Conceptual Blends Across Image Macro Genres

Abstract This paper analyzes image macros via the theoretical framework of conceptual blending on the example of online responses to 2011 and 2012 doomsday predictions. To date, conceptual blending has shown promise in the analysis of a variety of internet memes; however, most studies have been limited to one type of meme or one blending model. This project adds to the discussion in two ways. First, it presents a thematic analysis – image macros responding to end of the world predictions – thereby covering a broader variety of image macro genres. Second, it modifies blending models according to the needs of the different genres. The discussion explores the extent to which the three different image macro genres require different blending models and the implication for future study.


Introduction
For nearly twenty years, the theory of conceptual blending has been applied to analyze a variety of verbal and visual phenomena. During this time, Fauconnier and Turner's (2002) original four-space model has been modified and adapted to better explain the pragmatic meaning of the emergent blend. This study continues this line of research into applications of the theory of conceptual blending to analyze the relatively recent phenomenon of online image macros. While this is a growing area of research among scholars of language and culture, most studies are either focused on cultural analysis or on the creative verbal-visual mechanisms behind one meme genre.
The current paper is unique in that it is constrained by a thematic study of Polish responses to end-of-the-world predictions in 2011 and 2012, rather than a type of image macro. As a result, it looks at three image macros genres. Moreover, rather than forcing the image macro to the blend,

Shala Barczewska
it modifies the blend according to the need of the genre. As a result, it is able to comment on the applicability of different blending models to each.
The paper begins with a presentation of meme theory, its use in the Internet world and defends the study of Internet memes as language in use. This is followed by a review of conceptual blending and some of the criticisms and adaptations to the theory to date. The third section presents the research questions and describes the methodology employed. The fourth section analyzes the three types of image macros used in this study -Advice Animals, character-based templates, and demotivators -according to the information required for a successful blend. The final section summarizes the results and provides suggestions for future research.

Internet memes and image macros
The term meme was coined by Richard Dawkins in his 1976 book, The Selfish Gene. Since then the term has itself become a meme (cf. Dynel, 2016) and has developed polysemic meanings, including popular use to describe certain aspects of online behavior. Before diving into the use of meme to describe online behavior, it is important to first look at Dawkins's original definition: a noun that conveys the idea of a unit of cultural transmission, or a unit of imitation. [...] Examples of memes are tunes, ideas, catch-phrases, clothes fashions, ways of making pots or of building arches. [...] memes propagate themselves [...] by leaping from brain to brain via a process which, in the broad sense, can be called imitation. [...] it is responsible for the origins of language, art, religion, science, and other singular human feats, and [...] is as indispensable for basic everyday thought as it is for artistic and scientific abilities (Dawkins, [1976(Dawkins, [ ] 2006).
In other words, a meme is any aspect of culture that is passed from person to person or generation to generation. The feasibility of meme theory has been debated within the physical sciences and humanities and here is not the place to argue for its (in)accuracy. Rather, we follow the current trend in internet meme studies by "us[ing] the meme as a prism for understanding certain aspects of contemporary culture without embracing the whole set of implications and meanings ascribed to it over the years" (Shifman, 2013, pp. 362-363).
Borrowing the term, while leaving the controversial baggage aside, is not only common practice in the field of Internet meme studies, but is also necessary because, as has been noted by Shifman (2014) and Wiggins (2019), among others, there are significant differences between Dawkins's use of the term and the online phenomenon discussed in this paper. These become apparent when we compare Dawkins's explanation of meme above with the following definitions of Internet memes: -a "piece of culture, typically a joke, which gains influence through online" transmission (Davidson, 2012, p. 122); -content that is shared and has changed or evolved while being passed on to others (cf. Börzsei, 2013; Know Your Meme); -"units of popular culture that are circulated, imitated, and transformed by individual Internet users, creating a shared cultural experience in the process" (Shifman, 2013, p. 367); -"any artifact (a film, spoof, rumor, picture, song, etc.) that appears on the Internet and produces countless derivatives by being imitated, remixed, and rapidly diffused by countless participants in technologically mediated communication" (Dynel, 2016, p. 662). Several differences stand out. First, Internet memes are physical objects, whereas Dawkins's memes are abstract ideas. Second, as Wiggens (2019, p. 3) observes, Internet memes are "a genre of communication, not a unit of cultural transmission" (italics in the original). Moreover, Wiggens (2019, p. 8) points to the "discursive aspect of memes" and the "complex and multifaced ways in which content is created and spread, etc. online," which he claims cannot be accounted for in Dawkins's concept based on transfer via imitation.
In terms of scholarly interest, Internet memes have been identified as a valid source of linguistic and cultural data. While Börzsei argues that "the Internet meme is also a [...] genre" as "Internet memes have become a part of vocabulary for Internet users" (Börzsei, 2013, p. 2, 24). Shifman (2014, p. 118) identifies a variety of genres of online memes: (1) Genres that are based on the documentation of "real-life" moments (photo fads, flash mobs). These genres are always anchored in a concrete and nondigital space.
(2) Genres that are based on explicit manipulation of visual or audiovisual mass-mediated content (reaction Photoshops, lipdubs, misheard lyrics, recut trailers). These genres-which may be grouped as "remix" memesoften reappropriate news and popular culture items. Such transformative works reveal multifaceted attitudes of enchantment and criticism toward contemporary pop-culture.
(3) Genres that evolved around a new universe of digital and meme-oriented content (LOLCats, rage comics, and stock character macros).

