In the late 1970’s, linguistics researchers began to realize something interesting about our day-to-day conversation: we tend to use metaphors far more frequently than anyone had recognized. The reason it took so long to notice is that most of the metaphors we use are deeply engrained in our thinking. So deeply, in fact, that they’re almost unrecognizable for what they are. When we say things like “he’s in trouble again” or “I was on edge” we’re actually using the physical patterns of containment and relative centrality to convey the abstract concepts of distress and apprehension. In other words, we use aspects of concrete bodily experience to communicate metaphorically about abstract things. But the fascinating thing is not that we use these kinds of metaphors to express ourselves. That’s just a byproduct. The really profound discovery was that these kinds of metaphorical processes are central to the way we think.

Previously, linguists had considered metaphor a peripheral, largely rhetorical linguistic phenomenon. You know, something that playwrights and poets used to add a little descriptive color. “The moon’s a balloon” and “Juliet is the sun,” etc. As they investigated further, these researchers came to understand that not only is our use of metaphor pervasive in everyday conversation, but most of the metaphors we use are based on aspects of bodily interaction with the world. Further, these metaphors operate at a preconscious level- they are immediately and directly understood by our System 1 intuition engine. The simultaneous discovery of what has come to be known as conceptual metaphor by George Lakoff and Michael Reddy has had a profound impact on our understanding of how we understand the world. But before I get into that, it might be helpful to start with a little background on what metaphor is and how it works.

Metaphor is a special case of the more general cognitive phenomenon of analogy. Formally, analogy is the process of using the features of some well-understood source domain to understand some less well-understood target domain. Elements of the source domain are first projected onto elements of the target. Then characteristics of the source are used to make inferences about aspects of the target. According to Gentner’s structure mapping model of analogy, only the structural relationships among the mapped elements play a role in the inferencing process. In other words, the specific attributes of the source and target elements are ignored. So, for example, in Rutherford’s famous insight relating the structure of a hydrogen atom to that of the solar system, he was making an analogy in which the Sun is conceptually mapped to the nucleus of the atom and a planet is mapped to an electron. This enabled him to infer that electrons orbit nuclei in the same way that planets orbit the sun. In this analogy, the fact that the sun is huge and hot while protons are tiny and (relatively) cold is irrelevant. What’s important is the systematic structural relationships among the mapped elements, in this case the previously understood orbital movements of the planets vs. the inferred orbital movements of electrons.

Metaphor differs from analogy in that the source and target domains are conceptually (that is to say, relationally) dissimilar†. “A hydrogen atom is like a tiny solar system” is an analogy because the structural characteristics of the source and target domains are very similar (even though we know today that hydrogen atoms are nothing at all like solar systems). “Juliet is the sun” is a metaphor because the source domain (the sun) is conceptually very different from the target (a person). What makes this statement meaningful is our astounding ability to make the connection between the overwhelming physical impact of the Sun’s radiance and the overpowering emotion than Romeo feels in Juliet’s presence. Very different conceptual domains, but an almost identical cognitive mechanism.

Conceptual metaphors differ from linguistic metaphors like these in that they are so entrenched in our minds that they function as first order cognitive processes. Additionally, most of them are based on bodily (or “embodied”) patterns of physical experience like movement, object manipulation, seeing, and eating. It’s thought that patterns of experience like these become entrenched in our minds during early childhood. Later, we repurpose these patterns as metaphorical source domains for use in abstract thought. In this way, we’re able to understand and reason about highly abstract phenomena like time and learning in terms of concrete and familiar patterns like distance (“our deadline is close“) and eating (“please digest Chapter 6 before next class”).

