More than 70 million people worldwide have “pinned” personal image collections since Pinterest’s launch in 2010. This book explores the unique sense-making behaviors of independent user-curators actively contributing to an evolving social digital collection. Image naming practices are examined using a matrix including Panofsky’s three strata of iconological subject matter and a range of Wittgenstein’s language game attributes. The exploratory approach taken in this book is especially helpful for information professionals seeking a springboard for further discussions on emerging user needs in digital image curation as well as for researchers interested in user naming behavior in social media collections.
TABLE OF CONTENTS
ACKNOWLEDGEMENTS
CHAPTER 1 INTRODUCTION
Translating images into words
Collecting material memories
Large institutional image collections: Language and control issues
Pinterest launch and growth
Pinterest affordances
The login grid
Social collecting: The emergence of ‘user-curators’
Expanding collections by “following”
“A crazy human indexing machine”: Pinterest as a search mechanism
Unique user behaviors when naming in Pinterest
Analyzing the words used in names: Wittgenstein’s language games
Collecting the language used in names: Panofsky, Rosch and Shatford Layne matrix
Panofsky’s three strata of subject matter
Rosch’s three levels of categorical abstraction
Shatford Layne’s image attributes
Developing the Panofsky, Rosch and Shatford Layne matrix
Statement of the problem
Purpose of the study
Significance of the study
Research questions
Definitions of terms
Assumptions
Limitations of the study
Summary
CHAPTER 2 REVIEW OF THE LITERATURE
Visual categorization in image collection indexing
Social tagging and folksonomy
“Big, messy, organic” data sets
Visual categorization and interindexer consistency
Automated annotated image data
Triads of visual categories: Basic, subordinate and superordinate
Two stage (primary versus secondary) subject matter categories
Defining image attributes
Shatford Layne’s image attributes
User behavior in image naming
Image name iconology: Tools for assigning meaning
Iconclass
Wittgenstein’s rule-guided language-game analysis
Observed existing non-user attitudes related to the pinterest site in general
Pinterest Is (a) only used by women, (b) reducing its importance
Pinterest is a threat to feminism
Pinterest is primarily for selling products, principally to women
Pinterest should be studied and discussed like other “social media” sites
CHAPTER 3 MATERIALS AND METHODS
Introduction
Data collection approach
Data collection method
Image collection
Name collection
Data analysis: Panofsky/Rosch/Stratford Layne matrix
Wittgenstein’s Rule-Guided Language-Game Analysis : Observed Forms In Pinterest
Semantic analysis of pin names
Methodological issues
Scope and limitations
Expected results
Summary
CHAPTER 4 ANALYSIS OF DATA, RESEARCH FINDINGS, AND DISCUSSION
Alpha data collection
Final data collection
Research findings and discussion
Pin Name Distribution: Panofsky’s Strata of Subject Matter or Meaning
Expectations
Findings: Primary names
Findings: Intrinsic names
Findings: Secondary names
Rosch’s three levels of categorical abstraction
Types of Pinterest language games
Story-telling:
Nonlinguistic language games
Rules and determinacy in Pinterest language games
Most commonly observed language games related to pin naming in Pinterest
“Private” language: Codes in Pinterest names
Nonsense
New surface grammar construction
Unexpected Findings Related To Re-Searching Pins
CHAPTER 5 SUMMARY AND CONCLUSIONS
Summary of findings
Implications of research findings
Forcing precision
Challenges in Pinterest research
Recommendations for Future Research
Cunning Intelligence and Social Collecting
Conclusion
APPENDIX A ALPHA DATA SET
APPENDIX B: FINAL DATA SET
APPENDIX C: USER STATISTICS 2012 -2014 PINTEREST USER STATISTICS
APPENDIX D PIN SELECTION
APPENDIX E KAMATH’S BOARD COHERENCE
APPENDIX F ALL PIN NAMES: SECONDARY, PRIMARY, INTRINSIC
REFERENCES
ACKNOWLEDGEMENTS
I owe a large debt of gratitude to Dr. Brian O’Connor, who can be relied upon to always deliver one more interesting idea to delightfully complicate matters. I would also like to extend fond appreciation to both P. Tooley and K. Moses, consultants extraordinaire. And of course, deepest thanks to David Sutcliffe, for never giving up.
LIST OF TABLES
Table 1 Panofsky’s Three Strata of Subject Matter or Meaning 29
Table 2 Examples of Panofsky’s Three Strata 31
Table 3 Comparing Chracteristics: Taxonomies, Folksonomies and Social Curation 35
Table 4 Rosch’s Basic Image Category 39
Table 5 Rosch’s Superordinate Image Category 40
Table 6 Rosch’s Subordinate Image Category 40
Table 7 Shatford Layne Images Attributes Used In This Project 43
Table 8 Example: Applying Shatford Layne’s Image Attributes to an Image 44
Table 9 Applying the Panofsky/Rosch/Shatford-Layne Matrix 46
Table 10 Pinterest User Examples of Limited Social Interactions 62
Table 11 Final Search Terms 66
Table 12 Alpha Search Terms 75
Table 13 Alpha Data Available in Appendix A 75
Table 14 Final Data Set 76
Table 15 Example of Matrix Items Collected in Name Samples 80
Table 16 Storytelling Pin Names 81
Table 17 Non-Linguistic Pin Names 82
Table 18 Examples of Family Resemblances in Pin Names 83
Table 19 Examples of Commonly Observed Pinterest Language Games 85
Table 20 Private Language Codes Pin Names 86
Table 21 Observed Intrinsic Surface Grammar 88
Table 22 Observed Grammar Construction Examples 89
LIST OF FIGURES
Figure 1. Robert Cornelius, head-and-shoulders [self-]portrait, facing front, with arms crossed. Approximate quarter plate daguerreotype, 1839. LC-USZC4-5001 DLC. 1
Figure 2. Percentage change in unique visitors 6
Figure 3. Screen shot of the Pinterest home feed drop down link. 9
Figure 4. Pinterest activity and library technical services 10
Figure 5. "You don't have to talk to anyone." 12
Figure 6. "We're Pindred spirits." 12
Figure 7. “Time on Pinterest” 15
Figure 8. "A person on Pinterest" 15
Figure 9. Pinterest affordance: The Pin It button 15
Figure 10.Van Straten’s proposed revision of Panofsky’s three strata 32
Figure 11. Iconclass keyword search example 49
Figure 12. Wittgenstein (1958) Philosophical Investigations 51
Figure 13. Pinterest default display: Example of “endless scroll” 67
Figure 14. Findings by Panofsky strata 77
CHAPTER 1 INTRODUCTION
illustration not visible in this excerpt
Figure 1 . Robert Cornelius, head-and-shoulders [self-]portrait, facing front, with arms crossed. Approximate quarter plate daguerreotype, 1839. LC-USZC4-5001 DLC.
The invention of the photograph gave us what Oliver Wendell Holmes called a mirror with a memory. This gave us a way to see surface structures of objects; thus, we can now see what Robert Cornelius looked like in the autumn of 1839 (See Figure 1). However, surface appearance may not be the only way one chooses to represent oneself or what one thinks. New technologies enable manipulations in new ways, ways that can be accomplished individually and for individual needs. Naming practices within this new environment can also be individual.
O’Connor, B. C. 2014. "Selfies and Public Knowledge"
Founders Lecture in Proceedings of DOCAM 2014
I pin what reflects me. This is me. If you see what I am doing, you will see the real me. I can see the real me. This is what I am thinking about at that time.
Pinterest allows people to connect with others in an authentic way. This is who I am for real. Look what I can do, what I did. You can do this too, if you want.
When I go back and look at all the stuff I’ve pinned, it kind of tells about me, of myself, it comes together. I can’t explain it . . . you have to see it. I don’t have to explain that to someone. It is just there in the pictures.
Interview results from “Why do college students use Pinterest?” Sashittal, 2014.
Translating images into words
A classic problem which continues to challenge information scientists involves the process of representing images using words. Word-based language does not necessarily provide adequate descriptions of visual experiences and the issues of transmedial translation continue to complicate investigations into how people communicate their reactions to visual stimuli. The variability of language itself contributes to a degree of information loss when visual encounters are rendered into words and this project is rooted in that dilemma: How do people share their individual interpretations of a visual experience when their representational tools are word-based?
Collecting material memories
One way people have attempted to represent (and potentially preserve) their thoughts, experiences and memories throughout history is by creating and handing down hybrid collections combining both images and words, compiled to reflect self-selected aspects of themselves. Designating and saving representative illustrations, likenesses and written language into material remembrances continues to satisfy a basic human hunger. Examples of this urge to compile and perpetuate assemblages of personal meaning include Greek hypomnema (personal notebooks), 15th century Italian hodge-podge books and 17th century commonplace books (Curtis, 2011).
Modern examples of this type of meaningful cultural “material memory” collection tend to be divided between (a) personal collections of privately expressive documents (photo albums, family bibles, daily diaries, scrap books) and (b) public collections of culturally valued images, usually designated as either artistic archives with presumed didactic value (museum collections, for example) or commercial commodities purchased for consumption only by an approved audience (corporate graphic art archives or municipal police mug shot catalogs).
Large institutional image collections: Language and control issues
The challenging relationship between language and images can be observed on a grand scale in large traditional image collections, in particular collections paid for and accessed by institutions such as museums or corporations which have traditionally been expensive to create and maintain, requiring sizable budgets to absorb the direct and indirect costs of curation and access.
Because the expenses related to maintaining large collections of physical images have traditionally been greater than most individuals could afford (with a few historic exceptions), a majority of large public image collections have relied on institutional funding– and have been subject to institutional controls
Given the costs of curating large image collections, it is not surprising that the assumed use of a large institutional image collection would eventually become a factor in determining the complexity and semantic density of the indices provided (and the language involved). The needs of the users of large institutional image collections have intermittently been analyzed either formally or informally by image curators charged with providing access, although aiding the work of the collection user by providing accessible language has not necessarily had the highest priority in every instance.
Public art museums are only one example of the financial expense historically associated with large culturally valued image collections. Other more pedestrian image collections such as metropolitan mug shot binders and corporate graphics archives also tend to acquire both cost and value as artifacts which may not be based on any quantifiable data directly correlated to either the images or the original use. In order to remain economically viable, curators must assume that a public user of an art museum has a finite set of recognized and specific needs, just as the users of metropolitan police mug shot binders and the users approaching the corporate graphics archive are assumed to have a terminable set of needs when viewing those image collections.
The resulting institutional image retrieval systems, based on language generated by curatorial notions of how users might approach any given large image collection, have historically produced varying results, sometimes providing effective image retrieval for users and sometimes only increasing the internal ease of use of the collection for the curators themselves. Constructing efficient descriptive inventory listings tends to be of paramount historic focus for curators of large institutional image collections, while improving retrieval measures for non-curatorial collection users frequently becomes a secondary benefit of maintaining a well-ordered inventory. The subtle and fluid ways in which people may be using language as they encounter images in large controlled collections is challenging to capture and difficult to interpret, so the focus of these collections has tended to remain on effective subject-driven inventories.
In the past, the high costs of large institutional image archives virtually guaranteed that control of these collections would remain within organizations who could (a) afford the expenses of maintaining the images and (b) train the curators to inventory, index and provide access using institutionally-approved indices and vocabulary.
Pinterest launch and growth
The creation of sizable digital image collections is no longer exclusively controlled by officially-sanctioned institutional curator/gatekeepers. Large public non-institutional digital image collections are a reality, and ordinary people have begun creating and managing their own private image collections, using language in interesting ways in the process.
Pinterest ( http://www.pinterest.com ) is a free web site which describes itself as “a beautiful visual discovery tool.” (Madrigal, 2014). Since initial launch in 2009, seventy million users have created personal image collections using the site’s minimalist platform, staying logged in for periods averaging up to 40 minutes per visit, with the intention of creating and managing their own image collections (Palis 2012). Average web site visit times are notoriously difficult to verify, but an April 2014 Agbeat report showed Pinterest users remained on the site longer than on any other social media site except Youtube. (Agbeat, 2014)
Pinterest reached the 10 million monthly unique U.S. visitors milestone more rapidly than any other site previously monitored (TechCrunch, 2012) and became the third largest social network in the United States in March 2012 (Experian, 2012). Analysts estimate that Pinterest had approximately 7.5 million monthly visitors in December 2011 before jumping to 11.7 million in January 2012 (Pew Reports, 2013). As seen in Figure 2, traffic between January 2012 and February 2012 increased from 11.7 million unique visitors in January to 17.8 million in February, representing an unusually large change (a 52% increase in one month) for a relatively young site (Walker, 2012).
illustration not visible in this excerpt
Figure 2 . Percentage change in unique visitors
By July 2013, Pinterest reached 70 million registered users worldwide (Semiocast, 2013 ) with 24.9 million unique monthly U. S. desktop users reported in September 2013 (comScore, 2013). Through July 2014, Pinterest users have “pinned” 30 billion images on 750 million “boards” (Madrigal, 2014). Appendix C contains Pinterest user statistics from 2009 through 2014.
Pinterest more than doubled its international audience in 2013, expanding to include 31 languages (Frier, 2014), and the company announced plans to launch in ten additional countries before the end of 2014 (Brustein, 2013). Horowitz (2013) found that international users could potentially surpass the aggregate number of American users by the end of 2015, based on current international user growth rates.
As of May 2014, Pinterest reported receiving a total of $764 million in funding from investors who valued it at $5 billion, making it one of the most valuable venture-capital-backed startups in the world (MacMillan, 2014).
Pinterest affordances
Affordances are the aspects of interactivity within an interface which suggest available activities to users (Hocks, 2003). The affordances offered by Pinterest include the ability to fine tune the new images automatically displayed at login, selectively “follow” (collect) images and collections from other users and use an assortment of mechanisms to freely browse, “like”, share on other social media sites, email to other users, download, comment and name images in real time, during any curating session.
Unlike online image archival sites such as Flickr (http://flickr.com ), or real-time photo chatting apps such as SnapChat (http://snapchat.com ), Pinterest is not primarily designed as a image storage site or a content delivery platform, but rather a revolving exhibition of imagery related to each user’s personal interests. The stated mission of Pinterest is to "Connect everyone in the world through the 'things' they find interesting.” (Cold Brew Labs, 2012).
After creating and naming new empty “boards” to hold acquired images (“pins”), the new user-curator selects one image at a time from the login grid and views it on its originator’s board. Next actions can include repinning the image to a board in their own collection, liking the image, sharing the image via various tools, commenting on the image to the original poster, or disregarding the image and returning to browsing the login grid, alternatively drilling into selected category postings.
The login grid
The basic Pinterest© user interface is the login grid, composed of the most recently uploaded random images from all users. This display is automatically presented to every logged in user visiting the http://www.pinterest.com URL.
The login grid was designed by Evan Sharp, one of the site originators and an architecture student who admits to being fixated on the possibilities of an aesthetically pleasing interface: “It’s a visual product about beautiful images of meaningful things… The way you draw something is intricately tied to how good your solution to a problem is or how well the product you ship turns out. I am very, very obsessed with this idea” (Allen, 2014, p. 13).
The role of aesthetics when measuring user engagement with content is an ongoing debate and Pinterest provides an example of a successful minimalist approach. Tufte (1983) states that “The best graphics are about the useful and important, about life and death, about the universe. Beautiful graphics do not traffic with the trivial” (p. 177). As of 2014, there are no ads, instructions or unneeded text on the Pinterest login grid: only row after row of scrollable images, updated continuously. The relative starkness of this main display grid remains a unique feature of Pinterest, and has been credited by the site’s originators with much of the sites visual appeal: “The grid is the thing that got us big. Pinterest is about browsing through objects and picking out the ones that are meaningful to you. And what the grid does is facilitate your ability to go through objects in an efficient way. Our job is to put the right objects in front of you to start with” (Madrigal, p. 6).
Fine-tuning the Home Feed
The home feed screen automatically updates itself every time a new “followed” image is uploaded by another user. The navigational link to return to the home feed is included at the top of every page, on the drop down menu which provides available pre-populated categories, as shown in Figure 3.
illustration not visible in this excerpt
Figure 3. Screen shot of the Pinterest home feed drop down link.
User-curators retain control over what they see by customizing this home feed and can choose to be exposed only to those collections they are interested in following. The distinction between viewing “everything” on the login grid (every random upload from every user in real time without filtering) and viewing the personalized home feed (only the collections intentionally selected by that user for display on that particular home feed) is a central editing tool for effective Pinterest collection development “When you open up Pinterest, you should feel like you’ve walked into a building full of stuff that only you are interested in. Everything should feel handpicked just for you,” (Chafkin 2012, p. 93).
Because collecting images is the purpose of Pinterest, misunderstanding the basic mechanisms for image selection is a user oversight which limits Pinterest to a critical degree. Pinterest users who fail to take advantage of home feed filtering (which automatically occurs as soon as images from other users are followed) may have erroneously concluded that the randomly unfiltered flow of indiscriminate images on an public login grid is all that Pinterest contains. Using a Facebook analogy, assuming that the unfiltered Pinterest login grid reflects all available content within Pinterest is similar to opening a Facebook account but then failing to add friends. The login page on either site rarely delivers value without some level of personalization and interaction on the part of the user. Pinterest users principally customize their home feeds by finding and following the images of others.
