LITERARY UNIVERSALS WEBLOG: A series of informal observations and conjectures aimed at fostering more reflection on and discussion about cross-cultural patterns in literature.
Patrick Colm Hogan, University of Connecticut
I recently read Bradley Irish’s overview of debates concerning universality in emotions. Irish’s treatment of the subject is careful, scholarly, and well-argued. However, there are necessarily some points that he could not develop adequately in a short overview of such a large body of work. Perhaps the most consequential of these concerns the ways in which the type of theory one adopts has consequences for what counts as a universal. For example, he quotes a number of linguists to the effect that “the idea of a single underlying linguistic system different only in surface realization seems increasingly unlikely” (Majid 432). Majid is here alluding to Noam Chomsky’s view of universality. One of the linguists Irish quotes as representing this “general challenge to universalism” is Stephen Levinson (from a co-authored work; see Evans and Levinson), despite the fact that Levinson is no enemy of universalism, as is shown by his (excellent) 2025 book, The Interaction Engine, which puts forth a range of universals (admittedly, very different from Chomsky’s). I do not advocate a Chomskyan account of language or of universals. Even so, I believe that both Majid and Levinson are mistaken in their contention that Chomsky-type universalism has been (almost) disproven. I say this primarily because of the way universality, as well as evidence and counter-evidence bearing on universalism, are defined in a Chomskyan or similar theory. In order to clarify some of what is at stake in different types of theories of universals, I will go through two illustrative cases—Chomsky and Joseph Greenberg, who articulate very different types of universalistic theory.
Theories and Evidence
Unlike many authors outside of the field, linguists are aware that there are different types of universalistic theory about language. They know, for example, that Joseph Greenberg and Noam Chomsky articulated such theories and they were very different from one another. But it is often unclear that even linguists fully understand the exact nature and consequences of such differences. For example, all universalists acknowledge that candidates for universals might have apparent exceptions. How significant the exceptions are, however, is not self-evident. It is a function of the theoretical premises underlying one’s research program. For example, neither Majid’s nor Levinson’s comments, as quoted in Irish, has clear consequences for either a Chomskyan or Greenbergian universalist program in linguistics.
To most readers, it may seem that the case is straightforward: Universalists predict that all languages have property p and are falsified by even a single language that lacks this property. But, despite the popularity of (this popularization of) Popperian falsificationism, that would be very bad science, for two reasons. First, any theory that deals with statistical likelihood does not support falsification by a single disconfirming instance. (As Nigel Fabb reminded me, in some cases, this situation may occur because a language develops in such a way that two generally valid universals cannot both be instantiated in the language; in other words, the application of one universal principle may sometimes interfere with the application of another.) Second, any statistically significant deviation from random distribution requires explanation and is likely to prove relevant for understanding the field in which it occurs. For example, there is a “cross-linguistic preference for suffixing over prefixing” (Culbertson, Smolensky, and Legendre), such that test subjects “learning an artificial language with suffixes identifying word categories outperformed those learning a language with prefixes.” This is not rendered irrelevant to an account of language structure and language acquisition simply because there are languages with prefixes (though of course the account of language structure and acquisition will, in all likelihood, be different than it would have been if there were no languages of the latter sort).
We could just as easily use a universal posited by Levinson, such as regular turn taking, with a limited amount of time (as little as 200 milliseconds) between turns, indicating the extreme rapidity of the transition and thus the refined sensitivity we have to signals of turn ending (see Levinson 34-45). If we happened to find a language in which this is not the case—in which, say, any given speaker could speak as long as he or she wished, without particular justification—that would hardly render the overwhelming predominance of turn taking irrelevant to the study of language. Or consider the metaphorical mapping of time onto space (specifically, the subject-centered version of space, which is handled by parietal cortex, rather than hippocampal, objective spatial relations [see Clark, Boutros, and Mendez 43]). This yields such metaphors as “A bright future is before us,” or “Our best days are behind us.” Languages often stress a mapping which links the future with what is in front of us and the past with what is behind us. It caused a great kerfuffle a few years back when linguists noted that Aymara (a South American, First Nations’ language) apparently reverses this—as does English, when we say that “old age is creeping up on me” or, following Marvell, “at my back, I always hear,/Time’s wingèd chariot hurrying near” (“To His Coy Mistress,” ll.21-22). But even without the English parallels, the existence of the apparently less common pattern in Aymara hardly means there is nothing significant about the evidently more common pattern.
