on formulaic language, handed down and preserved by religious specialists in books,
which unfortunately no longer survive. So important was precise duplication of
formulae that magistrates recited prayers following a priest’s dictation. Such precautions
were especially important in situations where the archaic language of a traditional
prayer might even mean that the priests themselves did not understand it, as
Quintilian remarks of the ancient Salian hymn to Mars (Institutiones1.6.40). Failure
to follow the prescribed format was thought to impair the effectiveness of a prayer,
as shown in an incident described by Livy (41.16.1). At the annual Latin festival,
when the magistrate from the town of Lanuvium neglected to name the Roman
people in the list of beneficiaries of divine favor, it was necessary to perform the
ceremonies over again.
Scholars have often interpreted such emphasis on words to be indicative of a
magical quality in certain ritual behaviors, in that the effectiveness of the ritual seems
to be dependent on human technique rather than on divine power. But magic is no
longer seen as a useful hermeneutic category, for its distinction from religion, fre-
quently informed by polemical ends, is difficult or impossible to define with any fixity.
It also seems hopelessly bound to a positivistic view privileging modern religions over
so-called primitive magic. A more productive approach focuses on the techniques by
which language acquires efficacy.
The Latin word carmencrystallizes the Roman understanding of the power of words.
Romans applied the term to both prayers and hymns, but also to magical incanta-
tions. Putnam (2001: 133) defines the word’s original sense as “a verbal utterance
sung for ritualistic purposes.” The dominant feature is a style of expression charac-
terized by formulae, redundancy, and rhythm. So prominent was the rhythmical qual-
ity that by the late second century bcthe word carmenreferred primarily to poetry.
Closely related to the root of the verb “to sing” (*can), the word carmenfrequently
appears with verbs based on that root (Ernout and Meillet 1959– 60: 101). For the
Romans, at least formally, there was no distinction between prayer and spell and
poetry and song; all were intimately linked to one another.
Prayer as Performance
Austin’s theory of performative language provides a fruitful approach to understanding
the ancient emphasis on the power of ritual words. In Austin’s sense of the word,
prayers are performative, that is, speech acts that perform actions. Roman prayers
were not simple locutionary acts, addressing statements to the gods or describing
the human position vis-à-vis the gods. Rather, prayers were the performance of peti-
tion or promise or thanks based on an accepted convention in ritual contexts. This
performative quality is perhaps most clearly understood in prayers of thanksgiving.
In saying the words “I give thanks that,” the worshiper explicitly performs the act
of thanksgiving. While a gift or sacrifice often accompanied the prayer, the function
and efficacy of that offering was dependent on the words of thanks. Furthermore,
it is significant that prayers of thanksgiving always made mention of the divine act
that merited the offering of thanks. This is the act of praise that is so closely bound
236 Frances Hickson Hahn