Pseudo-Articulatory Representations (PARs) have been proposed and discussed in relation to speech processing with results reported for both synthesis and recognition. PARs are derived from linguistic specifications of articulatory activity which are both abstract and idealized. The abstractions and idealizations permit the linguistic generality to be distinguished from the articulatory reality; this is what we need in speech processing. PARs attempt to retain the linguistic generality whilst also gaining some realism through adoption of continuous articulatory feature values; the latter permits mapping to acoustic values. PARs promise a way of mapping acoustic (and other) data onto linguistic specification of speech (and vice versa), and although enough progress has been made to demonstrate the concept various problems remain for further study.