Social collecting: The emergence of ‘user-curators’
Zarro and Hall (2012) define Pinterest as a “social collecting” site, and describe how users become “user-curators” and “patron-curators” (p. 2). This user-centered perspective allows comparisons of pinner activities to traditional library service tasks as shown in Figure 4, but Zarro and Hall (2012) also note that “the cataloger and patron roles are one and the same in the social collecting model” (p. 3).
illustration not visible in this excerpt
Figure 4. Pinterest activity and library technical services
The ability to arrogate cataloging authority is presumably not the central reason that millions of people create image collections on Pinterest each month. User-curators appear to employ Pinterest to collect and share concepts, large and small, which take the form of images linked either to other Pinterest collections, to sites outside Pinterest or to uploaded images from their personal collections. User-curators do not appear to be seeking people. Rather, they are seeking ideas.
Although all Pinterest content is captured and uploaded by the members of the community, and all content is public, Pinterest users cannot be defined as purely “social” users. Typical social site activities (which usually involve direct personal interactions between users such as chatting, liking or commenting) are not as pivotal to the Pinterest experience as the indirect, nonpersonal action of repinning images. Unlike genuinely social-based users such as those on Facebook or Twitter, Pinterest users tend to focus on creating and maintaining a personal image collection, rather than interacting with other users. The central purpose of Pinterest is to share images, not necessarily to make friends or connect with other people. It is common for Pinterest users to have no direct communication with other users at all. As shown in Figures 5 and 6, a series of humorous pins has been widely circulated within Pinterest itself, acknowledging this characteristic:
illustration not visible in this excerpt
Figure 5. "You don't have to talk to anyone."
illustration not visible in this excerpt
Figure 6 ."We're Pindred spirits."
Pinterest co-founder Evan Sharp emphasizes the powerfully “non-social” aspects of the site, particularly when asked about similarities between Pinterest and other “purely social” sites: “Pinterest isn’t about friendships. It’s not a messaging app like most of these big startups. It’s about culture, for lack of a better word” (Summers, 2014).
In their study on college students using Pinterest, Sashittal and Jassawalla (2014) note that “The focal cognitive process of Pinterest usage is not a conversation with others; it is a soliloquy. Pinterest users are not telling others about how interesting they are; they are engaged in primarily defining for themselves, their deeply held, authentic interests” (p. 25). The data in their 2014 study emphasized valuing ‘authenticity’ as a motivation for using Pinterest, and contrasts this quest for authentic self-exploration with the surface-focused ‘popularity contest’ aspects sometimes apparent on Facebook and Twitter: “College students use Pinterest because the process of pinning and posting photographs on their pages, developing visual narratives and a deeply personal curated list is an experience of authenticity; a process that is closely aligned with the discovery, definition, development of an authentic sense of self. This experience stands in sharp contrast to one related to posing, posturing, or positioning oneself for the validation of others” (p. 8).
Expanding collections by “following”
Despite the lack of emphasis on direct interaction between pinners, the most powerful method of developing a large and personalized digital image collection includes finding and following other users who are focused on similar topics. “Following” is done by selecting an interesting image, and clicking that image to return to the originator's board. By visiting the originator’s related boards, the user-curator can review the full collection of images posted by this originator and explore both their archives and other images posted by additional people who follow this originator. A new user-curator may discover that a fellow pinner has no further image boards of interest or they may discover rich resources, both of fellow pinners who have related collections and of boards full of related imagery. The number of pins collected within each board is displayed on every pinner’s profile page, so a new user-curator may decide if they are interested in following an active board on a given topic (which may involve hundreds or even thousands of images). Any user may be unfollowed or re-followed at will, and any number of boards may be followed or unfollowed, without loss of related pins.
This process of branching through other curated boards on related images is one of the most powerful tools provided to the Pinterest user-curator. “Going down the rabbit hole” when visiting another pinners boards opens a variety of pathways to new search vocabulary, similar collections and peripherally related topics. For example, a general search on the terms “Claude Monet” in September 2014 yielded several thousand images, all of which link (among other things) to reproductions of Monet paintings, biographical information on the artist, an essay on how the human eye processes UV light, photographs of the village of Vétheuil where Monet painted in 1880, a blog on gardening at Giverny, a free cross stitch pattern based on the painting Garden with Irises, an article on the new Claude Monet rose in the New York Botanical Garden, and a Claude Monet Word Search Worksheet for a home school unit on French Impressionism. Each of these diverse links, in turn, leads the user-curator forward to new boards and additional pinners, which contain further new materials, tied to additional images and links.
This richness of related content partially explains why Pinterest user-curators typically spend hours on each visit, versus minutes on Facebook or Twitter. As seen in Figures 7 and 8, a subcategory of recognizing how quickly time flows past while pinning has emerged, with contributors wryly noting skewed perceptions of time when they are involved in a curating session:
illustration not visible in this excerpt
The additional affordances of linking out to source images and using browser plug-ins to speed pinning reportedly encourage site-wide user behaviors that do not appear to be duplicated on this scale in other free public digital image collections. Hocks (2003) notes in particular that the Pinterest browser “plug-in” called the Pin It Button shown in Figure 9 allows for an intensive and amplified layer of interactivity, because users can continue to interact with Pinterest even when they are not on the site. (p. 55)
illustration not visible in this excerpt
Figure 9 . Pinterest affordance: The Pin It button
Arguably, both images and language are being curated on Pinterest. While users are not required to create textual information for their images, the user-curators observed in this project are using language in their image names, presumably to annotate the content for themselves but also to attract other pinners interested in similar ideas: providing tags within the search tools provided on site, as well as illustrating, amplifying and creatively expressing each user-curator’s views. The layers of meaning added to the imagery with language (intended and inadvertent) contribute to the fascination many Pinterest users profess with the site. The depth and variation of the messages, both visual and verbal, available within the collections might explain why this site has so quickly become absorbing for millions of users on multiple levels.
“A crazy human indexing machine”: Pinterest as a search mechanism
Representatives of Pinterest have become consistent public missionaries for the concept that user-curators increase the depth and value of the site content through their independent use of language while collecting. Co-founder Sharp calls the site a “human indexing machine”:
HTML is the architecture of the web and it is about the presentation of text. It’s Hyper Text Markup Language. And if you’re Google and you’re trying to index that world of text information, you’re really great at text because that’s what the code on the Internet does. It marks up text. But if you want to get at objects or the things on web pages, we think you need humans to go in and do that for you . So we think of Pinterest some days as this crazy human indexing machine. Where millions and millions of people are hand indexing billions of objects—30 billion objects—in a way that’s personally meaningful to them. (Madrigal, p. 3)
In a 2014 interview, another co-founders of the site explained that
’Search’ for most people is web navigation, stitching together the human information on web pages. Or search is a tool for answering questions. We weave them together, but you could decompose those tasks on Pinterest in an interesting way if you were interested in solving search as a problem… [and] there’s a whole world of search and discovering [on Pinterest] that’s about the [search] process itself. And that’s an interface driven experience: How users self-describe their interests over time, rather than just the search technology we have today. (Madrigal, 2014, p. 8)
Unique user behaviors when naming in Pinterest
A striking affordance of Pinterest is the opportunity for each user-curator to name and re-name, to categorize and re-categorize, increasing the layers of possible meaning available to all viewers and allowing a level of interpretive expression and cognitive association not possible in the static physical archives of the traditional art museum, the police mug shot binder collection or the corporate graphic archive. The complicated, innovative, expressive ways user-curators have evolved to use language within their collections, on all levels, have become part of the fun.
While Pinterest is often referred to as a social media site, with public member collaboration producing the core of the image content, the process of creating pin names on the site has evolved into a personally expressive form of communication across the population of users. A core finding of this project confirmed this basic urge toward independent customization: Pinterest user-curators are not generally interested in applying any existing, predefined naming categories to their collections.
The intensity of this creative, highly personalized naming activity is not exclusively focused on providing efficient image retrieval. Rather, users appear to be embedding meaning in the file names they create, adding one more layer of interest and expression to the way they present their Pinterest collections. Carefully crafted names become part of the meaning behind the concepts. Pin and board names are frequently entwined with the concepts being staged and might include puns, word art, alliteration, malapropisms, spoonerisms, obscure words, rhetorical excursions, oddly formed sentences, ASCII art, emoticons and double entendres. Unique uses of upper and lower case fonts are found, as are abbreviations and malformed sentence/word phrases, designed to convey an intended meaning of either a pin or a board.
O’Connor and Greisdorf (2008) note “…[O]ften the only messages available to the image collector are the intended messages based on the history and circumstances surrounding the creation of the image” (p. 78). This statement leads to the questions that sparked this project: What happens to the meaning of an image when the history and circumstance of its creation are no longer available to the collector? Considering individual Pinterest users as curators of their own large image collections, how significant is the naming of an individual image when examining the overall structure of such large uncontrolled image compilations?
Analyzing the words used in names: Wittgenstein’s language games
Examining how user-curators manipulate language when creating names for images in their collections highlights the particular slipperiness of defining “meaning” in language. Biletzki and Matar (2014) note that “Traditional theories of meaning (in the history of philosophy) were intent on pointing to something exterior to the proposition which endows it with sense” (p. 207).
This view – that the inherent message of a word is predetermined by some force outside the user – is dissolved by Wittgenstein’s’ later work on language games, a specialized way to think about active language use, involving the recognition of the layers of influence at work when language is constructed, including the “natural history” of a given environment, the “forms of life” in which language may or may not be required and the circumstances at play during any particular human activity. A language game can include giving orders, describing the appearance of an object, constructing an object from a description (a drawing), reporting an event, forming and testing a hypothesis, making up a story, telling a joke, cursing, greeting, and praying (Wittgenstein, 1958, p. 11-12). Such language game activities have evolved some generally recognizable steps and conventions, both stated and unstated, related to the activity at hand, and extending to the kinds of language usually used during each type of action.
The core of language games involves activity. Blair (2008) notes that “What defines us as humans is not so much a common linguistic ability, but a common ability to engage in many simple and complex human activities. We can imagine people without language but not people without shared activities” (p. 163). Wittgenstein demonstrates that shared activities (rather than some hidden historical substructure) form the foundation of working language and clearly states, “For a large class of cases of the employment of the word ‘meaning’—though not for all—the meaning of a word is its use in the language” (Wittgenstein, PI 43). Blair (2006) reiterates this use-based theory of meaning in language:
When we use words in a particular way that conveys our meaning unambiguously we understand this usage, not because the words have some common essential meaning to them, but because we share the activities or practices in which the words are used. (p.167)
This action-oriented view of language will be referenced when analyzing the collected pin names in this project. Since language cannot be independent of the context and circumstances of its use, the words chosen for pin names may reflect some patterns and practice unique to the process of “collecting Pinterest images.” Exploring the language game of “ naming Pinterest images” (in this particular, limited sense, always remembering the other intricacies of Wittgenstein’s language game additional requirements) will provide a point of reference when observing how user-curators construct language within their image collections.
Collecting the language used in names: Panofsky, Rosch and Shatford Layne matrix
In order to begin analyzing the Pinterest names collected for this project, it was necessary to construct a matrix of the language chosen by user-curators. Three separate approaches were combined into one matrix: Panofsky’s strata of subject matter, Rosch’s levels of categorical abstraction and Shatford Layne’s divisions of image attributes. A brief summary of these approaches follows. A more detailed review follows in Chapter 2.
Panofsky’s three strata of subject matter
1. Primary subject matter (“What is depicted?”) can be described using elemental language (animals, people, settings) and does not require the viewer to have any knowledge of the culture related to the image.
2. Secondary subject matter (“What is the story?”) notes the literary and cultural themes, concepts and allegories intentionally depicted in an image. This level demands some specific cultural knowledge related to the image on the part of the viewer.
3. Intrinsic content (“What does this all mean?”) is the information available in an image representing the historical environment, including intentional (and unintentional) symbolical values related to the specific characteristics, technique and culture of the image and its creator.. Finding meaning in images on this level requires relatively in-depth knowledge of the culture and environment which produced both image and creator.
Rosch’s three levels of categorical abstraction
Rosch proposed three levels of categorical abstraction which users may employ when associating selections of “basic level objects” with the realities of actual observed environments.
Rosch’s basic image category is the most “inclusive” layer of classification because images here share the highest number of common attributes. A basic image category may include a wide variation of images which are all unique from one another, but which all fit multiple common requirements of being identifiable as a car or a chair based on a high number of common “car” or “chair” attributes.
Rosch’s superordinate image category is one level more abstract than the basic category. Images within this category commonly share only a few attributes. For example, images within the category of ‘vehicles’ (superordinate to cars) tend to have fewer common attributes than do images within the category of ‘ cars’ (the basic category).
Rosch’s subordinate image category contains images which are subsets of the basic category. These individual images tend to share many overlapping, predictable attributes with other member images in this distinct category. If ‘vehicle’ is the superordinate, and ‘car’ is the basic category, then ‘1969 Chevrolet Camaro RS’ would be an example of a subordinate category.
Shatford Layne’s image attributes
Shatford Layne developed a system of specific attributes of any given image which can be used to determine the types and density of meaning associated with that image: biographical attributes (how and where an image was created, including how it has been used, sold or changed), subject attributes (what an image is of - which can be concrete and specific - or what an image is about - which can be abstract and generic), exemplified attributes (characteristics of the image format, not related to subject matter) and relationship attributes (how this image is related to others, such as playing the role of a preliminary sketch or a final draft).
Developing the Panofsky, Rosch and Shatford Layne matrix
The combination of this particular set of strata, abstractions and attributes into one specific matrix for analyzing meaning in naming activity is unique to this project. While all of these tools are routinely used as independent analysis mechanisms, combining these particular tools in this specifically limited matrix occurred as a natural offshoot of attempting to isolate the language being used in this study. Creating a matrix using a combination of Panofsky’s subject matter categories, Rosch’s levels of abstraction and Shatford Layne’s attributes provided a framework to begin examining Pinterest image names, and to analyze the density and complexity of the language being used by user-curators when naming images in their large, personal digital image collections.
Statement of the problem
The creation of sizable digital image collections is no longer exclusively controlled by officially-sanctioned institutional curator/gatekeepers. Large public non-institutional digital image collections are a reality.
In traditional institutional service models, the keepers of image collections were trained in complex and detailed systems to enable them to identify, store and locate images. The approaches being used by non-professional social image collectors (who presumably have limited formal training in collection development or indexing when managing large digital image collections) have yet to be studied in the online environment.
Purpose of the study
The goal of this project is to increase understanding of the specific naming behaviors present in an image collection when the categorization vocabulary and subject descriptors are uncontrolled. Other types of information-based behaviors are simultaneously taking place within Pinterest, of course, including various forms of browsing, seeking and tagging. The purpose of this study, however, is to observe and capture the forms of human behavior most closely related to image naming activity in particular, and thus the findings from this project are offered to stimulate new thinking and research related specifically to Pinterest image naming practices and not as generalizable theory.
Significance of the study
Greisdorf and O’Connor (2001) detailed the ultimate inability of language to universally translate visual experiences and concluded that “No individual or small group of individuals, no matter how professional or rule intensive the approach, could ever capture a full panoply of impressions evoked by an image” (p. 7).
By observing the characteristics of Pinterest’s relatively non-ruled based approach to image naming in action, this project explores the language practices of Pinterest user-curators, isolating a sample of image names and considering where these names fit within a matrix of Panofsky’s subject matter categories, Rosch’s levels of abstraction and Shatford Layne’s attributes. The types of words chosen, the number and format of the characters selected, the linguistic constructions applied to each name when individually organized by each user-curator and the patterns which emerge throughout the sample give a small but unique snapshot of human language behavior during digital image curation.
Research questions
The two research questions in this project run parallel with the two language exploration techniques selected for observing Pinterest naming behavior.
Research Question 1 centers on the Panofsky/Rosch/ Shatford Layne matrix in an effort to isolate the language being selected in pin naming. Assigning the collected sample of names to the matrix provides a way to detach and extract the resulting language, allowing the words to remain separate from the related images. The specific question under consideration is: Where does the language used in creating image names in Pinterest tend to fit within the Panofsky/Rosch/ Shatford Layne matrix?
Research Question 2 concentrates on the facets of Wittgenstein’s language games which were observed in this sample. The question posed is: Which aspects of Wittgenstein’s language games including grammar construction were visible in the selected sample?
Definitions of terms
- Panofsky’s three strata of subject matter or meaning:
- Primary: Natural subject matter, described as the form of the image or subject, using factual information based on practical experience, requiring only a basic familiarity with ordinary objects and events.
- Secondary: Conventional subject matter, described in specific themes, concepts, stories and allegories which require some insight into historical conditions, history of types and literary sources.
- Intrinsic: Symbolic values which are culturally specific, interpretive or non-contextually defined and involve intuition, personal psychology or knowledge of cultural symbols.
- Pin: Visual bookmark intended to link back to the originating site, created by uploading original content or “re-pinning” from existing Pinterest collections. Pins are named by each user-curator, and the name can be the same as the originating pin, different from the originating pin or blank.
- Board: Collection point for pins, created and named by each user-curator.
- Pinner: User-curator who creates a personal digital image collection by uploading new images, pinning existing images from web sites and/or repinning images from other pinners.
Assumptions
Pinterest was selected to exemplify large digital social image collections in this project based on the number of participants and the increase in the number of users from 2012 to 2014. The site is assumed to be stable and available for public use through the expected timeframe of this project.
It is important to note that as of September 2014, all image posting and naming activity is public on Pinterest. All images are fully viewable as part of the larger site, and the implication is that all pinners are participating, voluntarily, in the larger community. This sense of community is maintained even when some pinners are collecting intensely personalized images with no defined meaning beyond their individual private messages, while other pinners are collecting images gleaned from mass media, advertising or merchandising, targeted at an audience of hundreds or thousands.
