In 2006 Alaska Sen. Ted Stevens infamously described the Internet as a “series of tubes.” He spoke in opposition to the concept of network neutrality, instead supporting the argument that large cable and telecommunications companies should be allowed to charge a premium to Amazon, Netflix, eBay, and other Internet businesses that require reliable Internet service. His comments quickly went “viral” and he was mocked by Jon Stewart, Stephen Colbert, and many others. Ironically, “tubes” is not a bad metaphor for some aspects of the Internet. Experts often use “pipes” to describe connections between computers and routers, and liken different pipes’ diameters to the varieties of network bandwidth. As we all know, this is how simple metaphors work: We use a familiar or shared concept (pipes) to help our listeners understand something new (bandwidth).
What got Stevens into trouble was relying on the metaphor for further details about how information moves across the Internet. His comments demonstrate his understanding of the pipes metaphor, but they reveal no grasp of the concept of a data packet or the process of packet switching. He said (rightly) that “the Internet is not something that you just dump something on. It’s not a big truck. It’s a series of tubes.” And then (wrongly), “if you don’t understand, those tubes can be filled, and if they are filled, when you put your message in, it gets in line and it’s going to be delayed by anyone that puts into that tube enormous amounts of material.” As an example, Stevens referred to an email (he called it an “Internet”) that allegedly took several days to reach his office. He seemed unaware that the principle of network neutrality, against which he was arguing, would eliminate network bias and ensure that everyone’s “Internet” got delivered equitably. But what was most frustrating to many of us listening to these remarks was the knowledge that one more metaphor would have helped the senator immensely: the postal metaphor, according to which Internet traffic is broken into a numbered series of “envelopes,” sent down that series of tubes, and reassembled in the correct order once delivered to our computers.
Conceptual blend theory can help explain what went wrong for Stevens . Whereas the common definition of metaphor suggests a single input source that helps explain the target (pipes help explain bandwidth), the theory of blends holds that figurative language draws from multiple input sources to create a new, blended space. The blended space selectively borrows properties from each input source to create new meaning. An example in the work of Mark Turner and Gilles Fauconnier describes a newspaper account of a catamaran in 1993 that was trying to beat a San Francisco to Boston sailing record set by a clipper in 1853 . The newspaper reported that as it went to press, the catamaran was “barely maintaining a 4.5-day lead” over the clipper. This is an example of a blend because it presents something that never existed (a race between the catamaran and the clipper), and it does so by selectively borrowing properties from each of the two inputs (from 1853 and from 1993). It also recruits structure from the frame of a “race” that never occurred; the frame is perfectly understandable to the reader because of prior experiences with such contests. The theory of blends helps us understand that when we use metaphor to convey meaning, we often are dealing with multiple input spaces rather than just one, and with frames of reference that are deeply embedded in our assumptions about the shared experiences of our listeners. Our use of figurative language to explain how to use our technologies will be effective only when our listeners (or readers or users) apply the appropriate (and disregard the inappropriate) properties from the various input spaces, and when they really have had the antecedent experiences required by the frames of reference we employ.
Blends can be seen as a conceptual space in which understanding does or does not occur. In describing the Internet as a series of tubes, Stevens created a blend that correctly highlighted the similarities of the tubes/pipes input domain, namely, the properties of length, diameter, joints, etc. He created appropriate structure for the blend by drawing upon a universal “conduit” frame in which information travels from one point to another. But he also applied inappropriate properties from the pipes input: for instance, the idea that pipes can be filled to bursting or clogging, or that they present a single path from A to B. His blend was overdetermined by the “pipes” input to the detriment of his overall understanding. (A similar phenomenon is evident in Stevens’s rhetorical question, “Ten movies streaming across that, that Internet, and what happens to your own personal Internet?” “Personal Internet” may be a beautiful oxymoron, but it is also the result of a blend that is overdetermined by the “personal computer” input, itself an interesting blend). If we mocked Stevens’s comments, it is mainly because we have access to other inputsmany of them also blended constructs such as “World Wide Web,” “store-and-forward packet switching,” or “routers”that help us better conceptualize how Internet traffic is managed.
