These numeration systems are usually broadly classified as being either positional or nonpositional. However, this dichotomous classification is somewhat coarse. To see this, consider the Roman numerals. These are often cited as the standard example of a nonpositional numeral system. However, if the Roman numerals were truly nonpositional, then we would expect IV and VI to represent the same value, which they do not. On the other hand, Roman numerals certainly do not form a positional system; they appear to be a broadly nonpositional system with some positional features. Conversely, there exist positional systems which display some of the features of nonpositional systems—specifically, the Babylonian sexagesimal [9, Chapter 26] and Maya vigesimal [9, Chapter 28] positional systems. Unlike the Hindu-Arabic numerals, each of these systems can have more than one symbol per position. The Babylonian system has dedicated symbols for 1 and for 10 which are then combined additively to make up the value (B59) in any given position. Thus, if we confine our attention to a single position, we find a nonpositional numeral system at work. This is evidence that the Babylonian sexagesimal positional system evolved from an earlier Sumerian decimal nonpositional system (which we will see in the next section). We find a similar phenomenon with the Maya system: there are dedicated symbols for 1 and for 5, which are used to make up the value in each position, hinting at the earlier use of a quinary nonpositional system. Clearly, a finer classification than the traditional positional-versus-nonpositional scheme is called for. The existence of such ‘‘mixed’’ numeral systems is acknowledged in [1], and a simple system of classification appears in [9, p. 429], into positional, nonpositional, and so-called hybrid systems; details can be found in [10, pp. 34–63]. The classification in the present paper will be rather different from that of Ifrah. The goal of this paper is to obtain a finer classification for nonpositional systems. Note that my analysis and classification will be of the mathematical and symbolical structure of numeral systems; a linguistical analysis of numeral systems has already been given in [15], for example. I will use the abbreviation ‘‘NNS’’ throughout to mean ‘‘nonpositional numeral system’’. Throughout this article, the word ‘‘number’’ will refer to a quantity, whilst the word ‘‘numeral’’ will describe a symbol. Thus, for example, ‘‘2’’ is a numeral denoting the number 2. However, since the Hindu-Arabic numerals will not be subject to analysis, there is no danger of confusion in our using these to denote numbers. Nonpositional numerical notations are by far the most common throughout history. They are characterised by the feature that the position of a particular symbol within the representation of a number is irrelevant. We may regard this as a consequence of the ‘‘additive’’ nature of such systems: juxtaposition of symbols denotes addition. For example, if h and j stand for 1 and 10, respectively, in a