Show Less
Open access

In the Beginning was the Image: The Omnipresence of Pictures

Time, Truth, Tradition


Edited By András Benedek and Ágnes Veszelszki

The authors outline the topic of visuality in the 21st century in a trans- and interdisciplinary theoretical frame from philosophy through communication theory, rhetoric and linguistics to pedagogy. As some scholars of visual communication state, there is a significant link between the downgrading of visual sense making and a dominantly linguistic view of cognition. According to the concept of linguistic turn, everything has its meaning because we attribute meaning to it through language. Our entire world is set in language, and language is the model of human activities. This volume questions the approach in the imagery debate.

Show Summary Details
Open access

#time, #truth, #tradition. An Image-text Relationship on Instagram: photo and hashtag (Ágnes Veszelszki)

← 138 | 139 →

Ágnes Veszelszki

#time, #truth, #tradition. An image-text relationship on Instagram: photo and hashtag

This chapter aims to examine the relationship between image (photo) and text (hashtag), on the photo sharing social network Instagram, through the examples #time, #truth, #tradition.

1.  What Is a Hashtag?

A hashtag is a type of label or metadata tag primarily used on social networking websites and microblogging services. It makes it easier for users to find content of the same topic. Hashtags are created by inserting the hash character (#)1 in front of a word or unspaced phrase.

The tagging system was initially popularized by Twitter2, and later taken over by Instagram, Facebook and other social networking websites as well. Now hashtags trend beyond social networking sites and digital communication into other media like television, print and they even appear in spoken conversations. The expression itself has not only become a verbalized form in orality but has also triggered changes in nonverbal communication: the hashtag sign can be shown by using the index and middle fingers from both hands laid over each other, rather than saying “hashtag” (Parker 2011; Kamer 2013). Because of its widespread use, the word hashtag was added to the Oxford English Dictionary in 2014.

Metadata (in a relationship with images) has the “ability to append linguistic signs to an image (or other data object), to facilitate its classification, archiving, retrieval and indicate provenance (authorship, ownership, conditions of use)” (Rubinstein–Sluis 2013). Searching on a hashtagged word on a social networking website will display all content so labelled on that website. With this sorting and searching function, however, hashtags not only connect different content or thematic blocs, but also connects users having similar fields of interests (“hashtag ← 139 | 140 → serves as […] a symbol of membership of a community,” Yang et al. 2012). It is often used as an instrument by social movements (for example, with the hashtag #bringbackourgirls Nigerian students directed attention to the fact that 270 girls were kidnapped from their school in 2014). Hashtags are important for social sciences too, as they make it possible to search and analyse opinions from different individuals about an event (Rambukkana 2015). Marketing professionals have also recognised the potential in hashtags (Ting et al. 2015: 16): they can use these labels to measure the effectiveness of promotion in social media.

In addition to its above-mentioned basic functions, hashtags are also used to abbreviate messages and to add some kind of a stylistic touch to what is being said. “Hashtags are deictic, indexical – yet what they point to is themselves, their own dual role in ongoing discourse” (Rambukkana 2015). The metacommentary written after the hash can be used to indicate the user’s attitude toward the hashtagged content (an apparently serious text can be made funny or ironic by adding the right label to it; cf. Parker 2011). In its broader sense, the hashtag may be taken as the marker of epistemological modality in relation to the text.3

Hashtags go unregistered and uncontrolled, so users can create as many tags as they want and use them the way they want to. Still, there are collections of advice on how to make hashtag campaigns. According to Cohen (2015), a good hashtag (that is suitable for marketing purposes) is unique, distinctive, easy to remember but not too general, short but meaningful, not too abstract and includes no slang elements. It is often advised that the initials of multiple word hashtags should be capitalised. Users should also make sure that there is only one way to spell out their hashtag (“if a hashtag can be misunderstood, it will be misunderstood,” e.g. #SusanAlbumParty vs. #SusAnalBumParty; Slegg 2014) and check whether it has been used earlier for other purposes (Hill 2012).