In conceptual metaphor theory, each source domain is characterized by a hierarchically organized schematic structure or frame (also sometimes referred to as a kinesthetic image schema). The components or “slots” of a frame are called semantic roles. Semantic roles are structurally and often dynamically related. Concepts in the target domain are mapped to semantic roles in the frame. Together, the semantic roles and relationships serve to organize knowledge about the target domain such that it can be understood. Frequently, more than one frame is used at a time, in which case the assembly as a whole is called a metaphor complex. A few examples should make these concepts crystal clear:

Conceptual Frame Semantic Roles Linguistic Expression
Containment A Container with an Interior and an Exterior, Contained Objects in love; file it away for later; he’s an outsider; I’m not into it
Path Following A Starting Point, a Destination, a Contiguous Path between the two, and a Path Following Entity follow the instructions; this is going nowhere; connect the dots
Blockage A Path, a Moving Entity, and an Obstruction he’s in my way; don’t get hung up; the legislation is blocked
Rough/Bumpy-Smooth A Surface, Movement over the Surface, a Moving Entity smooth sailing; hit some bumps; went through a rough patch
Center/Periphery An Entity, a Periphery, and a Center on edge; fringe element; central role; borderline illegal

Interestingly (and naturally), our use of conceptual metaphor extends to the domain of thought itself. Lakoff and Johnson (1999) point out that our conceptions of thinking seem to revolve around the following three primary metaphors:

  • The Mind Is [conceived as] A Body
  • Mental Functioning Is [conceived as] Bodily Functioning
  • Ideas Are [conceived as] Objects of Bodily Functioning

These serve as the basis for a huge variety of related metaphors, including the following four key subclasses:

  • Thinking Is Moving
Submetaphors: Ideas are Locations; Communicating is Leading; Understanding is Following
Examples: reach a conclusion; go off on a tangent; go step-by-step; do you follow me?
  • Understanding Is Seeing
Submetaphors: Ideas are Things Seen; Communicating is Showing
Examples: see what I mean; point of view; shed light on; provide a clear description
  • Thinking Is Object Manipulation
Submetaphors: Ideas are Objects; Communication is Sending; Understanding is Grasping
Examples: turn it over in your mind; toss the idea around; you gave me an idea
  • Thinking Is Eating
Submetaphors: Ideas are Food; Communication is Feeding; Understanding is Digesting
Examples: raw facts; half-baked idea; digest it; don’t swallow that; feeding them propaganda

< >
Note that there is more than one way to skin a cat, as it were. Communication, for example, can be construed as Showing, Sending, or Feeding. The specific frame employed is determined according to context and/or the intent of a communicator.

So, why is this important? Well, among other things, conceptual metaphor theory helps explain the power and utility of abstract data visualizations. We find such visualizations useful because they graphically depict relationships among abstract concepts in concrete, spatial terms. Abstract concepts like comparative approval ratings become instantly understandable when represented in terms of concrete physical patterns such as the relative lengths of bars in a bar chart. Similarly, pie charts enable us to construe proportional values in terms of the relative sizes of the components of a (single) physical object, i.e. in terms of the Part-Whole frame.

Beyond meaning construal, conceptual metaphor theory also helps explain the interactive aspects of computer-based visualization tools. For example, zooming to reveal “details on demand” mimics our experience of being able to see increasing detail in objects as we physically approach them. Some visualization tools (e.g., those that present dynamic graph layouts) even go out of their way to simulate interaction with physical systems. If Thinking Is Object Manipulation, and visualization systems serve as a kind of external memory, then perhaps interactive visualization tools are useful exactly because they enable us to generate, store, and interact with concepts just as though they were physical entities. Food for thought.

Perhaps more importantly, a recognition of the role that conceptual metaphor theory plays in visualization construal (let’s call it, say, graphical conceptual metaphor theory for now) provides a principled foundation for designing new, more effective data visualizations. More specifically, the catalog of conceptual metaphor frames developed by the linguistics community can serve as a source of design patterns for visualization engineers. I’ll flesh out this strategy in more detail in the coming months. In other words, I’ll show you what I mean.

[zotpress items=”HP79UQ6N,59ZW54V6″ style=”association-for-computing-machinery”]

†Gentner considers metaphor and analogy to be largely the same thing. I’m basing the definition given here on Lakoff and Johnson’s. In this reading, L&J’s “analogy” roughly corresponds to Gentner’s “literal similarity” and their “metaphor” corresponds to Gentner’s “analogy.”