In 1914 the Danish grammarian Otto Jespersen coined the term 'metanalysis'. That's rather a dry start to a piece on what is a lively and intriguing facet of the English language. To find out what prompted Jespersen to believe that we needed a new word, let's bring in a stage prop - the humble orange.
Many sources will tell you that oranges were originally called 'noranges' and that 'a norange' migrated to being called 'an orange'. Well, like so much folk etymology, that's not true, but there is a germ of truth in it - there never has been a word 'norange' in English, but there very nearly was.
The climate in England doesn't qualify it as an orange-growing area and the fruit were first imported there in the 14th century. Oranges originated in South-east Asia and when they arrived in Persia and Spain they were given the names 'narang' and 'naranja' respectively. As they got nearer to England, and hence nearer to requiring a name in English, they lost the 'n'. This happened on their journey through France, where they were known as 'pomme d'orenge'.
In English, the indefinite article may be 'a' or 'an', depending on whether it is followed by a word which starts with a consonant or a vowel. When the consonant is an 'n', we may run into the 'a norange'/'an orange' confusion. It was this displacement of a letter from one word to another that Jespersen took an interest in and named 'metanalysis'. medieval words like 'a napperon', 'a nuncle' and 'a nadder' could easily be confused in everyday speech with 'an apron', 'an uncle' and 'an adder' - and they were. The earlier forms aren't now used.
The misaligning of word boundaries can go the other way too, with the 'n' being added rather than lost. The best known examples of that are 'nickname' and 'newt', which were originally 'an eke-name' and 'an ewt'.
It's easy for us to see these examples now as errors, but bear in mind that the changing of words based on confusion about where words start and end took place before dictionaries or even printing and reading were commonplace. When we come across new words now it is just as likely that we see them in print as to hear them spoken. If we had to rely on speech alone we might now be coining mutations like 'an erd' or 'a Niphone'.