Although there are now vast numbers of digital images available via the Web, it is still the case that not enough is known or understood about how humans perceive and recognise image content, and use human language terms as a basis for retrieving and selecting images. There is an increasing belief that the difficulties of image management and description should be led and defined by the needs of users and by their information seeking behaviours. The Focusing Images for Learning and Teaching – an Enriched Resource (FILTER) project is investigating, through an online image description exercise, the ways that users describe different types of images. Through analysis of the exercise results, FILTER hopes to obtain an understanding of the ways in which people describe images and the factors that influence their approaches to image description and thus appropriate metadata. Preliminary analysis of data indicates that there is little consensus on the use of terms for image description or on categorisation of images into 'types'.