A “secret board” project was launched during December 2012 which allowed each user to create three non-public boards. This tool is still available as of September 2014 but the support pages indicate current issues are limiting the expansion of this service. Since the stated goal of Pinterest is to allow users to share images and the default instructions for all basic Pinterest activity continue to define all pins as being publicly viewable, the assumption can be made that all default activity on Pinterest will remain public.
Limitations of the study
Pinterest user-curators can choose to remain relatively anonymous in terms of reported demographic data. Very little individualized information about user-curators (such as gender, age, native language, educational background or online experience) can be deduced from normal Pinterest site activity.
Pins can be deleted or edited by user-curators at any time. Once data collection has been completed, it is necessary to create a static reference copy, since pins may be removed or changed at any time without notice on the site.
Summary
This project considers how independent user-curators are adapting language while naming their images in personal digital collections within the social collecting site Pinterest, where no controlling vocabulary is enforced or provided. Self-curated image collections like Pinterest would seem to allow an opportunity for user-curators to break free from the traditional constraints of the pre-defined vocabularies assigned by institutional content gatekeepers.
Pinterest user-curators appear to create collections as a collaborative expressive exercise, as a shared communication device and, frequently, as a private creative outlet thematically aimed at no other audience beyond themselves. Understanding how this personalization influences the way the images are categorized by the user-curator may lead to better methods for users in other image collections to contribute additional value to the collection in the form of meaningful image naming language, as well as reducing factors which appear to discourage existing users from contributing to the naming process in other large digital image collections.
CHAPTER 2 REVIEW OF THE LITERATURE
Visual categorization in image collection indexing
Research on the methods used by curators to efficiently index visual images has been shaped by the human ability (and frequent inability) to communicate experiences with (and perceptions of) visual stimulation (Rose, 2001, p. 43).
Attempts at analyzing human abilities to perceive and interpret visual stimuli have produced myriad academic landmines and hotly disputed, closely-held lexical theories revolving around the semantics of “meaning” (Mirzeoff, 2006, p. 18). For the purposes of this project, the intriguing but eternally complex issues related to defining terms such as “visual culture” and “meaning” have been carefully skirted, since a clear and noncontroversial set of tools is needed to collect and sort the language used by Pinterest user-curators. As a final note on the semantics and semiotics entrenched in this project, it is interesting to note that Mirzeoff (2006) defines visual culture as “ the product of the collision, intersection and interaction between capital’s picturing of the world and that which cannot be commodified or disciplined” (p 66).
Since economic factors determined the existence of many large institutional image collections in the past, it is no surprise that the focus of image collection research in the twentieth century was generally directed toward increasing the “efficiency” of search and retrieval activities (Gombrich, 1999, p. 299). Those responsible for managing large institutional image collections traditionally focused on the tools needed to provide identified users with specific levels of image retrieval speed and perceived accuracy (Hibler, Leung & Mwara, 1992).
Image indexing research evolved into considering how people looked for images: the language they used, the ways they organized their thinking, and/or the paths they tried when the image was not easily described by ordinary language (Reed, 1972 ; Shatford, 1986; O'Connor, O'Connor, & Abbas, 1999; Shatford Layne, 2002).
Creating a practical system to identify visual objects requires a wide range of interdisciplinary tools. Previous attempts have included aspects of cognitive psychology, library sciences, art history, content-based retrieval, semantics, semiotics, physiology and optics, among other fields (Jaimes & Chang, 2000; Hollink et al., 2004; Rorissa, 2005; Rorissa & Iyer, 2008).
Oyarce (2012) further explored the related Greisdorf and O’Connor (2008) concept of cognitive synthesis and verbal expression, re- naming this tangled user experience the “perception-conception interplay” and observing how subconscious memories and experiences add to the influences affecting every user’s reaction to any given image. (p. 9)
Despite the known limitations of quantifying the visual experience, the act of categorizing what viewers perceive (and can communicate) when confronted with a particular image has been broken down into a variety of measurements, always rooted (with varying degrees of consensus) in what might constitute a more successful image retrieval system. Panofsky’s three strata of subject matter
In 1939, the German art historian Erwin Panofsky introduced a controversial approach to analyzing the symbolic forms identified in Renaissance art. His ideas are the basis for much of modern iconology, having been challenged (and refined) by art historians for decades. Panofsky’s core proposal as shown in Table 2 suggests three distinct levels of meaning (some possibly unintended by the creator) which may be identified within an image.
Table 1
Panofsky’s Three Strata of Subject Matter or Meaning
illustration not visible in this excerpt
Elsner and Lorenz (2012) note that Panofsky’s approach advocates these three levels of meaning in a work of art, and, further, includes “the three levels of interpretation needed to elicit them” (p. 485). The interpretive levels include the following:
- Primary subject matter (“What is depicted?”) can be described using elemental language (animals, people, settings) and does not require the viewer to have any knowledge of the culture related to the image. Panofsky labels this interpretation of primary subject matter as ‘pre-iconographical description’ within the three levels.
- Secondary subject matter (“What is the story?”) notes the literary and cultural themes, concepts and allegories intentionally depicted in an image. This level demands some specific cultural knowledge related to the image on the part of the viewer. Panofsky labels this level of finding meaning as the ‘iconographical analysis’ of an image.
- Intrinsic/symbolic content (“What does this all mean?”) is the information available in an image representing the historical environment, including intentional (and unintentional) symbolical values related to the specific characteristics, technique and culture of the image and its creator. This level of interpretation is Panofsky’s "iconographical synthesis’. Modern interpretations of this symbolic level of meaning include McAllister’s (2013) definition of “visual reasoning” as literal depictions of the objects of the reasoning, as well as those characteristics which constitute “metaphorical” depictions of objects. (p. 29)
The matrix of Panofsky’s strata of meaning in images was first applied to examples of symbolism in classical, medieval and Renaissance art in the early twentieth century. Since 1955, when Panofsky’s lectures were published in English for the first time, this matrix has been used to examine a wide variety of fine art images and is valuable for art history students who wish to investigate the historical and cultural details within images from unfamiliar environments and time periods. Moxey (1986) notes that
The system of checks and balances that characterizes Panofsky's iconological method has proven to be the door through which it has become possible to essay an interpretation of works of art that does justice to their complex historical particularity. [This] method still offers the discipline one of the most sensitive approaches to the understanding of the art of the past. (p. 272)
Panofsky’s matrix has continued to be used when deciphering visual metaphors in the form of allegorical symbols such as the personifications of moral virtues and human attributes found in ancient, Renaissance, and Baroque painting and sculpture. The matrix can also provide a useful way to describe simpler contemporary images as shown in Table 2.
Table 2
Examples of Panofsky’s Three Strata
illustration not visible in this excerpt
At its most elemental, iconology is the study of logos (the words) of icons (the images). Iconology has been defined as the “notation of imagery” and the “rhetoric of images”: ways of studying the tradition of writing about pictures, combined with looking at “the ways in which images seem to speak for themselves” (Mitchell 1986).
Iconology is not only the identification of visual content, but also includes the analysis of the meaning of visual content. Panofsky described his new approach as “the branch of the history of art which concerns itself with the subject matter or meaning of works of art, as opposed to form” (Panofsky 1972).
Van Straten (1986) notes that iconology should not be seen as an all-comprising method or approach toward art objects for several reasons, including the fact that Panofsky believes there are categories of subjects within the visual arts which have no "secondary" subject matter. He proposes a “revised scheme” which introduces several variations the original model as detailed in Figure 10.
illustration not visible in this excerpt
Figure 10.Van Straten’s proposed revision of Panofsky’s three strata
Woo (1994) notes that iconology as an interpretive tool has a variety of limitations, including the built-in problems of using written text to describe visual objects. Additionally, Panofsky’s symbolic/intrinsic level of interpretation contains a variety of pitfalls for traditional index creation, specifically for individual catalogers assigned to identify meaning in particular images within a large non-personal image collection. When trying to assign symbolic or intrinsic meaning to an image, direct correspondence between a complex concept and a specific term is generally not well-defined. Woo suggests that traditional indexing vocabulary itself has further limitations, since large professionally indexed corporate image collections “do not attempt to interpret ‘symbolic values’ and thus there is no available vocabulary for it” (p. 5).
Social tagging and folksonomy
Traditionally, there has been a divide between the people who generate information and the people who consume it. This divide still exists on many levels, of course, but individuals can now sometimes choose to simultaneously generate and consume information, to become both creator and audience, interchanging the role of cataloger with that of patron by actively indexing their own personal collections, using their own choice of language in the process.
This duality of roles available to the social digital image collector is rooted in the ability of a single user to assign a meaningful text label to a distinct online item. Naming (and renaming) “is a means of restructuring reality. It imposes a pattern on the world that is meaningful to the namer” (Olson, 2002, p. 4). As new information sources became more widely distributed in the 1990’s, users began to assign their own identifiers (widely referred to as “tags”) which Wichowski (2009) notes “unwittingly gave rise to a new information organization system” known as social tagging or folksonomy (p. 3). Folksonomies were seen as one approach diverging from traditional classification, allowing users to create relatively brief pieces of text associated with a specific item in real-time, based on a decentralized cooperative view of the user community.
Quintarelli (2005) notes that as the World Wide Web expanded, classification schemes were needed which could adapt to increasingly unstructured and nonhierarchical collaborative collections. Folksonomies were a vital part of the emergence of metadata (information about information made available by the creator of publicly shared materials) which became an contributing factor to the contemporary user’s ability to freely name and add meaning to social collections. For the sake of clarity in this project, the term metadata is limited to the more rigorously controlled back-end content activities such as citation analysis, link structure studies, and recommendation systems (such as Amazon’s customer reviews) (Mathes, 2004). In contrast to a focus on pure metadata, systems implementing variations of folksonomy tagging, including Pinterest, tend to highlight a relatively unrestricted vocabulary as well as a generally decentralized and collaborative view of direct and personal collection management.
A comparison of the characteristics of traditional taxonomies, folksonomies and Pinterest’s social curation process as shown in Table 3 highlights some important differences among these approaches:
Table 3
Comparing Characteristics: Taxonomies, Folksonomies and Social Curation
illustration not visible in this excerpt
“Big, messy, organic” data sets
Since the first folksonomy was observed, researchers have been intrigued by tagging behavior. Wichowski (2009) suggests that tags conform to power laws, where a few tags are used by a large population of users. Mathes (2004) notes that tags on particular types of folksonomies (such as http://www. del.icio.us ) are primarily from the users of documents that were written by someone else, while tags on other types of folksonomies (specifically Flickr) are primarily used by individuals to manage their own digital images, with the majority of users tagging photos they created themselves. Tonkin et al (2008) found that people seem to use different tags if sharing content with a community as opposed to identifying content for self–use later. Both Cattuto (2006) and Schifanella et al (2010) attempted to map some universal tagging behavior activity patterns but concluded that “Uncovering the mechanisms governing the emergence of shared categorizations or vocabularies in absence of global coordination is a key problem with significant scientific and technological potential” (Cattuto, p. 1464).
Mai (2011) introduces the entrepreneurial aspects of do-it-yourself tagging and suggests that encouraging this kind of innovative user activity adds the unique advantage of allowing “a plurality of viewpoints and opinions” while continuing to provide an overarching organizational framework. (p. 7) Kim, Breslin, Chao and Shu (2013) propose that allowing users the ability to tag increases the strength of ties between group members and creates an “object-centered sociality” which “mediates the ties between them and serves to indicate why people affiliate with others or participate in communities.” (p. 252)
Dismissing folksonomies and collaborative social image naming practices has a long history among catalogers concerned with the effort and time needed for creating and maintaining viable user-generated tag systems, but Dotsika (2009) states, “Against all odds and the belief that collaborative tagging is useless and chaotic, it [tagging] has proved to be effective for organizing personal and corporate information, blog searching, facilitating innovation and enabling the discovery of marginalized information such as in the area of the so-called long tail” (p. 409).
Shirky (2005) describes tagging as a more “organic” way to handle information, and suggests that “the strategy of tagging-free-form labeling, without regard to categorical constraints-seems like a recipe for disaster, but as the Web has shown us, you can extract a surprising amount of value from big messy organic data sets” (p. 44).
The numerous ways in which Pinterest user-curators appear to be adapting language to create names for their image collections, especially in the midst of the big, messy, organic data sets that comprise Pinterest, seems to support the user-curator attraction for categorizing “marginalized” content, even if the categorization is invented by each user-curator for their own collecting purposes.
Visual categorization and interindexer consistency
One measure of visual categorization efficiency is the degree of interindexer consistency: how frequently the index terms chosen by indexers overlap. Shatford Layne (1994) summarized various research done on interindexer consistency when working with image collections and concluded that “There will be interindexer consistency on certain aspects, perhaps the principal and more objective aspects, of the subject of an image, but that there will be less consistency on secondary and “subjective” aspects” (p. 585). Somewhat less optimistically, Winget (2004) claims that “Providing subject access tends to be too complex from an inter-cataloger consistency standpoint” (p. 88). Little current research has been published examining interindexer consistency in large uncontrolled public digital image collections, although applying Panofsky’s matrix of image meaning should allow a limited examination of interindexer consistency as a byproduct of data collection in this project.
Automated annotated image data
Non-human content identification in image indices has thus far not been proven to be the most effective method to increase the usefulness of a large image collection to a given user. Hanbury (2008) compares methods of improving the automated metadata generation for images, including automated image annotation and object recognition, and then notes that “Automated content description and annotation algorithms being developed cannot yet be expected to perform at the same level of detail as a human annotator.” It is possible that the user-curator pin naming language games developing in Pinterest could eventually provide clues to a more flexible or inclusive human-based method to investigate image identification as it evolves.
Cognitive economy and perceived world structure
One goal of effective visual categorization is to supply viable information to a user with a minimum of effort. Rosch and Lloyd (1978) reinforce Panofsky’s first level of subjective meaning: “There is generally one level of abstraction at which the most basic category cuts can be made” (p. 5) and then examine the aspects of image categorization in detail, equating categories with the number of objects that are considered equivalent, examining how users perceive structures in the real world, and suggesting the principle of “cognitive economy”:
The task of category systems is to provide maximum information with the least cognitive effort… Thus maximum information with least cognitive effort is achieved if categories map the perceived world structure as closely as possible…These two basic principles of categorization, a drive toward cognitive economy combined with structure in the perceived world, have implications both for the level of abstraction of categories formed in a culture and for the internal structure of those categories once formed. (p. 82)
Triads of visual categories: Basic, subordinate and superordinate
Rosch (1978) proposes three levels of categorical abstraction which users may employ when associating selections of “basic level objects” with the realities of actual observed environments.
The basic image category as shown in Table 4 is defined by Rosch as the most “inclusive” layer of classification because images here share the highest number of common attributes. A basic image category may include a wide variation of images which are all unique from one another, but which all fit multiple common requirements of being identifiable as a car or a chair based on a high number of common “car” or “chair” attributes.
Table 4
Rosch’s Basic Image Category
illustration not visible in this excerpt
The superordinate image category is one level more abstract than the basic category, as detailed in Table 5. Images within this category commonly share only a few attributes. Rosch (1978) uses the example of vehicles and furniture to show how these more abstract categories allow fewer shared attributes among member images. Images within the category of vehicles (superordinate to cars) tend to have fewer common attributes than do images within the category of cars (the basic category).
Table 5
Rosch’s Superordinate Image Category
illustration not visible in this excerpt
A subordinate image category as shown in Table 6 contains images which are subsets of the basic category. These individual images tend to share many overlapping, predictable attributes with other member images in this distinct category.
Table 6
Rosch’s Subordinate Image Category
illustration not visible in this excerpt
Rosch (1978) summarizes the three levels of categorical abstraction: “Very few attributes are usually listed for superordinate categories (‘furniture’). Significantly greater numbers of attributes are assigned to basic level objects (‘chairs’). Subordinate level objects ( ‘black yew splat-back George II 1740 Windsor armchair’) do not have significantly more attributes assigned than do basic-level objects.”
In a study conducted by Rorissa and Iyer (2008), user assignment of image category labels was found to generally be generic, interpretive and to belong to the superordinate to the basic level.
In this project, patterns emerged in the pin names collected showing few image names had characteristics of the generic superordinate category (‘furniture’). Significantly greater numbers of primary pin names are assigned to basic level objects (‘chairs’) while similarly larger numbers of secondary pin names fit into the more specific, detailed subordinate levels.
Two stage (primary versus secondary) subject matter categories
Wingett (2004) suggests that viable image indexing might be accomplished using only two basic divisions: “primary” subject matter (objective description including “form, color, and pattern of visual images as a representation of the real world”) and “secondary” subject matter (“identifying cultural symbols based on the prior identification of primary subject matter.”) The similarities in this two-part approach to Panofsky’s first two tiers are noted by Wingett. (p. 4)
Markey (1983, p. 211) proposed a similar two part “primary-secondary” indexing scheme as did Krause (1988, p. 10) who applied the terms “soft” and “hard” to the secondary and primary designations.
Jaimes and Chang (2000) propose a ten-level structure to provide a systematic way of indexing images, but their extensive approach ultimately reverts to binary evaluations of meaning based on “syntax” (the descriptions related to color, texture and other “primary” attributes of an image) along with “semantics” linked to “objects and events” (p. 156).