Certain tags most likely go hand in hand with others. Co-occurring, multiple hashtags are called in this paper hashtag chains (on the relationship of hashtag chains and phrasems cf. Veszelszki 2016). Analysing various contents shared on Instagram (Cohen 2015), the highest interaction is triggered by 11 or more hashtags, so on this platform the more hashtag someone uses, the more efficient he or she will be in building community. Hashtag chains typically relate to the most popular topics of amateur photography (travel, touristic sights; food and drink; fashion; selfie ‘self-picture’ and work), and their structure prototypically follows ← 140 | 141 → the user name, text, hashtags order: It’s a classic mille feuille but with a twist…the flavors of bananas! #recipeoftheday #syntaghthshmeras #millefeuille #napoleon #dessert #sweets #deliciousrecipes #recipes #food #foodlover #foodblogger #foodtricks #instafood #instagood #instadaily #instalover #instamood #foodpics #bananas. Or: Cica szelfi #selfi #selfie #szelfi #cat #macska. As the second example shows, hashtags are often made of spelling variants or synonyms of the same word. Using multiple hashtags in the middle of the text makes comprehension quite difficult (e.g. Very #cute #lovely #kitten #cat #kitty! I so #love it!).

Based on empirical research, hashtags are classified into three main groups (Csire 2015; with my own observations and examples):

  1. Like-hunter tags are added to images to boost the number of views (e.g.: #followme, #likeforlike; #instadaily, #instalove; #photooftheday). They do not form an integral part of the image, their only aim is to obtain new likes and followers and that way they lead to a certain microcelebrity which “is a mind-set and a collection of self-presentation practices endemic in social media, in which users strategically formulate a profile, reach out to followers, and reveal personal information to increase attention and thus improve their online status” (Marwick 2015: 138, according to Senft 2013).
  2. Thematising tags usually relate to the topic of the photo, how it was made (black-and-white, no-filter [#nofilter]), its atmosphere, or sometimes the persons on the picture (#girls, #friends, etc.), objects (#food, #cake) or events (#wedding, #conference), so this way they contribute to the searchability of similar contents.
  3. Contextualising tags complete posts with the user’s emotions, mood or the circumstance of taking the photo, such as: #cominghometothis, #annapostedit, #grüsseausdertaiga. Such hashtags are much more characterised by uniqueness: they do not want to enhance searchability but to add extra verbal information to the visual content. For this reason, they are one-time, non-recurrent texts.

As one of the main functions of hashtags is to facilitate searching, their presence is relatively constant on the internet. This stability4, however, can only be true ← 141 | 142 → to like-hunter and thematising hashtags, because contextualising hashtags are unique and casual.

2.  Corpus and Method

The present study aims to describe the hashtag as a new, image-bound, minimalistic type of text. For the empirical analysis, five hypotheses (presented at the results) were formulated which combine the potentials of qualitative and quantitative analyses.

The examination was carried out on the Instagram photo sharing site. Highfield and Leaver outlined emerging methods to study uses and activity on the image-sharing app and social media platform Instagram. They found that “the importance of tagging on Instagram […] has conceptual and practical links to the hashtags employed on Twitter (and other social media and ‘Web 2.0’ platforms), with tags serving as markers for the main subjects, ideas, events, locations, or emotions featured in tweets and images alike” (Highfield–Leaver 2015). “So far, textual content has dominated social media research, particularly for large-scale analyses; this is in part due to the ease of collecting and processing text in comparison to images. […] Preliminary research into Instagram has offered large-scale analyses of images from specific locations” (Highfield–Leaver 2015, mentioning Hochman–Manovich 2013). With the hashtag function the images of Instagram also include “fragments of metadata” (Wendt 2014: 37).

As a start, I selected three keywords that matched the theme of the VL-conference (#time, #truth, #tradition) and used them as hashtags to search related content. The results were sorted using two methods: first, by choosing photos with the highest number of likes (indicating popularity) and second, by selecting the most recent Instagram photos on the day of examination (25 October 2015). For each keyword 100 photos were saved together with labels. This corpus was later completed with a sample containing another 100 items selected from the latest photos, serving as control corpus, to balance (at least to a certain degree) the bias caused by the keywords. In line with Instagram’s privacy policy, only publicly posted images were examined. This 400-item corpus was analysed according to the aspects specified in the hypotheses. The analysis was based on a sample (what I refer to – according to Rocheleau–Millette [2015] – as ‘small data’), so its results may not necessarily correspond to the results produced by automatic analyses of big data. Using a different sampling method can produce different results. After presenting the data, I used a mixed methodology, taking advantage of both ← 142 | 143 → qualitative and quantitative data collection and analysis methods. This method is accepted as valid by Highfield and Leaver, who deal with Instagram research methodology. They found that “Studying data through qualitative methods is critical to understanding how social media are used in non-standard ways, identifying practices that might easily be missed through automated analyses. […] For visual platforms such as Instagram, this is even more critical, for the textual and graphic components of a post each offer key information, and analysis needs to take into account both aspects” (Highfield–Leaver 2015).

