Notes on the semantics of compound words

The following are some of my notes for my paper on ὑδροποτέω at SBL2018 in Denver. These represent an initial sketch of reasons why it is illegitimate to simply define a compound on the basis of the meaning of its constituent parts. That is: why ὑδροποτέω should not simply be interpreted and, in turn, translated as “I drink water.”

Compounds are complicated. They are formally complex, involving wide variation in their morphological/lexical formation. These formal complexities introduce their own series of semantic challenges.

In the literature, a number of accounts of compound formation have been put forward. Bisetto and Scalise (2005) is fairly widely adopted and is used here. Their analysis typology or compounds involves two levels of categories. The first is morphosyntactic in nature, considering the formal relationship of the compounds relative to their constituent lexical categories and syntax. Subordinate compounds describes those where the head takes an argument, as in taxi driver or cutthroat, where both taxis and throat would otherwise function as complements of the head verb: drive a taxi, cut a throat. Attributive compounds involves property attribution to the compound head: blue cheese, spam mail, and our blackboard/whiteboard examples above. Finally, coordinate compounds are those where the lexemes share the same lexical category and maybe conceived as having two heads rather than one: painter poet (Scalise and Bisetto 2011)—though we expect that one lexeme in the compound can receive inflectional morphology for the compound as a whole (i.e. *painters poet is ungrammatical pluralization).

The second type categorization is semantic in nature, pertaining to the lexical categorization of the compound relative to its constituent parts. Each of the three types (subordinate, attributive, and coordinate) can be also categorizes as either endocentric or exocentric. Endocentric compounds are those where the compound function as a hyponym of the compound’s head. A hyponym is a term that is superordinate to another term: the word driver is superordinate to taxi driver, board is superordinate to blackboard. Exocentric compounds are those that lack this categorial quality. A pick pocket is a subordinate compound because pocket would function as the object of the verb pick, but it is exocentric because there is no superordinate category of “pickers” wherein a pick pocket is a type. Similarly, a killjoy has a syntactic relationship: John killed all the joy in the room, but kill is not a hyponym of killjoy. Here are a few other examples:

Subordinate    copyeditpickpocket
Attributive      blue cheeseblue collar
Coordinate      singer-songwriternortheast
A classification of types of compound words

Copyediting is a type of editing, blue cheese is a kind of cheese, singer-songwriters are types of both singers and songwriters, but this is not true of pickpocket, blue collar, or northeast.

So that is a useful formal or typological context for how we can categorize different compounds, but what about their semantics. Here, they are arguably even more complicated. Compounds, as a class of words/lexemes, feel simple at first blush, but that simplicity is deceptive. The definitions of introductory texts on morphology, such as Bauer (2003) where a compound is presented as, “the formation of a new lexeme by adjoining two or more lexemes,” feeds this sense of deceptive simplicity. The naïve linguistic assumption is that the formal composition is simple in nature and the meaning of the compound is merely the sum of its individual parts. These categories are not merely formal/structural, however. They also implicate an iconicity in their semantics. Coordinate compounds involve a relative equality in their internal morphology that makes it difficult to attribute structural headedness to one lexical component over the other. For the compound singer-songwriter, both lexical categories are fully activated, and both evoke their respective frames. As a result, coordinate compounds are also the most transparent in their semantics. This is still true for northeast, albeit to a lesser degree. The mental representations for north and for east are in conflict with each other. The compound does not mean its constituent parts, but rather compromise between the constituent parts: north and east are reconciled in a cardinal direction relatively spaced between the two.*

*And of course, this is then complicated by the homonym of deictic/directional northeast: definite Northeast, which usually appears with the article. The meaning of Northeast is even more culturally bound to and constrained by more specific geographic frame, one that varies depending on where one lives in the world and the cultural encyclopedic knowledge tied to the geographic location.

A word of caution might be useful here. The semantic reality here is not one of compositionality. This is not how transparency should be understood here. Even with highly transparent compounds, the meaning is not the sum of the parts singer-songwriter should not be viewed as being +singer and +songwriter. As I use the term, semantic transparency involves the degree to which the mental frame of a portion of the compound is activated: the frame for singer is activated here just as much as the frame for songwriter is. And even then, the total meaning of the compound is still larger than the sum of its parts. Singer-songwriter activates additional cultural knowledge on its own, separate from both singer and songwriter. These include a particular set of instruments that are expected to be involved: likely an acoustic guitar, piano, or keyboard. Similarly, the style of music of a singer-songwriter is not likely to be, say, death metal—a fascinating compound in its own right. Semantic transparency must be understood within the domain of frame semantics and mental representations.

A little further down the scale of semantic transparency are endocentric attributive and subordinate compounds. These compounds also create a sense of transparency as well, that could be naïvely interpreted as componentiality. This is especially true when one or more parts are high frequency lexical items on their own. Let us examine blue cheese. This cheese is conventionally recognized as a shade of yellowish white specked blue across its surface and internal structure. And from that simple statement the compound might be said to be componential in its semantics given its lexical parts. The specks of blue distinguish it from other cheeses, certainly. And equally important: blue cheese is very much a cheese. Since endocentric compounds imply relationships within larger category, they also feed the idea of semantic componentiality. Yet again, the meaning of blue cheese meaning is still more complicated than merely a cheese that is blue. For one, it isn’t always blue and even when it is, it never particularly blue. Additionally, for anyone who has encountered and eaten blue cheese, it seems improbable that the color is the first thing that activates neurologically. Taste, smell and its dry crumbly texture are the far more salient features of the cheese. Aspects relevant to its creation and the molds/fungi involves are also central to its meaning structure. Blue cheese is defined by characteristics of the referent that are independent of color. Indeed, in a sense, the more you define the compound on the basis of the manner of its production, as perhaps the cheese maker would, the more blue becomes a sort of accidental byproduct.