The Greenbergian tradition in linguistic universalism treats this issue straightforwardly. (For a fuller discussion of these points, see Comrie.) Some phenomena are statistical, while others appear to be more absolute in their conformity to laws (i.e., they exhibit more uniform behavior)—not only in linguistics, but in every field where one might seek universals (such as literature). In consequence, Greenbergians distinguish absolute from statisticaluniversals. Moreover, some patterns occur only in contexts where other patterns occur. These are implicationaluniversals. An example of an implicational universal is as follows: If a language has verb-subject-object word order, then it has prepositions, as opposed to postpositions (see Comrie 17; note that implicational universals may themselves be absolute or statistical). When a number of implicational universals cluster together, you have typological universals. Claims regarding all of these types of universals are in principle subject to a statistical significance test. Thus, they are as falsifiable—or at least as open to rational discussion, thus logical and empirical criticism and defense—just as claims are elsewhere in the social and biological sciences.
Chomsky’s approach to universals partially overlaps with that of Greenberg in that he in effect makes use of the idea of implicational universals. But he does so in such a radically different context as to give even those points of overlap a very different significance or function in the overall, theoretical system. Since Chomsky developed numerous theories in the course of his career, I will focus on one–the “Principles and Parameters” (P&P) framework–both because I find it particularly appealing and because it is adaptable to different specific theories, at least to the extent of illustrating the problems with a “simple falsificationist” view of universals. First of all, Chomskyan universals—in the sense of universals that enter into his theories–are absolute. Moreover, this is because they form the core of every language and variation in languages is confined to the periphery (as suggested in the quote from Majid). However, that core and periphery are not simply what we think of intuitively as central and marginal. For example, to a great extent, the sound-meaning pairings that define vocabulary may be categorized as peripheral. This is because the core or universal part of language is defined by a specific, explanatory function. It is, specifically, what allows us to learn language. The universals in Chomsky-type theories are just that subset of universals (in the Greenbergian sense) that form our innate Language Acquisition Device (LAD). Thus, the elements of that LAD are shared by all speakers of all languages. But that does not mean that they appear in all languages in the same surface form. Indeed, they could not appear in all languages because, in a P&P framework, the elements are principles with specified (i.e., specifically listed) variables. Learning a language consists largely in learning which variables to select from each list. For example, the “head directionality parameter” specifies that standard word order in a language can be head first or head last. (A “head” is “the word that a phrase is built around” [Baker 247].)
I should point out that the P&P framework is much more complex than I have indicated. For example, the parameters are not set at random, but in a definite order. Only after the head directionality parameter is set can certain other parameters be set. Indeed, setting one parameter one way typically sets some other parameters, in effect preventing the child from setting them. (For a lucid treatment of some of these complexities, see Baker.) Even in this simplified form, however, I hope it is clear why the head directionality parameter counts as a universal in (this particular) Chomskyan theory, even if there seem to be exceptions to its instantiation. Those exceptions should be handled by principles and parameters which apply later, later both logically and developmentally (as can be argued on the basis of acquisition data).
I also hope it is clear, without entering into technical details, that the P&P framework or other Chomsky-type theories of universals can result in what Imre Lakatos called a “degenerating” research program. For example, it could develop in such a way as to require an increasing number of ad hoc principles which involve no generalizations of further linguistic patterns, but serve merely to allow for individual anomalies. An extreme and unrealistic, but illustrative case would be principles which referred to single languages as exceptions—say, a disjunctive parameter, such as “Is Aymara/is not Aymara,” with a simple statement of Aymara’s unique characteristics following from the former alternative. As such ad hoc principles multiply, the theory decreases in its explanatory power. Thus, these types of theories are open to rational adjudication. However, they are not open to simple falsification by the discovery of apparently anomalous properties in a particular language. In short, the intuitive case for the simple falsification of either Greenbergian or Chomskyan universals is not valid.
Before addressing literary universals, it is worth mentioning three other differences between Greenbergian and Chomskyan universals. (For a concise representation of the key differences, see Figure 1.) Again, Chomskyan universals serve to explain language acquisition—both what children acquire and in what order they acquire it; as part of this, they are also recruited to explain recurring errors made by children in the course of language acquisition. But how might the universals be explained? As they are innate, the obvious way in which the LAD universals would be explained is by evolution, though Chomsky has himself stressed that they could have developed by other means (cf. his comments on evolution in On Nature and New Horizons). In any case, Greenbergian universals are in principle open to any sort of explanation. Most importantly for our purposes, not being innate, they may have developed through convergent social development and thus be open to “culturalist” explanations. Moreover, as such, they may serve to explain various social or cultural developments as well. (Of course, all universals to some extent rely on innate features in some respect—for example, universals that concern visible objects rely to some extent on features of the human eye. However, in itself, that reliance is trivial.)
