Sir

Julio M. Ottino in his Commentary “Is a picture worth 1,000 words?” (Nature 421, 474–476; 2003) divides images into two categories: those that convey data and those that illustrate scientific ideas. He defends the practice of image manipulation as sometimes being a necessary part of the process of discovery, yet expresses concern about the blurring of the line between fantasy and reality in scientific illustration.

It may be that the two categories are not that distinct. The European Space Agency's programme Innovative Technologies from Science Fiction for Space Applications assumes that even the most fantastic illustrations may be a useful stimulus to science. On the other hand, images that purport to convey factual data may convey something else entirely. Many observers, for example, were convinced that they could see little human beings in the sperm images produced by the sixteenth-century microscopist Antoni van Leeuwenhoek, and they recorded this observation as fact, presumably influenced by their beliefs.

Today, we can and should capture not just an image but information documenting the process of image creation itself, from the original unmodified data to the final web-ready or journal-ready artwork. If questions arise about the interpretation of an image, we need to be able to go back to the raw data or, if this is impossible, at least to have a full record of what was done to it, and why. For example, in performing video-enhanced contrast microscopy, one always subtracts a digital background 'mottle' image from the live video stream to obtain the mottle-free video frames that are recorded, and viewers need to be informed that background subtraction has been carried out. Similar arguments relate to the point spread functions used in preparing the deconvolved fluorescence images mentioned in the article.

Of course, as in the days of pen-and-ink illustration, scientists should still consider the purpose served by each image in their publications, and should make these objectives clear to their readers. Scientists today have the additional responsibility of recording the processes by which images are created, so that these can be accurately replicated. Currently, such information is usually held only in the laboratory from which the image came, if it is recorded at all.

The BioImage Database Project (http://www.bioimage.org), part of ORIEL (Online Research Information Environment for the Life Sciences; http://www.oriel.org), will be a searchable database of multidimensional images of biological specimens. From its outset, we have felt it essential to acquire not just high-quality source images, but also the various images derived from them, and detailed metadata documenting the process of their creation. We believe that this approach addresses many of the issues raised by Ottino, and should be far more widely adopted.