All of these two-stage indexing systems (objective description followed by interpretive observations) neglect the third step Panofsky proposes: the recognition of “deeper” intrinsic, cultural-historic symbols and concepts, including “essential human tendencies” and “representations not explicitly intended by the image creator” (Panofsky, 1939, p. 77). Identifying the intrinsic meaning of an image name may not prove viable within the limits of this project but an attempt to identify this level of meaning will be made, if only to further highlight which types of image iconology seem to continue to elude quantification.
Defining image attributes
A central difficulty in understanding how human image perception occurs is rooted in human language itself. Both written and spoken words have proven to be a barrier to accurate descriptions of what people think they see.
Yoon and O’Connor (2010) note that because images are not easily represented with words, there can be no “simple algorithmic relationship between images and words” (p. 761).
Studies related to how users appear to interact with images highlight the difficulties of limiting human visual responses to pre-defined terms. A variety of studies have evolved attempting to delineate how humans interpret and react to visual stimulation, particularly when “similarity” of images must be detected and weighed by searchers. (Beach, 1964; Tversky, 1977)
Rosch and Lloyd (1978) state that users will apply attributes based on the way they view their current environment: “One influence on how attributes will be defined by humans is clearly the category system already existent in the culture at a given time” (p. 4).
Shatford Layne’s image attributes
Shatford Layne (1994) proposes a matrix for examining the specific attributes of any give image as shown in Table 7:
Table 7
Shatford Layne Images Attributes Used In This Project
illustration not visible in this excerpt
The Shatford Layne matrix provides a wealth of combinations for analyzing meaning in images. Not every attribute exists in every image, but Pinterest user-curators may be combining aspects of these attributes as they create original names for their images. For example, analyzing meaning in an image using Shatford Layne’s image attributes as illustrated in Table 8 provides the following information:
Table 8
Example: Applying Shatford Layne’s Image Attributes to an Image
illustration not visible in this excerpt
Creating a matrix using a combination of Panofsky’s subject matter categories, Rosch’s levels of abstraction and Shatford Layne’s attributes provides a framework to begin examining Pinterest image names, and to analyze the density and complexity of the language being used by user-curators when naming their large, personal digital image collections. The Panofsky/Rosch/Shatford-Layne matrix is used as a tool for identifying meaning in a pin name, in a way similar to the individual approaches traditionally used to describe meaning in images as shown in Table 9. (Given the uncontrolled nature of image naming within Pinterest, it is probable that any selected pin name may reflect a range of properties from the Panofsky/Rosch/Shatford-Layne meaning matrix. Since image retrieval is not necessarily the main purpose of pin name creation in Pinterest, it is possible that user-curators are evolving particular language patterns and devising personalized naming systems which may not become apparent even after extensive observation of naming activity.)
Table 9
Applying the Panofsky/Rosch/Shatford-Layne Matrix
illustration not visible in this excerpt
User behavior in image naming
Before large numbers of people had frequent access to online digital image collections, researchers were limited in the ways they could observe image file naming behavior. Previous studies collected and classified user image naming behaviors while users attempted activities such as retrieving pictures based on text narrative, captioning images, and annotating still photographs (Shatford, 1984; Hibler, Leung & Mwara, 1992; O'Connor, O'Connor, & Abbas,1999; Schreiber, Dubbeldam, Wielemaker, & Wielinga, 2001; Hollink, 2004; Hanbury, 2008).
Because digital image user-curators increasingly need to name their images outside of (and sometimes in place of) traditional static indexing formats (including flexible social media tools such as YouTube playlists and Pinterest boards) indexers who work exclusively with digital image collections have started to consider the implications of crowd-sourcing of search entomologies and other more collaborative approaches to constructing indexing tools (Harpring, 2010; Feinberg, 2012).
Sandhaus and Boll (2010) considered how the semantic web might provide searchers with more options to retrieve images, specifically photographs and commercial images which may need to be accessed repeatedly or in high numbers. However, even in the presumably more flexible environment of digital image collections, the contrasting needs of the user versus the indexer remains an ongoing issue. Harpring (2010) notes a specific problem between vocabularies intended for digital image retrieval “to accommodate nonexpert searches” and vocabularies used for indexing, in which the assumption is that “warrant, correct usage, and authorized spelling of terms” is the over-riding concern of the indexer. (p. 81)
Image name iconology: Tools for assigning meaning
Greisdorf and O’Connor (2008) state:
The problem with discussing meaning in association with images is that multiple definitions apply to the term. Meaning in the context of image engagement and complexity can stand for (1) the intended message of the image, (2) the expressed message of the image or (3) the signified message of the image…Often the only messages available to the image collector are the intended message based on the history and circumstances surrounding the creation of the image, or the expressed messages attached to the image as communicated by its creator and/or its critics. (p. 79)
For a variety of reasons, user-curators in Pinterest may not have access to the intended message, the expressed message or the signified message of the image creator when they name their images within their collection. Upon discovering that the history and circumstances of a collected Pinterest image are not available, how might the Pinterest user-curators assign meaning to an image?
This leaves the assignment of meaning to any given image almost entirely in the hands of the user-curator, who is not subject to controlled vocabularies, naming conventions or even the constraints of providing retrieval access for other users.
So where might a user-curator conceivably look for meaningful language to describe images? Traditional iconological tools exist for identifying symbols in fine art. Reference databases used by image collectors when identifying meaning in images include Groves Art Online, Oxford Art Online and the iconographic database Iconclass.
Iconclass
Within art history research, the evolution of large iconographic databases has encouraged the development of indexing terms related to fine art imagery. Iconclass (“a multilingual classification system for cultural content”) is a database used by researchers for a systematic overview of subjects, themes and motifs in Western art. The project began in the 1950s and after six decades of gradual technical evolution, the Iconclass 2100 Browser launched in 2009. As of 2014, the system contains 450 “basic” categories broken into ten “main” categories. There are approximately 28,000 hierarchically ordered definitions, with each containing a unique “notation” along with a text description of the iconographic subject. The Iconclass index contains roughly 14,000 keywords used for locating the notations, such as the example shown in Figure 11.
illustration not visible in this excerpt
Figure 11 . Iconclass keyword search example
Iconclass is generally used for academic projects such as classifying the master print collections of the Gemäldegalerie, Berlin and the German Marburger Index but the tools have also been useful outside of pure art history, including on sites like Flickr. (RKD, 2009)
Iconology indices such as Iconclass are interesting practical examples of the strengths and weaknesses of a system constructed from words when used to organize and describe particular aspects of a given set of images (Couprie, 1978, p. 34). It is possible that new, adaptive uses will be discovered for such extended text systems when applied to large, international public digital image collections. However, when millions of images from cultures unfamiliar with the Western canon of visual art analysis are suddenly included in a collection, will such a narrowly constructed index still have value or will Iconclass choose to adapt in some other way?
Elkins (1999) suggests the problems with these kinds of systems are based in “the dual sense of pictures” in which viewers are “conflicted about what they take pictures to be.” Writing about images is basically broken into two opposing components in this view: writing that describes an image as a “pure art object” or writing that allows an image to be a “substitute for writing” , which then makes the image a “carrier of determinate meaning” (p. 110).
Wittgenstein’s rule-guided language-game analysis
Wittgenstein’s rule-guided language-game analysis is “a specific way of looking at linguistic practices as operations governed by a set of discrete concepts that the analysis must seek to express” (Xanthos, 2006, 212). Although Wittgenstein provided no single definition of his term “language game”, a generally accepted central aspect of this concept concerns socially shared ways of using semiotic signs, of signifying and of representing. Wittgenstein used the examples shown in Figure 12 to illustrate the sense of "the multiplicity of language-games":
illustration not visible in this excerpt
Figure 12. Wittgenstein (1958) Philosophical Investigations
Blair (2006) expands Wittgenstein’s concept of language games by emphasizing how closely all human language is “defined” by use. Language games evolve from activities that require a particular sort of communication. Over the course of his career, Wittgenstein changed his views on the importance of “determinacy” or the precision by which meaning can be defined. Wittgenstein’s final view suggested that determinacy (such as strict inflexible permanent glossaries, specialized vocabulary lists and detailed definitions) were not only not vital for meaning to be shared, but were probably not even needed. Normal language, as used by ordinary people sharing particular tasks, is capable of carrying all of the meaning that is required to support the activities involved: “We can make language very precise if we want, not by bringing out some kind of hidden logical underpinning, but by looking at the context, circumstances and practices in which language is used” (p. 17).
The idea that language is “not so much a collection of “meanings” but something that can be used to do things” (Blair, 2006, p. 221) helps explain how Pinterest user-curators are evolving pin naming “rules” based on day-to-day activities and practices:
If we want to understand the meaning of a sentence we must look at how it used. This is the most basic level of analysis that we have in language…We cannot generally reduce ordinary language to more primitive components of meaning without losing some of the meaning that emerges from its use… Language needs no central authority to control usage. It needs only day-to-day interactions of its native speakers to establish and retain its meaning. (Blair, 2006, p. 14.)
Pinterest user-curators appear to be creating language games as part of the activities involved in pin name creation. Analyzing the pin names generated may reveal some of the types of language games generated by the activity “pinning an image” and “naming a Pinterest pin.”
Observed existing non-user attitudes related to the pinterest site in general
A variety of publicly published opinions from non-Pinterest users were observed during data collection for this project during 2014. Some highly visible attitudes toward Pinterest in 2014 included aspects of these four perceptions:
1. Pinterest is (a) only used by women, thereby (b) reducing its technological sophistication and importance when compared to “real” technology sites used by other demographic segments online.
2. Pinterest is a threat to feminism.
3. Pinterest primarily exists to sell products, principally to women.
4. Pinterest should be studied and discussed as if it were similar to other “social media” sites including Twitter or Facebook.
While this project is not focused on examining these attitudes in detail, it is important to note that these kinds of reactions to Pinterest existed as of October 2014. Although the Pinterest site itself does not appear to be blatantly oriented toward any single demographic, contains no commercial mechanisms (shopping carts, wish lists, credit card sales) and shares few observable characteristics with Facebook or Twitter in either content, user base or delivery approach, the emergence of these attitudes about the site are important to consider and warrant a brief discussion in this literature review.
Pinterest Is (a) only used by women, (b) reducing its importance.
The two aspects of this attitude which require examination are the claim that women are the principal users of Pinterest and the related claim that technology used by women is inherently less sophisticated than technology used by other demographic groups.
Claims in 2011 which implied that Pinterest was globally used primarily by women became entwined with the mythology of its record-setting growth. This continuing perception (combining a previously untapped market discovering a new “killer app”) may have contributed to unavoidable, chronic and exaggerated misconceptions about what Pinterest is, and what people typically do on the site.
Despite vigorous promotion as a direct marketing tool for women, irrefutable evidence that any one particular demographic comprises the principal user of Pinterest can be difficult to find. Determining anything specific about Pinterest users from self-defined profiles is challenging since pinners retain a high degree of anonymity. Pinterest does not require (or encourage) users to reveal gender identity, and users can choose to present a relatively blank personal profile., displaying only a self-generated user name. Users are not required to self define themselves in any way, and can create elaborate image collections with essentially no identifiers beyond their required user name, which can be purely nonrepresentational and even nontextual.
Verifiable attempts at harvesting reliable data about users (including gender) from their names, activities or self descriptions appear to have had relatively limited results. For example, in Mittal's dataset of over 3 million users, less than 18% included profile descriptions of any kind (information such as age, marital status or contact data): “From our user profile dataset of 3,323,054 users, we found that only 17.73% of users had profile descriptions. The description field is where users reveal private details such as age, marital status, personal traits, email IDs, phone numbers, etc” (2014). Mittal then attempted to study those Facebook users who had both indentified their gender and linked their Pinterest accounts to Facebook, but this connection is suspect, as any similarities in activities between Facebook and Pinterest users remain undefined, and using gender as the only identifier between the two sites does not provide any measurable set of characteristics related to user behavior.
While attempting to prove that one gender uses Pinterest more than another, Moore (2014) mapped user-provided Pinterest names to US Census Bureau data, stating “Pinterest doesn’t share gender data publicly, but they do share users’ names. About 75% of users supply a name that maps to a name as recognized by the US Census Bureau. We mapped name data to census data to arrive at gender.” This is another intriguing but questionable approach to identifying user characteristics, since no information is given regarding the number of names mapped or the basis for determining which names were irrefutably gender-specific. Additionally, as in other social media platforms, an unknown percentage of user names appear to be generic, invented, nonsensical or non-content-based ( such as User123 or SwimTeam2014).
The perception that women are the main users of Pinterest, whether accurate or not, leads to the dismissal of the site by some non-users. Tekkobe (2014) notes that “real” technology users have “reinforced the hegemonic technology narrative that women only consume technology, while men make technology, arguing that Pinterest is ‘what happens when you empower people not to create, but to share’ (p. 382).
Along with raising the question of what constitutes valuable or “real” activity online, Tekkobe examines what happens when content and activity on a particular site is judged by nonusers as less relevant or less “technical”, observing that the role of arbitrator (deciding whether Pinterest is a valid application of networked technology) has been voluntarily assumed, by default, within a set of self-defined “technical” Internet users. In a tongue-in-cheek “attack” on Pinterest in 2012, a tech blogger at the industry-watch site Complextech.com announced his opinion that Pinterest is “ The Most Regrettable Social Network Yet.” While this blog entry is clearly aimed at creating an artificial “controversy” for a particular commercial site, the blogger repeatedly emphasizes the personal aspect of Pinterest which most devalues it in his opinion: “On Pinterest, one merely co-opts and shares images. This, in a soft light, could be viewed as a kind of generosity. But the focus here is as much on the pinner as it is on that which was pinned” (Ugwu, 2012).
When discussing whether or not Pinterest’s content is a worthwhile use of the technologies’ affordances, Tekkobe states “These voices uncritically position themselves as arbitrators of the value of Pinterest as a social networking site, and the worthiness of the site content as saved and shared by the Pinterest community. These privileged voices assess Pinterest as a community of women who indulge in silly feminine daydreams rather than engage in the serious work of valuable content creation” (p. 5).
Examples of this level of dismissal appear in some research related to Pinterest. A small number of ostensibly credible academic research reports contain drastically simplified summations, undefined assumptions and remarkably small samples given the enormous user population:
Authors of this work found that females on Pinterest make more use of lightweight interactions than males.(Mital, 2013, p. 2)
Our participants prefer pins that catch their eye, are easily understandable, or are in a particular style. (Linder, 2014, p. 9)
Our analysis was based on a partial subgraph of the Pinterest network, and suggests that Pinterest is a social network dominated by “fancy" topics like fashion, design, food, travel, love etc. across users, boards, and pins…Since there is not much prior work on Pinterest, we do not have enough academic literature to claim that our dataset is representative of the whole Pinterest population. (Mital, 2013, p. 11)
Ultimately, the issues to consider in the case of who uses Pinterest are (a) why one largely undifferentiated demographic (“women”) have been strenuously promoted as the principal users of a site which does not emphasize or volunteer any form of user data identifying that demographic and (b) how this assumption affects the evaluation of the site by various arbitrators of technical and cultural value.
Pinterest is a threat to feminism
Media focus on Pinterest’s reported use by women seems to have encouraged political reactions from various groups, both demonizing the site and extolling its expected commercial potential. As the number of Pinterest users grew in 2012, a perception of misogyny, principally rooted in claims of negative body image stereotyping, began to surface among various potential and existing user groups. Machirori summarizes one aspect of this perception of Pinterest’s genderfication:
Pinterest is a social media site that has largely leveraged itself through appealing to women’s perceived normative domestic pursuits, such as cookery and fashion. It has come under fire from some feminists for peddling ‘kitchen porn’, placing unrealistic expectations of domesticity and beauty on women and therefore reinforcing patriarchy though ‘trivializing’ women’s interests and catering narrowly to the private sphere of women’s interactions. (Linder, 2014, p. 4)
It is possible this equating of Pinterest with “kitchen porn” is based on the first experience a new user may have when encountering the unfiltered main grid as a pinner for the first time. When a new user initially opens Pinterest, they view all images most recently posted by all users. This “open” flood of all posting activity by all users is not fine-tuned to the curator and is not limited to any particular topic or board, but is a fully randomized real time snap shot of all posting activity taking place at that instant. It is possible for a new user to assume this flow represents all of the content available in Pinterest, when in reality this open login view represents only an uncontrolled random snapshot of all data being uploaded at a given moment. This uncategorized flow of unrelated images is immediately refined as soon as the new curator chooses to follow any given pinner’s images.
Additionally, studies which support the ‘kitchen porn’ theory of Pinterest content appear to base their conclusions on surveys of the most “popular” pins or users, reducing the complexity and depth of 70 million user experiences to the top eight pinners, for example. Simplifying a multifaceted image collection site, particularly one using 31 languages and including millions of curator-users, by reducing usage to “popular” participants suggests that the use of any collaborative site can be evaluated by averaging the heaviest users. This averaging approach does not take into consideration the size, depth and relevance of the Pinterest curator-user community and seems to provide a “quick and dirty” method of supporting preexisting conclusions about Pinterest content, as well as Pinterest users.
Machirori suggests that Pinterest user-curators are being manipulated “by men” and details the perception that women remain content consumers, ( a pejorative role) while men retain the title of content owners (a more desirable position to attain):
The arguments against women’s wholesale uptake of Pinterest echo the body of western feminist rhetoric that places a premium on women’s movement from more private and domestic spheres of interaction into more public, male-dominated and politicized spaces. The debate is therefore not only about whether women own social media and technological innovations. But it is also about what they are using them for. Indeed, have Facebook, Pinterest and other sites provided the emancipatory cyberfeminist promise for women to explore the fluidity of their identities? Or have they merely served to further entrench women’s position on the margins of public discourse? In essence, it appears that a limited range of interests and pursuits have been packaged and marketed to women, by men, so much so that the dominant use and consumption of social media lies with women, while ownership and innovation remains the preserve of men” (p. 112).
