One of the most basic terms of grammar is the term root, referring to the indivisible content-bearing parts of words, e.g. the first element in dog-s, sunn-y, famili-ar-iz-ation. The term originally comes from Hebrew grammar and has been known to European grammarians since the 16th century – but for a long time, it mostly referred to the original elements of languages and was long associated with etymological speculations (this association is still felt in von der Gabelentz 1891: 188). An early synchronic work that clearly opposes roots (“radicals”) to affixes is René de Saussure’s (1919) work La structure logique des mots (René was Ferdinand’s brother and was primarily a mathematician and interlinguist, so his book did not become influential; but see now Anderson & de Saussure (2018) for a new appraisal).
At least since Nida (1949: 82), the term root has been part of the common vocabulary of general grammar:
But what exactly is a root? In other words, how should the term be defined? In Haspelmath (2012; 2021a: 41) I proposed a very simple but somewhat “too radical” definition: I said that a root is a morph that denotes an object, an action or a property. Here I revisit this definition and propose a somewhat narrower definition that corresponds more closely to our intuitions. (However, it should be noted that definitions of commonly used grammatical terms almost never capture our intuitions exactly, because these intuitions tend to come from salient exemplars, not from definitions.)
Before getting to the revised definition (“roots as solitary contentful morphs”, see below), let us look at a few definitions of the term in the literature – specifically, definitions of root in glossaries of morphology textbooks:
Haspelmath (2002: 274): root = a base that cannot be analysed further – i.e. a base that consists of a single morpheme
Harley (2006: 288): root = the morpheme conveying the main meaning in a word. In cats, cat is the root. In teacher, teach is the root. In economics and economy, econom- is the root.
Aronoff & Fudeman (2011: 269): root = the basic form from which another is derived by internal change or by addition of inflectional or derivational morphemes.
Booij (2012: 323): root = the part of a word after the removal of all affixes
Three of these definitions (2001/2011/2012) define root with respect to possible (inflectional or derivational) derived forms, and they ignore the possibility of roots that do not allow any further affixation (or derivation). But many languages have nouns, adjectives or even verbs which do not allow inflection or derivation, e.g. goulash in English (there is no plural form *goulashes, and no derived verb *goulashize) – these definitions thus exclude such forms (which is not what we want). Perhaps even more importantly, two of the definitions make reference to notions like “affixes” and “inflectional/derivational morphemes”, which probably have to be defined in a way that makes reference to roots (as in Haspelmath 2021b) – so the notion of ‘root’ is very basic and needs to be defined in still more basic terms. My 2002 definition is based on the notion of a “base”, which is defined in terms of the abstract notion of “morphological operation”. (But what is an “operation”? This seems very analysis-dependent, so in more recent work, I have come to prefer definitions in terms of kinds of forms, as in Haspelmath 2020.)
Harley’s definition of root (“the morpheme conveying the main meaning in a word”) does not exclude goulash-type words and avoids reference to notions such as “affix” or “base”, but it is vague in that it may not be clear which morph conveys the “main meaning” of a word. Moreover, what is a “word”? Harley does not define this notion, and like “affix”, it probably has to be defined with reference to “root” (see Haspelmath 2023).
A problem that all four definitions have is that they do not consider roots in compounds, e.g. black-bird and cat-fish (mentioned by Nida in the excerpt cited above). Attempting to use Booij’s definition, if we remove the plural suffix from blackbird-s, we get blackbird, but this is not a root – it is two roots. Attempting to use Harley’s definition, what would the “main meaning” of a compound like flower-pot? It’s a kind of pot, but is this really the main point of this compound? (In Czech, a flowerpot is called květin-áč, a simple suffixed derivative of květina ‘flower’, with no root meaning ‘pot’.)
My 2021a definition of root (‘a morph that denotes an object, an action, or a property’) is better than Harley’s in that it makes more precise what is meant by “main meaning”, but it of course relies on understanding what an object, an action and a property are, and it does not cover nouns which denote abstract concepts (beauty, soul) or events (war). The latter issue is not too serious from a comparative perspective (because it does seem to be the case that objects, actions and properties are the shared core of nouns, verbs and adjectives across languages; Croft 1991), but here I address another problem: Some affixes clearly denote actions or objects. I am referring, in particular, to causative affixes as in Lithuanian tirp-in-o ‘melted (tr.), caused to melt’, derived from tirp-o ‘melted’. Instrument noun affixes are also naturally regarded as referring to an implement, e.g. Latin -culum (e.g. incerni-culum ‘sieve’).
Moreover, some languages have “lexical affixes” (e.g. Mithun 1997; Bischoff 2011). This is particularly well-known from Salishan and Wakashan languages. In the following examples from Makah (Southern Wakashan; Davidson 2002: 93), the element -’ax̣s means ‘canoe’ but is not the ordinary word for ‘canoe’ (this is čapac).Are these elements roots or not? What does “lexical” refer to in this term? Are they affixes that do not have “grammatical” meaning like other affixes?
While the “lexical affixes” of the Pacific Northwest languages may appear “exotic”, similar phenomena exist in English, where the neoclassical compounds have given rise to a lot of discussion: Elements such as geo-, astro-, neuro– do not occur on their own, but only in compound-like forms. What should we do with them?