My aim was not only to describe the image-text relationships (the semantic and semiotic connection points – or with other words: the immediate context – of these keywords) on Instagram, but also to complete the time-truth-tradition theme of the conference with a new semiotic aspect, namely with what visual and verbal signs connect to these keywords – on one of the most popular photo sharing sites of our time.

3.  Results

Searching on the hashtag #time on the date of examination resulted in 24 million posts. Not surprisingly, many of the pictures depicted a clock or watch, so this sub-corpus contained a much higher share of advertisements than the search results of the other keywords. The hashtag #truth was used in more than 29 million posts, dominated by religious content (Buddhist texts, Bible and Quran quotes), wisecracks, aphorisms, sophisms and inspirational messages. As a characteristic feature, these texts were not attached to a picture but a picture was organised around them. The keyword #tradition resulted in 2.2 million Instagram pictures, primarily depicting traditional, often folk costume, seasonal preparations and the architecture or culture of exotic countries.

After removing repetitions, altogether 2784 different hashtags were added to the corpus out of the total 4966. On the list by frequency of occurrence, the three keywords are followed by the rather general #love, #life, #happy #instagood, #family, #beautiful, #followme, #photooftheday and #smile tags. The gesture of showing something is not only represented by the pictures but is also reflected in the hashtags. The significance of publicly sharing something is primarily emphasized by possessive structures (e.g.: #myart, #mybag, #mybirthday, #myboy, #mylife, #mysore; #ourdisneyadventure) or by the Hungarian word muti… [‘show me’] (#mutimiteszel [‘show me what you eat’]). ← 143 | 144 →

As regards the hypotheses, I found the following:

3.1  Hypothesis 1: Instagram users will generally tag their photos with eleven hashtags, where a hashtag chain is expected to contain synonyms and spelling variants of the same forms.

The 400 pictures contained altogether 4966 hashtags, that is an average of 12.41 hashtags per picture. This figure is a little bit higher than the expected eleven but it supports Cohen’s (2015) views who recommended at least 11 hashtags per picture to trigger the most user interactions.

Hashtag chains usually contain multiple words having the same semantic field (e.g.: #time, #hours, #sundayfunday, #timeless, #timepass, #timepieces; #teatime). Various morphological forms of the same word may appear next to each other: derived words (e.g.: #health, #healthy; #cook, #cooking); plural and singular forms (#fitgirl, #fitnessgirl, #fitnessgirls); clipping (#sis, #sisters; #chillin, #chilling). The elements of a collocation may also appear together (e.g.: #happy #time; #life #time). Sometimes the same word appears in different languages in a hashtag chain (e.g.: #time #zeit #idő #tiempo). Mistyped forms do not count as intentional word forms (e.g.: #autmn, #autum instead of autumn).

3.2  Hypothesis 2: The major part of hashtags will be in English, but at least half of the examined photos will contain hashtags in two languages (particularly in the case of users for whom English is a foreign language).

The language of the hashtags was English in almost all of the cases (there were some rare instances in the corpus of Russian, Chinese, Korean, Italian, Arabic, German and Hungarian tags but almost always next to English counterparts). Naturally, the fact that the three keywords were in English strengthened this trend. The greater presence of German tags was partly due to the fact that the English tradition and the German Tradition are formal equivalents. In most of the cases, however, hashtags were given in English even if the text accompanying the picture is in another language (e.g.: Возможно, вы не знаете, но я желаю всем доброй ночи. Каждую ночь!!! #jasmin #jasminshor #smile #love #night #sweet #dream #my #time).