This type of political rhetoric, particularly when broadly applied to a largely uncontrolled public image collection site, does not appear to be based on any observable behaviors of user-curator. Based on the data collected for this project, using the publicly available site resources in 2014, there was no observable “limited range of interests and pursuits packaged and marketed to women” in terms of vocabulary, categorization or content. The undefined open Pinterest tool set is available to all users and contains no discernible political or commercial messages. All users default to generic undefined categories until they intentionally self-label themselves, their pins or their collections. Pinterest, as a web entity, promotes no apparent or conscious focus or agenda in site design, language use or tools.
It is possible, given the commercialized media surge surrounding the site, that researchers have approached Pinterest without questioning who actually uses the site, additionally inferring that detectable usage patterns can be accurately based on only (a) gendered definitions or (b) the usage patterns of the most “active” users, based on number of pins, reducing the complex activity of millions of self-directed, non-socially oriented user-curators into a few simplified “average user” categories.
Pinterest is primarily for selling products, principally to women
Compared to many ordinary commercial sites, including Facebook and Twitter, the Pinterest interface itself is not well designed for selling products. While “popular” random rankings on the initial “everything” upload page may display a predictable number of images related to weight loss, cute shoes and recipes for cheese biscuits, the public forum of the login page does not reflect the content each user-curator chooses to recognize. Every user automatically customizes which pins they view (or do NOT view) as soon as they begin to participate by pinning and following. Additionally, as of 2014, there are no uncontrolled “posts” from vendors or from Pinterest inserted into users activities (such as ads posing as “news” items which automatically occur within Facebook news feeds). In fact, this uniquely reduced intrusion from outside commercial interests allows user-curators to fine-tune their displays to include only images they are interested in, to a remarkable degree. Being “ad-free” has been both a revered and a denigrated state for Pinterest since it’s launch, and apparent attempts to place more blatant purchasing tools on the main public Pinterest landing screens have so far failed.
In an interesting twist, a project launched in 2014 is attempting to “crawl” Pinterest to allow data extraction from Pinterest’s millions of users. Four researchers from the University of Toronto have developed the SerpentTI analytic system, specifically to “extract” user data from Pinterest. They are deploying more than 200 processes across a cluster of 16 machines to handle each of the different crawling tasks. As of July 2014, SerpentTI systems have crawled over 3 billion pins, and can update profiles of 96 million boards in under 45 days.(Cheng et al, 2014). Their published description of the project includes suggestions on how this data might be used commercially, including harvesting user data based on expressed interests, pinning “authority” and other implied demographics. It will be intriguing to observe if the data collected by these systems contains viable commercial contact information, or whether the unpredictable nature of language use in Pinterest will stymie these types of aggressive bulk crawling and extraction.
Pinterest should be studied and discussed like other “social media” sites
Pinterest is fundamentally different from other social media sites. It shares few traits with Facebook or Twitter, for example, although it is regularly discussed as if it were the same style of user experience. Although all content on Pinterest is provided by other members of the community, Pinterest users are only as social as they prefer to be, and can tightly control not only what they view, but what information they chose to reveal about themselves. Linder (2013) notes:
Despite the public nature of boards, Pinterest users do not feel scrutinized as they pin. They are more interested in the pins themselves than where they came from, or who found them. This contributes to the feeling of anonymity in Pinterest users, which serves to dampen the kind of extrinsic motivation that is detrimental to creativity. (p. 5)
The purpose of Pinterest is not to make friends, but to share images. The high level of anonymity makes Pinterest different from other community-driven sites, and contributes to its uniqueness for users. While Pinterest images are available for everyone to search and share, and are posted by other members of the community, there is no incentive for users to connect or interact with each other. This suggests that curators primary impetus is to create and enlarge their own image collections: “Comments on Pinterest are rare, usually occurring among friends and family. Social actions mostly go unnoticed, removing inhibitions typically experienced when authoring social media” (Mital 8).
“Finding” other people and then forming social connections (beyond those which lead to additional collections to be repinned) is not a central focus of activity, since ordinary social connections can be formed in many alternate sites, whereas sharing images in a concentrated way can only be done on Pinterest. The closest related “social” sites are image-based services such as Flickr and image sharing sites such as Imgur, although neither of these sites begin to rival Pinterest in user loyalty or ongoing growth rates.
Studies of Pinterest behavior based on usage statistics show that few users participate in “liking” or “commenting” on images (available tools which allow communication directly between users) but a high percentage “repin”, (which allows a user to add the selected image to their personal collection). Mittal uncovered some intriguing aspects of the Pinterest dataset analyzed in 2013 as shown in Table 10:
Table 10
Pinterest User Examples of Limited Social Interactions
illustration not visible in this excerpt
Zhong et al. (2014) concludes that Pinterest users value the social aspect of the service principally in terms of how it helps them find people with similar tastes in pictures: although new Pinterest users tend to try the “friend finder” tool to copy close friends they know from established source networks like Facebook, when they discover new friends on Pinterest with shared visual preferences, they tend to prefer those new Pinterest users with similar tastes (p. 312).
CHAPTER 3 MATERIALS AND METHODS
Introduction
This chapter describes the data collection and analysis method used along with a discussion of the methodological issues involved, including scope and limitations, the expected results and a summary.
This study is exploratory and descriptive in nature, using the Panofsky/Rosch/ Shatford Layne matrix as a framework to organize data, while Wittgenstein’s language-game analysis provides a central structure for thinking about the data captured within the matrix.
Based on Crotty’s three assumptions, a constructivist worldview is taken:
- Human beings actively develop meaning as they engage with their world.
- Context and setting is central to understanding behavior.
- Meaning is most efficiently generated from data collected in the field (Cresswell, 2014, p. 9).
Essentially, such a constructivist worldview suggests that Pinterest users may be adapting language to suit their needs, that the specific environment provided by Pinterest may be spurring particular types of user behaviors and that the most valuable information in this study may be gleaned from the user language collected, rather than from any outside interpretation or analysis.
An exploratory, descriptive approach was selected in order to identify and compile approximately 700 pin names, followed by assignment of the language used in each name to a strata of the Panofsky/Rosch/ Shatford Layne matrix. Language game analysis was then completed, with conclusions proposed based on the combined results of the matrix assignments and the language game observations.
A focus on qualitative research methods in this study will allow information to emerge from text directly generated by Pinterest participants, in the “natural setting” of Pinterest itself. Data collection will occur on the live site without a need for interviews or predetermined specific questions. Any interpretations of the meaning of the data, including themes or patterns that emerge, were made from the data sets, maintaining a central focus on observing how people were using language when naming visual images in large personal digital collections. The context of the unique community being studied (Pinterest) was integral to the user behavior being explored.
Data collection approach
The process used to collect the pin names for this project was made up of these steps:
1. Create 18 unrelated search terms, broken into six unique sets of Panofsky’s three strata of meaning.
2. Search Pinterest using each of these 18 terms, capturing 40 images for each term.
3. For each search term, save all related images, pin names and creator names
4. For each search term, compile all pin names and save into a spreadsheet.
5. Note examples of language games including puns, word art, alliteration, malapropisms, spoonerisms, obscure words, rhetorical excursions, oddly formed sentences, ASCII art, emoticons, double entendres, unique uses of upper and lower case fonts abbreviations, and malformed sentence/word phrases
6. Interpret any patterns or themes using Wittgenstein’s rule-guided language-game analysis
7. Suggest potential conclusion: How do the pin names collected correspond to each of the levels in the Panofsky/Rosch/Shatford Layne matrix?
Data collection method
Pinterest is a public site and users are routinely made aware that all activity is socially shared. This public aspect of the research site allows observation of random activity to potentially yield a full spectrum of user behavior.
Because this project is exploratory in nature, a relatively small sample size was developed and the intentionally restricted sample size did not warrant controls for intercoder reliability.
The researcher was the primary instrument in data collection, rather than any remote mechanism. Observation of activities at the research site was achieved using 18 English search terms to collect a cross section of non-repeating images. The search terms were organized as six independent data sets , containing three search terms per data set, with each term purposefully selected to represent an approximation of one of Panofsky’s primary, secondary or intrinsic levels of meaning. The search terms selected are shown in Table 11.
Table 11
Final Search Terms
illustration not visible in this excerpt
Image collection
Each term was used as a search trigger in the default public Pinterest search window, capturing forty images for each search. The first forty non-repeating images produced by each search term were compiled, along with the pin creator information for later verification. The observational protocol for the alpha data set consisted of populating Word documents with all images captured under each search term. The observational protocol for the subsequent beta data set consisted of capturing the search results in a set of individual Pinterest boards, restricted to pins collected during this project. Additional field notes in Word were compiled while conducting observations during both data collection procedures.
The limit of forty images per search term was chosen since one “endless scroll ” Pinterest default display at 1200 x 800 resolution tends to yield approximately four rows of ten images each, as seen in Figure 13.
illustration not visible in this excerpt
Figure 13. Pinterest default display: Example of “endless scroll”
Duplicate images were discarded until all forty images for each of the 18 search terms were unique. Duplicate creators (pinners) were discarded until all images were created by unique users. Unique images with no names were discarded.
Name collection
For each of the 18 search terms, forty returned images were saved and compiled, along with the creators user names for verification purposes. For each of the saved images, the text from the related pin names was compiled into a spreadsheet.
Each collected image has two potential user-curator-designated names: a board name - assigned entirely by the user although Pinterest provides a set of default board topics to adapt or ignore - and an individual pin name (unique to that image, always assigned by the user-curator). Both names can be left blank by the user-curator. All non-blank pin names were saved in their entirety for each search result and the language used in each pin name was compiled for analysis into Excel worksheets.
The inclusion of board names (as well as pin names) in this project was determined to exceed the available time constraints for this study, but the analysis of board names related to the captured pin names may be revisited in the future as an additional data resource.
Data analysis: Panofsky/Rosch/Stratford Layne matrix
Each pin name was assigned to one level of Panofsky’s strata of meaning, based on the language used in the name.
If Rosch’s basic, subordinate and superordinate image categories or Shatford Layne’s biographical, subject and relationship/role attributes were strongly apparent in a pin name, that name was annotated with these related characteristics.
Wittgenstein’s Rule-Guided Language-Game Analysis : Observed Forms In Pinterest
Throughout the data analysis phase of the project, each of the compiled pin names was examined for characteristics matching particular types of “language games.” Aspects of Wittgenstein’s rule-guided language-game analysis were applied to each pin name, revealing examples of puns, word art, alliteration, rhetorical excursions, oddly formed sentences, ASCII art, emoticons and malformed sentence/word phrases. Additionally, instances of storytelling, personal comments, nonlinguistic or nonsensical names and abnormal word use were noted as potential additional language game types.
Semantic analysis of pin names
Blair (2008) emphasizes one of Wittgenstein’s central premises: words and their understood meanings are directly connected with the activity in which the word use occurs. “Meaning and grammar are not independent in natural language. Language does not operate as a kind of calculus” (p. 137).
To discover how meaning and grammar might be connected to specific pin names, a range of high level semantic analysis tools were used. In this project, each pin name was considered in terms of surface grammar, including whether or not the use of the words seemed to be under “normal circumstances”, whether a grammatically correct pin name became meaningless without its attendant image, whether the pin name taken out of context became nonsensical or misleading, whether the pin name shares a personal response to an image and whether the pin name comments on the subject natter of the image. Applying these semantic analysis tools to the 720 collected pin names produced a range of findings. The resulting research report includes the specific names tabulated and a description of any found data patterns including any patterns of language games present in the complete data set. The report is available in Appendix F.
Methodological issues
Following an attempt at an extremely large scale Pinterest study in 2013, Gilbert et al. noted that obtaining a truly random Pinterest sample is not possible without an application programming interface (API) from Pinterest, which would allow researchers to actually “drill” into the live site for large numbers of data samples:
The way we obtained a sample of Pinterest data was fairly labor-intensive and doesn’t offer a guarantee of randomness. For example, the fact that the average pinner in our sample had 1K pins suggests that we were sampling from the high end of the activity distribution. While we believe our results still stand, we obviously would prefer to obtain a random sample. Clearly the best way for researchers to be able to obtain appropriate data samples would be for Pinterest to publish an API. (Gilbert et al., 2013, p. 6)
A Pinterest API is not available as of September 2014. Naturalistic observation does not allow for scientific control of variables, so control for extraneous variables was not possible.
Scope and limitations
Pinterest users can delete or rename images at any time, and can also remove their active account at any time. This required saving all observed images and related data for future reference outside of the observed live Pinterest feeds.
Selecting effectively random samples of pin names without retrieving unmanageable numbers of duplicated names required manual analysis of a larger sample than the proposed 40 images per search term. Some search terms yielded relatively large numbers of duplicated names, and required additional rounds of searching to produce unique names.
Pinterest users can choose to remain anonymous in terms of reported demographic data so limited information related to age, gender, education or income can be deduced from categorization activity. This has no direct impact on this project, since the user-curators remain anonymous through the data collection, but the lack of demographic data on Pinterest in general has given rise to some misperceptions about the site. (See topics for further research below)
Expected results
Based on observation, the collected pin names in this project were expected to correspond to a widely dispersed variety of levels of the Panofsky/Rosch/ Shatford Layne matrix. Pinterest user-curators were expected to reflect the disparate user population with a range of aesthetic and linguistic interests, and the pin names created by this diverse group were expected to provide examples of varying strata of meaning and differing approaches to language game creation, as well as demonstrating diversified categories of abstraction.
It was expected that names which were factual, recognizable and did not require specialized knowledge (the strongest positive correlation to Panofsky’s category of “Primary”) would occur most often in pin names based on “Primary” search terms.
It was also expected that names which rely on a theme, a literary allusion, specialized knowledge, formulas, allegories or other layers of meaning beyond the immediately factual and recognizable (the strongest positive correlation to Panofsky’s category of “Secondary” ) would occur most often in pin names based on “Secondary” search terms.
Intrinsic names (which required a more specialized cultural knowledge to decipher) included symbolic, culturally specific, interpretive, historically defined or non-contextually defined words and may indicate an attempt on the part of the user-curator to provide a relatively sophisticated message. Names which were categorized as “intrinsic” were expected to be difficult (or impossible) to understand when separated from their attendant images, and this category of name was expected to make up a smaller percentage of overall names, since creating these meaning-dense names presumably requires greater effort on the part of the user-curator.
Summary
Using an exploratory, descriptive approach , this project was designed to shed light on the way individual Pinterest users are adapting language as they name their image collections online.
Data collection occurred on the live site in 2014, which involved gathering pins and associated pin names based on eighteen search terms. The language used in each name was then examined and assigned to a strata of the Panofsky/Rosch/ Shatford Layne matrix. Language game analysis was completed, and findings were based on the combined results of the matrix assignments and the language game observations.
Pin names were expected to provide examples of varying strata of meaning and differing approaches to language game creation, as well as demonstrating diversified categories of abstraction.
Lack of an API was challenging in terms of re-finding previous images and uncovering user details. Collecting visual data from a live site required manual archives of both text and images, to assure future availability of project data.
By observing the characteristics of Pinterest’s relatively non-ruled based approach to image naming in action and by exploring the types of words chosen, the number and format of the characters selected, the linguistic constructions applied to each name when individually organized by each user-curator and the patterns which emerged throughout the relatively restricted sample, a small but unique snapshot of human language behavior during digital image curation was captured.
CHAPTER 4 ANALYSIS OF DATA, RESEARCH FINDINGS, AND DISCUSSION
Alpha data collection
Between January and March 2014, the first round of data collection took place, with 120 unique Pinterest images and the associated names captured using three search terms.
Search terms: For the first round of data collection, three search terms were needed to represent each of the three iconological levels on Panofsky’s matrix. This was a problematic exercise, as noted earlier in the discussion on language and meaning in imagery. Even limiting the search terms to those in modern colloquial English provided little assurance that such words would yield the needed range of pin names necessary for comparisons across Panofsky’s spectrum of meaning.
However, because one of the central goals of this project is to analyze how curators self-name images in large collection, three search terms were needed to begin image collection at a even the most rudimentary level.
Assuming that these three terms might be refined during the full data collection phase, three search terms corresponding to Panofsky’s three levels of meaning were eventually selected.
Primary search term: A single, reasonably cogent noun was preferred as this term was required to produce images which were factual and recognizable, and did not require the viewer to have any knowledge of the culture related to the image. The choice for the primary image search term in the aloha data collection was ‘tree’.
Secondary search term: Here a search term was needed which would yield images containing a theme or literary allusion, or required specialized knowledge, formulas, allegories or other layers of meaning beyond the immediately factual and recognizable. Again, determining a search term which would yield enough images across a spectrum of meaning was challenging. A variety of terms were tested before ‘American Civil War’ was chosen as the secondary alpha search term. Also tested as a secondary search term (but discarded due to high rates of duplicated naming language) were ‘American West’, ‘New England’ and ‘American South’.
Perhaps the most challenging image search term to select was that used to produce symbolic or ‘intrinsic’ images. Panofsky defines this level of meaning as being culturally specific, interpretive or non-contextually defined. For the alpha data collection, the name of an individual contemporary artist (American photographer Saul Leiter) was chosen for the intrinsic level search term.