Shala Barczewska
This study focuses on three image macro genres. According to Know Your Meme, image macros are "captioned images that typically consist of a picture and a witty message or a catchphrase", which, apart from reaction Photoshops, belong to the third category described above. Know Your Meme also includes demotivational posters and edited pictures, both of which will be discussed in this paper.
In her analysis of meme composition and imitations, Shifman (2013, pp. 367-368) suggests "isolat[ing] three dimensions of cultural items that people can potentially imitate: content, form, and stance". While it is not possible to completely separate these elements, recognizing their centrality to meaning creation and propagation will be helpful in the analysis that follows.

Conceptual blending theory
The theory of conceptual blending was first proposed by Fauconnier and Turner (2002). It explains the cognitive processes they believe to be behind the complex nature of human thought and creativity. In its simplest form, conceptual blending can be diagrammed in a four-space model similar to Figure 1. 1 The generic space is said to provide the shared structure of the two input spaces. The emergent blend combines elements of the two input spaces; however, the result is not simply a compilation, but a new creation, which itself may then be an input space for future blends.

Conceptual blends across image macro genres
According to Fauconnier and Turner (2002, p. vi), this process "is responsible for the origins of language, art, religion, science, and other singular human feats, and [...] is as indispensable for basic everyday thought as it is for artistic and scientific abilities". Indeed, their book and numerous publications demonstrate how the theory can be applied, among other things, to art, math, poetry, metaphor, clocks, and grammar. However, as has been discussed in Barczewska (2017, p. 427), there have been several criticisms of the original diagram proposed by Fauconnier and Turner. These include: -the absence of a "cultural-historical component" often inhibits understanding of the "true causes" and "real conceptual structure" of the items under investigation (Glebkin, 2013(Glebkin, , p. 2408 2 ; -lacking clarity as to how the receiver is able to understand "the intended meaning of that utterance", that is the implicatures (Pérez Hernández, 2002, p. 185). In order to address these different issues, adaptations to the blend have been proposed, several of which are discussed in Barczewska (2017). Here we will mention three that are relevant to our analysis.
The first was proposed by Brandt and Brandt (2005). They suggest removing the generic space on the grounds that they view it to be superfluous. Then they give specific roles to the two input spaces: a presentation space and a reference space. In the case of the utterance that surgeon was a butcher, the example analyzed by Brandt and Brandt in their paper, the generic notion of a butcher would be in the presentation space and the specific surgeon in question would be in the reference space. These two spaces are projected (or blended) to create a virtual space in which the "surgeon is seen as a butcher" (Brandt & Brandt, 2005, p. 229). A semiotic space is also added which would include all the relevant information about the given discourse event: the expressive act, the context of a recent surgery and the pheno world -or general knowledge -that would include butchers and surgeons. This semiotic space provides both situational relevance and argumentation relevance for the given event; here the latter would be a force-dynamic image schema of an agent harming a patient. It is only when the relevance space(s) and the virtual space undergo the blending process that the meaning space emerges. Coulson and Oakley (2005) simplify Brandt and Brandt's model. They retain the labels given to the two main input spaces: presentation space and reference space. However, they claim that a grounding box that includes the participants, forum, and circumstances is sufficient to contain the additional information needed to build the blend and grasp its implicatures. Coulson andOakley (2005, p. 1514) apply their diagram to the expression In France, the Lewinsky affair wouldn't have hurt Clinton. They argue that the blend itself does not tell us whether the lack of concern is a good or bad thing. However, once the statement is grounded both in a particular discourse and a particular type of newspaper or magazine, it is possible to understand whether the authors are suggesting that (a) a French Clinton would be free to focus on fighting global terrorism or (b) the French lack moral clarity. When it comes to analyzing metaphorical statements, two blended spaces are employed. The first provides a hyper-literal reading of the metaphor, the second interprets this hyper-literal reading according to the information in the grounding box, which, in this example, would include the participants and the setting. While it is admitted that contextual grounding plays a key role in metaphor interpretation, recent research has cast doubt onto the assumption that metaphors are processed literally before they are understood metaphorically.

