- The paper demonstrates a morphology-first classification system for Mapudungun verbal roots based on suffix combinatorics.
- It applies finite-state methods to validate morphotactic diagnostics, highlighting causative suffixes as key markers.
- The study reveals labile verb behavior and challenges rigid syntactic-based valency models, improving computational analysis.
Valency Classification of Mapudungun Verbal Roots: Morphotactic Evidence
Introduction
This paper presents a systematized approach to the classification of Mapudungun verbal root valency, exclusively via the morphotactic mechanisms of the language. Root valency—transitive, intransitive, or labile—is defined through the possible and restricted combinations of specific suffixes in the verb complex, rather than semantics or syntactic argument realization. The motivation is both theoretical, for understanding Mapudungun verb systems in greater morphosyntactic detail, and practical, to refine the finite-state computational analyser Düngupeyüm.
Theoretical Background
Canonical approaches to valency in Mapudungun, most notably Smeets (2008), have classified verbal roots primarily on observed usage. However, Mapudungun’s agglutinative and polysynthetic typology presents pervasive context-driven valency alternations, as evident from prior observations (cf. dedios 2013, Zúñiga 2006, Golluscio 2010, Zúñiga 2024). This renders semantics and argument structure insufficient on their own for robust morpholexical categorization. Suffixal behavior—specifically, which affixes can combine with which roots—offers a more reliable diagnostic since core suffixes are highly constrained by root valency (2604.00789, Chandía, 11 Feb 2025).
Methodology
A review was conducted of all roots classified as verbal in Smeets (2008), first confirming true verbal status and then exhaustively investigating allowed affixations. Affix–root combinatorics were extracted from grammars, dictionaries (including corpus lexicography), and naturalistic texts (notably Mosbach 1930). For each root, its ability to combine with a suite of morphotactically diagnostic suffixes was logged, focusing particularly on suffixes with known valency-altering or valency-diagnostic functions, such as causatives (-(ü)m, -(ü)l), the transitiviser/applicative -tu, and others.
Morphological Diagnostics of Valency
Causative Suffixes
- -(ü)m is impinging: it occurs almost exclusively with intransitive roots (notably unaccusatives), productively transitivizing them. The restricted productivity and strong selectional constraints make -(ü)m a prime diagnostic for the intransitive classification of roots (see Smeets 2008, Golluscio 2007, Longkon 2011).
- -(ü)l is broader, attaching to both intransitive (especially inergatives/labile roots) and transitive roots in a less restrictive manner. However, not all authors agree on precise boundaries: in some subdialects and speakers, -(ü)l may act as an applicative or benefactive in addition to a causative, and double/multiple causativization is attested.
- Both causatives display distinctly regular morphophonological effects that serve additionally as confirmation in finite-state specification.
Other Core Suffixes
- -tu (transitiviser/applicative) displays polyfunctionality, occasionally increasing valency (making a monoargumental root biargumental), but also serving repetitive, restorative, antipassivizing, or denominal verbalizing functions. The ambiguity and context-dependent behavior limit its usefulness for root-classification absent further information.
- Directionality suffixes -fi (direct), -e (inverse) are gold markers for transitive forms—but their occurrence does not discriminate transitivized intransitive (labile) roots from inherent transitives, and their presence is regularly due to broader morphosyntactic context.
- Aspectuals -nie, -künü (‘progressive persistent’, ‘perfect persistent’), stative –(kü)le: While Smeets (2008) suggested some can only combine with intransitives, empirical evidence in the paper shows clear violations of these claims—either due to noncanonical root status or prevalent verbal compounding and argument incorporation.
Labile Verbs
A significant finding is the pervasiveness of labile (ambitransitive) verb roots. These combine with suffix sets of both canonical transitives and intransitives, with morphosyntactic context dictating alternations. This is addressed computationally in Düngupeyüm by allowing dual classification, at the cost of introducing controlled analysis ambiguity.
Suffix Interaction and Morphotactic Rules
A major contribution is the formulation—tested computationally—of rules that block or allow co-occurrence of causatives and other valency-increasing suffixes with specific root subclasses. For example, -(ü)m and -(ü)l are forbidden with non-verbal roots except for a controlled set of denominal or adjectival derivations; labile roots admit causative suffixes only in intransitive senses; direct/inverse markers are checked for validity against the base root valency after all valency-modifying operations.
Empirical Observations and Contradictory Data
The thorough example survey, drawing from both the main reference grammar and primary oral corpora, exposes flaws in several grammaticalizations held as strong constraints by the reference literature. Suffixes with supposed strict root selectional properties (e.g., -(kü)le) are evidenced with roots of “opposite” valency, usually through compounding or reflexivization structures. Similarly, some aspectual suffixes that supposedly only attach to one class are shown to participate in forms with transitivized or labile bases.
The morphological analyser was thus adjusted to prefer strict rules only where counterevidence is lacking and fall back to ambiguous or dual analysis for labile and polyfunctional roots/suffixes.
Implications for Finite-State Morphological Analysis
By explicitly encoding root valency and the compatible morpheme slots in the FST analyzer, the ambiguity of analysis is substantially reduced for typical and productive forms. Allowing well-motivated ambiguity for the labile set ensures the analyzer does not falsely reject attested alternations. The morphotactic criteria here replace less reliable argument-structure heuristics, yielding a more robust, internally consistent computational morphology and part-of-speech assignment framework (Chandía, 2021, 2604.00789).
Table: Effectiveness of Morphotactic Diagnostics
| Suffix/Class |
Diagnostic Power |
Main Limitation |
| -(ü)m (Causative) |
High (intransitive) |
Non-productive for some roots |
| -(ü)l (Causative) |
Moderate/Labile |
Polyfunctionality, compound ambiguity |
| -tu (Applicative) |
Low |
Polyfunctionality |
| -fi/-e (Direct/Inverse) |
Moderate |
Only at post-suffix level |
| -nie, -künü, -(kü)le |
Variable |
Multiple exceptions, compounding |
| Labile Root category |
Essential |
Introduces ambiguity |
Rules derived from this morphotactic mapping are now foundational for the FST system architecture.
Typological and Theoretical Implications
The methodology and findings are instrumental for cross-linguistic research on valency-changing morphology in polysynthetic/agglutinative languages. The deep mapping between suffix distribution and valency argues for a morphology-first approach to argument structure in computational grammars. The empirical demonstration of valency fluidity, labile verb regularity, and pervasive polyfunctionality challenges any a priori rigid classification in Mapudungun and similar languages.
Future Directions
- Further automated extraction and unsupervised clustering of root–suffix compatibility from large oral corpora would refine root class inventories.
- Investigation into diachronic drift in causative/applicative allomorph distribution could clarify some "contradictory" co-occurrence facts.
- Extension to denominal, adjectival, or other non-verbal root classes will refine FST architecture for word class assignment.
Conclusion
The paper operationalizes valency classification of Mapudungun verb roots purely by the language’s morphotactic constraints and provides robust rules for computational analysis. Causative morphemes (-(ü)m, -(ü)l) offer the most precise diagnostics for intransitive status; labile roots demand dual analysis but with careful constraint on causative co-occurrence. Broader suffixes such as -tu, directionality, and aspectuals, provide valuable but often contextually ambiguous evidence. The approach strengthens both theoretical understanding and practical language technology for Mapudungun (2604.00789).