As previously stated, formulae are specific elements of language that are retrieved whole from the mental lexicon at the moment of speech production. In this sense, a number of proprieties distinguish formulae from words: “wide-ranging intraphrase flexibility”, “mandatory affective and attitudinal nuances”, “considerable length” (Van Lancker Sidtis 2012a: 358). The last point probably should be further refined on other occasion. Van Lancker Sidtis refers to them with a term “formuleme”, which is an item stored in the mental lexicon, a conglomerate of features relating to meaning (social and affective connotations), linguistic form (grammatical, phonetic, prosodic) and pragmatic contingencies (ibid.).The interest of researchers in formulae dates back to the mid-19th century, when the first studies of aphasiac patients revealed that some parts of language remain intact in their case (Wray 2002). Concepts interpreting the fixedness of idioms began to appear, e.g., in Bobrow and Bell’s (1973) model idioms are treated similarly to lexemes in memory. According to them, the analysis mechanism is as follows: literal interpretation is attempted first, and if unsuccessful, a switch is made to select an idiom form the lexicon. Further studies (Swinney and Cutler 1979) employing visual classification latencies did not confirm this theory, on the contrary, they revealed that idioms were processed even faster than non-idiomatic material (Wood 2015).

In one of the first reports in the neurolinguistic field, Winner and Gardner (1977) noted that patients with right hemisphere damage showed problems with a picture matching task involving formulaic metaphors. In the late 80s, a neuropsychological protocol defined a new term – pragnosia (Nelson et al. 1989), a deficit in the social use of language in persons with right hemisphere damage: