Last month I gave a lecture to the MSc Science Communication course at the University of the West of England, Bristol, on the Visual Communication of Science. I began talking about the images that scientists (specifically, molecular biologists) generate during the research process, through the use of drawing as a research tool, to a discussion of the ways that images might be used to communicate stories in science.
Images are visual representations of things – these may be objects, concepts, people, viruses, etc. (Are there examples of images that do not represent anything? Perhaps there are some examples in art?) Representation (re-presentation) is a much-discussed word in the Philosophy of Science. Nelson Goodman provides a good discussion of representation in his book about symbols, ‘The Languages of Art’ (1976, p.5):
The plain fact is that a picture, to represent an object, must be a symbol for it, stand for it, refer to it; and that no degree of resemblance is sufficient to establish the requisite relationship of reference. Nor is resemblance necessary for reference; almost anything may stand for almost anything else. A picture that represents—like a passage that describes—an object refers to and, more importantly, denotes it. Denotation is the core of representation and is independent of resemblance.’
Goodman discusses representation as distinct from resemblance. An image might resemble an object, and it follows therefore that the object will resemble the image; it is a reflexive, reflective, symmetric relationship. However, the same is not true for representation: an image might represent an object, but the object does not then represent the image. ‘Representation’ infers a time component: for an image to represent an object it will have been created after the object.
The example given in the Philosophy of Science class that I attended was that of Picasso’s portrait of Gertrude Stein. The ‘object’, the woman Gertrude Stein, is all aspects of Gertrude Stein, including her past, present and future selves, and thoughts about her by other people projected onto her. It is the complete set of attributes. The image, Picasso’s portrait, portrays only a portion of these attributes, chosen by the artist. The image therefore represents a specific version of the object. Gertrude Stein the portrait resembles Gertrude Stein the woman, and vice versa. However, Gertrude Stein the woman does not represent Gertrude Stein the portrait; the object possesses many more attributes than the image. Gertrude Stein the portrait only carries the attributes of Gertrude Stein the woman chosen by Picasso; he chose how the portrait would present her. Indeed, the head was completed in the absence of the sitter and reflects Picasso’s recent encounter with African, Roman, and Iberian sculpture. When someone commented that Stein did not look like her portrait, Picasso replied, “She will.”
Goodman goes further to suggest that resemblance is in fact not necessary for representation, which is particularly evident in scientific images. Following my Visual Communication lecture I ran a one-hour practical session in which the students explored some of the messages from the lecture in more detail. For one of the tasks I had printed out a large number of different types of images that are generated by molecular biologists, either from my own work or from published papers, distributed them so that each group had a selection of different kinds of image, and asked the students to discuss them based on elements I had covered in the lecture. The discussions went well but I next asked the students to arrange the images on a scale from ‘figuration’ to ‘abstraction’, which was not as straight forward as I had originally thought.
As previously discussed, the images generated by scientists (molecular biologists, although may extend to other areas of science) can be grouped into three categories. I have renamed these as in the original post I neglected to acknowledge that all images generated by scientists are representative.
1. Raw Data / Primary Images.
These are the images generated as part of the research process. They have not been manipulated, they are ‘objective’. Multiple images might be generated by a single experiment, and a single experiment would be repeated three or more times. Together, the single experiment and its repeats might uncover a single property of a system, for example of a protein’s function.
This first image above, made using the technique of thin layer chromatography, is an abstract representation of the phospholipids present in a number of extracted lipid samples. It would not resemble them if one were to take a photograph of the lipids in their native form.
This second image, a ‘photograph’ taken using immunofluorescence microscopy, might be thought of as being less abstract and more figurative. However, the cells here are artificially coloured using antibodies raised against specific proteins of interest and conjugated to fluorescent dyes. The detected components are depicted relative to one another in space; a huge quantity of information has been removed as a consequence of its not being detected. This image is a snapshot of the system in time, and may have been captured for its aesthetic properties rather than in a truly objective way.
In these types of image, raw data has been manipulated or processed in some way. Grey bands on western blot or TLC images, for example, might be subjected to densitometry, the results of several experiments averaged, and the data plotted on a graph. Diffraction patterns generated by X-ray crystallography are analysed to determine the positions of the atoms in a protein in space and plotted using molecular visualisation software. This processing of data requires that the researcher chooses a means by which to best show the data – different types of graphs may lay emphasis on different aspects of the data. In this way bias might be brought into the reporting of scientific discoveries.
Mechanistic diagrams contain multiple layers of information, derived from many experiments, often by multiple researchers. They contain information that is already known and published in the wider literature, the results of new experiments, and hypotheses to be tested in future experiments. They commonly appear in review articles and as the final, summary figure in primary research papers. They are an illustration of a system, employing standard symbols, and components at different scales or levels of organisation. The figure below was taken from a review article about the involvement of a particular hormone system (the renin-angiotensin-aldosterone system) in the progression of heart disease, and contains elements at the molecular, cellular and tissue levels.
Clearly all of these types of image represent (re-present) some aspect of the relevant system. The amount of information that each holds varies, and increases in the direction from raw data to mechanistic diagrams. The degree to which the image resembles the system also varies, but this does not depend on the image type but varies on an image by image basis. Is a scale of resemblance equivalent to a spectrum from from figuration to abstraction, where figuration closely resembles the system as it is seen with the human eye, or with the most basic of light microscopes, and abstraction where the image shares no similarity with the visual nature of the system the image represents? In the images above, the immunofluoresence microscopy image shows the most resemblance to the cells themselves (cardiac fibroblasts) (most figurative) since the image was made using a camera, and the thin layer chromatography image shows the least resemblance (most abstract). Interestingly, the mechanistic diagram exhibits both abstract and figurative elements.