Frost provides a strong, refreshed case for the idea that reading is “parasitic on language,” a correction on the claim by Mattingly (Reference Mattingly, Kavanaugh and Mattingly1972) that reading is “parasitic on speech” (Snowling & Hulme Reference Snowling, Hulme, Snowling and Hulme2005, p. 397). Similar expressions of the idea that language rather than speech is the reference point for reading are found in Perfetti (Reference Perfetti2003) and Seidenberg (Reference Seidenberg, McCardle, Miller, Lee and Tzeng2011), among others.
Frost's focus on language provides a reminder: Successful skilled reading enables the language system to take over from the visual system with astonishing speed. It does this because the orthography has somehow managed to enable the language to be “seen” through the print.
Language is more than speech, and orthographies are more than the spellings of phonemes. The big issue is how to understand why writing systems have come to work the way they do. Alphabetic writing systems are not generally notational systems for speech but notational systems for language: the morphology as well as the phonology. One general statement of this tradeoff is, “Alphabetic writing picks out the phonemic level … and then makes various adaptations to the morphological level” (Perfetti Reference Perfetti2003, p. 22). (See also Perfetti & Harris [under review] for a fuller treatment of the language-writing connection that draws on both Frost's target article and Seidenberg [2011].)
Frost's claim, however, is stronger than this “various adaptations” idea: He writes, “orthographies are structured so that they optimally represent the languages' phonological spaces, and their mapping into semantic meaning” (sect. 3, para. 1). The more memorable rendering of this claim is that “every language gets the writing system it deserves” (sect. 3). It is worth a brief digression to note that both the underlying idea and the literal form of the claim have resonated for other scholars. Frost attributes the quote to Ignatius Mattingly. Another source is M. A. K Halliday, who, in a 1976 lecture, observed that the development of writing was an incremental process over long periods of time. “In the course of this process (unlike the conscious efforts, which are often subject to the fads and fashions of linguistics of the time), a language usually gets the sort of writing system it deserves” (Halliday Reference Halliday and Webster2003, p. 103, reprinted from Halliday Reference Halliday1977).
Recently, Seidenberg also has expressed this claim in some interesting detail by noting a correlation between language typologies and writing systems: Complex inflectional morphology begets shallow orthography. Seidenberg captured the tradeoffs writing makes between exposing phonemes and exposing morphemes with the “grapholinguistic equilibrium hypothesis” (Seidenberg Reference Seidenberg, McCardle, Miller, Lee and Tzeng2011). At the equilibrium point, “spoken languages get the writing systems they deserve” (p. 169).
The writing system that is deserved might not be one that is literally optimal in the sense defined by Frost. Indeed, the optimization hypothesis is elegantly startling in its implausibility. Perhaps optimization algorithms could be run on the phonological spaces of a sample of a couple dozen languages to see what the theoretically optimal orthography should be – that is, if one could also map the morpheme semantic spaces as well. Underlying optimization is the assumption that writing systems “learn” over time, by analogy to network models that self-correct in response to environmental input. Applied to writing systems, this self-correcting network would modify spellings in response to feedback that pushes the network to spell more in relation to morphology or to phonology. It is an intriguing idea that might work, if there were sufficient tolerance about spelling variability to allow the “votes” of usage to lead to increasing stability. There are strong conservative forces at play in writing, and while some change can and does occur, its reach is checked by these forces. Depending on the net result of change forces and conservative forces, orthographies will wind up considerably short of optimality.
Of course, the optimality hypothesis may be taken to mean that, by now, all orthographies have reached the equilibrium state proposed by Seidenberg and are now optimal. But to avoid test by presumption (optimality is asserted to have resulted in the world's current distribution of writing), it remains to develop optimization algorithms that will reveal the equilibrium point for any realistically large vocabulary for a given language.
To return to the weaker claim of adaptability: Adaptations can occur. Some are gradual, as when Chinese characters moved from more iconic to more abstract over time. (Similar changes occurred in Middle Eastern writing systems.) Some are rather sudden, as when King Sejong invented an alphabet for Korean in 1446. It is interesting for the adaptations argument that this pure Hangul alphabet, with one-to-one mapping of graphs to phonemes, gave way to a certain degree of one-to-many mappings in order to allow words that shared root morphemes but not exact pronunciations to be represented by a single spelling (Perfetti Reference Perfetti2003).
In the absence of an enlightened monarch, attempts at sudden change (spelling reform) meet strong resistance. In 1906, Andrew Carnegie's Simplified Spelling Board made recommendations for English spelling change with mixed success. While thru has become a competitor to through, the recommendations to spell tapped as tapt and wish as wisht died the death of the morphologically opaque.