As the sampling was done by a Hungarian linguist, so after English, the second most frequently used language was Hungarian, followed by Spanish, German and Russian. Languages with less than ten occurrences, single-occurrence and therefore unclassifiable forms (#bezel), nouns (#Berlin), numbers (#1956), interjections ← 144 | 145 → (#braaaap) and vague, uncommon abbreviations (#bzh) were all classified into the Other category. Nearly 90% of all hashtags were in English (if only different tags are considered, and repetitions are deleted, this rate is 82%), so non-English tags accounted for only 10% (571 items). This partly supports my hypothesis, as English hashtags indeed dominate the international corpus, but not every second picture is accompanied by a non-English hashtag.

3.3  Hypothesis 3: The most characteristic pattern of photo subtitles will be a connection of text and independent hashtags.

The most typical picture-text pattern is the following: the picture is followed by non-linked text which can be taken as a subtitle and then – as further textual content – the hashtags (Table 1). Subtitles can be a single word, a sentence, a poem, a quote, and sometimes, though relatively rarely, even a very long text. If the post is shared on Instagram, subtitles sometimes appear as hashtags. This may take the specific form that instead of a series of independent hashtags all syntactic units or all words of a phrase or sentence are marked with a # sign. Sometimes more important elements and keywords are emphasized as hashtags within the subtitle text (and this may be further combined with @ signs marking persons).

Table 1: Photo subtitling patterns on Instagram

no text, only image
only hashtags #awesome #time #sunday #bff #love #music #followback #Delhi #sdamarket
only non-linked text[No occurrence in the corpus, as only hashtagged pictures were collected. Still, there may be one example: Christmas]
non-linked text + independent hashtags (the most typical form)Found this old photo from over 25 years ago. Scary how quickly time passes! #time #passingtime #teenager #teenageyears
hashtags within a text (keywords are emphasised by hashtagging)Our latest #mashreads book selection is #FatesandFuries by Lauren Groff, a love story told from two very disparate #perspectives. Based off of the themes of the #book, for this week’s MashReads social challenge, we want to see examples of juxtapositions you find in the wild such as light and dark or two contrasting textures. Be sure to tag your photos with #mashreads to enter. #NYC #PhotoChallenge (Image: @tylertronson) ← 145 | 146 →
coherent text with hashtags + list of extra hashtags #cloudporn with #lighthouse at old #port of #Chania #Crete #Creta #Greece #travelgram #instatravel
hashtagged words in expressions, collocations #sweet #dream #my #time #crazy #time
each word in a text is hashtagged#life #is #like #a #car #if #you #keep #going #in #reverse #you #will #not #get #to #your #destination #but #if #you #speed #up #and #do #not #take #the #time #to #enjoy #your #surroundings #you #might #get #lost

3.4  Hypothesis 4: The vast majority (90%) of the hashtags will fall into the thematising or like-hunter categories, while unique hashtags will represent only a small minority (10%). A large number of hashtags is expected to refer to the manner the photo was created.

Analysing the corpus, the originally planned three categories (thematising, like-hunter, unique hashtags) had to be completed with another three: hashtags indicating group identity, hashtags describing how the picture was created, and hashtags that did not fit into any category and were meaningless for outsiders without a context, though not necessarily containing unique words (e.g.: #N16, #is, #and, #get).

Thematising context-marker hashtags dominate the sample (83%). Similarly to posted pictures, these were characterised by seasonality: keywords like autumn, fall, halloween, pumpkin appeared in such high numbers as the sample was taken in late October.

A user’s popularity is measured with the number of its followers and thus the number of user interactions it can achieve. In the case of the Instagram profile of companies, products and brands, the large number of followers enhances the efficiency of advertising and helps reaching the highest possible number of potential customers. For this reason, there were many (55 different types, 230 with repetitions) hashtags which encouraged following in several forms (e.g.: #followme, #follow, #follow4follow, #followback, #followmeplease). Similarly, hashtags encouraging users to like or share the post (#like, #likesplease, #likeitplease, #likeme, #likemyphoto; #pleaseshare) also aim to popularise the posted information reaching as many users as possible, and are often based on mutuality: like is offered for a like or a follower (#like4like, #likesforlikes, #likeforfollow). Hashtags reflecting on the ← 146 | 147 → phenomenon of hashtagging can be taken as meta-hashtags.5 Hashtags referring to joining to a community also serve the spreading of the content (e.g.: #poetrycommunity, #PoetsofInstagram; #instaitalia).

