Some thoughts on roots and variation

This is a very belated post sharing some thoughts about the potential for crosslinguistic variation in the content of roots. It was inspired by the special session on roots at WCCFL 39, hosted virtually by UArizona in April 2021; I wrote this post then, but promptly forgot about it until now, when I unearthed it from my drafts folder. I’ve now lightly edited it and added links to the references, but it’s mostly unchanged from April.

There were a number of talks and posters at WCCFL 39 investigating and making claims about the properties of roots, many (if not most) within Distributed Morphology (DM). Following Halle and Marantz (1993 et seq.), in DM the assumption has been that roots are acategorial: they have no syntactic category of their own, but (obligatorily) unite with categorizing functional heads in the syntax. It is these categorizing heads, usually abbreviated v, n, a, etc that are responsible for category-specific behaviour.

This idea has led to many insightful and fruitful research questions. But there are several ongoing debates about the properties of these acategorial roots, several of them on display in the WCCFL special session:

• Are roots individuated in syntactic representations, or does syntax know nothing more than “there’s a root here”?
• If roots are individuated, is this done by their form, their meaning, or by an arbitrary syntactic identifier?
• Can roots be associated with syntactically-visible formal features?

The exchanges in these debates seem generally to presume that whatever the answers are, we expect those answers to be stable across languages. It’s this presumption that I want to push on a bit here.

As context for what I want to say about roots, let me first make some observations about the types of variation that are often proposed in another domain, the structurally “higher” projections associated with temporal meaning, in tense and aspect systems.

Now, a lot of formal semantic work tends to assume—sometimes implicitly—that even though not all languages have overt tense systems, they all have some kind of syntactic-semantic tense that plays a role in the composition of clauses—though the denotation of said tense may vary, and in some languages it may be underspecified.

But there have also been various proposals of greater variation “under the hood” in how languages do (or don’t) represent tense. Ritter and Wiltschko (2014), for example, propose that all languages may use Infl to “anchor” clauses to the utterance situation, but that some languages do this via (non-)coincidence in time (tense), while others may do so via location or participants—other deictic aspects of contexts of situations.

Pancheva and Tonhauser (2019) make a quite different proposal that languages without morphosyntactic tense do not merely have underspecified semantic tense, but are actually best modelled as lacking tense entirely, with clausal temporal interpretations being derived purely by aspect.

Beyond variation in what dimensions of temporal meaning are represented in syntax (and subsequently interpreted by semantics or realized by morphology), there may also be variation in how the features that encode temporal meanings are bundled. Cowper and Hall (2013) propose that the bundling of tense and aspect changed over time in English, in the transition from the “passival” (the house was building) to the contemporary progressive passive (the house was being built). They argue that at earlier stages English clauses bundled voice and aspect distinctions on a single functional head, only later separating them out. This proposal sits within a broader approach to features in syntax, proposing that features are constructed by learners in the course of acquisition on the basis of observed contrasts in language, mapped onto divisions in the nonlinguistic input. These nonlinguistic divisions are imposed by systems of human perception and cognition (building on the Contrastivist Hypothesis of Dresher (2009) for phonology).

What does any of this have to do with roots? Well, if we find this type of variation in structurally higher functional domains, shouldn’t we also expect it in lower domains? In the history of Distributed Morphology in particular, there seems to have been a general assumption that roots are a radically different kind of object than functional heads are: that they have special “Encyclopedic” meanings, that they aren’t subject to competition-based insertion, that they aren’t individuated syntactically, and so on.

These special properties have to some extent melted away, or at least have become open for debate. There seems to be active debate on whether roots are individuated in the syntax, that they can be individuated syntactically, etc. I feel like I’ve even seen proposals that functional heads can have Encyclopedic content (though I couldn’t tell you where). So what special properties of roots remain? Unlike purely functional elements, roots unquestionably have some kind of non-grammatical meaning—meaning that is not encoded in the types of contrastive formal features to which syntactic structures can be sensitive.

And yet! Do we have strong reasons to think that non-contrastive content is necessarily be quarantined within a distinct syntactic object with the label “√”, that has the same set of properties in all languages? I’m not sure we do—at least not if we think contrastive syntactic features can be bundled differently in different languages.

This leaves interesting space for crosslinguistic variation. It could be that roots are bundled with categorial information in some but not all languages, or that they are bundled with fixed atoms of argument structure or event structure in some languages but not in others. Indeed, we might expect some languages to grammatically encode divisions of event structure (e.g. Russian, St̓át̓im̓cets), while others exhibit the same fundamental distinctions in interpretation (plausibly due to properties of how humans conceptualize events), but with no evidence of grammaticalization in the same ways (e.g. English).

But all in all, I wonder if we aren’t moving away (in Distributed Morphology) from such a strict division between roots and functional elements. Is it still DM without that division? I think it probably is!

References

Credit for header image: Diego Delso, CC BY-SA 3.0 https://creativecommons.org/licenses/by-sa/3.0, via Wikimedia Commons

Cowper, E., & Hall, Daniel C. 2013. Syntactic change and the cartography of syntactic structures. Proceedings of NELS 42. GLSA. (preprint)

Dresher, E. 2009. The contrastive hierarchy in phonology.. Cambridge University Press.

Halle, M., & Marantz, A. 1993. Distributed morphology and the pieces of inflection. In []The View from Building 20](https://mitpress.mit.edu/books/view-building-20), eds. Ken Hale and Samuel Jay Keyser. MIT Press.

Pancheva, R., & Zubizarreta, M.L. 2019. Temporal reference in the absence of tense in Paraguayan Guaranı́. Talk given at NELS 50, held at MIT in October 2019. (abstract) (proceedings available on Amazon)

Ritter, E., Wiltschko, M. 2014. The composition of INFL. Natural Language & Linguistic Theory 32:1331–1386. https://doi.org/10.1007/s11049-014-9248-6