Another difference is that Chomsky-type universals can be articulated in part through the examination of one language—say, English. This is because the LAD is the same for all children. Insofar as one is able to formulate the principles and parameters needed to learn English, one has formulated some of the principles and parameters needed to constitute the LAD. Though Chomskyan universals cannot be fully articulated without recourse to all available language data, initial articulation can rest on one language. In contrast, Greenberg-type universals must observe the constraints of genetic and areal distinctness far more rigorously. In other words, they must draw on languages that have independent lineages (genetic distinctness) and that have not influenced one another through contact (areal distinctness).
I have drawn on a non-innatist version of P&P theory as a model for aspects of my accounts of descriptive ethics (see Literature and Moral Feeling), authorial creativity and simulation (see How Authors’ Minds Make Stories), and other phenomena. But the applicability of a Chomsky-type theory to non-linguistic problems seems limited as the initial formulation of the theory is so narrowly tailored to issues of language. Indeed, I take it to be fairly clear that Chomskyan universals as such (thus, not merely as general models) are on the whole too restrictively defined to be of much direct use in literary study (exceptions to this would include some of Paul Kiparsky’s and Nigel Fabb’s work). In contrast, the Greenbergian stress on statistical universals and the possibility of social or cultural explanation fits the interests of literary scholars much more clearly than would a stress on acquisition. Acquisition of, for example, narrative appears to come down to the understanding of particular causal sequences and distinguishing agency from efficient causality. These do not appear to be problems unique to literature, but rather appear to be the sorts of thing developmental psychologists would study. More generally, whatever one thinks of the idea that language is a modular (thus, relatively independent) function of the human brain, there is clearly much more reason to think of language as modular than to think of literature as such. Thus, when literary theorists, such as Gerald Prince, drew on Chomsky-type theories to model narrative, they had reason to consider models from linguistics, but (I believe) erred in choosing a Chomsky-type theoretical framework.
I should mention one more thing about Chomsky and Greenberg. In distinguishing these two types of theories, I passed over what they have in common. One part of this commonality is important for discussing universals. It is also commonly misunderstood, or not considered at all. To clarify this point, I need first to draw a terminological distinction between disciplinary description and explanation on the one hand and ultimate or underlying description and explanation on the other. By “disciplinary description and explanation,” I mean description and explanation of some particular set of phenomena as they appear to us within their field of study—say, linguistics with its references to phonological, morphological, and syntactic properties. By “underlying description and explanation,” I mean description and explanation in terms of some appropriate non-disciplinary mechanisms—usually material-biological or social; for example, description and explanation in terms of particular sequences of neuronal activation in particular regions of the brain associated with identifiable functions. At some point, the disciplinary descriptions and explanations should be identifiable with some underlying material-biological description and explanation. For example, in one version of his syntactic theory, “Government and Binding” theory, Chomsky takes it as a datum that “I wanna win the race” is grammatically fine, but that *”Who do you wanna win the race” is ungrammatical (see, for example, Lectures 181 and Rules 159-160). His explanation for this is (very roughly) that, in underlying structure, the question has an earlier version, “You want who to win the race.” The “who” moves to the front, but leaves a “trace” between “want” and “to.” That trace marks any contraction of “want” and “to” as ungrammatical. What does this have to do with disciplinary and underlying explanation? Well, the trace-blocking account of wanna-contraction (a disciplinary explanation, dealing as it does with familiar elements of language, such as words) should presumably have some correlate to the trace and some correlate to the movement of “who,” along with correlates to the other familiar, language elements in the underlying explanation of “wanna”-contraction. If such correlates are lacking, either the trace-blocking account is mistaken or the underlying account is inadequate to the phenomenon of “wanna”-contraction.
