Metaphors are intriguing puzzles. How can it be possible that language users understand an expression which joins two or more previously unconnected objects and which is, in most cases, literally false? The character of this process of interpretation is largely obscure. Black (1976) has suggested that in metaphor like 'A is B', commonplaces that are generally associated with one object (the vehicle, e.g. 'wolf'), are projected on another object (the topic, e.g. 'man'). Appealing though this interaction view may seem, it does not tell exactly which associations are transferred, nor why. In the example 'Man is a wolf', it is apparent that only some of the associated commonplaces are relevant for the understanding of the metaphor (especially [aggressive]) and others are not (f.e. [hairy]) A systematic explanation of this selection process is lacking. In an earlier paper (Van Klink and Royakkers 1994), we offered a logical reconstruction of analogy comprehension by means of a step-by-step model of complexity reduction based upon a theory of similarity that was introduced in cognitive psychology by Tversky and has been further developed within the framework of a theory of metaphor by Ortony and Kittay. Our focus was on relatively new, 'living', metaphors in ordinary discourse, not on 'dead' metaphors, literal comparisons and metaphors in literary, scientific or other specific uses of language. In this paper, to begin with, we apply the model developed so far to reasoning by analogy in legal discourse. This results in an elaborated model of legal reasoning that provides criteria for evaluating any given argument by analogy. Next, we will discuss other formalisations of reasoning by analogy in legal theory and make a comparison between these formalisations and our model. Finally, we address the question in which respect(s) a general theory of metaphor can profit from studying legal reasoning by analogy. (Bart van Klink & Lamber Royakkers)