Usually, unique hashtags are not only made of set phrases or collocations (#goodfood, #goodgirl, #goodlife, #goodvibes) but they are also created by writing a longer syntactic unit, even a full sentence, as one word. The sample comprised 154 unique hashtags (e.g.: #gottadowhatyougottado, #StillAKidAtHeart, #tellmewhatyoudask, #whereisthebagelemoji). Although they are not totally unique, imperatives and good advices constitute a separate group (e.g.: #thinkaboutit, #riseup, #neverstopexploring, #livethemoment).

Meta-information about creating the pictures was expected to be much more dominant in the corpus: only 50 types (with 102 occurrences) were found in this category (e.g.: #vscocam, #photoshoot, #nofilter, #blackandwhite, #canon; cf. Schrey 2015).

3.5  Hypothesis 5: The examined pictures are expected to be mostly about people, followed by objects and animals in order. As a sub-hypothesis, selfies are expected to represent 20% of the pictures depicting people (cf. Wendt 2014).

The biggest thematic group contained indeed photos of people (138 out of 400). Of this, 43 evidently or admittedly (according to the hashtags) qualifies as selfie. This rate, however, surpassed my expectations: 31% of the pictures taken of people are self-portraits. Surprisingly though, only 11 of the 43 selfies reflected explicitly on the act of taking a selfie with the hashtag #selfie.

The second largest group of pictures depicted objects (furniture, books, etc). Probably, the keyword #truth resulted in the high number of posts displaying texts (memes, quotes) as images. Further categories deal with food and beverage (often hashtagged as #foodie by analogy with selfie), nature, built environment and animals. Surprisingly, the traditional composition of a person posing with a famous touristic site in the background was quite rare in the sample (note that this would have been one of the most dominant composition if the keyword #travel or #travelgram had been used). ← 147 | 148 →

4.  Conclusion

I used three keywords matching the theme of the VLL conference to describe the relationship between pictures and hashtags, these new-type minimalistic texts which are similar to subtitles but have many different functions compared to them. Quantitative and qualitative analysis of small data was carried out to test the validity of five hypotheses.

The first hypothesis, which expected to find many spelling variants and synonyms among the hashtag chains consisting of an average of eleven hashtags, was mainly confirmed: the pictures in the corpus were accompanied by an average of 12 hashtags, which however, indeed included synonymous words and morphological variants. The second hypothesis, which expected an overwhelming dominance of English hashtags, was clearly justified with a 90% rate. The third hypothesis relied on previous observations when expecting that Instagram subtitles mostly follow the pattern of “unlinked text + independent hashtags”. The fourth hypothesis was based on the results of former hashtag analyses. According to my results the three categories (thematising, like-hunter, unique hashtags) should be completed with at least three further categories (hashtags marking group identity, hashtags indicating how the picture was created and unclassifiable hashtags). The highest number of items belonged to the category of contextualising hashtags which verbally represent the content of the photo, while technical meta-information relating to the taking of the photo was not so characteristic than it seemed to be prior to the analysis. The analysis of the fifth hypothesis applied a semiotic approach, focusing on the connection between the content of the photos and the accompanying hashtags. More than one third of the pictures depicted people, and of this, one third qualified as selfie though seldom indicating this fact with the #selfie hashtag.

As it has been emphasized earlier, my results on frequency of occurrence are valid only to this corpus and were meant to complete the theme of the conference. However, hashtag types, subtitle patterns and the very strong co-occurrence of words having the same semantic field in a hashtag chain are results that can be generalised. Further examinations may deal with the co-occurrence of certain elements in hashtag chains from a semantic aspect and the grammatical analysis of neologisms and word formation methods in Instagram-related texts. ← 148 | 149 →


Cohen, David (2015): SurePayroll: Infographic: Hashing Out Hashtags. 23. 04. 2015.

Csire, Tímea (2015): #selfie. A hashtagek használata az Instagramon. E-nyelv Magazin 2015/2.

Highfield, Tim – Leaver, Tama (2015): A Methodology for Mapping Instagram Hashtags. First Monday 20/1.

Hill, K. (2012): #McDStories: When A Hashtag Becomes A Bashtag. Forbes. 24. 01. 2012. [not available].

Hochman, Nadav – Manovich, Lev (2013): Zooming into an Instagram city: Reading the Local Through Social Media. First Monday 18/7.

