One view (from Russell) suggests that any name is associated with a description which is true of a unique individual and it is via this description that the name refers to that person. Another view (Kripke) sees reference as involving an initial act of naming to which later uses of a name are connected by a causal chain linking one user or use to another. Each user succeeds in referring to the same thing because their use of the word is appropriately causally connected to the first use. So for example if I use the name ‘JRR Tolkein’ I succeed in referring to the author of LotR not because that name is attached to a description that uniquely fits him, but because I learned the name from someone who learned the name from someone…who was present at the christening of JRRT.
The Madagascar example was proposed by Gareth Edwards as a problem for this causal account, because it is a real-world case of reference shift via error (as opposed to reference shift via deliberate reapplication of a name such as calling a cat ‘Lenin’). The story goes that Marco Polo was the first European to learn and use the name ‘Madagascar’, but he applied it to the large island off the east coast of Africa while in fact the users of the term from whom he acquired it used it to refer to part of the mainland. It is assumed that Marco Polo intended to use the name as they did, but he made a mistake about what they intended. His mistake then led to the modern use of the word to refer to the island, not the mainland. The question for the causal theory of reference is: how can the name ‘Madagascar’ as used today refer to the island (as it clearly does) when its causal history leads back ultimately to the naming of a different place altogether?