Over five separate sessions throughout January and February 2014, forty random non-repeating Pinterest images were captured using each of three search terms, based on the first displaying occurrence of each given term on the main indexing page at http://www.pinterest.com on that date as shown in Table 12. Duplicate images and images without names were discarded. The default display language used for pin names was English. (As of April 2014, Pinterest allows the use of 21 languages.)
Table 12
Alpha Search Terms
illustration not visible in this excerpt
After forty unique images were captured for each search term, all images were exported to Word files, to preserve the visual image along with the board and tag names and originator data. Frequency of words used in names was calculated, and language game analysis was applied to the collected words to determine the level of semiotic play. The full collection of data collected in the alpha data set as shown in Table 13 are available in Appendix A.
Table 13
Alpha Data Available in Appendix A
illustration not visible in this excerpt
Final data collection
Over a range of dates between April and September 2014, fifteen additional unique Pinterest search terms were explored, capturing 40 images per search term within selections of primary-secondary-intrinsic search term sets. Including the images previously collected in the alpha data set, the total data set contains 720 total image names (18 x 40) as shown in Table 14.
Table 14
Final Data Set
illustration not visible in this excerpt
Each of the 40 images matching each of the 18 search terms was downloaded , along with that image’s associated pin name and creator name. The pin names are compiled in Appendix. B. Each pin name is assigned to an entry in the Panofsky/Rosch/ Shatford Layne matrix. Any Wittgenstein-related “rules”, instances of new grammar construction or apparent language games observed were annotated. Pin names with matrix assignments are available in Appendix F.
Research findings and discussion
From the sample used in this project, 6% of names corresponded to the primary strata of subject matter, 37% of names corresponded to the intrinsic strata and the majority of names (57%) corresponded to the secondary strata, as shown in Figure 14.
illustration not visible in this excerpt
Figure 14 . Findings by Panofsky strata
Pin Name Distribution: Panofsky’s Strata of Subject Matter or Meaning
Expectations
The pins names collected in this study were harvested from search terms rooted in Panofsky’s three strata of subject matter. This was done both to ensure a wide and balanced spectrum of search terms, encompassing as broad a grouping of language types as possible, while also attempting to verify Panofsky’s approach: three divisions of meaning are often available in a cultural artifact, and a viewer can usually isolate at least one specific strata of meaning from any given example.
It was expected that the names which resulted from a search on a specific strata would reflect a similar level of meaning: names resulting from searching primary terms would probably have a large proportion of primary names, while names resulting from searching secondary terms seemed likely to include secondary-level meaning. Intrinsic names were expected to make up a smaller percentage of overall names, since creating these meaning-dense names presumably requires greater effort on the part of the user-curator.
Findings: Primary names
Although one third of all triggering search terms were considered primary, less than ten percent of all names collected met the criteria of the primary strata. Contrary to expectations, only 6% of the collected names in this project described natural or factual subject matter requiring little or no specific cultural knowledge on the part of the reader (the ‘primary’ strata).
This suggests that one characteristic of naming activity in Pinterest may be the user-curators urge to supply more than the bare minimum of information in pin names, regardless of the subject matter. This finding implies that user-curators appear to be willing to create names with some depth of meaning or at least avoid reverting to the most primitive default of a primary object noun, even when the pin subject is relatively simple. This finding seems to contrast behavior on Pinterest with other social image sites, particularly Flickr, where non-user-created, generic default image labels predominate. Pinterest user-curators observed in this project seem to include more than just the basics, and they do this despite the complexity of the pin image content.
Findings: Intrinsic names
More than a third of all names collected in this project (37%) require in-depth knowledge of the culture and environment which produced both the name and the image (the intrinsic strata). In fact, names designated as intrinsic were generally difficult or impossible for a reader to visualize without the associated image: 263 of 716 names required a degree of intuition, personal psychology, familiarity with related cultural symbols and/or insight on the part of the reader to make any sense of the name, and significantly more effort when the associated image was not available to add context.
The complexity of the names in the intrinsic strata suggests that user-curators may be investing thought and creativity in the process of naming their pins, and may be evolving new surface grammar rules during the collection process. The intrinsic selections had a high percentage of names which were purely textual (quotes, puns, jokes, riddles), names which corresponded to few or none of Rosch’s levels of categorical abstraction levels and names which did not follow traditional surface grammar rules.
Findings: Secondary names
The secondary strata contained more pin names that either the primary or intrinsic strata.. More than half of the collected names (57%) were included as requiring some specific cultural knowledge on the part of the reader to interpret. These names tended to describe specific themes, concepts, stories or allegories. The majority of secondary names collected (413 of the 716 ) required some familiarity or insight related to the associated image. Examples and descriptions of the matrix items collected in the name samples for this project are shown in Table 15.
Table 15
Example of Matrix Items Collected in Name Sample s
illustration not visible in this excerpt
Rosch’s three levels of categorical abstraction
Few image names had characteristics of the generic superordinate category (‘furniture’). Significantly greater numbers of primary pin names are assigned to basic level objects (‘chairs’) while correspondingly similar numbers of secondary pin names fit into the more specific, detailed subordinate levels.
Types of Pinterest language games
Story-telling: One of the language games used within the naming of Pinterest images involves the act of telling a story. The language used for creating the pin name in this particular game formulates “a story”, as shown in Table 16, and often use a variety if identifiable “parts” such as a first-person narrator, a setting, background, audience, tone and time period.
Table 16
Storytelling Pin Names
illustration not visible in this excerpt
Nonlinguistic language games
One of the most common subsets of language games played within Pinterest names are nonlinguistic , which contain words which may not be clearly understood outside of the involved activity. “How do you like your salad?” is a question directly related to the non-linguistic language game of ‘eating’ and might seem nonsensical outside of the activity of naming pins related to eating. A characteristic of this kind of language game is that grammatically correct sentences do not make sense outside of the given context, such as the examples shown in Table 17.
Table 17
Non-Linguistic Pin Names
illustration not visible in this excerpt
“Family resemblances” within Pinterest language games
In describing the usefulness of language games as an analytical tool, Biletzki and Matar (2014) describes the “rule-governed character of language.” While such “rules” within the naming of Pinterest images are generally not the traditionally accepted systems of definitions and penalties associated with other more traditional word-based games, there seem to be observable “family resemblances” in the language choices made, which may reveal patterns of usage when examining the collected data set. (p. 3.3) Two examples of family resemblances include using word or ASCII art and the preference for numbered-list-type titles as shown in Table 18.
Table 18
Examples of Family Resemblances in Pin Names
illustration not visible in this excerpt
Rules and determinacy in Pinterest language games
In order for language games to have meaning, the activities involved must be recognizable to the observer. “Determinacy” exists without dictionaries or controlled vocabularies, because the activity itself suggests the appropriate word choices and definitions. However, if an observer is not able to identify the concept being illustrated, the language games related to that concept will not make sense. For example, unless the observer recognizes the concepts related to spinning yarn from alpaca wool, the language games (instructions, notes and detailed descriptions of process using standard alpaca wool spinning terminology) related to that activity may seem meaningless.
“Form of life” embedded language game rules in Pinterest
Additionally, some language game rules are embedded in what Wittgenstein refers to as “the form of life”, making specific instructions unnecessary within the parameters of the activity. “Don’t eat the chess pieces” is not usually a specific chess-playing game rule among adult human beings, because those who understand the form of life ‘chess games’ already have this information. Within Pinterest, some unspecified “form of life” rules appear to be focused on avoiding typographical gibberish and remaining within the most common language (English, Spanish, French or any of the 21 other available versions of the Pinterest URL used for the original login)
Most commonly observed language games related to pin naming in Pinterest
The most commonly observed language games related to naming pins in the collected sample include making puns, creating word art, using alliterations, delivering rhetorical excursions, constructing oddly formed sentences, including ASCII art and emoticons, odd formatting including unique uses of punctuation or upper and lower case fonts, and the apparently intentional use of malformed sentence/word phrases. Examples of the most commonly observed language games related to naming pins in Pinterest are shown in Table 19.
Table 19
Examples of Commonly Observed Pinterest Language Games
illustration not visible in this excerpt
“Private” language: Codes in Pinterest names
Biletzki and Matar (2014) notes that words used in language games have to meet “public standards and criteria of correctness” to be considered part of a working “game” (p 62). Wittgenstein introduced the question of “private” language in which “words … are to refer to what only the speaker can know - to his immediate private sensations …” (PI 243), but Biletzki and Matar (2014) affirms that this kind of limited private usage should not be considered a genuine, meaningful, rule-governed language. The limitations of such private codes lies in their restricted scope among a wide body of users: “The signs in language can only function when there is a possibility of judging the correctness of their use, “so the use of [a] word stands in need of a justification which everybody understands” (PI 261). The image names collected in this project appear to contain some examples of private codes, where the immediate linguistic definition of the text is not apparent, as shown in Table 20.
Table 20
Private Language Codes Pin Names
illustration not visible in this excerpt
Nonsense
It is interesting to note that even among the most complex names at the intrinsic level, the number of nonsensical names was relatively small. Since the intrinsic category was selected as the default level for all non-textual names, including pure ascii entries and all word art, some loss of literal meaning was expected in this strata. There were 50 names considered nonsense in the intrinsic strata of 263 names, more than in any other strata but still only 7% of the overall sample. None of the names considered nonsensical were pure gibberish. All nonsensical names (aside from emoticons, word art and ascii art) were either grammatical fragments or phrases which did not have an immediately recognizable meaning. For example the name ‘LIFE | FLY’ is considered nonsensical, since no specific meaning can be assigned to this name, but the letters do form recognizable English words and the name was apparently created to assign meaning of some kind. to the associated image.
New surface grammar construction
One specialized kind of Pinterest naming activity involves improvising new “surface” grammar rules. Surface grammar usually applies to correct syntactic and semantic usage including spelling, word order and subject verb agreement. Biletzki and Matar (2014) details how Wittgenstein’s language games allow users to shift the ‘normal’ requirements of word or sentence construction to fit specialized circumstances. “Grammar, usually taken to consist of the rules of correct syntactic and semantic usage, becomes the wider—and more elusive—network of rules which determine what linguistic move is allowed as making sense, and what isn't” (p. 3.5).
Primary Grammar: Perhaps not surprisingly, 75% of the resulting primary pin names (which described natural or factual subject matter requiring little or no specific cultural knowledge) fit in to either in Rosch’s “basic” level of categorical abstraction or Shatford Layne’s “subject” attributes. No primary names were nonsensical. 6 names contained ascii art.
Intrinsic Grammar: The widest variety of surface grammar variations occurred in the intrinsic strata. Intrinsic surface grammar variations occurred 347 times and included the examples shown in Table 21.
Table 21
Observed Intrinsic Surface Grammar
illustration not visible in this excerpt
The examples of observed grammar construction in Table 22 demonstrate the range of inventiveness involved across all strata in the Pinterest pin naming process:
Table 22
Observed Grammar Construction Examples
illustration not visible in this excerpt
Unexpected Findings Related To Re-Searching Pins
When the three original alpha data sets [Trees, America Civil War and Saul Leiter] were collected in February 2014, the pin names and related images were captured in screenshots taken live from Pinterest. All images and words were saved in both Word docs and in Excel worksheets. However, Pinterest boards were not created at that time to contain the images: only screen shots of the live search results were captured.
When the fifteen additional beta data sets were collected in August 2014, all images and names were captured and saved in both Word docs and in Excel sheets. Additionally, all pin images were saved in their own Pinterest boards. [ See http://www.pinterest.com/tamisresearch/ ] Since the three alpha data sets [Trees, America Civil War and Saul Leiter] were to be included in the final analysis, it became useful (four months later) to recreate those original data sets in Pinterest, as independent boards.
The process of re-finding these exact pins, by the identical originators, a second time, four months later, uncovered some interesting traits about user-curator behavior.
1. Pinners tended not to change their own name, but they did change the names of boards and pins randomly, across all topics. It was discovered (when trying to recreate pins captured four months earlier) that the most effective way to locate a known pin was to use the pinner name. The next most effective way was to use the pin name itself. The matching image could infrequently be found by searching on the pin name, but the related pinner/board data had sometimes either disappeared or changed. A more robust search tool would be helpful for this type of research but this raises the question of whether actual user-curators would have any need for it. Since the goal of Pinterest is to share images, the expectation would be that searching should remain focused on an image and not on an originators name.
2. Occasionally, the name of the image, pin and board was so common that it was not practically possible to quickly isolate one individual needed image. (Quote marks do not seem to be delimiters in the Pinterest search algorithm, although misspellings can cause zero returns.) For example, the image associated with a pin named “Palm Tree in Moonlight” pinned by Dianne Henry in April on a board named “Photos” [P1. Image 30] was still visible on August 15, but no longer viewable under that pin or pinner name. (There were more than 20 pinners named Dianne Henry in August 2014.) The original image now displayed using other pin and user-curator names but did not appear to be connected directly to the original pin name, board or pinner. Another example of this was a pin named “Christmas tree farm, Wisconsin” captured on a board named “Trees” by a pinner named Mary Howard. When that pin name was entered as a search in August, the associated image appeared under numerous (more than ten) pin names and pinners, but not under Mary Howard.
3. Some pins were entirely missing: either no longer in Pinterest or no longer readily findable (given the time constraints of this project) under the saved pin name, pinner name or board.
4. Pinners randomly changed the locations of boards and pins. For example, the pin “Tree” on the board “Tree” by Suneel Sethi was not practically able to be located, since the board “Tree” was no longer displayed on Suneel Sethi’s gallery. (25 boards, 19,636 pins). Looking through 19, 636 pins for “Tree” was not viable given the time constraints of this study.
5. “Find on screen” tools in Chrome, Firefox, Safari and IE behave unpredictably in Pinterest. Being able to “find” one word on a screen containing over a thousand scrolling images would aid in searching, but browsers return unreliable results with the FIND tools, missing instances of exact matches while including inexplicable items. Chrome “find on page” filters seemed most powerful, while Internet Explorer 9 had the weakest searching capabilities, in terms of locating text on a page or returning faulty matches.
In summary, trying to recapture collected images with their original names after more than 120 days presented the problems of names being changed indiscriminately, names being so common that they returned unviably large search results, images being deleted, users changing locations of images and browser search tools performing inadequately when asked to find within large scrolling Pinterest displays.
CHAPTER 5 SUMMARY AND CONCLUSIONS
Summary of findings
People have been collecting blended forms of images and words throughout history. Large public image collections in the past have typically been financed and controlled by organizations who could afford to support professional curators and specialized vocabularies. Now that personal digital image collections have become increasingly available, non-professional image collectors have begun to adapt language to fit their evolving personal image collections. The goal of this project is to increase understanding of the specific naming behaviors present in a personal digital image collection when categorization vocabulary and subject descriptors are uncontrolled.
This project isolated the language of Pinterest user-curators by assigning collected image names to a matrix of Panofsky’s subject matter categories, Rosch’s levels of abstraction and Shatford Layne’s attributes, as well as by examining the naming language created in terms of Wittgenstein’s surface grammar including the related aspects of language game construction.
Research Question 1 asked how the language used in creating image names in Pinterest tended to align within the Panofsky/Rosch/ Shatford Layne matrix. From the sample of 722 names used in this project, 6% of names corresponded to the primary strata of subject matter, 37% of names corresponded to the intrinsic strata and the majority of names (57%) corresponded to the secondary strata. Few image names in any of the Panofsky strata had characteristics of Rosch’s generic superordinate category, while the majority of primary pin names aligned with basic level objects and more than half of the secondary pin names corresponded with the subordinate categories.
These findings based on the matrix assignments suggest that the names in this sample contain relatively complex levels of meaning, based on the specific cultural knowledge required to interpret the naming language created. Even where the related trigger images were obtained from intentionally selected primary search terms based on relatively simple, broadly intuitive representation (for example, moon, bird and tree were used as primary search terms) the naming language isolated in this project was predominantly complex, required relatively advanced cultural knowledge to interpret and included both personalized responses and widely varying user-added information.
The findings from the matrix suggest that Pinterest users are adapting language to suit their needs and one characteristic of this naming behavior is the urge to supply more than the bare minimum of information in pin names, regardless of the density or simplicity of the associated image. User-curators appear to be willing to invest time to create names with a relative depth of personal meaning and to assign these names despite the complexity or overtness of the content of the associated image.
Research Question 2 concentrates on the aspects of Wittgenstein’s language games which were observed in this sample of pin names. The question posed asked which aspects of Wittgenstein’s language games were observed in this sample of image names and which patterns (if any) became visible during analysis.
The complexity of the language games observed, including story-telling, personal comments and rhetorical excursions, suggest that user-curators include the process of inventing meaningful names for their pins as an integral part of the curation process. The sharing of clever, innovative and/or personalized pin names is part of the enjoyment of building, sharing and maintaining a large personal digital image collection. Based on the intricacy of the examined surface grammar, Pinterest user-curators in this sample are investing time, thought and creativity in the process of naming their pins, including evolving new surface grammar rules as an accepted step in the collecting process. The traditional “burden” of assigning representation to images has become an enjoyable and accepted part of Pinterest image curation activity, based on the complexity and quantity of information being voluntarily provided by user-curators in the sample collected in this project.
The numerous ways in which Pinterest user-curators appear to be adapting language to create names for their image collections, especially in the midst of the big, messy, organic data sets that comprise Pinterest, seems to support the user-curator attraction for categorizing “marginalized” content, even if the categorization is invented by each user-curator for their own collecting purposes.