Shala Barczewska
A third adaptation gives a special place to phraseology. Omazić (2005, pp. 47-50) begins with the utterance Bagdad is Bush's blue dress, which she unpacks using Fauconnier and Turner's four-space model. She neither removes the generic space nor specifies its contents. With her next example, This article describes the read 'Blue Dress' in Bush's closet, Omazić argues that a third input space should be added containing the idiom skeleton in someone's closet, which is implicitly referenced. She explains: The emergent structure in the blended space inherits the idiom structure from input three, and lexical projections from input spaces one and two: blue dress is projected from input space one, Bush from input space two. The blend inherits the structure and meaning from input three, and different meaning elements (knowledge of world affairs and culture) from inputs one and two (Omazić, 2005, p. 50). In this description, Omazić assigns to the input spaces information that Brandt and Brandt (2005) place in their semiotic space and Coulson and Oakley (2005) place in their grounding box. As the study that follows will demonstrate, adaptations from all three of these models are helpful to unpack both implicit and explicit meaning of image macros. Moreover, it will be important to recall Shifman's (2013, pp. 367-368) observation that, for the purpose of study, memes can be divided into content, form, and stance. Hence, it should not be surprising to find these three elements represented in the different mental spaces projected into the blend. Moreover, Shifman has observed, each genre will emphasize these different elements of imitation to different degrees.

Method of research
As mentioned in the introduction, the source material in this study came from a collection of image macros addressing two different end-of-theworld predictions: the end of the Maya calendar on Dec 21, 2012, and Harold Camping's end of the world predictions for May and October in 2011. This database was selected and has been analyzed as part of a research project into Polish cultural values and responses to the 2011/2012 apocalyptic visions (Barczewska & Barańska, 2019). During this study, a variety of blending mechanisms were observed that required further study. Previous studies on Internet memes tend to focus on one type of image macro rather than comparing various genres across a topic. By controlling for the topic, but not the type of image macro, it is possible to discuss how different types of memes rely on different patterns of conceptual blending.
The research questions for the current study are as follows: -What elements must be included if we are to analyze an image macro as a blend? -Does this differ depending on the type of image macro? (Advice Animals, character-based image macros, demotivators) The corpus of Internet memes was collected by searching Google Images for the terms koniec świata memy ['end of the world memes'] for Dec 1-Dec 31, 2012 and May, Oct 2011 Google. They were selected for representativeness and appropriateness. 3 The previous study identified the presence of foreign (i.e., non-Polish) templates, but few direct translations of the lexical content. The adaptations often included elements specific to Polish culture and require in-group knowledge to fully understand the intended message.
To answer these questions, selected Advice Animal image macros, character-based image macros, and demotivators from the corpus are analyzed according to the theory of conceptual blending. In doing this, we take into consideration the characteristics of different meme genres presented in Section 2 as well as the variations and adaptations to conceptual blending discussed in Section 3.

