Here we paused to review what we were doing. With each new combination and reordering of the images, we were forming an immediate impression of their “fit.” With each change, the force of attraction between the two images grew stronger or weaker. Meanwhile the quality of attraction between them kept varying, too: some images “go together” because of their colors, others because of the figures represented, others because an emotional response, and so on. Although there would have to be many different ways for the images to fit together, we seemed to be evolving a certain repertoire of combinations. Could we draw up a typology of relationships between images? That was an easy question: yes, it seemed possible. Would a typology of relationships be different from a typology of images themselves? That was a hard question: maybe, maybe not.

Here we arrived at a crucial theoretical point. “Do you think it is possible, in principle, to draw a connection between any two images in this stack?” It seemed like an easy question. Everybody nodded.

“Why? Why do we assume that we can always connect any two images? Is it because we think that we can always invent a connection? Or because we think that the connection is already there, and it’s just a matter of finding it?”

Those were the only choices, and I asked for a vote. About two thirds of the group said that our ability to connect images was based on our ability to invent our own connections; the other third insisted that our work consisted in finding the connections that already existed. In other words: the larger group thought that our ability to connect images derives from imagination; the other group thought that it is grounded in intuition.

“Let’s treat the distinction between imagination and intuition as fundamental,” I proposed, although we did not have the time to elaborate on the historical and philosophical consequences. For our purposes, “imagination” would be understood as the fabrication of something new, and therefore as a kind of “transcendence” of the two initial images by a third image or idea. By contrast, “intuition” would be understood as the discovery of an “immanent” path between the first two images, the revelation of a link that must have already existed. (Once the distinction was stated in these terms, various people began to supply philosophical and theological references to amplify it.) For the moment, however, we needed to name this distinction in order to sort through the various ways in which we seem to choose “spontaneously” between stronger and weaker pairings. A great deal of knowledge, habit, and sensitivity comes instantly into play every time we encounter multiple images, and we were trying to slow down the process in order to see how it actually works.

A line from Godard fit the moment perfectly: “In order to see one image, you need two.” Or to be more categorical about it: we can never see just one image, because are always looking to see what is around it, what comes before it and after it, what other images are summoned up or blocked from view. It became easier to conceive of the way every image belongs to immense webs, endlessly sorting themselves into new configurations.

Yet the realization that images are always tangled up with each other—whether that tangling happens “in our heads” or “out in the world”—raised a new question. “What does the first image know about the second?”

It is a disorienting question: we are likely to say that the first image never “knows” what the next image will be. But that is clearly not true when we are watching a movie or a TV show. In the vast majority of cases, in fact, each image is meticulously crafted to prepare for the next one. In general, cinema and television *never* require us to construct relationships as complex and uncertain as the ones we were building between the postcards. Our capacity to “think through images” is usually confined to a very narrow range of tasks.

Now I divided the stack of postcards among small groups of people, giving each group about a dozen cards. They had to pick three of them to make the “best” sequence or set. The room buzzed as they sorted through the possibilities. Finally, each group placed their chosen cards on the table in front of them. Some chose to present a simple left-to-right sequence, while others configured their cards in some idiosyncratic cluster like panels on a comic book page.

The first round of the postcard game produced some remarkable combinations: movie posters and portrait photographs; strange constellations of artworks; cartoons clashing with political slogans; landscapes alongside cute animals; baroque interiors set off against desert wastes, connected by a fragile-looking window. It did not particularly matter whether the players recognized that a certain picture portrayed Jean-Paul Sartre, or that another as a painting by Dürer, or that a sixteenth-century image was sitting beside a nineteenth-century one. Or rather, such recognitions might matter, but only if the other images could draw out and inflect the relevant aspects. A statue of Joyce in Dublin was quite directly reinforced by a painting by George Grosz. When that combination was extended to a study by Vladimir Tatlin, a whole set of questions about modernism rose into view. When that trio was extended again to a piece of medieval Chinese calligraphy, those questions gained a new axis and dimension. Rather than forming a closed set, such sequences constructed an open-ended questionnaire that invited further testing, research, selection and expansion.

Each of the postcard montages deserved to be examined, tweaked and elaborated at length. Each seemed like the sketch for a long train of thought. But it would have taken many hours of discussion to sort through all of them. For the time being, all we could do was to remain at the level of first impressions and simply ask which sequences or sets worked best. After showing all of them around the room, there was a fairly strong consensus about the “best” ones, and the fact that we could reach consensus so quickly and easily seemed significant. It might indicate that we shared common habits or even prejudices, or on the contrary it might suggest that we would be able to use such images to teach and learn from each other more readily than we had previously suspected.

We tried another round of the game, this time aiming to compose a “concept” in three images. As examples of concepts, I mentioned “happiness,” “beauty,” and “truth.” This time, the composition process went very quickly, and yet the results were not very satisfying. The images composing “beauty” were indeed beautiful, and the images of “happiness” were unmistakably cheerful in subject or tone. But the images tended to become subordinate, as if they could be no more than partial illustrations of a single word. This phase of the game would have to go through many more rounds before we could reach the point where the images proved strong enough to force the concepts to grow new dimensions.

More than two hours had passed and we were out of time. It was not possible to begin the third experiment, where each group would compose a picture of “the world” using postcards. So instead, as a final challenge, I offered another two-sided choice. Would they prefer to sort through the stack of postcards in search of the very best picture? Or would they prefer to use as many of the postcards as possible, perhaps even all of them? In other words: would their construction of a “world picture” tend toward fewer and fewer images, perhaps leading to one all-encompassing image? Or would it tend to spread out across many images, perhaps even toward all images?

As we debated these alternatives, I added one more possibility: perhaps “the world” is precisely what escapes every connection we make.