The target article makes a plea for the development of a universal model of reading “which outlines the common cognitive operations involved in orthographic processing in all writing systems” (Abstract) – that is, a model that is based on the universals of reading. But are there any universals of reading?
It is clear that there are universals of language. There are two kinds of language universals: formal and substantive (Chomsky Reference Chomsky1965). A few examples of each will illustrate the difference. Probably the most discussed formal linguistic universal is recursivity (a.k.a. “discrete infinity”). Recursive procedures create outputs that suffice for their own reapplication. Recursive procedures account for the creativity of human language, yielding potentially unbounded expressions, for example, possessives (Bob's mother's aunt's nephew's…) and verbs with sentences as complements (Bob told mother that aunt Betsy believes her nephew Ben said…). Another formal universal is structure dependence. The sequences of words that are produced and understood by language users are hierarchically organized, as opposed to being organized in a linear fashion, like beads on a string. For example, no human language derives a question from a statement by pronouncing the last word of the statement first, and even adults find it difficult to learn to play the part of a puppet that responds to other people by omitting the first three words of whatever they say (Read & Schreiber Reference Read, Schreiber, Wanner and Gleitman1982). Although structure-independent operations are conceptually quite simple, human languages and their learners never adopt them (see, e.g., Crain & Nakayama Reference Crain and Nakayama1987; Smith & Tsimpli Reference Smith and Tsimpli1995).
Substantive universals provide inventories of linguistic types, including the syntactic and semantic primitives of human languages. For example, all human languages have phrases that contain nouns and verbs. Verbs take noun phrases as their “arguments,” and each argument noun phrase is associated with a unique thematic role (e.g., agent, theme, experiencer). Some verbs take a single argument (e.g., sleep, dine), some take two arguments (e.g., donate, meet) and some take more than two arguments (e.g., give, put). Turning to semantics, all languages express negation (e.g., no, not), and all languages use adverbial quantificational expressions (e.g., all, always). Finally, when two or more logical expressions appear together in sentences, different “scope” relations among these expressions are computed. This yields semantic ambiguities: for example, All planes do not carry pets.
Linguistic universals make important contributions to language acquisition. Universals like structure-dependence reveal that certain linguistic operations, although logically possible and computationally simple, are never found. It follows that children are not expected to try out such operations in the course of language development. This explains, in part, the universal mastery of human languages by young children; that is, the fact that every normal child is able to rapidly acquire any human language by age 3 or 4, without formal instruction or carefully sequenced input. This includes both spoken and visual-gestural (sign) languages. Notably, universal mastery is not paralleled in children's attempts to master writing systems.
Another relevant observation about linguistic universals is that they are contingent facts about human language. Human languages could have evolved differently. Human languages might have consisted of lists of word sequences, rather than sentences generated by a recursive procedure; human languages might not have included adverbial quantifiers; and logical expressions might have been interpreted in linear order, rather than being assigned different scope relations.
Are there reading universals, in the way in which there are language universals? The target article claims this to be true. In contrast, we take the view that there are no reading universals. One problem is that it is not clear what Frost means by “reading universal.” Sometimes the idea seems to be that reading universals are general properties of writing systems; at other times, the idea seems to be that reading universals are general properties of the cognitive information-processing systems used for reading. We consider both of these ideas.
As the target article states, the definition of “writing system” is: any system by which visual symbols can be used to express meanings and pronunciations. The world has had hundreds, probably thousands, of writing systems. Can one identify any properties of this collection of systems that are contingently true of all of them, in the way that, as we have noted above, there are various features of language which are contingently true of all human languages? Note that one cannot reply to this challenge by saying “All writing systems represent meaning on the page” or “All writing systems represent pronunciation on the page” because these statements are not contingently true; they are true by definition of the term writing system. Can Frost offer a short list of properties which all of the world's writing systems just happen to have, as one can do with all the world's languages? We doubt that such a list can be compiled. But in the absence of such a list, the claim that there are reading universals in the sense of universal features of writing systems has no justification.
Frost might instead mean, by his claim that there are reading universals, that the cognitive information-processing systems that are used by readers of all the world's writing systems happen to have certain features in common. But, again, to justify such a claim one would have to produce a list of those properties that contingently hold for every reading system regardless of which writing system it is used for. We doubt that this kind of list could be compiled either.
In sum, then, our view is that whichever of these two different senses of “reading universal” is adopted, there are no reading universals. All that need be done to refute our view is to provide a list of some of these reading universals, just as we have provided a list of some of the universals of language.
But if there are no universals of reading, then Frost's desire for the development of a universal model of reading can never be fulfilled.