Results and analysis
The analyses that follow apply the theory of conceptual blending to three different types of image macros identified in the end-of-the-world corpus: Advice Animals (Section 5.1), character-based image-macros (Sec-tion 5.2), and demotivators (Section 5.3). Moreover, it will demonstrate, by means of the input spaces involved, that character-based memes differ from Advice Animals -although the line between them may be blurry at times. The aim in these sections is to present the various image macros as a complex blend or a series of complex blends. The discussion at the end of this section asks whether some of this data represents shared knowledge that would be better placed in a semiotic space or a grounding box (cf. Brandt & Brandt, 2005;Coulson & Oakley, 2005).

Advice Animal image macros
The first type of image that we will analyze is commonly referred to as Advice Animals; however, this type of image macro has expanded to include a number of other images -both human and nonhuman. Moreover, while the traditional Advice Animal template involves a cutout picture (usually just a face) on a multi-colored pinwheel, not all are so designed. Some retain the central image of an animal face but change the background. For instance, in the case of Stoner Dog, the background image usually involves marijuana leaves and smoke around the dog's face. Others, such as Evil Toddler, a photograph of a young boy on a green field laughing in a sinister way, are pictures that have come to represent a given speech act. All Advice Animal templates involve distinct elements which could be analyzed as comprising at least three input spaces: (a) The image macro visual -its original context, relevant cultural information e.g. Capt. Picard, Willy Wonka, Conspiracy Keanu, Socially Awkward Penguin, Bad Luck Brian, Scumbag Steve; (b) The speech act (force and function) of that image macro series e.g. expressing surprise at the 'stupidity' of others' behavior/decisions as in the case of Capt. Picard's facepalm; (c) Textual expectations -often but not always involving both a specific syntactic structure and a specific type of commentary. Moreover, elements (a) and (b) can be viewed as ad-hoc categorization in which the image represents a type of, for example, a highly unlucky person or an incredibly selfish teenager. This template is then adapted and modified by a variety of internet users. The result is a new blend comprising the template, the text, and contemporary cultural concerns.
These input spaces loosely correspond to Shifman's (2013, pp. 367-368) three dimensions of culture that people can imitate in editing and sharing memes: form, content, and stance. In the first blend, (a) can be understood as corresponding to form, (b) to stance, and (c) to content.
Textual information added by the creator of a specific image macro would also add to the context and review the creator's stance on the issue at hand. Any and all of these can be -and likely are -the results of blends themselves and it is for this reason that the term loosely is used, as they are not exact matches, and even Shifman herself does not draw a clear delineation between these dimensions.
According to Dynel (2016), Advice Animals are best understood as a cycle of visual-verbal jokes based on two genres of canned jokes: the one-liner and the riddle. However, the role of the picture in interpreting the meaning or humor of the image macro varies according to the text. Irrespective of the centrality of the image in conveying meaning, she demonstrates that Advice Animals "operate in accordance with the incongruity-resolution framework, being interpreted sequentially but materializing nonlinearly across verbal and visual modes of presentation in terms of the setup and punchline" (Dynel, 2016, p. 684).