Blends are a natural part of how the human mind works. Turner and Fauconnier argue that blends have defined human thought since the advent of language, approximately 50,000 years ago. So when we use figurative language to help people understand new technologies, we need to be aware of how blends function. Moreover, we must avoid metaphors that inadvertently encourage blends that impede usability.
The concept of an e-book or e-reader has already become almost meaningless as a result of marketers’ failure to consider how their metaphors will play out as blends in the minds of users. In this metaphor customers are expected to apply what they know about traditional books and electronic media to a new product, the electronic book. The blend that results borrows known properties of books (they have numbered pages that can be turned; they are shelved with other books in libraries, etc.), and known properties of digital texts (they can be delivered electronically; they are full-text searchable; etc.). The resulting blend should then define for the user the imagined properties of the new device. In an ideal situation, the result is improved usability, with the blend representing those properties of each source that are relevant to understanding the new product. Just as important, the blend should not represent inappropriate properties from the inputs. Nor should it import any properties from input sources that are altogether inappropriate. The process was nicely illustrated in a 2007 Charlie Rose interview with Jeff Bezos, Amazon.com founder. Rose imagined Bezos’s creative process in developing the Kindle: “When I first heard about this, my instinct was that the beginning of this for you was that, always being interested in books, and certainly understanding the digital world and electronic books, you had looked at the success of the iPod and said to yourself, why can’t we do something like this for books?”  Here Rose is creating a new mental space in which the properties of the Kindle can be understood in terms of multiple input sources (books, digital media, the iPod). For his part, Bezos immediately “corrects” Rose’s blend by explaining ways in which the Kindle is not like an iPod (it doesn’t require syncing with a computer, for instance).
But the e-books blend is not altogether effective, for at least two reasons. First, many properties from the known spaces, traditional books and digital media, are not adequately supported by e-book devices. Something as simple as page numbering is difficult to support on a device that lets users change font size. Something as natural as recommending a passage to a friend becomes difficult and awkward (“Check out page 36, FONT SIZE 4!”). So it is easy to find user support sites where baffled new e-book device owners trade secrets on how to repaginate their texts, or how to use device workarounds (bookmarks, “locations”) to recover some aspect of book reading that they assumed would be part of this new reading experience. Even the manufacturers’ official support sites are full of instructions on how to accomplish traditional-book tasks that most users would expect to be primary “properties” of a digital book, like how to view pictures; how to open and read a book; how to set and use bookmarks; and how to put a book into your library . Likewise, with Digital Rights Management protections, e-books on the Kindle and the Sony Reader do not carry the defining property of digital texts; one can not make perfect copies of them whenever one wishes. Instead it is now more difficult to loan a book to a friend. This is merely to say that it is difficult to know what properties from these two input sources should be projected into the e-book blend, and which should be left behind. Those users who import irrelevant or misleading elements from the “traditional books” or “digital media” inputs will produce blends that make the devices more difficult to use and understand.
A second reason the e-book blend fails conceptually is that it has been used to describe products ranging from the Amazon Kindle and Sony Reader; to Web-based multimedia versions of college textbooks accessed through “reader” software; to The New York Times’s “Reader,” which is simply a repackaging of seven days of newspaper content synced to one’s desktop or laptop computer. This promiscuous use of the book metaphor degrades its utility in explaining what this new product category is and how it works. Just how meaningless the blend has become is evident in the Yogiism expressed in the Rose-Bezos interview, as Bezos explains why the company needed its own hardware device: “We’ve been selling e-books for a long time; nobody’s been buying e-books” The solution to this problem? Another kind of e-book (the Kindle) that further confuses real readers (humans) who are trying to develop a conceptual understanding of this new product space.
Our use of metaphors in the way we speak about technology reduces usability when it produces blends that contain inappropriate properties from the input sources. Novice users are still baffled by the process of moving files between two remote computers because we insist upon using the vertical spatial metaphor of up- and downloading. Up and down might be called “image schemata”: constructs of human experience that our brains naturally comprehend because they have always been part of human experience . They contribute known properties (movement from high to low, or low to high; restrictions of gravity, etc.) to any blend that forms when they are used metaphorically, as when we speak of downloading video from a website or uploading content to a server. Because nothing is really moving “up” or “down” when data moves on the Internet (except in the case of satellite transmissions), novice users have difficulty understanding these processes and avoid using the devices that employ them. Listen to Rose describe one of his favorite features of the Kindle: “What would be great about developing an electronic book, a digital book? It would be if you did not have to go through the downloading process, it was all self-contained.” Users rarely express this level of trepidation when moving files from one local directory to another. Even “sending” email, and “attaching” files, makes more sense to such users, because these expressions more accurately draw from metaphors that suggest horizontal movement. We would never speak of “uploading” an email to a friend, even though, protocols aside, the process is technically very similar to adding a new page to a Web server.
