David Gil
This paper introduces a distinction between two kinds of semantic compositionality, simple and relational, and argues that simple compositionality is phylogenetically prior to relational. The two types of compositionality are defined as follows:
- Simple Compositionality
The meaning of a complex expression is determined solely by the meanings of its constituents. - Relational Compositionality
The meaning of a complex expression is determined by the meanings of its constituents and also by its structure.
A general characterization of the distinction between simple and relational compositionality ispresented below:
| compositionality | semantics | form |
| none | X Y | X Y |
| simple | A ( X, Y ) | X Y, compositional licenses |
| relational | A ( X, Y ) constraints | X >, compositional licenses, constructional signs |
The above table represents the possible outcomes of the combination of two meaning-bearing elements X and Y. The first row represents the case in which X and Y fail to combine to form a new meaning-bearing element X Y, and there is accordingly no compositionality, while the second and third rows represent the two types of compositionality, simple and relational.
The semantics column makes reference to the association operator A. In its monadic form, the association operator corresponds in its interpretation to familiar genitive or possessive constructions. In its dyadic guise, in the formula A ( X, Y ), it means ‘entity associated with X and Y’. The association operator is the sina qua non of semantic compositionality, reflecting the fact that whenever two signs are brought together, the result is a new sign whose meaning is connected to the meanings of the two constituent signs. As suggested in the semantics column, associationality is present in both simple and relational compositionality. However, whereas simple compositionality consists entirely of associationality, relational compositionality involves additional constraints that narrow down the range of possible interpretations of the combination of X and Y.
The form column distinguishes between two kinds of elements that may be added to the simple combination of X and Y: compositional licenses and constructional signs. Compositional licenses are the glue that connect X and Y; some examples include spatial or temporal proximity, contiguity, and intonational phrasing. Constructional signs have more specific functions, narrowing down the otherwise wide range of interpretations made available by the association operator; their presence is what distinguishes the more highly constrained relational compositionality from the basic, unconstrained simple compositionality. Constructional signs come in a variety of flavours, including linear order and a wide variety of so-called grammatical items, characterized by an array of properties that include phonological reduction, abstract meaning, and reduced discourse prominence.
With these definitions in hand, this paper compares compositionality in human languages to that reported in various animal communication systems including those of bees, Putty-nosed monkeys, Diana monkeys, Campbell monkeys, and (captive) apes. In human languages, relational compositionality is widespread, though in some contexts simple compositionality is also observed. In contrast, compositionality in animal communication systems is mostly or exclusively of the simple type. The facts are argued to support an evolutionary trajectory from simple to relational compositionality, with instances of simple compositionality in contemporary human languages representing living fossils.