Figure 2. A conceptual blending diagram of the Conspiracy Keanu template
We would like to propose that this is successful, in part, because Advice Animals or template-based memes begin with an image and a corresponding speech act. For example, one such template has a picture of Keanu Reeves from the movie Bill and Ted's Excellent Adventure. In the picture, Reaves looks a bit perplexed; as a result, the meme has been named "Conspiracy Keanu" and image macros developed using this template have Keanu questioning some aspect of daily life as if it were a conspiracy. These questions usually begin with What if.... Unpacking these memes requires a blend similar to the one sketched in Figure 2. Here, to facilitate visualization, the emergent space -the image macro template -is placed in the center. The three input spaces are the visual (e.g. a picture of Captain Picard or So-cially Awkward Penguin) which represents a member of a category through a metonymic link. One of the salient emotions from this category is selected to comprise the second input space.
One of the memes in this corpus (Figure 3)   Another example is the Philosoraptor: a picture of a green dinosaur head on the pinwheel background typical of traditional Advice Animals. The dinosaur appears to be scratching his chin and usually asks If...then questions focused on an aspect of a contemporary issue and thereby revealing an obvious discrepancy.
Following the conceptual blending diagrams suggested in Figure 2, the Philosoraptor Advice animal could be diagramed as presented in Figure 4. In this corpus, the Philosoraptor meme was used to question the wisdom of assuming that the end of the Maya calendar signaled the end of the world by referencing other calendars that ended at different time periods, e.g., the calendars on the Nokia phone or the Windows computer operating system. Two Philosoraptor image macros from the corpus are presented in Figure 5 and Figure 6. When it comes to unpacking a specific Advice Animal, the template acts in a way similar to the idiom skeleton in someone's closet in Omazić's (2005) analysis. This frame or template, including its pragmatic meaning and illocutionary force, are blending along with the new linguistic content and contemporary cultural information to convey a particular message or joke. Figure 7 illustrates what this might look like for the Philosoraptor image macros. In diagraming the processes necessary for meaning creation, it is worth noting that an input space providing an accepted and culturally relevant scale also must be added. The result in the emergent blend not only helps the Internet user make sense of the meme but also provides the meme creator's stance on a given issue. In the case of the Philosoraptor, the meme creator seems to be suggesting that it is foolish to equate the end of a calendar with the end of the world.

Character-based image macros
Character-based image macros differ from Advice Animals in that, although they may have a well-known person at the center, neither the grammatical structure of the text that accompanies the picture nor its illocutionary force is predetermined. For example, Chuck Norris is a common persona for template-based image macros. There is no set text, and it is his persona, or imagined persona, rather than a predetermined syntactical structure that facilitates the expectations regarding the meme's meaning. To continue with this example, Chuck Norris memes usually highlight his well-known martial arts prowess, which has been exaggerated and developed by the Internet community via a list of "Chuck Norris facts" such as When Chuck Norris does push-ups, he doesn't push his body up, he pushes the earth down. These facts, which began as part of an Internet discussion board, have also appeared on T.V. and now comprise their own webpage (chucknorrisfacts.net). 8 Memes with Chuck Norris found in this corpus depict him either canceling or causing the end of the world (see Figure 8).
A schematic drawing of the conceptual blend involved in creating and understanding Chuck Norris memes is presented in Figure 9 below. The levels of blending for character-based memes appear to be much simpler than those employed for Advice Animals. Perhaps this is because the textual expectations and jokes are located within the Internet version of the celebrity rather than the grammatical or phraseological structure of the text that accompanies them. In the case of Chuck Norris, he is charged with both causing and canceling the end of the world. Other celebrities/politicians given this privilege include Putin, MacGyver, and Radom's "Chytra baba z Radomia" ('sly/greedy woman from Radom'). 10 The Polish MacGyver solves the apocalypse with a shoelace and paper, as opposed to the iconic duct tape in American culture ( Figure 10). "Chytra baba z Radomia" is a purely Polish Internet persona. The meme developed from a photo of a woman is reaching for a bottle of soda on the table at Radom's Christmas party for the less fortunate. Subsequent memes have her taking a similar bottle (which has been photo-shopped in) and a variety of other items from unexpected individuals. 11 In the meme found for this project, her negative role is recast and she becomes a hero for saving the planet from the meteorite about to catch or block it ( Figure 11).  Piekot (2012, p. 192) notes that "each demotivator should, at least in theory, negate some real value present in the dominant culture and should comply with very restrictive genre norms". He continues on the same page to list these norms as:

