[6 May 2010]
PopMatters Multimedia Editor
Google does not really talk about all of the rules that govern the ranking of images (and thus, why particular images appear in the order that they do) in Google Image Search queries. In a video provided by Google, Peter Linsley encourages webmasters to be aware of how they title images to produce better searches. The closer that a term or terms come to describing the content of an image, the more likely it will be well ranked, it seems.
Also, Linsley suggests that additional html tags and the context surrounding images that pertain directly to them can also produce better results in garnering Google’s spiders’ attention to particular images (“Images – Webmaster Tools Help”, Google: Webmaster Central, 2010). In addition to these vaguely described but known quantities, exists a host of speculation on the part of users and webmasters about how hit counts and the volume of links to images might drive images up in those rankings. While webmasters are no doubt interested in how results are ranked, so that they can get their images noticed, what discussions of the search engine often don’t consider is the interpretation of those results by the user.
At first blush, a Google Image Search seems like an arbitrary mixture of signs related to search terms, but given that those who name files and talk about images are (whether intentionally or not) assigning values to these signs by naming them in the way that they do, interesting shifting patterns of meaning emerge as results pages are viewed. It’s possible that signification emerges through a consensus that is built by calling attention to how an image is seen by viewers, and as a result of these underlying rules, a definition by consensus emerges.
The interesting element that I allude to concerning consensus derives from this idea that image rankings are governed by the way that naming conventions and the way that image context effect such ranking. In other words, if a user titles an image file, they are in part defining the way that they see that image, and thus, how (or if) we see it. Labeling an image “Lindsey-Lohan.jpg” is an obvious descriptor and a quite different one than “gold-scarf.jpg”, though either one might appropriately describe the same, single image of Lyndsey Lohan in a gold scarf.
Whether the former or the latter is the one chosen is determined by the person naming the file, and their interest in the image. That interest is passed along in some sense to the user of an image search engine, since what they want is also governed by searching for a desired image of “Lindsey Lohan” or “scarves”. Additionally, words that carry a value judgment have even more ramifications on the way that the person naming the image and the user viewing it potentially feel about the image or how they will potentially “see” the image. Something like “slut.jpg” is going to produce a very different exchange of meaning between the publisher of the image and its viewer when encountered.
Thus, like emergent narratives that seemingly arbitrarily occur within video game worlds, Google Image Search may also seem quite arbitrary in the way that it pulls images out for our perusal. Nevertheless, underlying those images are both intended and unintended signification that has the potential to emerge out of consensus around particular ideas and the way that a group of users offering up those images perceive them.
A number of examples might be helpful in illustrating this idea. I should note that all of these examples provided here are based on searches done in about mid-April of 2010. Thus, a similar search may, at the time you’re reading this article, produce very different results. However, as we will see, the changing nature of the results page speaks directly to the interesting kind of liquidity of the collective consensus that in part informs Google Image Searches.
In addition to the obvious and utilitarian sorts of results that one might expect (images of maps of the United States, North America, and South America), much of the imagery that greets the user who enters “america” as a search term in Google Image Search is of a rather emblematic, and more specifically, patriotic quality. There is an awful lot of red, white, and blue on the page, usually in the form of rippling flags. The images that accompany the more obvious patriotic signifiers suggest a great deal about the self image that Americans have of themselves, especially assuming that the publishers of these images are associating them with their sense of “america” by naming them as such.
At the time of my recent search of the term, following the first image of a “map” is in fact a rather noble looking bald eagle set against one of those aforementioned rippling flags. More ennobling still is the dearth of images of Captain America, a muscle-bound image of good representing liberty and strength that sprawls across page after page of this particular search query’s results. Best of all in my search revealed an image of Captain America that appeared in the second row of the search (and actually the first of many Captain America images) that contained the good Captain squaring off against Hitler. The notion that the embodiment of American strength and liberty should be represented as battling the popular embodiment of 20th century evil suggests an American self image that holds itself in a rather high regard.
Super heroes are a common enough result in generalized search queries performed in English, though. Interestingly, while a search for the word “man” produces an initial image of figures of man from classic art as well as a black and white photo of a well muscled, shirtless male, images from Marvel comics abound, too—especially depictions of Spider-Man and Iron Man. Now certainly, part of the reason for such results concern the popularity of both figures in recent film and that their names include the word “man”, but the super hero as a gross exaggeration of masculinity serves as a reasonably archetypal ideal of masculinity, maybe more particularly for the Americans that are tagging and naming these particular images that Google Image Search is homing in on.