The reality is that if asked what blue cheese was, no native speaker of English would reply “a cheese that is blue.” Instead, blue cheese would be described in terms of its smell, flavor, texture, and its physical constituency. Blueness To the extent that it is blue, that blueness plays very little role in its definition. Still, if we encountered someone who had no experience of blue cheese, show them a picture and said, “This is blue cheese,” their likely response would probably be something along the lines of, “That makes sense. It has some blue in it.” This would be a fairly natural response. The compound is motivated by aspects of its referent, even though one of those aspects is not particularly salient for the referent. But this semantic transparency is still, in a sense, a trick. The relationship is asymmetric. If a person who has not encountered blue cheese were asked, “What is blue cheese?” The answer, “Cheese that is blue,” is not a particularly accurate response, despite the fact that it does touch on one aspect vaguely relevant to the referent.

This is an important point: compounds do not mean their constituent parts; rather they mean something more specific and more complex, even when they are readily related to the semantics of their constituent parts. Essential aspects of a compound’s mean are not derivable from those parts. Still, the head of the compound, cheese is fully relevant to the semantics of the in the compound. This one component of the compound, as the head of its internal morphological structure, is close related to the word’s meaning. It could be that for, endocentric compounds such as this one, a compounds head lends itself to semantic componentiality, even if the attributive or subordinate portion of the compound does not. But even here, perhaps the situation is not as simply as it seems at first glance.

Consider the English compounds blackboard and whiteboard. These are also highly transparent, but again their meanings are more complicated than ‘black board’ and ‘white board’. As before, if ask what a blackboard was, no native English speaker would reply, “a board that is black.” Instead, the blackboard would be described in terms of its use, materials, and the larger cultural frame that it exists within: chalk, erasers, matte black color, educational settings, etc. Similarly, a whiteboard would be described/defined in terms of its frame: dry erase markers, being most prominent. The same asymmetry exists as well. The fact that the referents of blackboard and whiteboard are called by these terms is less than surprising, given the descriptive facts about them, but given those same facts, their synonyms chalkboard and dry erase board are no more surprising either.

To complicate things further, the head of these compounds are not like the head of blue cheese. The category of cheese is a far more restrictive category, and as result, the frame that is activated by the head is far more detailed and, in turn, salient for the meaning of the compound as a whole than would otherwise be the case. But this is not the case with the head board. The frame that board activates far less defined in terms of its mental representation. Arguably the frame is limited to perhaps a single nominal aspect operator: shape: rectangular, solid, and rigid.

Indeed, it is the frame that the compound activates matters most for its meaning, rather than the constituent pieces. In that context, compounds create the illusion of componential transparency only to the extent that they activate the same or similar frames as their constituent parts.

From this we can summarize a few things:

  • Componentiality can deceptive.
  • Semantic transparency is a product the frames, not +/- feature assignment.
  • Formal distinctions can be defined by the salience of the frames a compound activates.
    • The lexical parts of coordinate compounds activate their frames equally.
    • Attributive & subordinate compounds give more salience to frames activated by the head
    • Exocentric compounds are less semantically transparent than endocentric compounds.
  • Compounds activate referents high in specificity and complexity.
  • Their referents are necessarily distinct from those of their non-compound phrasal relatives.
    • Compounds involve a higher degree of semantic specificity (the blackboard).
    • Phrasal expressions with the lexical parts tend toward generic senses (a blackboard).**

Compounds do not mean their constituent parts; rather they mean something more specific, more complex, even when they may be readily connected to their constituent parts. The interpretation of compounds, then, must begin with the frames in which they appear.

**When the constituent parts are homonyms with the compound, this is easily seen if one uses a periphrastic construction or replaces the head with a synonym: “a board that is black” or “a black plank.” Both of these activate a separate mental representation than black board would.


Bauer, Laurie. 2003. Introducing linguistic morphology. Washington, DC: Georgetown University Press.

Benczes, R. 2006. Creative compounding in English. Amsterdam/Philadelphia: John Benjamins

Bisetto, Antonietta and Sergio Scalise. 2005. “The classification of compounds.” Lingue E Linguaggio IV.2 319-332.

Bisetto, Antonietta and Sergio Scalise. 2011. “The classification of compounds.” Pages 34-53. The Oxford handbook of compounding. Edited by Rochelle Lieber and Pavol Štekauer. Oxford: Oxford University Press.

Koch, P. 1999. Frame and contiguity: On the cognitive bases of metonymy and certain types of word formation. In K.–U. Panther & G. Radden (Eds.), Metonymy in language and thought (pp. 139–167). Amsterdam/Philadelphia: John Benjamins

Lieber, Rochelle. 2011. “A lexical semantic approach to compounding” Pages 78-104. The Oxford handbook of compounding. Edited by Rochelle Lieber and Pavol Štekauer. Oxford: Oxford University Press.

Lieber, Rochelle and Pavol Štekauer. 2011. The Oxford handbook of compounding. Oxford: Oxford University Press.

Leino, J. 2004. Frames, profiles and constructions. In J.-O. Östman & M. Fried (Eds.), Construction Grammars: Cognitive grounding and theoretical extensions (pp. 89–120). Amsterdam/Philadelphia: John Benjamins.

Moore, K. 2011. Frames and the experiential basis of the Moving Time metaphor. Constructions and Frames, 24, 80–103

Sweetser, E. 1997. Compositionality and blending: semantic composition in a cognitively realistic framework. In Proceedings of the Fifth International Cognitive Linguistics Conference (pp. 129–162). Amsterdam: ICLA.