Conceptual blends across image macro genres
-a black background signaling that it is a demotivator, -a headline in all caps that should not exceed one word, -a photo or possibly a film or drawing, -a caption "very often in the form of a punch line or sharp rejoinder, explains the meaning of the photograph." These guidelines could be diagramed as a Conceptual Blend (Figure 12). However, while these norms do a fairly accurate job of describing demotivators in the English language speaking world, they fail to cover the variety found within the Polish Internet community. First of all, the blackand-white demotivator frame is frequently used for material that cannot be classified as a demotivator. Examples on demotywatory.pl include: -pleas for social activism, such as a picture of a man just after donating bone marrow and several paragraphs explaining his decision and encouraging others to do the same; -encouragement, such as how to be a good boyfriend/husband; -commentary on sporting events or political issues.
While these images have the traditional black frame and two colors of text -one usually larger and shorter than the other -they do not all fit Piekot's (2012) description in terms of their pragmatic role. Nevertheless, many of those identified in the corpus do, at least loosely, fit Piekot's norms and a play on words or a punch line in the baseline. This is often accomplished through a reinterpretation of the phrase end of the world, which results from a blend of various definitions of the words end and world as diagrammed in Figure 13 and/or the phrases idiomatic meaning of "not the worst thing that could happen". Thus, in the prototypical end-of-the-world demotivator, expectations regarding the headline "Koniec Świata" ('End of the World') are undermined by the textual content that follows. The visuals support either the traditional understanding of an apocalyptic event with a picture of one or several images of a catastrophe or, as in Figure 14, a picture emphasizing the macros subversive message.
Here we will analyze four demotivators. The first two deride the excitement and panic over the Maya apocalypse and remind the Internet user that whether or not the apocalypse happens as predicted, the end of our own personal world is coming.
The first example ( Figure 14) features a black-and-white picture of an elderly man with a cane looking at announcements of recent deaths outside a church. It reminds the viewer of the shortness of life, personalizes the apocalypse and reverses the order of the cause and effect suggested by the doomsday prediction. Thus, it is not the end of the world that will bring about an individual's death, but death that will bring about the end of each individual's participation in this world.
The second example ( Figure 15) is much more colorful, with four pictures of disasters often associated with the end of the world or mass deaths: an atomic bomb, a tsunami, an alien invasion, and a meteorite striking Why do you believe that for you it will end spectacularly?' 15 the planet. This demotivator relies on the same reversal of perspective; however, it also accuses doomsday proponents of an egotistical outlook. This twist might cause viewers to question their own perspectives and motivations. Both of these demotivators require the viewer to construct two hypothetical timelines. The first comprises a normal birth-to-death cycle, the second is similar but interrupted by an apocalyptic catastrophe. In a blend similar to Fauconnier and Turner's (2002, pp. 40-43) Buddhist Munk diagram, both the life hit by catastrophe and the life lived to old age meet at the same point.

Conceptual blends across image macro genres
The other two appear to have been created after December 21, 2012, as they comment on the failure of the apocalypse to materialize. The meme in Figure 16 posits a worse alternative: Monday. This accesses a wide variety of cultural connotations and jokes regarding the beginning of a typical Western work week. This demotivator is de facto one motivator embedded in another as the first half of the textual content is on a picture of a mushroom cloud explosion with MONDAY assuming the font of the macro's headline. Both the end of the world and Mondays are placed within the scalar frame of BAD THINGS, with Monday identified as worse than an apocalypse. The outer black frame interprets the significance of this scalar frame, which results from blending elements of cultural knowledge (the failure of the Maya prediction, the fiveday workweek) and the concept of a demo version from the domain of music or computers. In the emergent blend, the hyperbolic comparison between the stress of Monday and the annihilation of an apocalypse provides comic relief while enabling Internet users to feel a sense of comradery in their weekly suffering.
In the example in Figure 17, the idiomatic reading of "end of the world" is employed. The graphic in the center of the demotivator is a Maya calendar, and the text is placed at the bottom.

Conceptual blends across image macro genres
Here, the key phrase end of the world comes at the end of textual content. The graphic serves as a metonymic link to one of the most salient aspects of Maya culture and the motivation behind doomsday predictions. The first lines reiterate the presupposition underlying contemporary interpretations of the calendar: the Maya people were an advanced civilization with the knowledge that may exceed that of the 21st century. Hence, if their calendar ends, it must signify something. However, the closing line forces a reinterpretation of end of the word according to its idiomatic meaning.
The result of the clever play on words actually turns the demotivator into a motivator and ironically can even be viewed as promoting a growth mindset approach to life.
To summarize, although Polish demotivators do not always follow Piekot's (2012) guidelines regarding textual layout, those discussed here preserve the genre's sarcastic tone. In the first two demotivators discussed, this tone is critical and serious, whereas in the latter two the message is light-hearted and humorous, if not encouraging. Indeed, in all four of the motivators, the Internet user is challenged, but not necessarily demotivated as being reminded of one's eventual death can be viewed as an opportunity to use the time remaining wisely.