And yet the problem is replicating itself today as we speak of “cloud” computing in the naive hope that this will mean something to our audiences. Bezos made a casual use of this metaphor in his interview with Charlie Rose: “When you make margin notes or highlight text,” he explains, “that also is stored and saved, and it’s saved on the server side too, it’s saved in the Internet cloud at Amazon so you can never lose those notes and marks.” Rose did exactly what most users do when we speak in such terms: He moved on to the next question, not wanting to reveal ignorance about what a “cloud” computer might be. Of course, he knows what a cloud is, and what a computer is, but he was unable to “run the blend” because there are very few properties of a cloud that could actually contribute to his understanding of this technology. Indeed, the most salient property of a cloud, its transience, is not something we would want users to associate with our data-storage solutions.
The current proliferation of touch interfaces will provide repeated opportunities to think through the challenges of blend theory. When Steve Jobs introduced the iPod features of the iPhone at MacWorld 2007, he began with the unremarkable claim that we can now “touch our music.” The phrase elicited silence from the audience, as Jobs’s listeners tried to make sense of this expression. They all knew what it meant to hold a CD and page through liner notes, and most of them probably remembered the look and feel of an LP album cover. They all knew what listening to music on an iPod Classic was like, the “touch” of smooth plastic and its revolutionary click wheel. But these two source spaces don’t converge easily into a blend that describes what it is like to tap, flick, and swipe through songs and Cover Flow artwork on the iPhone/iPod Touch, and I’ve never heard anyone talk about “touching your music” since then. The figure of speech doesn’t work, which is to say, it doesn’t help people understand how to operate the new interface.
Metaphors have been problematic in the field of interaction design ever since the Macintosh had us throwing good content into the trash can. Alan Cooper has written that metaphors are bad for interaction design for at least three reasons: They don’t scale well, they presume shared (antecedent) experiences that might not really exist, and they impede digital projects with the constraints of the physical analogs upon which they are often based . His superb advice is to focus instead on idiomatic design: interactions (such as right-click or pointing with a mouse) that don’t come naturally and don’t presume shared experiences, but that are extremely easy to teach and learn. Most “touch” interactions will fall into this category. And the expression “cloud computing” is really an idiom, a “colloquial” metaphor that makes sense within the context of technical manuals and the community of IT professionals, but not to others, to whom the expression is confusing. Our literary minds are immensely receptive to figurative language. Knowing about conceptual blends can help us think a little more deeply about how users might be interpreting the figurative language we use to explain how our devices work.
1. For a treatment of blends as a design principle, see Manuel Imaz and David Benyon. Designing with Blends: Conceptual Foundations of Human-Computer Interaction and Software Engineering. Cambridge: MIT Press, 2007.
5. In their 1998 essay, “Metaphors We Surf the Web By,” on how novice and advanced users perceive movement toward information on the Web, Paul P. Maglio and Teenie Matlock define image schemata as “basic pre-conceptual structures that arise from our embodied experience [and] shape both metaphorical and non-metaphorical thought.” <http://www.almaden.ibm.com/u/pmaglio/pubs/meta-4surf.pdf>
Charles Hannon is associate professor and founding chair of the information technology leadership department at Washington & Jefferson College in Washington, PA. He teaches courses in human-computer interaction, the history of information technology, data presentation, and project management, among others. He is the author of Faulkner and the Discourses of Culture. More recently, he has published widely on the role of educational technologies in higher education. His current book project is Usable Devices: Mental and Conceptual Models, and the Problem of Contingency.
©2009 ACM 1072-5220/09/0500 $5.00
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee.
The Digital Library is published by the Association for Computing Machinery. Copyright © 2009 ACM, Inc.