The overall findings from the pin names examined in this project support the idea that a majority of the naming behaviors and language-based activities being conducted within Pinterest do not tend to fit within any clearly defined patterns of pre-categorized meaning. User-curators do not seem to expect or rely on predefined subject categories or naming vocabularies, and the levels of engagement, creativity and personalization displayed during naming behaviors on Pinterest exceed the expectations of a traditional shopping experience or content storage site.
Implications of research findings
The observed naming behaviors in this project imply that providing pre-defined authoritative subject categories to users of large personal digital image collections is not efficient or necessary, since users tend to immediately create their own personalized naming conventions, independent of any outside authority.
The most valuable information gleaned in this study is rooted in the user language collected, rather than in any particular analysis results extracted from the relatively small data set. Given the uncontrolled nature of image naming within Pinterest, it is probable that any given pin name examined in this project may reflect a range of surface grammar and language game patterns which will continue to manifest themselves in the future, and which may only become apparent upon examination of future pin name patterns. Dissection of specific levels of meaning for any given individual pin name was not the goal of this project. Rather, the attempt to identify the range of language being used to convey personalized meaning assigned by user-curators when creating pin names provided an opportunity to document which types of image iconology seem to be evolving in large personal digital image collections.
Forcing precision
Confirming Wittgenstein’s approach rooted in the inefficiencies of dictionaries, definitions and forced precision, this study proposes the best way to observe future user-curator naming behavior in large digital image collections including Pinterest will continue to be observing and collecting live naming activities as user-curators evolve them. Any attempts at “forcing precision”, in terms of creating required categories or subject restrictions would seem to be impractical, given the rapidly shifting environment and user expectations of the current Pinterest population. As Blain notes: “Precision is not a result of the process, but a requirement for it. Blurriness can be important by itself. Removing blurriness does not always create clarity” ( pg 88 ). Users appear comfortable adapting grammar and language games to create meaningful pin names for their collections and the need for specialized naming tools or vocabulary does not seem apparent based on the names examined in this project.
Challenges in Pinterest research
If you are listening in via social media you're probably only hearing a very particular group, representative at best only of those with similar demographic characteristics. Until you know and understand those demographics, it's important not to extrapolate too much from the data.
Noreena Hertz, University College London, October 22, 2013 (Sashittal, 2014)
A review of published “research” related to Pinterest begins illustrates several challenges related to Pinterest studies. American businesses caught the scent of a new way to interact with potential customers when Pinterest originally began to attract media attention in 2012. Promotion of Pinterest as a revenue generating site began, hyping the site as a 21st century way to sell to women.. For example, a brief survey of all books on Amazon and in WorldCat in January 2014 related to Pinterest reveals titles limited to web marketing, e-commerce, entrepreneurship and small business multilevel marketing. Even Facebook and Twitter books are given a broader set of topics within Amazon in 2014, including communication and culture as possible related subjects. But Pinterest so far appears to be the exclusive domain of sales and commerce: find new customers, sell more products, and (sometimes explicitly, sometimes subliminally) – sell more to women.
Studies published between 2012 and 2014 on Pinterest users as members of social networks inadvertently raise questions about viable methods of collecting information about the user-curators of large digital image collections.. For instance, data collection from the five most “popular” (frequently followed) pinners is used to form a tag cloud of the top 100 terms present in their “popular” pin descriptions: “We picked the initial seeds for our data collection process as the top 5 most followed users on Pinterest. We understand that this technique suffers from bias, and the sample taken is not completely random” (Mittal p. 9). Since Pinterest is NOT primarily a “social” network, using this “most popular” method (while interesting) does not begin to plumb the depths of the daily activity percolating throughout Pinterest.
Because there are no reliable tools available to count pin views as of October 2014, basing a conclusion about how often pins are viewed on the percentage of repins and likes would seem to be problematic. For example: “The low percentage of repins and likes shows that there is a limited set of pins that get popular, and that a majority of pins go unnoticed” (Linder 2). Aside from the undocumented assumption that “getting popular” is a goal of Pinterest user-curators, the fact that a pin has not been repinned does not conclusively indicate it has gone unnoticed. User-curators interact in a variety of ways with a plethora of images during any image curation session and may return, rename, resave, relike and repin at any point. This dispersed activity would indicate that basing in-depth user analysis on how often a given set of user curators have repinned a given image seems limited, at best.
Han (2014) also attempted to explain and/or predict user behavior by mapping the patterns of “repinning.” “This subsection analyzes pin propagation patterns based on the 32 Pinterest-defined categories and the top 20 sources which are sorted in terms of the number of corresponding pins” ( p. 5). The challenge in this approach seems to be relying on the predefined topics provided by Pinterest as true indicators of user intent. The predefined categories are not generally used by a majority of observed pinners across most studies reviewed for this project, and basing generalized conclusions about user intent within such a large and demographically undefined population based on the activities of a sample, even when collecting millions of data points, does not reveal much about the user reasoning, aims or expectations.
Large numbers of both interactions and users within Pinterest simultaneously provide appealing research possibilities and complex challenges. As a large digital image collection, Pinterest provides scanty demographic detail on users, making forming conclusions about Pinterest user behaviors, whether based on millions of aggregated steps or collated from several hundred manually collected image names, demanding. The lack of specificity related to users can lead to difficult-to-quantify generalities, such as “We observed that the most common interests were in line with the most common professions (like artist, designer, cook, photographer) mentioned by the users“ (Mittal p. 4).
It becomes clear that targeted research based on observable Pinterest user behavior is scant. The available large scale Pinterest studies as of 2014 appear to be rooted in existing and possibly unsupported assumptions about user demographics and reveal sometimes subtle biases not tied to directly observable user behaviors. Sashittal notes
“Media habits vary with demographics; a study of a defined demographic segment versus the general population is more likely to produce actionable insights. Industry experts caution against formulation of social media strategies based on a general understanding of heterogeneous population of users. Instead, they strongly advocate for understanding the motivations and behaviors of narrow demographic segments and tailoring strategies based on this learning” (p. 2).
Despite the gold rush mentality surrounding Pinterest, reliable unbiased data about verifiable user behavior remains difficult to isolate.
Recommendations for Future Research
Cunning Intelligence and Social Collecting
One aspect of social image collecting (and Pinterest activity, in particular) which could benefit from future research is an understanding of cunning intelligence: an undirected, unfocused style of exploration applied to shifting, transient environments (such as those embodied by large digital image collections). Understanding how users accommodate the unexpected and adapt to unanticipated or accidental information discovery could help increase a social collector’s ability to swiftly navigate through complex and changing layers of information. The roots of this type of unstructured wandering run deeply through human culture and have been extensively notated by Detienne and Vernant (1978) in their work on metis, a kind of “practical intelligence” that appears throughout Greek myths:
There is no doubt that metis is a type of intelligence and of thought, a way of knowing; it implies a complex but very coherent body of mental attitudes and intellectual behavior which combine flair, wisdom, forethought, subtlety of mind, deception, resourcefulness, vigilance, opportunism, various skills, and experience acquired over the years. It is applied to situations which are transient, shifting, disconcerting. and ambiguous (p. 322).
The ever-shifting environment of the online social image collector would seem to require this kind of polymorphic ability to adapt and repurpose both content and language. The cunning intelligence of metis may be one of the traits allowing a social image collector to thrive in a shifting and unexpected set of circumstances, to comfortably evolve the new grammar, naming rules and language game constructions which appear to be integral parts of the image naming process within Pinterest. Future research examining how user-curators are navigating through large personal digital image collections might uncover applications of cunning intelligence as aids in attempts to evolve even more intuitive naming systems and adapt even more expressive language games.
An additional consideration for future Pinterest researchers centers on identifying the difficulties inherent in observing and recording activity on a site which is, by definition, constantly changing. User-curators appear and disappear, while images are added, deleted, named and renamed, with innovative language use and shifting user interests weaving the entire collaboration into an enormous living ongoing reflection of the users and their ideas. As of October, 2014, the most effective (but time-consuming) method of genuinely capturing a slice of this activity involved screen captures and archival copies of the pin images and names being studied. Perhaps a more flexible method of capturing both the words and the images, simultaneously, could allow for a wider more comprehensive sample of user-curator behaviors in the Pinterest environment.
Conclusion
The aim of art is to represent not the outward appearance of things, but their inward significance.
Aristotle
An ancient question continues to resurface: how can we best describe our interpretation of a visual experience when our most basic representational tool is word-based language? The variability of language itself presents obstacles to adequately translating images into any form of shared “meaning”.
Which is most valuable, then, the words or the images? Given any real life situation with limited resources, should more emphasis and energy be focused on the image or on the representation of the image? This is an insoluble paradox, of course, since words themselves are only marks on paper, flickers of electrons or temporary noises, and can never duplicate the intensity and nuances of even the simplest visual image. And yet, the mute image, separated from context and isolated without frame of reference can melt into neutralized abstraction without some explanation, connotation or description, by either the creator or the individual experiencing the artifact in real time.
Pinterest user-curators appear to create collections as a collaborative expressive exercise, as a shared communication device and, frequently, as a private creative outlet thematically aimed at no other audience beyond themselves. Understanding how this personalization influences the way images are categorized by the user-curator may lead to improved methods for users in other image collections to contribute additional value to the collection in the form of meaningful image naming language, as well as reducing factors which appear to discourage user-curators from contributing to the image naming process.
Researchers interested in user behavior related to image naming must recognize the possibility that as large public image collections adapt to evolving user needs, the choice to become independent of any institutional vocabulary or authority will allow collection users to assume a larger role in the image attribute assertion process. “It is likewise our hope that taking some of the assertion making responsibility off the shoulders of the cataloguers and putting it into the hands of the users of the system will generate a more dynamic system that is more richly representative of both the images and the user requirements” (O’Conner & O’Connor, 1999).
Currently, the assignment of meaning to any given Pinterest image lies almost entirely in the hands of each individual user-curator. These active image collectors are not presently subject to controlled vocabularies, naming conventions or even the constraints of necessarily providing retrieval access for other users. The relatively unrestricted naming activity on Pinterest offers a glimpse of both the strengths and weaknesses of a user-driven naming system, allowing a greater depth of meaning and personalization for each individual user-curator, but shifting the responsibility for providing search efficiency and relevance for the entire collaborative collection onto the shoulders of community members (who thus far do not appear motivated to organize their personal collections using predictable categories or traditional vocabularies.)
Will Pinterest user-curators eventually have to adopt a more controlled set of naming conventions, in order to retain relevance and accessibility for the billions of images flowing through the site?
Or will user-curators continue to evolve independent private systems of naming, relying on their own grammar construction and language games to capture the meaning they are building in to their ever-expanding personal digital image collections?
Just as we understand that word-driven language can never produce a direct translation of the meaning of an image, we must also acknowledge that words are frequently the only method available to humans to express our most profound reactions to what we see.
And we apparently hunger for both the words and the images, the showing and the telling, the visceral pleasure of seeing but also the intellectual shiver which accompanies sharing the right words to genuinely name our visions.
APPENDIX A ALPHA DATA SET
Data collected: January – March, 2014
illustration not visible in this excerpt
276 words used in 40 pin names: Search term = tree
illustration not visible in this excerpt
282 words used in 40 pin names: Search term = American Civil War
illustration not visible in this excerpt
138 words used in 40 pin names: Search term = Saul Leiter
illustration not visible in this excerpt
APPENDIX B: FINAL DATA SET
Final Data Set
Collected August 2014
All pins with associated names are available here:
http://www.pinterest.com/tamisresearch/
Login user name: research@tamisutcliffe.com
Password: research
illustration not visible in this excerpt
illustration not visible in this excerpt
illustration not visible in this excerpt
All Primary Pins: Tree, Bird(s), Man, Water, Woman, Moon
illustration not visible in this excerpt
APPENDIX C: USER STATISTICS 2012 -2014 PINTEREST USER STATISTICS
illustration not visible in this excerpt
Retrieved from http://www.comscore.com/Insights/Blog/comScore-Releases-Top-50-US-Multi-Platform-Properties-for-September-2013
illustration not visible in this excerpt
Retrieved from http://www.comscore.com/Insights/Press-Releases/2014/4/comScore-Media-Metrix-R-Ranks-Top-50-US-Desktop-Web-Properties-for-March-2014#
illustration not visible in this excerpt
illustration not visible in this excerpt
APPENDIX D PIN SELECTION
illustration not visible in this excerpt
Primary pins :
Abbildung in dieser Leseprobe nicht enthalten
Secondary pins:
Abbildung in dieser Leseprobe nicht enthalten
Intrinsic pins:
illustration not visible in this excerpt
APPENDIX E KAMATH’S BOARD COHERENCE
illustration not visible in this excerpt
illustration not visible in this excerpt
APPENDIX F ALL PIN NAMES: SECONDARY, PRIMARY, INTRINSIC
illustration not visible in this excerpt
REFERENCES
AgBeat (2012). Pinterest users’ time on site nearly matches YouTube. Retrieved from http://agbeat.com/social-media/pinterest-users-time-on-site-nearly-matches-youtube/
Allen, E. (2014). Evan Sharp: Pinterest. The Designer Founders book series. Retrieved from https://www.designerfounders.com/interviews/evan-sharp/
Anscombe, G. E. M., & Anscombe, E. (2001). Wittgenstein’s philosophical investigations: The German text with a revised English translation 50th anniversary commemorative edition. Malden, MA: Blackwell Publishing.
Beach, L. R. (1964). Cue probabilism and inference behavior. Psychological Monographs, 78, 582-588.
Belkin, N. J. (1980). Anomalous states of knowledge as a basis for information retrieval. Canadian Journal of Information Science. 5, 133-43.
Biletzki, A. & Matar, A. (2014). Ludwig Wittgenstein . The Stanford encyclopedia of philosophy (Spring 2014 Edition), Edward N. Zalta (ed.) Retrieved from http://plato.stanford.edu/archives/spr2014/entries/wittgenstein/
Blair, D. (2006). Wittgenstein, language and information: Back to the rough ground. (Vol. 10). Springer.
Brustein, J. (2013, Oct). Why investors love Pinterest so much. Business Week. Retrieved from http://search.proquest.com/docview/1475401034?accountid=7113
Campbell, T. P. (2013). The Metropolitan Museum of Art: An important message from the director. Retrieved from http://www.metmuseum.org/about-the-museum/now-at-the-met/from-the-director/2013/important-message#
Cattuto, C., Loreto, V., & Pietronero, L. (2006). Semiotic dynamics and collaborative tagging. Proceedings of the National Academy of Sciences. 104 (5), 1461-1464. Retrieved from http://www.pnas.org/content/104/5/1461
Chafkin, M. (2012). Starring Ben Silbermann as the pinup kid. Fast Company, 169, 90-94; 146-147.
Chang, S., Kumar, V., Gilbert, E., & Terveen, L. G. (2014). Specialization, homophily, and gender in a social curation site: Findings from Pinterest. In Proceedings of the 17th ACM Conference on Computer Supported Cooperative Work and Social Computing (pp. 674-686). Baltimore, MD: ACM.
Cheng, A., Malit, M., Zhang, C., & Koudas, N. (2014). SerpentTI: Flexible analytics of users, boards and domains for pinterest. In Proceedings of the 2014 ACM SIGMOD International Conference On Management Of Data (pp.1075-1078). ACM. Retrieved from http://dl.acm.org/citation.cfm?id=2594518
Cold Brew Labs, Inc. (2012) Pinterest terms of service. Retrieved from http://pinterest.com/about/terms/ .
comScore. (2013). comScore releases top 50 U.S. multi-platform properties for September 2013. Retrieved from http://www.comscore.com/Insights/Blog/comScore-Releases-Top-50-US-Multi-Platform-Properties-for-September-2013
comScore. (2014). comScore Media Metrix® ranks top 50 U.S. desktop web properties for March 2014. Retrieved from http://www.comscore.com/Insights/Press-Releases/2014/4/comScore-Media-Metrix-R-Ranks-Top-50-US-Desktop-Web-Properties-for-March-2014#
Couprie, L. D. (1978). Iconclass, a device for the iconographical analysis of art objects. Museum International (Edition Francaise), 30, 194 -198.
Cresswell, J. W. (2014). Research design. Thousand Oaks, CA: Sage.
Crotty, M. (1998). The foundations of social research: Meaning and perspective in the research process. Thousand Oaks, CA: Sage.
Dervin, B. (1992) From the mind’s eye of the user: The sense-making qualitative- quantitative methodology. In Jack D. Glazier and Ronald R. Powell (Eds.) Qualitative research in information management. (pp. 68-70). Englewood, CO: Libraries Unlimited.
Detienne, M., & Vernant, J. P. (1978). Cunning intelligence in Greek culture and society. Hassocks: Harvester Press.
Dotsika, F. (2009). Uniting formal and informal descriptive power: Reconciling ontologies with folksonomies. International Journal of Information Management, 29, 407-415.
Elkins, J. (1999). The domain of images. New York: Cornell University Press.
Elsner, J. & Lorenz, K. (2012). The genesis of iconology. Critical Inquiry, 38 (3), 483- 512. Retrieved from http://www.jstor.org/stable/10.1086/664548
Experian Marketing Services (2012). Digital Marketer: Benchmark and Trend Report 2012. Retrieved from http://go.experian.com/forms/experian-digital-marketer-2012?WT.srch=PR_EMS_DigitalMarketer2012_040412_Download?send=yes
Feinberg, M. (2012). Personal digital collections as creative expression. Webcast sponsored by the Irving K. Barber Learning Centre and hosted by the School of Library, Archival, and Information Studies. Retrieved from http://hdl.handle.net/2429/43528
Frier, S. (2014). Can Pinterest be found in translation? Bloomberg Businessweek. Retrieved from http://www.businessweek.com/printer/articles/202692-can-pinterest-be- found-in-translation
Gilbert, E., Bakhshi, S., Chang, S., & Terveen, L. (2013). I need to try this: A statistical overview of Pinterest. Proceedings of the Special Interest Group on Human-Computer Interaction Conference on Human Factors in Computing Systems. Minneapolis, MN: ACM.
