A Companion to Chomsky. Группа авторов
rel="nofollow" href="#uaa242a8e-3ba5-554f-966f-2936e0c2e251">Chapter 15 of this volume).
With this as a backdrop, it should be added that Chomskyan generative grammar was extended fairly early on to situations and phenomena that do not fit the description in the quote above. Slabakova (Chapter 14 of this volume) shows this for second language learning, where substantial and crucial work especially started to emerge in the 1980s. However, multilingualism more generally has become much more prominent (cf. Kupisch, Soares, Puig‐Mayenco, and Rothman, Chapter 15 of this volume), in particular work on heritage languages (Polinsky 2018; Polinsky and Scontras 2020 make this vivid) and also work on code switching as an attempt to differentiate different formalisms. We will provide one example from the area of code switching and grammatical gender.
There has been a lot of discussion of how grammatical gender should be analyzed in the literature. Issues range from the nature of gender assignment to the syntactic location of gender features in the syntactic representation (see, among many others, Corbett 1991; Kramer 2016). The traditional assumption is that a gender feature is simply part of the lexical representation of a word. However, recent theoretical work has questioned this idea and also the idea that words are the grammatically relevant unit. Instead, words are epiphenomenal, a result of the syntactic derivation. In particular, Distributed Morphology has promoted the idea that the smallest units in a syntactic derivation are roots and functional elements that among others provide roots with a syntactic category (Marantz 1997; Alexiadou 2001; Arad 2003; Embick and Marantz 2008; Embick 2015). As an example, consider (29). In (29a), a root merges with a v to form a verb, and in (29b), the same root merges with an n to form a noun.
1 (29)[v v √ROOT ] b. [n n √ROOT ]
A root has to be merged in order to ensure that a category is assigned. Embick and Marantz (2008: 6) label this the categorization requirement. In her comprehensive approach to grammatical gender, Kramer (2015) argues that n is the locus of gender assignment. If true, that means that gender is syntactically assigned. This has important ramifications for understanding language mixing involving grammatical gender. For instance, in American Norwegian, a heritage language of Norwegian spoken in the United States (see Haugen 1953), speakers easily assign grammatical gender to nouns borrowed from English, i.e. masculine, feminine, or neuter gender; (30) provides some examples, where the English items are in boldface (Grimstad, Riksem, Lohndal, and Åfarli 2018, p. 198).
1 (30)a.en blanketa.M blanket‘a blanket’b.ei nursea.F nurse“a nurse”c.et crewa.N crew“a crew”
As Grimstad, Riksem, Lohndal, and Åfarli (2018, 198) point out, the genders on English items are not identicial to the translational equivalents in Norwegian. The question is what determines gender assignment for English words in American Norwegian. Arguing that speakers have one lexical representation with, say, blanket carrying a masculine gender feature, and one without such a feature, simply pushes the problem back: How come speakers have acquired two lexical representations where one has grammatical gender and the other does not? Instead, Grimstad, Riksem, Lohndal, and Åfarli (2018) make the case that gender is assigned syntactically. Following Kramer's (2015) specific implementation, gender features would be on n (cf. (29b)), meaning that if n has a gender feature and the root is an English root, then we get e.g. en blanket, as in (28a). The structures in (31) illustrate what this would look like for the three nouns in (30).
1 (31)[n n[MASCULINE] √BLANKET ][n n[FEMININE] √NURSE ][n n[NEUTER] √CREW ]
The major advantage of this analysis is that if gender were specified on the roots themselves, then multiple lexical entries for each root would be required. In contrast, the distributed approach in (31) allows the same root to be combined with any gender‐flavor of n, and the data from American Norwegian support this variability since several speakers assign a different gender to the very same root. There are certain patterns in how gender assignment works, e.g. phonological similarity with Norwegian words has been argued by Haugen (1953) to be a factor. Space does not allow us to expand on this analysis any further, but it demonstrates how data from multilingual speakers have become rather important in developing formal analyses of, say, grammatical gender and what the units of word formation actually are; see Alexiadou and Lohndal (2018) for further discussion.
3.7 Conclusion
The ongoing development of generative grammar is a testament to Chomsky's extraordinary ability to set the research agenda and continuously ask new questions and thereby sharpen our tools so that we get an ever‐better understanding of the unique human ability for language. In terms of the larger conceptual issues, there have been some changes over the years. An example is the change from Government and Binding to the Minimalist Program when it comes to approaching Universal Grammar: Do scholars assume that Universal Grammar is whatever it needs to be to account for the data, or do they instead assume that it should be as minimal as possible, in particular to get a possible handle on the evolutionary origin of the human ability for language? With the shift to the latter position, Chomsky has also positioned generative grammar better for interactions with research into cognition more generally.
Endnotes
1 1 This chapter was written as part of the international research project MultiGender at the Centre for Advanced Study at the Norwegian Academy of Science and Letters in Oslo during the academic year 2019–2020, where both authors were fellows. We are grateful to an anonymous reviewer for very helpful comments on a previous version, and to Nicholas Allott for very detailed feedback. AL 554/8‐1 (Alexiadou) is also hereby acknowledged.
2 2 Chomsky (1965, 4) emphasizes the importance of developing explicit theories. In doing so, he contrasts generative grammar with traditional grammars. The latter provide a lot of structural information, but they do not offer mechanisms that can account for the “regular and productive syntactic processes.” Such mechanisms are implicit as these traditional grammars presuppose that speakers and listeners have such knowledge. That is, speakers know where to put the finite verb in a yes/no question, but an explicit theory should explain the rule that governs how the finite verb can in Can flying eagles swim? becomes the first constituent. Importantly, trying to be explicit also raises new questions, such as why Can eagles that fly swim? is a question about the swimming ability of flying eagles, not the other way around. Such new questions have continuously led to novel generalizations about languages and language in general.
3 3 Since then, two additional questions have also been put on the agenda (cf. Boeckx 2006):(i)How is that knowledge implemented in the brain?How did that knowledge emerge in the species?We will not discuss the first of these further here, but see Zaccarella and Trettenbein (Chapter 20 of this volume). We briefly discuss the latter in Section 3.5.
4 4 It is important to note that “knowledge” incorporated into mental systems need not be true. However, when it comes to each person's internal grammar, the question of truth does not really make sense. Furthermore, it is unclear whether knowledge needs to be believed. For sure, it does not have to be consciously available or integrated with general reasoning. Lastly, it does not have to be stored and retreivable (i.e. represented in the philosophers” sense): It may be simply embodied/instantiated by the workings of a particular system (here the I‐language). See Allott and Smith (Chapter 34 of this volume).
5 5 Note that in principle the linguistic theory should help explain any linguistic