Note that this means one cannot simply take up a non-disciplinary concept from neuroscience–or any other area of cognitive or social science–and use it to explain an area such as language or literature, unless there is some way of translating that nondisciplinary account into the linguistic or literary account and vice-versa. The disciplinary account constitutes our fundamental explanation of the phenomena under consideration. The material account is, so to speak, an explanation of our disciplinary explanation. Actually, the same point holds for social explanations or any other accounts we would hope to be underlying explanations. For example, it is undoubtedly the case that language has developed in such a way as to facilitate cooperation. However, saying this does not constitute an explanation of any linguistic phenomena. If we want to explain language by cooperation, we need to explain how cooperation accounts for the preference for suffixes. Similarly, our acquisition and use of language have no doubt benefited from the mirror neuron system. But saying this does not even begin to isolate syntactic, morphological, phonological, or even lexical phenomena, to identify patterns within or across languages, or to say anything at all specific to language. (For a detailed criticism of the uncritical appeal to mirror neurons, which seems particular common in literary study, see Hickok.) This is not to say that underlying explanations must cohere into the single, unified system. Again, that is what Chomsky is seeking in at least some iterations of his linguistic theory. But in a Greenbergian account, an underlying explanation could form a patchwork of unrelated “local” theories. The key point is simply that, in each theory, there must be a point at which the entities named in the underlying account map directly onto the entities named in the disciplinary account in order to guarantee that the two accounts are treating the same objects and are explaining the same phenomena. Of course, any such mapping is bound to be highly complex. For instance, a word will be mapped onto some areas of cortex governing its pronunciation, other areas governing its meaning, and so on. The point is simply that loose, general correlations may provide us with areas to explore and they may be fruitful in that regard, but as long as they remain general correlations, they tell us nothing whatsoever about universals in the disciplinary field we are studying—language, narrative, literature, or whatever.
WORKS CITED
Baker, Mark. The Atoms of Language: The Mind’s Hidden Rules of Grammar. New York: Basic Books, 2001.
Chomsky, Noam. Lectures on Government and Binding: The Pisa Lectures. Dordrecht, Neth: Foris, 1981.
Chomsky, Noam. New Horizons in the Study of Language and Mind. Cambridge: Cambridge UP, 2000.
Chomsky, Noam. On Nature and Language. Cambridge: Cambridge UP, 2002.
Chomsky, Noam. Rules and Representations. New York: Columbia UP, 1980.
Clark, David, Nashaat Boutros, and Mario Mendez. The Brain and Behavior: An Introduction to Behavioral Neuroanatomy. 2nd ed. Cambridge: Cambridge UP, 2005.
Comrie, Bernard. Language Universals and Linguistic Typology. Chicago, IL: U of Chicago P, 1982.
Culbertson, Jennifer, Paul Smolensky, and Géraldine Legendre. “Learning Biases Predict a Word Order Universal.” Cognition 122.3 (2012): 306-329.
Evans, N. and S. Levinson. “The Myth of Language Universals: Language Diversity and Its Importance for Cognitive Science.” Behavioral and Brain Sciences 32.5 (2009): 429–448.
Hickok, Gregory. The Myth of Mirror Neurons. New York: Norton, 2014.
Hogan, Patrick Colm. How Authors’ Minds Make Stories. Cambridge: Cambridge UP, 2013.
Hogan, Patrick Colm. Literature and Moral Feeling: A Cognitive Poetics of Ethics, Narrative, and Empathy. Cambridge: Cambridge UP, 2022
Irish, Bradley. The Universality of Emotion: Perspectives from the Sciences and Humanities. Cambridge: Cambridge UP, 2025.
Kiparsky, Paul. “The Role of Linguistics in a Theory of Poetry.” In Essays in Modern Stylistics. Ed. Donald C. Freeman. London: Methuen, 1981.
Kiparsky, Paul. “On Theory and Interpretation.” In The Linguistics of Writing: Arguments Between Language and Literature. Ed. Nigel Fabb, Derek Attridge, Alan Durant, and Colin MacCabe. New York: Methuen, 1987.
Lakatos, Imre. “Falsification and the Methodology of Scientific Research Programmes.” In Criticism and the Growth of Knowledge. Ed. Imre Lakatos and Alan Musgrave. Cambridge: Cambridge UP, 1970.
Levinson, Stephen. The Interaction Engine: Language in Social Life and Human Evolution. Cambridge: Cambridge UP, 2025.
Majid, A. “Current Emotion Research in the Language Sciences.” Emotion Review 4.4 (2012): 432–443.
Prince, Gerald. A Grammar of Stories. The Hague: Mouton, 1973.