Gombrich, E. H. (1999). The uses of images: Studies in the social function of art and visual communication. Boston: Phaidon Press.
Greisdorf, H., & O’Connor, B.C. (2002). Modeling what users see when they look at images: A cognitive viewpoint. Journal of Documentation, 58 (1), 6-29.
Greisdorf, H. F. and O’Connor, B.C. (2002). “What Do Users See?” Proceedings of the 65th ASIST Annual Meeting, 39, 383-390.
Greisdorf, H & B. O’Connor. (2008). Structures of image collections: From Chauvet- Pont-d’Arc to Flickr. Libraries Unlimited.
Han, J., Choi, D., Chun, B. G., Kim, H. C., & Choi, Y. (2014). Collecting, organizing, and sharing pins in Pinterest: Interest-driven or social-driven? Proceedings of SIGMETRICS '14: The 2014 ACM International Conference On Measurement And Modeling Of Computer Systems (pp. 15-27). Austin, TX: ACM. Retrieved from http://dx.doi.org/10.1145/2591971.2591996
Hanbury, A. (2008). A survey of methods for image annotation. Journal of Visual Languages and Computing, 19 (5), 617 – 627.
Harpring, P. (2010). Introduction to controlled vocabularies: Terminology for art, architecture, and other cultural works. Los Angeles:Getty Research Institute.
Hastings, S. K. (1999). Evaluation of image retrieval systems: Role of user feedback. Library Trends, 48 (2), 438.
Hibler, D., Leung, C.H.C. & Mwara, N. (1992). Picture retrieval by content description. Journal of Information Science, 18 (2), 111 – 119.
Hocks, M.E. & Kendrick, M. R. (2003). Eloquent images: Word and image in the age of new media. Cambridge, MA: MIT. Retrieved from http://tinyurl.com/nt3dsse
Hollink, L. et al. (2004). Classification of user image descriptions. International Journal of Human-Computer Studies, 61 (5), 601–626.
Horowitz, J. (2013). Pinterest now has 70 million users and is steadily gaining momentum outside the US. The Next Web. Retrieved from http://thenextweb.com/socialmedia/2013/07/10/semiocast-pinterest-now-has-70-million-users-and-is-steadily-gaining-momentum-outside-the-us/#comments
Jaimes, A. and Chang, S.F. (2000). A conceptual framework for indexing visual information at multiple levels. IS&T/SPIE Internet Imaging, 3964. Retrieved from http://www.ee.columbia.edu/ln/dvmm/publications/00/ajaimes-spie00_internet.pdf
Kamath, K. Y., Popescu, A. M., & Caverlee, J. (2013). Board coherence in Pinterest: Non-visual aspects of a visual site. In Proceedings of the 22nd international conference on World Wide Web companion (pp. 49-50). International World Wide Web Conferences Steering Committee. Retrieved from http://faculty.cs.tamu.edu/caverlee/pubs/kamath13www-poster.pdf
Kaufman, J. E. (2009). Troubles deepen for museums: Layoffs, budget cuts and cancelled shows. The Art Newspaper, 201. Retrieved from http://www.theartnewspaper.com/articles/Troubles-deepen-for-museums-layoffs-budget-cuts-and-cancelled-shows/17148
Kim, H., Breslin, J., Chao, H. & Shu, L. (2013). Evolution of social networks based on tagging practices. Services Computing, IEEE Transactions, 6 (2), 252-261. Retrieved from http://libproxy.library.unt.edu:2247/stamp/stamp.jsp?tp=&arnumber=6072202&isnumber=6522397
Krause, M.G. (1988). Intellectual problems of indexing picture collections. Audiovisual Librarian, 14 (2), 73-81.
Kuhlthau, C. C. (1991). Inside the search process: information seeking from the user’s perspective. Journal of the American Society for Information Science, 42 (5), 361-71.
Linder, R., Snodgrass, C., & Kerne, A. (2014). Everyday ideation: All of my ideas are on Pinterest. In Proceedings of the 32nd Annual ACM Conference on Human Factors in Computing Systems. (pp. 2411-2420). Toronto, Canada: ACM.
MacMillan, D. (2014). Pinterest valued at $5 billion. Wall Street Journal. Retrieved from http://search.proquest.com/docview/1524809810?accountid=7113
McAllister, J. W. (2013). Reasoning with visual metaphors. The Knowledge Engineering Review, 28 (3), 367–379.
Madrigal, A. C. (2014). What is Pinterest? A database of intentions. The Atlantic. Retrieved from http://www.theatlantic.com/technology/archive/2014/07/what-is-pinterest-a-database-of-intentions/375365/
Mai, J. E. (2011). Folksonomies and the new order: Authority in the digital disorder. Knowledge Organization, 38 (2), 114-122.
Machirori, F. (2013). Technologies and power: Dynamics in women’s public and private spheres. BUWA! A Journal on African Women's Experiences. Feminism and Culture, 2, 86-92.
Markey, K. (1983). Computer-assisted construction of a thematic catalog of primary and secondary subject matter. Visual Resources, 3, 16-49.
Mathes, A. (2004) Folksonomies: Cooperative classification and communication through shared metadata. Computer Mediated Communication, LIS5900CMC (Doctoral seminar), Graduate School of Library and Information Science. University of Illinois Urbana–Champaign.
Mittal, S., Gupta, N., Dewan, P., & Kumaraguru, P. (2013). Pinned it: A large scale study of the Pinterest network. Indraprastha Institute of Information Technology, Delhi (IIIT-D). Retrieved from http://precog.iiitd.edu.in/Publications_files/Pinterest-IKDD.pdf
Mittal, S., Gupta, N., Dewan, P., & Kumaraguru, P. (2013). The pin-bang theory: Discovering the Pinterest world. Retrieved from http://arxiv-web1.library.cornell.edu/pdf/1307.4952.pdf
Mitchell, W. J. T. (1986). Iconology: Image, text, ideology. Chicago: University of Chicago Press.
Mitchell, W. J. T. (1995). Picture theory: Essays on verbal and visual representation. Chicago: University of Chicago Press.
Moore, R. J . (2014). Pinners be Pinnin’: How to justify Pinterest’s $3.8B valuation. RJMetrics. Retrieved from http://blog.rjmetrics.com/2014/05/07/pinners-be-pinnin-how-to-justify-pinterests-3-8b-valuation/
Moxey, K. (1986). Panofsky's concept of iconology and the problem of interpretation in the history of art. New Literary History, 17 (2), 265-274. Retrieved from http://www.jstor.org/stable/468893
Neilson, J. (2011). How long do users stay on web pages? Retrieved from http://www.nngroup.com/articles/how-long-do-users-stay-on-web-pages/
O'Connor, B. C. (1992). Preservation and repacking of lantern slides within a desktop digital imaging environment. Microcomputers For Information Management, 9 (4), 209-24.
O’Connor, B. C. (1993). Browsing: A framework for seeking functional information. Science Communication, 15 (2), 211-32.
O’Connor, B. C. (2014). Selfies and public knowledge. Founders Lecture in Proceedings of DOCAM 2014. Kent, OH: DOCAM 2014.
O’Connor, B. C., Anderson, R. L. & Kearns, J. L. (2008). Doing things with information: Beyond indexing and abstracting. Westport, CT: Libraries Unlimited.
O’Connor, B. C. & Copeland, J. H. (2003). Hunting and gathering on the information savanna: Human information seeking behavior. Lanham, MD : Scarecrow Press.
O'Connor, B. C., & O'Connor, M. K. (1999). Categories, photographs and predicaments: Exploratory research on representing pictures for access. Bulletin of the American Society for Information. Science, 25 (6), 17-20.
O'Connor, B.C., O'Connor, M.K., & Abbas, J.M. (1999) User reactions as access mechanism: An exploration based on captions for images . Journal of the American Society for Information Science and Technology, 50 (8), 681-697.
O’Connor, B. C. & Wyatt, R. B. (2004). Photo provocations: Thinking in, with, and about pictures. Lanham, MD: Scarecrow Press.
Olson, H. A. (2002). The power to name: locating the limits of subject representation in libraries. Boston: Kluwer.
Ottoni, R., Pesce, J.P., Las Casa, D., Franciscani Jr., G. Meira Jr., W., Kumarguru, P., & Almeida, V. (2011). Ladies first: Analyzing gender roles and behaviors in Pinterest. Association for the Advancement of Artificial Intelligence. Retrieved from http://homepages.dcc.ufmg.br/~jpesce/wp-content/plugins/papercite/pdf/icwsm13_pinterest.pdf
Oyarce, S. (2012). In pursuit of image: How we think about photographs we seek. (Unpublished doctoral dissertation). University of North Texas: Denton.
Panofsky, E. (1939). Studies in iconology: Humanistic themes in the art of the Renaissance. New York: Harper & Row.
Panofsky, E. (1972). Meaning in the visual arts. New York: Harper & Row.
Palis, C. (2012). Pinterest traffic growth soars to new heights. Huffington Post. Retrieved from http://www.huffingtonpost.com/2012/04/06/pinterest-traffic-growth_n_1408088.html
Pew Reports (2013). The demographics of social media users: 2012. http://pewinternet.org/Reports/2013/Social-media-users.aspx
Quintarelli, E. (2005). Folksonomies: Power to the people. Paper presented at the ISKO Italy-UniMIB, Milan, Italy: June 24, 2005. Retrieved from http://www.iskoi.org/doc/folksonomies.htm
Reed, S. K. (1972). Pattern recognition and categorization. Cognitive Psychology, 3, 382-407.
RKD (Rijksbureau voor Kunsthistorische Documentation) aka Netherlands Institute of Art History. (2009). Launch of new Iconclass website and browser. Iconclass News, 10 November 2009. Retrieved from http://www.iconclass.nl/iconclass-news .
Rorissa, A. (2005). Perceived features and similarity of images: An investigation into their relationships and a test of Tversky's contrast model. Denton, Texas. UNT Digital Library. Retrieved from http://digital.library.unt.edu/ark:/67531/metadc4749/ .
Rorissa, A.,& Iyer, H. (2008). Theories of cognition and image categorization: What category labels reveal about basic level theory. Journal of the American Society for Information Science and Technology, 59 (9), 1383 -1392.
Rosch, E. & Lloyd, B., eds. (1978). Principles of categorization. Cognition and categorization. (pp. 27-48). Hillsdale, NJ: Lawrence Erlbaum.
Rosch, E. (1978). Principles of categorization. University of California, Berkeley.
Rose, G. (2001). Visual methodologies: An introduction to the interpretation of visual materials. London: Sage Publications.
Sandhaus, P.& Boll, S. (2010). Semantic analysis and retrieval in personal and social photo collections. Springer Science+Business Media. Retrieved from http://libproxy.library.unt.edu:2196/content/pt87804r8r721136/fulltext.pdf .
Schreiber, A.T., Dubbeldam, B., Wielemaker, J. & Wielinga, B. (2001). Ontology-based photo annotation. IEEE Intelligent Systems, 16 (3), 66 – 74.
Semiocast. (2013). Geolocation and activity analysis of Pinterest accounts. Retrieved http://semiocast.com/en/publications/2013_07_10_Pinterest_has_70_million_users
Sashittal, H. C. & Jassawalla A. R. (2014): Why do college students use Pinterest? A model and implications for scholars and marketers, Journal of Interactive Advertising. Retrieved from http://dx.doi.org/10.1080/15252019.2014.956196
Shatford, S. (1984). Describing a picture: A thousand words are seldom cost effective. Cataloging & Classification Quarterly, 4 (4), 13-30.
Shatford, S. (1986). Analyzing the subject of a picture: a theoretical approach. Cataloging & Classification Quarterl y, 6 (3), 39-62.
Shatford Layne, S. (1994). Some issues in the indexing of images. Journal of the American Society for Information Science, 45 (8), 583-588.
Shatford Layne, S. (2002). Introduction to art image access. In Baca, M. (ed). Art image access: Issues, tools, standards, strategies. Los Angeles: Getty.
Schifanella, R., Barrat, A., Cattuto, C., Markines, B., & Menczer, F. (2010). Folks in folksonomies: Social link prediction from shared metadata. Proceedings of the 3rd ACM Int’l Conf. On Web Search And Data Mining, (pp. 271–280). New York, NY: ACM. Retrieved from http://tinyurl.com/nkdeyx3
Shirky, C. (2005). Ontology is overrated: Categories, links, and tags. Clay Shirky's Writings: 2005. Retreived from http://www.shirky.com/writings/ontologyoverrated
Summers, N. (2014). Pinterest co-founder Evan Sharp on guided search, promoted pins, wearables, and more. The Next Web. Retrieved from http://thenextweb.com/socialmedia/2014/04/30/pinterest-co-founder-evan-sharp-talks-guided-search-promoted-pins-wearables/
Taylor, R. S. (1968). Question negotiation and information seeking in libraries. Journal of College and Research Libraries, 29 (3), 178-94.
TechCrunch. (2012). Pinterest hits 10 million U.S. monthly unique users faster than any stand alone site. Retrieved from http://techcrunch.com/2012/02/07/pinterest-monthly-uniques/
Tekobbe, C. K. (2013). A site for fresh eyes: Pinterest's challenge to ‘traditional’digital literacies. Information, Communication & Society, 16 (3), 381-396.
Tonkin, E. et al. (2008) Collaborative and social tagging networks. Ariadne (54) . Retrieved from http://www.ariadne.ac.uk/issue54/tonkin-et-al/
Tufte, E. R., and Graves-Morris, P. R. (1983). The visual display of quantitative information. Cheshire, CT: Graphics Press.
Tufte, E. R. (1990). Envisioning information. Cheshire, CT: Graphics Press.
Tversky, S. (1977). Features of similarity. Psychological Review, 84, 327-352. Retrieved from http://ruccs.rutgers.edu/forums/seminar1_fall03/Lila2.pdf
Ugwu, R. (2012) UnGoogleable: Why Pinterest is the most regrettable social network yet . Complex Networks Inc. Retrieved from http://www.complex.com/pop-culture/2012/03/ungoogleable-why-pinterest-is-the-most-regrettable-social-network-yet
Van Straten, R. (1986). Panofsky and ICONCLASS. Artibus et Historiae, 7 (13), 165- 181).
Walker, T. (2012). State of the US Internet in Q1 2012. Presented at ComScore Inc. State of the Internet: U.S. Quarter One 2012. Retrieved from http://www.comscore.com/Insights/Presentations_and_Whitepapers/2012/State_of_US_Internet_in_Q1_2012
Wichowski, Alexis. (2009). Survival of the fittest tag: Folksonomies, findability, and the evolution of information organization. First Monday, 14 (5). Retrieved from http://www.uic.edu/htbin/cgiwrap/bin/ojs/index.php/fm/article/view/2447/2175d
Wilson, T.D. (1999). Models in information behavior research. Journal of Documentation, 55 (3), 249-70.
Winget, B. (2004). Intellectual access to digital art-objects: Image attributes and art historical knowledge. Presented at the Visual Resources Association (VRA) 2004 Conference: Portland, Oregon. March 8 – 10, 2004. Retrieved from http://www.unc.edu/~winget/research/VisualInfo.pdf
Wittgenstein, L. (1958). Philosophical investigations (trans. G.E.M. Anscombe.) New York: Macmillan.
Xanthos, N. (2006). Wittgenstein's language games. Signo: University of Quebec. Retrieved from http://www.signosemio.com/wittgenstein/language-games.asp
Woo, J. (1994). Indexing: Playing in the fields of Postmodernism. Visual Resources: An International Journal of Documentation, 10 (3), 248-258.
Yoon, J.W. & B. C. O'Connor. (2010). Engineering an image-browsing environment: Re- purposing existing denotative descriptors. Journal of Documentation, 66 (5), 750 – 774.
Zarro, M. & Hall, C. (2012). Pinterest: Social collecting for #linking #using #sharing. JCDL '12 Proceedings of the 12th ACM/IEEE-CS joint conference on Digital Libraries. New York, NY: ACM. Retrieved from http://mikezarro.com/docs/Zarro_JCDL2012_Poster.pdf
Zhong, C., Salehi, M., Shah, S., Cobzarenco, M., Sastry, N., & Cha, M. (2014). Social bootstrapping: How Pinterest and last. fm social communities benefit by borrowing links from Facebook. In Proceedings of the 23rd International Conference on World Wide Web. (pp. 305-314). International World Wide Web Conferences Steering Committee. Retrieved from http://arxiv.org/pdf/1402.6500.pdf
Zhong, L. (2014). My pins are my dreams: Pinterest, collective daydreams, and the aspirational gap. (Doctoral dissertation, Massachusetts Institute of Technology). Retrieved from http://dspace.mit.edu/handle/1721.1/89975
- Arbeit zitieren
- Tami Sutcliffe (Autor:in), 2014, The Iconology of Pinterest, München, GRIN Verlag, https://www.grin.com/document/288985
-
-
-
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen. -
-
Laden Sie Ihre eigenen Arbeiten hoch! Geld verdienen und iPhone X gewinnen.