Whenever generative approaches claim that they can account for broadly cross-linguistic regularities, I try to pay close attention. In a recent short paper in Linguistic Inquiry, Moskal (2015) is concerned with the generalization that nouns may show number suppletion (e.g. Russian rebënok ‘child’, deti ‘children’), but almost never suppletion for case, whereas personal pronouns often show number suppletion (e.g. English I/we) as well as case suppletion (e.g. English I/me, we/us). This seems to be a robust observation,
even though it has never been documented thoroughly, as far as I know. (Moskal claims that her study is based on 79 languages, but it is not clear what kinds of data she got from these languages and how; she does not mention Brown et al. (2003), the Surrey Suppletion Database).
Her explanation appeals to a locality constraint in a Distributed Morphology (DM) framework, as well as to the idea that nouns, but not pronouns, include a “little n” node in their tree structure. Thus, a noun such as Russian deti ‘children’ would have the underlying structure [[[[√CHILD] n] pl] nom], while a pronoun such as us would have the structure [[[[√we] pl] acc]. In the DM framework, vocabulary insertion is constrained by locality in some complex way (involving notions like cyclic domains and actual vs. potential phase initiators), resulting in the possibility of vocabulary items like us, but the impossibility of a hypothetical Russian suppletive *benjam for the dative plural of ‘child’ (the dative node in [[[[√CHILD] n] pl] dat] is too far away from the root).
A more accessible account of the general approach is given by Bobaljik (2015), who also summarizes the results of his (2012) book, where he observed similar patterns in comparative and superlative suppletion: Quite a few languages have suppletion patterns of the type ABB (bad/worse/worst), but none have the pattern AAB – Bobaljik explains this by saying that the superlative feature is “too far away” from the root to allow a suppletive form to express it.
Now intuitively, locality does indeed seem to play a role, as Bybee (1985) had observed: The categories that tend to be involved in suppletion (plural, aspect, tense) are also the categories that tend to occur closest to the root. So for suppletion in comparatives and superlatives, Bobaljik’s approach has some plausibility, because indeed the superlative marker sometimes includes the comparative marker and is further away (e.g. in Persian: X, X-tær, X-tær-in). But why should case be further away in nouns than in pronouns? Moskal’s proposal is that this is because nouns contain an extra (“category-defining”) node n which pronouns lack, but she does not say how this is otherwise motivated.
I would ask why we need an explanation in terms of abstract structure if a perfectly plausible usage-based explanation is available: Suppletion has long been known to be driven by high absolute frequency (e.g. Ronneberger-Sibold 1988; Hippisley 2001; Vafaeian 2013; with antecedents going back to the 19th century), and this predicts that personal pronouns should show a stronger tendency for suppletion than nouns – for example, case suppletion (or suppletion for other categories like gender, clusivity or politeness). This explanation is needed anyway, because no structural account is available to explain why ‘child’ is more likely to show a suppletive plural than ‘rabbit’. A generative story for case suppletion would need to address this competing explanation if it wanted to derive support from the cross-linguistic distributions. Nobody has collected systematic data yet, but it seems to me that frequency is also the best predictor of gender suppletion (he/she, contrasting with nonsuppletive actor/actr-ess), and frequency additionally explains why interrogative pronouns (apparently) show much less case suppletion than personal pronouns. Moskal only talks about “pronouns”, perhaps implying that interrogatives are included in this category as well (or is there reason to think that they include a little n, like nouns?). Moreover, Vafaeian (2013) points out that nominal suppletion is often governed by possession, and one wonders how this might fit into Moskal’s system.
In footnote 3, Moskal mentions the differences between UG-based true universals and mere “typological generalizations” (following Kiparsky 2008). The frequency-based explanation presumably falls into the latter group, but she aims for true universals. It should thus be a problem for her to find occasional counterexamples, like the Russian noun god ‘year’, which has the suppletive form let in the genitive plural, i.e. a clear instance of case suppletion of a noun. This example is not mentioned, but instead Moskal mentions an exception from Archi and one from Lezgian. In Archi, ábt:u ‘father’ has the suppletive ergative form ummu, which Moskal explains by noting that this noun does not occur in the plural, i.e, it has the structure [[[√FATHER] n] erg], where the case feature is close enough to the root. This would seem to predict that case suppletion should occur freely in case-inflecting languages that lack nominal plurals, such as Kayardild or Yidiñ. It would also seem to predict that highly frequent mass nouns which do not have plurals (e.g. German Milch, Wasser) should have suppletive case forms. Neither of these predictions seem to be confirmed by any facts.
In Lezgian, the word for ‘water’ does have case suppletion, but unfortunately, Lezgian is not one of those languages where mass nouns lack plurals. Instead, Moskal’s “explanation” for the exceptional case suppletion in ‘water’ is that the singular node is “pruned” in the context of oblique case. This is puzzling, because it seems to make the analysis empirically vacuous: If any number node can be deleted in an ad hoc way, then this allows ad hoc suppletive case forms of nouns without limits, contrary to what we find. (Moskal merely adds a footnote where she says that “pruning seems to be a highly marked configuration”, but this is precisely what one would want to explain.)
Thus, I am not impressed by this particular attempt to refer to abstract DM-style structures to help explain broadly cross-linguistic patterns, leaving out any discussion of well-known effects such as frequency. However, Bobaljik’s (2012) discussion of comparatives and superlatives is much more solid, and much more difficult to find holes in. If his general approach could be shown to work elsewhere, it would begin to look convincing.
Bobaljik, Jonathan David. 2012. Universals in comparative morphology: Suppletion, superlatives, and the structure of words. Cambridge, Mass.: MIT Press.
Bobaljik, Jonathan David. 2015. Suppletion: Some theoretical implications. Annual Review of Linguistics 1(1). 1–18. doi:10.1146/annurev-linguist-030514-125157.
Brown, Dunstan, Marina Chumakina, Greville G. Corbett & Andrew Hippisley. 2003. Surrey Suppletion Database. University of Surrey. http://dx.doi.org/10.15126/SMG.12/1
Bybee, Joan L. 1985. Morphology: A study of the relation between meaning and form. Amsterdam: Benjamins.
Hippisley, Andrew. 2001. Suppletion, frequency and lexical storage. In: Andronis, Mary & Ball, Christopher & Elston, Heidi & Neuvel, Sylvain (eds) Chicago Linguistic Society 37: The Main Session. Vol. 1, 201–14. Chicago: Chicago Linguistic Society.
Kiparsky, Paul. 2008. Universals constrain change; change results in typological generalizations. In Jeff Good (ed.), Linguistic universals and language change, 23–53. Oxford: Oxford University Press.
Moskal, Beata. 2015. Limits on allomorphy: A case study in nominal suppletion. Linguistic Inquiry 46(2). 363–376. doi:10.1162/LING_a_00185.
Ronneberger-Sibold, Elke. 1988. Entstehung von Suppletion und Natürliche Morphologie. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 41: 453-462.
Vafaeian, Ghazaleh. 2013. Typology of nominal and adjectival suppletion. Sprachtypologie und Universalienforschung 66(2). 112–140.
Pingback: How can diachrony help explain typological distributions? | Diversity Linguistics Comment