I recently saw an interesting article that asks about biases in evolutionary biology introduced by the fact that it is studied by humans. In linguistics, there is no shortage either of biases introduced by big languages such as Latin and English, and all linguists are at least aware of this possibility. But are we taking the potential problem seriously enough?
I keep wondering about this, and I feel that a particularly egregious case is the idea that languages like Spanish have null subjects in sentences like (1b).
(1a) Marta lleg-ó ayer. [Marta arrive-3SG.PST yesterday] ‘Marta arrived yesterday.’
(1b) Llegó ayer. [arrive-3SG.PST yesterday] ‘She arrived yesterday.’
There is a clear way in which the subject is expressed in (1b): by means of the suffix -ó, which is the counterpart of English she in the translation. The verb may also take suffixes of the first or second person, as in (2a) and (2b).
(2a) Llegu-é. ‘I arrived.’
(2b) Lleg-aste. ‘You arrived.’
What is null here? Or alternatively, which pronoun is “dropped” here, so that it would make sense to call Spanish a “pro-drop” language? It seems that the answer does not come from Spanish, but from English, and from our spelling habits. If llegó (spelled as a single word) is matched with arrived (rather than she arrived), then the Spanish sentence has nothing compared to English she – it has a “null subject”. But would it occur to a linguist who is not biased by English to say that something is null (or was “dropped”) in llegó?
Maybe the problem is not so much English bias, but bias by the traditional terminology (which was established in the 19th century, largely by German scholars, whose native language was like English in the relevant respects). We simply do not have good traditional terms for the situation of Spanish, so it seems that we need some new terminology. In Haspelmath (2013), I proposed the term cross-indexing for the Spanish-type pattern, where a subject can be expressed either by an index (= a bound person-form), or jointly by an index and a conominal (as in 1a, where the subject is expressed by the index -ó plus the nominal Marta).
A new term may suggest that there is something special about the Spanish pattern, but in fact, there isn’t – the majority of the world’s languages are like Spanish!
Let us distinguish four types of languages with respect to subject expression: Spanish-type languages with cross-indexes that can be conominated (3a), Swedish-type languages with no bound person forms (3b), German-type languages (3c), where an index on the verb has to be combined with an independent pronoun, and Japanese-type languages, where the subject can indeed be null when it can be inferred from the context (3d).
(3a) Marta came-she / came-she (= Spanish-type languages)
(3b) Marta came / she came (Swedish-type)
(3c) Marta came-she / she came-she (German-type)
(3d) Marta came / came (Japanese-type)
Two world-wide studies have looked at the frequencies of these types: Siewierska (1999: 238) found that in her sample of 272 languages, only about 15% are of the Japanese or Swedish type and lack person indexing. About 70% of the languages are of the Spanish type, about 14% belong to a fifth type (“Marta came / came-she”) that I called pro-indexing, and only two languages, Dutch and Vanimo, are of the German type.
(Dryer (2005) also gives information on subject expression in the world’s languages, but he lumps German-type languages with Swedish-type languages, finding that these two types represent about 12% of his sample.)
For a language of the German (or Dutch/Vanimo) type, a description in terms of “agreement” of the verb with the subject is not unreasonable, because there is always a nominal whose features may be copied onto the verb. However, as Siewierska found, German-type languages are extremely rare. So is there any other justification for treating Spanish in terms of “agreement of the verb with a null subject”, other than the fact that German and English (as well as Dutch, also a big language with an imperial past) can be described in this way?
For my (2013) paper, I looked at a lot of papers, but I could not find any justification – linguists seemed to simply assume that there are only two possibilities for a Spanish-type language: null subjects controlling verb agreement, or alternatively person indexes with exclusive argument status (a view that used to go by the strange name “pronominal argument theory”). However, there is really no reason (other than lack of imagination) to exclude the simplest way of thinking of Spanish clauses like Marta llegó: in terms of double expression of the subject, both by a nominal and by a person index (Haspelmath 2013: §5.4; Corbett (2003) had earlier used the equivalent term multirepresentation).
In view of the long list of papers discussing various possibilities, it came as a disappointment to me that the introduction to the recent book “Null subjects in generative grammar” (Cognola & Casalicchio (eds.) 2018) does not even mention alternative conceptions (the book is behind a paywall, but the preprint of the introduction is available here). It does not cite Siewierska or Dryer, and although it cites Gilligan (1987), which was based on a smaller sample but arrived at very similar conclusions, it continues to present the world as if Spanish-type languages (and their many subvariants, especially in Europe) were somehow unusual. They even make the claim that “many languages with rich morphology are not null subject languages (like Icelandic)” (i.e. many languages with person indexing on the verb require an additional independent person pronoun), even though Siewierska (1999) had noted that such languages are extremely rare.
(It is true that Siewierska (1999: 239) mentioned a few other languages of the German type: Standard German, Swiss German, English, Icelandic, Faroese, Romansh, French, plus Anejom; but these are either closely related Germanic languages, or were probably influenced by Germanic languages at some point; the only independent case is Anejom, an Oceanic language of Vanuatu).
Is it perhaps too much to demand that linguists whose expertise is primarily in Romance and Germanic linguistics would read the typological literature? Maybe, and I don’t want to deny that the papers in the volume (judging by the discussion in the intro chapter) make novel contributions by considering additional aspects that were not part of Siewierska’s and Dryer’s comparative studies. However, I am very suspicious of the idea that one can usefully appeal to a rich UG without considering worldwide diversity and making sure that one’s claims have at least some initial plausibility. The intro chapter is brimming with terms like “C-domain”, “extended EPP-feature”, “Agree”, “Spec,I/TP”, which are opaque to most linguists and which make sense only if it is assumed that there is a very rich innate UG – an assumption that has little initial plausibility and that even Chomsky does not seem to make anymore. Thus, the value of these high-tech notions for a volume about various subtle effects in subject expression in the Romance and Germanic languages is not clear to me.