The target article makes a plea for the development of a universal model of reading “which outlines the common cognitive operations involved in orthographic processing in all writing systems” (Abstract) – that is, a model that is based on the universals of reading. But are there any universals of reading?
It is clear that there are universals of language. There are two kinds of language universals: formal and substantive (Chomsky Reference Chomsky1965). A few examples of each will illustrate the difference. Probably the most discussed formal linguistic universal is recursivity (a.k.a. “discrete infinity”). Recursive procedures create outputs that suffice for their own reapplication. Recursive procedures account for the creativity of human language, yielding potentially unbounded expressions, for example, possessives (Bob's mother's aunt's nephew's…) and verbs with sentences as complements (Bob told mother that aunt Betsy believes her nephew Ben said…). Another formal universal is structure dependence. The sequences of words that are produced and understood by language users are hierarchically organized, as opposed to being organized in a linear fashion, like beads on a string. For example, no human language derives a question from a statement by pronouncing the last word of the statement first, and even adults find it difficult to learn to play the part of a puppet that responds to other people by omitting the first three words of whatever they say (Read & Schreiber Reference Read, Schreiber, Wanner and Gleitman1982). Although structure-independent operations are conceptually quite simple, human languages and their learners never adopt them (see, e.g., Crain & Nakayama Reference Crain and Nakayama1987; Smith & Tsimpli Reference Smith and Tsimpli1995).
Substantive universals provide inventories of linguistic types, including the syntactic and semantic primitives of human languages. For example, all human languages have phrases that contain nouns and verbs. Verbs take noun phrases as their “arguments,” and each argument noun phrase is associated with a unique thematic role (e.g., agent, theme, experiencer). Some verbs take a single argument (e.g., sleep, dine), some take two arguments (e.g., donate, meet) and some take more than two arguments (e.g., give, put). Turning to semantics, all languages express negation (e.g., no, not), and all languages use adverbial quantificational expressions (e.g., all, always). Finally, when two or more logical expressions appear together in sentences, different “scope” relations among these expressions are computed. This yields semantic ambiguities: for example, All planes do not carry pets.
Linguistic universals make important contributions to language acquisition. Universals like structure-dependence reveal that certain linguistic operations, although logically possible and computationally simple, are never found. It follows that children are not expected to try out such operations in the course of language development. This explains, in part, the universal mastery of human languages by young children; that is, the fact that every normal child is able to rapidly acquire any human language by age 3 or 4, without formal instruction or carefully sequenced input. This includes both spoken and visual-gestural (sign) languages. Notably, universal mastery is not paralleled in children's attempts to master writing systems.
Another relevant observation about linguistic universals is that they are contingent facts about human language. Human languages could have evolved differently. Human languages might have consisted of lists of word sequences, rather than sentences generated by a recursive procedure; human languages might not have included adverbial quantifiers; and logical expressions might have been interpreted in linear order, rather than being assigned different scope relations.
Are there reading universals, in the way in which there are language universals? The target article claims this to be true. In contrast, we take the view that there are no reading universals. One problem is that it is not clear what Frost means by “reading universal.” Sometimes the idea seems to be that reading universals are general properties of writing systems; at other times, the idea seems to be that reading universals are general properties of the cognitive information-processing systems used for reading. We consider both of these ideas.
As the target article states, the definition of “writing system” is: any system by which visual symbols can be used to express meanings and pronunciations. The world has had hundreds, probably thousands, of writing systems. Can one identify any properties of this collection of systems that are contingently true of all of them, in the way that, as we have noted above, there are various features of language which are contingently true of all human languages? Note that one cannot reply to this challenge by saying “All writing systems represent meaning on the page” or “All writing systems represent pronunciation on the page” because these statements are not contingently true; they are true by definition of the term writing system. Can Frost offer a short list of properties which all of the world's writing systems just happen to have, as one can do with all the world's languages? We doubt that such a list can be compiled. But in the absence of such a list, the claim that there are reading universals in the sense of universal features of writing systems has no justification.
Frost might instead mean, by his claim that there are reading universals, that the cognitive information-processing systems that are used by readers of all the world's writing systems happen to have certain features in common. But, again, to justify such a claim one would have to produce a list of those properties that contingently hold for every reading system regardless of which writing system it is used for. We doubt that this kind of list could be compiled either.
In sum, then, our view is that whichever of these two different senses of “reading universal” is adopted, there are no reading universals. All that need be done to refute our view is to provide a list of some of these reading universals, just as we have provided a list of some of the universals of language.
But if there are no universals of reading, then Frost's desire for the development of a universal model of reading can never be fulfilled.