At a recent academic conference on “Thinking through Drawing,” two British rapporteurs addressed the audience in tandem, and both pronounced the word drawing with an epenthetic [r] between [aw] and [ing], so that the word consistently came out as [dráwring]. To contemporary American ears this pronunciation––called the “intrusive r–– sounded utterly alien, although it is common in British English as well as in the dialects of eastern Massachusetts (recall the speech of President John F. Kennedy) and is of a piece with the so-called “linking r” of I saw[r] it, etc.
The intrusive r, being unjustified orthographically and hence etymologically inauthentic, is not considered to be orthoepic, i. e., not part of the so-called Received Pronunciation (RP = “The King’s/Queen’s English”), and has in fact traditionally been regarded as a vulgarism, its use (but not that of linking /r/) banished by prescriptivists in England since the nineteenth century.
This phenomenon, traditionally subsumed under allophonic variation or automatic alternation, also goes by the name of sandhi, a Sanskrit word meaning ‘joining’ imported into general linguistics about a century ago. Phonological sandhi rules, as in the case of intrusive [r], beside their secondary binding function (“textual cohesion”) have two primary functions, systemic and (what has been called) “metaphonological.” When such rules produce distributions of distinctive feature values that are diagrammatic of the distinctive or allophonic value of the feature at issue, as well as of the markedness relations that obtain between different values of the same feature, they fulfill a systemic function, which is semiotically iconic. But beyond reflecting the structure of a distinctive feature system iconically, the distributional facts also constitute the raw material for language acquisition, since they are the data from which learners infer their phonology. It is in this latter sense that the sandhi rules may also be said to fulfill a metaphonological function.