(However, I agree, of course, with the editors’ claim, in §1.3 of their intro text, that “an incomplete or wrong theory is better than no theory, demonstrating, as it does, the energizing effect of a strong theory, which inspires researcher to seek out novel data in order to confirm or refute it” – but I would add that it is even better if the theory is formulated by using clear and commonly agreed terminology, and does not include too many moving parts)
Cognola, Federica & Jan Casalicchio (eds.). 2018a. Null subjects in generative grammar: A synchronic and diachronic perspective. Oxford: Oxford University Press.
Cognola, Federica & Jan Casalicchio. 2018b. On the null-subject phenomenon: An example of successful linguistic research. In Federica Cognola & Jan Casalicchio (eds.), Null subjects in generative grammar: A synchronic and diachronic perspective. Oxford: Oxford University Press.
Corbett, G.G. 2003. Agreement: The range of the phenomenon and the principles of the Surrey Database of Agreement. Transactions of the Philological Society 101(2). 155–202.
Dryer, Matthew S. 2005. Expression of pronominal subjects. In Martin Haspelmath, Matthew S. Dryer, David Gil & Bernard Comrie (eds.), The world atlas of language structures, 410–413. Oxford: Oxford University Press. https://wals.info/chapter/101.
Gilligan, Gary Martin. 1987. A cross-linguistic approach to the pro-drop parameter. Los Angeles: University of Southern California dissertation.
Haspelmath, Martin. 2013. Argument indexing: A conceptual framework for the syntax of bound person forms. In Dik Bakker & Martin Haspelmath (eds.), Languages across boundaries: Studies in memory of Anna Siewierska, 197–226. Berlin: De Gruyter Mouton.
Kokko, Hanna. 2017. Give one species the task to come up with a theory that spans them all: what good can come out of that? Proc. R. Soc. B 284(1867). 20171652. doi:10.1098/rspb.2017.1652.
Siewierska, Anna. 1999. From anaphoric pronoun to grammatical agreement marker: Why objects don’t make it. Folia Linguistica 33(1–2). 225–252.
Hello Prf. Haspelmath,
This is exactly what I have been thinking. Here is a snippet of what I was writing when I came across this post (; I guess I’ll have to reference this website as well):
“Throughout the history of linguistic analysis on Navajo and other related Na-Dene languages, we find a general tendency for linguists to be drawn to only specific areas of these languages. One of those areas is the so-called ‘verb’. Much of the work on Navajo has been devoted to describing, classifying, and decomposing all verbal component parts. Effectively, this is a logical outcome of the documentation projects that were initiated under the auspices of Edward Sapir (Hoijer 1938; 1945; 1946a; 1946b; 1948; 1949; Sapir 1915). To say the least, much descriptive and theoretical ink has been spilled over minute details concerning, for instance, verbal aspect or word formation processes (e.g. Hale & Keyser 2002: §4.2; Kari 1989; Midgette 1995; Smith 1991: ch. 12; Young 2000). Nevertheless, somewhere in all this mosaic of language description, or rather deconstruction, the essence of the Navajo ‘verb’ was lost, principally when this word class has been consistently investigated within the definitional criteria of Indo-European predicate phrases, viz English verbs. In many respects, terminological misnomers, theoretical blinders, and linguistic biases have distorted, in the strong Whorfian sense, the thinking of linguists in regard to Navajo data (see Haspelmath (2013: 210) on grammatical terminology)”
Dear Professor Haspelmath, I would like to point out the following things to avoid confusion:
1) our book focuses on null subjects in Generative Grammar (hence the title) – that’s why we had to narrow down our the scopes of our volume to this approach to language;
2) we think that your very interesting typology of the realization of the subject across languages is not so far from that generally proposed in Generative Grammar and by the authors of the volume (see especially Frascarelli on the presence of null topic in Italian/Spanish-style languages);
3) the term „UG“ only appears once in our paper (in a citation) and we use the categories CP/TP/VP/EPP etc in our chapter as purely descriptive devices to describe the sentence’s structure and position – and I think this is uncontroversial in Generative Grammar.
Thanks for these comments! I should indeed have a look at the Frascarelli paper. But I don’t think that working within generative grammar exempts one from the duty to examine all the relevant facts – if one is committed to the unity of science, a particular framework can be adopted only as long as it’s the best one. And I was very puzzled by your last comment, that you use CP/TP/EPP “as purely descriptive devices” – because for descriptive purposes, other concepts are MUCH simpler and more suitable. But this confirms my general feeling that linguistics is in a very confused state at the moment…
A question. I have argued that in Riau Indonesian, a language of your “Japanese type”, in a sentence like “Came.”, there is no evidence for an unfilled subject syntactic slot, and therefore Riau Indonesian is, in no sense of the word, a “pro-drop” language. I’m wondering: Are you familiar with any generative studies that provide explicit arguments in support of an unfilled subject position in some Japanese type language, rather than — as they usually do — just assuming it?
I think for Chinese, the assumption of an unfilled subject position has been standard since C.-T. James Huang’s work of the 1980s. The argumentation is characteristically complex (not to say convoluted), but it has generally been taken as providing actual evidence for the empty subject view (given certain prior assumptions that were never questioned).
Huang, Cheng-Teh James. 1984. On the distribution and reference of empty pronouns. Linguistic Inquiry 15.531-574.
Huang, Cheng-Teh James. 1989. Pro-drop in Chinese: a generalized control theory. In Osvaldo Jaeggli and Kenneth Safir (eds.), The null Subject parameter, 185-214. Dordrecht: Kluwer.