Discussion
In this section, complex blends were used to map the potential creative composition of image macros. It was recognized that certain image macros carry their own pre-set format and textual expectations which are blended in the creation of the stock-image -be it a demotivator or Advice Animal -and then become an input space in the understanding of subsequent memes created using the same template. This coincides with observations made by Bellar et al. (2013, p 19): Many memes require a certain level of knowledge about popular culture (such as recognizing popular media characters and what they represent (e.g. Captain Jack Sparrow) or Internet culture (recognizing a meme genre, e.g. Bad Luck Brian), to decode the elements of irony or humor used.
Thus, rather than treat the image macro templates as in input space, it would also be possible to treat predetermined elements of the meme as common knowledge and part of the grounding box (Coulson & Oakley, 2005) or a semiotic space (Brandt & Brandt, 2005). Of course, this does not deny that these elements are themselves the result of blends, but it would perhaps allow for an easier understanding of the stance of the meme's creator. In this case, that would include how he/she wants Internet users to view the doomsday predictions and how they should respond. For example, are the predictions just a joke, or also an opportunity to contemplate our own mortality. Alternatively, one could read the meme as a commentary on contemporary issues, such as politics, rather than as a position on one of the doomsday predictions. For instance, several of the character-based image macros commented on how the new year would save the country's leadership from having to break another promise and raise taxes in 2013.

Conclusion
This paper looked at the phenomenon of online image macros from the theoretical perspective of conceptual blending. The data for the project came from a thematic corpus analyzing Internet users' responses to the doomsday predictions by Harold Camping in 2011 and the end of the Maja calendar in 2012. This thematic focus allowed for the collection of a variety of genres of image macros. As a result, it is possible to discuss the differences and similarities in conceptual blending patterns involved in the creation and comprehension of Advice Animals, character-based image macros, and demotivators.
Firstly, the analysis of Advice Animals closely aligns with research into phraseological blends conducted by Omazić (2005). It has been suggested that Advice Animal templates can be treated as phraseological input spaces, which are themselves the results of blending. As these templates and their pragmatic meaning are usually well known to the Internet community, the templates could be seen as comprising the shared ground within Coulson and Oakley's (2005) grounding box or Brandt and Brandt's (2005) semiotic space.
A different blending model was required for character-based image macros. It was suggested that rather than a two-step blending process as in the case of Advice Animals, it is possible to analyze character-based image macros using just one blending diagram. This is because the meaning potential is found in the legend surrounding the character depicted in the image provided. The character is best viewed as belonging to an input space because the meaning of the image macro that emerges is a result of blending the given character's recognized persona with the textual content.
Demotivators within the Polish language Internet community appear to possess a wider range of stylistic potential than they do in the English language Internet community. While the black frame is standard, text size and textual expectations differ. The analysis of the demotivators in this corpus revealed two stages of blending in which the second stage requires the reader to reinterpret the previous blend according to the subversive subtext. This often requires simultaneously holding two interpretations, which is believed to increase the pleasure factor (cf. Coulson, 2000;Dynel, 2011).
During this process, the categories suggested by Shifman (2014) proved to be useful, as content, form, and stance all appeared in both the input spaces and the emergent blend. This study also identified additional elements involved in the blending process. The first is ad hoc categorization (e.g. in the case of Advice Animals), and the second is a frequent use of frames. Both relied on knowledge specific to a given cultural group as well as general knowledge. These aspects of successful meme creation seem not to have been given much attention in the literature on Internet memes and are worthy of further study.
N O T E S 1 http://markturner.org/blending.html 2 It was observed by the reviewer that this criticism follows the solutions discussed. In his paper, Glebkin mentions Brandt in his introduction; however, he does not discuss his work in detail, nor does he seem to find Brandt's solution satisfactory. Barczewska (2017) appears to be of the same opinion.