments, as indicated by their positive appreciation of contextually-relevant keywords such as idea and talk. The argument that users share (and see themselves as sharing) similar views is supported by the analysis of we and our, in that users overwhelmingly used these pronouns to refer to themselves and other users, and their common concerns. Finally, the frequency of intensifiers in the corpus suggests that graduation is a feature across the corpus; TED users share similar opinions and voice them enthusiastically.