Seri verbs: multiple complexities

Project Overview

Period of award

July 2017 - July 2020


Arts and Humanities Research Council

Many languages show seemingly arbitrary elaboration of their inflection. For example, most English nouns can take a plural ending, but it is not the same for every noun : compare bird-s, ox-en, and phenomen-a. Exactly which one to use is an additional fact that must be learnt and remembered. In this case that might not seem all that great a task, but there are languages which far outstrip English, adding an extra layer of seemingly gratuitous complexity to the already considerable machinery that languages employ for the expression of meaning. Within a general theory of language, such morphological complexity is rather the elephant in the room: it is far from clear what it’s doing there, and why it is taking up so much space. One of the most extreme examples of inflectional variation – the most extreme, we would argue – comes from Seri, a language isolate spoken by approximately 900 people on the Sonoran coast in Mexico. The verbs in Table 1 illustrate the basic parameters. Each verb has four forms, distinguishing subject number (‘s/he’ vs ‘they’) and event number (‘do something once’ vs ‘do something more than once’). There are different endings, as with the English examples just mentioned, but in place of the small handful of alternative endings we find in English, there are dozens. Table 1 shows three of the resulting inflection classes, but this is just the tip of the iceberg: according to the best information we have (Moser & Marlett 2010) there are over 250 inflection classes based on the form of these endings. And if we include other means of inflection found with some verbs, such as infixation or vowel changes, the number rises to over 500. This dwarfs anything we are aware of in any other language – for the sake of comparison, note that this is seven times the number of classes that French has (Stump and Finkel 2013), a language well known for the complexity of its verbal system. 



‘went away’

‘made sandals’


SG subject, SG event




SG subject, PL event




PL subject, SG event




PL subject, PL event




Table 1. Contrasting verbal paradigms


However, the complexity of this system involves far more than can be conveyed by a mere enumeration of the number of endings or the number of classes. First, membership in these classes is fairly evenly distributed across verbs in the language, as far as can be deduced from the evidence in Moser and Marlett (2010). This is in sharp contrast to a language like English, where there is a clear distinction between regularly inflected verbs such as talk~ talked or open ~ opened, and a small, internally diverse set of irregular verbs such as sing ~ sang or fly~flew. Thus a learner of English is generally on safe ground by picking the regular verb endings, but in Seri there is no dominant pattern, so that it is hard to make even an educated guess as to what the forms of a given verb would be.

A second complicating factor comes from the variable distribution of forms in the paradigm. The endings -oj and -olca in Table 2 illustrate what we mean by this. The verbs ‘was bad’ uses the ending -oj for plural subjects; the difference between singular event and plural event being indicated by the verb stem (tmiipl- vs. tmiipal-). Similarly, the verb ‘helped put’ uses -olca for plural subjects, and event number is marked on the stem. So far we should feel confident in saying that the endings -oj and -olca both mark plural subject, independent of event number. But now consider the verb ‘was hot’ – in this case the job of marking event number is shifted to the endings: -oj for singular events, and -olca for plural events. These endings thus sometimes contrast in function, and sometimes have the same function. And this sort of distribution is not just a quirk of these two endings – it is characteristic of most of the elements that make up the verbal paradigm. What this means is that we cannot assign a fixed function to the morphological formatives. Nevertheless, our preliminary research suggests that the system is not chaotic, and that the paradigm follows a strict morphological hierarchy. Whether this hierarchy reflects aspects of meaning or is morphologically autonomous is a question of considerable theoretical and typological interest, as we are not aware of parallel examples from other languages.



‘was bad’

‘helped put’

‘was hot’

SG subject, SG event




SG subject, PL event




PL subject, SG event




PL subject, PL event




 Table 2. Variable distribution of endings: -oj and -olca


A third complicating factor, at least for our understanding of the system, is meaning. Subject and event number marking on verbs is closely allied with number marking on nouns: nearly all the formatives across the two word classes are the same. Consider the paradigm of ‘hurried’ in Table 3: its [singular subject, plural event] form is built up in the same way as ‘dolphins’, [plural subject, singular event] is marked the same way as ‘herons’, and its [plural subject, plural event] form  is marked the same way as ‘cloths’.








SG subject, SG event






SG subject, PL event






PL subject, SG event






PL subject, PL event






 Table 3. Verbal and nominal number


Since the verbal paradigm draws from the same morphological resources used to mark plurality in nouns, we can at some level say that the verbal forms mark number, whether of subject or event. But while subject number is a notion which is fairly easy to grasp, the meaning of event number is far from clear. Cross-linguistically, number can be a property of events in a variety of ways, from simple quantification (‘cough once’, ‘cough many times’) to inherent characteristics of the verbal meaning (e.g. scatter in English implies a plural object, and wobble implies multiple instances of rocking back and forth). This can in turn interact with, overlap with or perhaps even be a component of ASPECT, the grammatical feature that tracks the course and composition of events over time, and is a central component of the verbal systems of many languages, as in English wrote vs. was writing. All of these properties appear to play a role in Seri verbs, but we simply do not know how predictably or how consistently; but we do know that event number is MORPHOLOGICALLY quite robust, being distinguished by the bulk of the verbal lexicon. Given how little we currently know about the semantic and syntactic principles underlying the system, it could be that at least some portion of the morphological complexity can be traced back to as yet unrecognized aspects of meaning or function, i.e. that variation in the form and distribution of inflectional markers has a functional explanation.

These data confront us with two major research opportunities. First, the morphology of verbal paradigms – in its proliferation of classes, in its unpredictability, and in the variable functionality of its forms – appears to have attained here a degree of complexity and opacity with few parallels in the languages of the world. A fuller account of how different speakers negotiate their way through this system, and converge on a shared representation adequate for communication, would be an important contribution to the study of language. Second, the meanings and functions subsumed under the rubric ‘event number’ represent a category which is the subject of lively ongoing research cross-linguistically, and whose status within a general theory of grammatical systems is still being defined. In the face of this we have formulated three research questions:


Q1: What strategies do speakers use to construct the verbal paradigm? 

Q2: Does the complexity have a semantic explanation? 

Q3: Is event number a grammatical feature?  

A better understanding of this system will be a significant contribution to our understanding of morphological complexity, both in its role as an autonomous component, and in its relationship to semantics at various levels (lexical, situational) and morphosyntactic features.