If we wish to explain pictures, in the sense of expounding them in terms of their historical causes, what we actually explain seems likely to be not the unmediated picture but the picture as considered under a partially interpretative description. This description is an untidy and lively affair. Firstly, the nature of language or serial conceptualization means that the description is less a representation of the picture, or even a representation of seeing the picture, than a representation of thinking about having seen the picture.
There was a countryside and houses of a kind appropriate to peasant country-people – some larger, some smaller. Near the cottages were straight-standing cypress trees. It was not possible to see the whole of these trees, for the houses got in the way, but their tops could be seen rising above the roofs. These trees, I dare say, offered the peasant a resting-place, with the shade of their boughs and the voices of the birds joyfully perched in them. Four men were running out of the houses, one of them calling to a lad standing near – for his right hand showed this, as if giving some instructions. Another man was turned towards the first one, as if listening to the voice of a chief. A fourth, coming a little forward from the door, holding his right hand out and carrying a stick in the other, appeared to shout something to other men toiling about a wagon. For just at that moment a wagon, fully-loaded, I cannot say whether with straw or some other burden, had left the field and was in the middle of a lane. It seemed the load had not been properly tied down. But two men were trying rather carelessly to keep it in place – one on this side, one on the other: the first was naked except for a cloth round his loins and was propping up the load with a staff; of the second, one saw only the head and part of his chest, but it looked from his face as if he was holding on to the load with his hands, even though the rest of him was hidden by the cart. And as for the cart, it was not a four-wheeled one of the kind Homer spoke of, but had only two wheels: and for that reason the load was jolting about and two dark red oxen, well-nourished and thick-necked, were much in need of helpers. A belt girded the drover's tunic to the knee and he grasped the reins in his right hand, pulling at them, and in his left hand he held a switch or stick. But he had no need to use it to make the oxen willing. He raised his voice, though, saying something encouraging to the oxen, something of a kind an ox would understand. The drover had a dog too, so as to be able to sleep himself and yet still have a sentinel. And there the dog was, running beside the oxen. This approaching wagon was near a temple: for columns indicated this, peeping over the trees . . .
This – the greater part of a description written by the fourth-century Greek Libanius of a picture in the Council House at Antioch – works by retailing the subject-matter of the picture's representation as if it were real. It is a natural and unstrained way of describing a representational picture . . . and one we still use. It seems calculated to enable us to visualize the picture clearly and vividly: that was the function of the literary genre of description, ekphrasis, in which it is a virtuoso essay. But what really is the description to be considered as representing?
It would not enable us to reproduce the picture. In spite of the lucidity with which Libanius progressively lays out his narrative elements, we could not reconstruct the picture from his description. Color sequences, spatial relations, proportions, often left and right, and other things are lacking. What happens as we read it is surely that out of our memories, our past experience of nature and of pictures, we construct something – it is hard to say what – in our minds, and this something he stimulates us to produce feels a little like having seen a picture consistent with his description. If we all now drew our visualizations – if that is what they are – of what Libanius has described, they would differ according to our different prior experience, particularly according to which painters it made us think of, and according to our individual constructive dispositions. In fact, language is not very well equipped to offer a notation of a particular picture. It is a generalizing tool.
– from Patterns of Intention: On the Historical Explanation of Pictures (Yale University Press, 1985) by Michael Baxandall