I really like the definition of root that is given by Bauer et al. (2013: 17): “A root is the centre of a word, a lexically contentful morph, either free or bound, which is not further analysable”. They evidently use the expression “contentful” in a sense corresponding to my ‘denoting an object, an action or a property’. The only thing that I would add here is that a root CAN be the (single) centre of a word, as it can also combine with another root to form a compound. So my proposal here is to exclude all contentful morphs that cannot occur without another contentful form as part of a free form:
A root is a contentful morph (i.e. a morph denoting an action, an object or a property) that can occur as part of a free form without another contentful morph.
Putting it informally, we can say that roots must be able to be “solitary” (occurring without another root-like form). This excludes causative affixes (like Lithuanian -in-), which denote actions but cannot occur without another contentful morph, and it also excludes Salishan and Wakashan “lexical affixes”, because these, too, have to occur with another contentful morph. It also excludes instrument-noun suffixes like Latin -culum.
Contentful morphs which must occur with another contentful morphs in a free form are often called “(obligatorily) bound roots” in English-language linguistics (Bauer et al. 2013; 18), but in many languages, all roots must occur with some affixes (e.g. most Russian noun roots, and most Italian verb roots). Thus, the fact that these elements are bound forms is not a remarkabl feature from a cross-linguistic point of view. Following the present proposal, they are contentful nonroot morphs (one could of course think of a new term for them, e.g. “nonsolitary contentful morphs”).
That roots should be defined in this unexpected way, with reference to a semantic criterion and a distributional criterion that makes reference to free forms, is probably surprising to many readers. The larger reason for this choice is the general philosophy behind my proposals for defining traditional grammatical terms (Haspelmath 2021a): They should be defined in such a way that the same criteria can be used for all languages. Semantic distinctions (contentful morphs) and free occurrence are highly general criteria, so they are suitable for defining highly general concepts such as root, affix and clitic. When describing particular languages, one should of course make use of language-particular criteria. Nida (1949: 82) lists a number of such criteria:
But for general linguistics, we must use the same criteria, because otherwise we run the risk of comparing apples with oranges. Nida says that “It is not always easy to distinguish between roots and nonroots”, but this is so only if one does not know which criteria to use. Cross-linguistic classification is never “difficult” if one knows the facts and if one has a generally applicable definition.
P.S. In this post, I completely ignore the more recent use of the term “root” as an abstract element in the Distributed Morphology framework (where people say things such as “affixes are roots”, e.g. Creemers et al. 2018). As I say above, I am interested in types of forms, not in types of abstract elements that one may posit.
Anderson, Stephen R. & de Saussure, Louis de (eds.). 2018. René de Saussure and the theory of word formation. Berlin: Language Science Press. (doi:10.5281/zenodo.1306472)
Aronoff, Mark & Fudeman, Kirsten Anne. 2011. What is morphology? 2nd edition. Chichester: Wiley-Blackwell.
Bauer, Laurie & Lieber, Rochelle & Plag, Ingo. 2013. The Oxford reference guide to English morphology. Oxford: Oxford University Press.
Bischoff, Shannon T. 2011. Lexical affixes, incorporation, and conflation: The case of Coeur d’Alene. Studia Linguistica 65(1). 1–31. (doi:10.1111/j.1467-9582.2011.01177.x)
Booij, Geert E. 2012. The grammar of words: An introduction to linguistic morphology (3rd edition). Oxford: Oxford University Press.
Croft, William. 1991. Syntactic categories and grammatical relations: The cognitive organization of information. Chicago: University of Chicago Press.
Davidson, Matthew. 2002. Studies in Southern Wakashan (Nootkan) grammar. University at Buffalo. (PhD dissertation.)
de Saussure, René. 1919. La structure logique des mots dans les langues naturelles considérée au point de vue de son application aux langues artificielles. Berne: Büchler.
Harley, Heidi. 2006. English words: A linguistic introduction. Malden, MA: Blackwell.
Haspelmath, Martin. 2002. Understanding morphology. London: Arnold. (https://zenodo.org/record/1236482)
Haspelmath, Martin. 2012. How to compare major word-classes across the world’s languages. In Graf, Thomas & Paperno, Denis & Szabolcsi, Anna & Tellings, Jos (eds.), Theories of everything: in honor of Edward Keenan (UCLA Working Papers in Linguistics 17), 109–130. Los Angeles: UCLA.
Haspelmath, Martin. 2020. The morph as a minimal linguistic form. Morphology 30(2). 117–134. (doi:10.1007/s11525-020-09355-5)
Haspelmath, Martin. 2021a. Towards standardization of morphosyntactic terminology for general linguistics. In Alfieri, Luca & Arcodia, Giorgio Francesco & Ramat, Paolo (eds.), Linguistic categories, language description and linguistic typology, 35–57. Amsterdam: Benjamins. (DOI: 10.1075/tsl.132.02has, https://ling.auf.net/lingbuzz/005489)
Haspelmath, Martin. 2023. Defining the word. (in preparation).
Mithun, Marianne. 1997. Lexical affixes and morphological typology. In Bybee, Joan L. & Haiman, John & Thompson, Sandra A. (eds.), Essays on language function and language type: Dedicated to T. Givón, 357–371. Amsterdam: Benjamins.
Nida, Eugene A. 1949. Morphology: The descriptive analysis of words (2nd edition). Ann Arbor: University of Michigan Press.
von der Gabelentz, Georg. 1891. Die Sprachwissenschaft, ihre Aufgaben, Methoden und bisherigen Ergebnisse. Leipzig: C. H. Tauchnitz. (https://langsci-press.org/catalog/book/97)