While a larger quantity of super heroes appear when querying “man” than when querying “woman”, nevertheless, images of Wonder Woman abound in this latter search, representing again an exaggeration both physically and morally of an American conception of woman. It’s also interesting that the first image that results from a query of “woman” is a painting of a pregnant woman. Once again, the most visually expressive qualities of the feminine (a chief biological distinction, the ability to carry a child) are ranked rather high by the search engine, but again, user contribution aids in the process of selecting out the most relevant images for Google Image Search to rank. This image of woman is generated partly in a seemingly arbitrary fashion, but also from a consensus of individuals tagging images of the pregnant woman and the superhero with a term that calls attention to their sense of the image’s femininity, be that Wonder Woman or a pregnant woman.
More provocative searches also yield interesting results and understandings of consensus on the ideas surrounding a term. Searching “cleavage”, again, results in a lot of what one would expect. The familiarity of faces attached to that cleavage is also probably unsurprising, both Salma Hayek and Scarlett Johansson make the first page of hits. The first figure of any notability that appears in the search, however, is Hilary Rodham Clinton. She appears twice before either of the infamously voluptuous stars, which seems surprising until you begin considering the way that Google Image Search ranks images according to the manner in which files are named but also due to the context surrounding them on the page.
Both images are derived from pages concerned with the “horror” of the possibility of the Secretary of State possibly showing cleavage in public. In addition to Google Image Search’s ranking, the naming of files and interests of the article also reveal a curious contentiousness in American interests in sexuality. While the images on the page reveal an interest in the female body as an object of sexual desire, the images of Clinton reveal a desire to also cluck our tongues at the female body (maybe especially the female body of a certain age or of a certain status) on display.
A consensus of some individuals that want to place boundaries on sexuality is as represented on the page as those who do not. A curious double-bind of the principles of the libertine and the Puritan that is observable in American life is on display in the search query results.
Consensus and perception can be seen emerging in searches though throughout the engine in words more obviously loaded and less clinically descriptive as my previous examples. When searching for the term “bitch”, I actually expected to see Clinton appear as, perhaps, the first person that I would recognize amongst the images that I would be presented with. I was making the assumption that perhaps Americans would be comfortable tagging these signs with a woman reviled by a significant chunk of the population with a term that would call attention to their displeasure with her.
Scanning the first page quickly, I didn’t see Clinton, nor did her image appear on the second page either. However, a familiar face did crop up on that second page, Kate Gosselin, making me reconsider who America currently might consider most indicative of “their bitch”. However, my expectation that “asshole” would result in an image of George W. Bush as the first notable and familiar face for “America’s asshole” was not unfounded.
If hit counts, links, and other factors additionally drive image rankings, a constant reconsideration of how America sees itself, sees men, sees women, sees cleavage, etc. should continually shift as a result of the kinds of queries made and the ways that images are represented by those who make them available for Google to spider change up their images. “Reading” those images is partly governed by the users’ perceptions of ideas and their sense of how patterns emerge amongst these images. This seems true in part because my observations about superheroes are partly governed by my own thoughts on the roles that super heroes serve as a kind of American pantheon and by the nature of the searches that I ran that happened to call up such images. In my defense, I should note that images of Superman appear when searching for “masculine”, so it probably is not purely coincidental that super heroes only crop up because their names include the specific term “man”. In the case of the Superman image, it is additionally one associated specifically with masculinity it would seem.
However, such interpretation is also governed by the way that a consensus of “namers” of images help to generate the patterns that we read. If enough folks keep labeling images of Kate Gosselin with the word “bitch”, that consensus will inevitably create a pattern that suggests how a group understands that individual. Thus, there may be good reason to recognize a consensus about what her image signifies: “bitchiness”.
This process of emerging signification is nothing new. Indeed, much of how signification occurs within a culture is related to consensus building about meaning. Attend enough funerals where everyone is wearing black, and the association of the color with the concept of funerary proceedings and death seems evident enough.
The process on the web though, like many other processes, may be accelerated by the rapid publication of web images, or it may solidify the meanings associated with these images even more concretely because it makes cultural consensus obvious (through its straightforward visuals), also as a result of these accelerated rates. Google Image Search has a tendency to reveal a possible lack of a purely arbitrary quality to this building of consensus. It’s the rules that define how images are prioritized in a Google Image Search that direct how these messages are sent and grouped. Far from arbitrary, the machine refracts meaning, governing how we “see” our ideas of ourselves in a visual form.