Kamer, Nimrod (2013): Brace Yourselves for the Proliferation of the ‘Finger Hashtag’. Wired. 06. 03. 2013.

Kim, Erin (2012): Twitter Unveils ‘Cashtags’ to Track Stock Symbols. CNNMoneyTech. 31. 07. 2012.

Marwick, Alice E. (2015): Instafame: Luxury Selfies in the Attention Economy. Public Culture 27/1: 137–160.

Parker, Ashley (2011): Twitter’s Secret Handshake. New York Times. 10. 06. 2011.

Rambukkana, Nathan (ed., 2015): Hashtag Publics. The Power and Politics of Discursive Networks. Frankfurt et al.: Peter Lang.

Rocheleau, Sylvain – Millette, Mélanie (2015): Meta-Hashtag and Tag Co-occurrence: From Organization to Politics in the French Canadian Twittersphere. In: Rambukkana, Nathan (ed.): Hashtag Publics. The Power and Politics of Discursive Networks. Frankfurt et al.: Peter Lang.

Rubinstein, Daniel – Sluis, Katrina (2013): Notes on the Margins of Metadata. Photographies June 2013: 151.

Schrey, Dominik (2015): Retrofotografie: Die Wiederverzauberung der digitalen Welt. MEDIENwissenschaft 01/2015: 9–26. ← 149 | 150 →

Senft, Theresa (2013): Microcelebrity and the Branded Self. In: Hartley, John – Burgess, Jean – Bruns, Axel (eds.): A Companion to New Media Dynamics. Malden, MA: Wiley. 346–354.

Slegg, Jennifer (2014): What Marketers Can Learn From These Hashtags Fails. The SEM Post. 01. 08. 2014.

Thode Hougaard, Tina (2014): #hashtags #metakommunikerende #nøgleord. In: Schoonderbeek Hansen, Inger – Thode Hougaard, Tina (eds.): 15. Møde om Udforskningen af Dansk Sprog. Århus. 149–170.

Ting, Hiram – Winnie Wong Poh Ming – Run, Ernest Cyril de – Lau Yin Choo, Sally (2015): Beliefs about the Use of Instagram: An Exploratory Study. International Journal of Business and Innovation 2/2: 15–31.

Veszelszki, Ágnes (2016): A hashtag mint új frazeologizmus? [Are hashtags new phrasemes?]. In: Bárdosi, Vilmos (ed.): Frazeológia – az emberi világkép tükrözője [Phraseology]. Budapest: Tinta. 147–158.

Wendt, Brooke 2014: The Allure of the Selfie: Instagram and the New Self-Portrait. Network Notebook #08. Amsterdam.

Yang, Lei – Sun, Tao – Zhang, Ming – Mei, Qiaozhu (2012): We Know What @You #Tag: Does the Dual Role Affect Hashtag Adoption? University of Michigan.

W1 = #OriginStory. Carnegie Mellon University. 2014.

1 In the English language, the # symbol is the number sign or the hash as Northern Americans call it (not to be confused with the sharp sign # in music, the abbreviation No. or the pound sign in North American English).

2 Its first appearance on the microblogging site Twitter is commonly associated with Chris Messina (W1), who used this sign in a message in 2007 the following way: “how do you feel about using # (pound) for groups. As in #barcamp [msg]?” [23 Aug 2007].

3 For example: “I can’t decide which series to watch this evening. #firstworldproblems”. The expression “first world problem” is a self-ironic reflection on the fact that for the user the mentioned issue is the greatest problem in life.

4 Examples of this fixed form include: #mik (acronym of Hungarian Instagram Community), #yolo (you only live once), #swag, #sundaymorning; #bestoftheday, #picoftheday, #worstnightmare, #grexit. The same need for stability may motivate the usage of joker suffixes, such as -porn and -gram (deriving from the word Instagram): coffee porn, foodporn, wordporn, lobster porn; latergram, travelgram, foodstagram; as well as prefixes, such as insta- (also deriving from Instagram): instafood, instatravel, instagood, instamood, instadaily, instasize, instahun, instanight (on further linguistic characteristics of hashtags: Thode Hougaard 2014).

5 An example: #lookatthecutekittenontherightalsoextralonghashtagsrule.