Finally, whether writing systems adapt to language optimally or only roughly, there is the question of what makes a theory of reading universal. Models that allow neuronal flexibly in coding graph order are not wrong just because languages vary in their tolerance for graph order. The neural resources for reading are modified through experience with specific languages and orthographies (Bolger et al. Reference Bolger, Perfetti and Schneider2005; Paulesu et al. Reference Paulesu, McCrory, Fazio, Menoncello, Brunswick, Cappa, Cotelli, Cossu, Corte, Lorusso, Pesenti, Gallagher, Perani, Price, Frith and Frith2000) and both brain circuits and reading behavior show “accommodation” (Perfetti et al. Reference Perfetti, Liu, Fiez, Nelson, Bolger and Tan2007) to the mapping principles and their orthographic implementation. The neural basis of alphabetic reading certainly can allow adaption to the specific demands of order-coding. Chinese complicates matters by placing weak demands on serial order of graph-to-speech mapping and strong demands on spatial relations. These variations do not rule out a universal mechanism that accommodates such differences.
It is intriguing to consider that these differences among languages and writing systems could fall out from a single high-level universal based on optimizing language-writing mappings. However, specific models that can actually account for these differences more directly, for example, by specifying reading's behavioral and brain adaptations to experience, have the major burden of explanation.
Frost provides a strong, refreshed case for the idea that reading is “parasitic on language,” a correction on the claim by Mattingly (Reference Mattingly, Kavanaugh and Mattingly1972) that reading is “parasitic on speech” (Snowling & Hulme Reference Snowling, Hulme, Snowling and Hulme2005, p. 397). Similar expressions of the idea that language rather than speech is the reference point for reading are found in Perfetti (Reference Perfetti2003) and Seidenberg (Reference Seidenberg, McCardle, Miller, Lee and Tzeng2011), among others.
Frost's focus on language provides a reminder: Successful skilled reading enables the language system to take over from the visual system with astonishing speed. It does this because the orthography has somehow managed to enable the language to be “seen” through the print.
Language is more than speech, and orthographies are more than the spellings of phonemes. The big issue is how to understand why writing systems have come to work the way they do. Alphabetic writing systems are not generally notational systems for speech but notational systems for language: the morphology as well as the phonology. One general statement of this tradeoff is, “Alphabetic writing picks out the phonemic level … and then makes various adaptations to the morphological level” (Perfetti Reference Perfetti2003, p. 22). (See also Perfetti & Harris [under review] for a fuller treatment of the language-writing connection that draws on both Frost's target article and Seidenberg [2011].)
Frost's claim, however, is stronger than this “various adaptations” idea: He writes, “orthographies are structured so that they optimally represent the languages' phonological spaces, and their mapping into semantic meaning” (sect. 3, para. 1). The more memorable rendering of this claim is that “every language gets the writing system it deserves” (sect. 3). It is worth a brief digression to note that both the underlying idea and the literal form of the claim have resonated for other scholars. Frost attributes the quote to Ignatius Mattingly. Another source is M. A. K Halliday, who, in a 1976 lecture, observed that the development of writing was an incremental process over long periods of time. “In the course of this process (unlike the conscious efforts, which are often subject to the fads and fashions of linguistics of the time), a language usually gets the sort of writing system it deserves” (Halliday Reference Halliday and Webster2003, p. 103, reprinted from Halliday Reference Halliday1977).
Recently, Seidenberg also has expressed this claim in some interesting detail by noting a correlation between language typologies and writing systems: Complex inflectional morphology begets shallow orthography. Seidenberg captured the tradeoffs writing makes between exposing phonemes and exposing morphemes with the “grapholinguistic equilibrium hypothesis” (Seidenberg Reference Seidenberg, McCardle, Miller, Lee and Tzeng2011). At the equilibrium point, “spoken languages get the writing systems they deserve” (p. 169).
The writing system that is deserved might not be one that is literally optimal in the sense defined by Frost. Indeed, the optimization hypothesis is elegantly startling in its implausibility. Perhaps optimization algorithms could be run on the phonological spaces of a sample of a couple dozen languages to see what the theoretically optimal orthography should be – that is, if one could also map the morpheme semantic spaces as well. Underlying optimization is the assumption that writing systems “learn” over time, by analogy to network models that self-correct in response to environmental input. Applied to writing systems, this self-correcting network would modify spellings in response to feedback that pushes the network to spell more in relation to morphology or to phonology. It is an intriguing idea that might work, if there were sufficient tolerance about spelling variability to allow the “votes” of usage to lead to increasing stability. There are strong conservative forces at play in writing, and while some change can and does occur, its reach is checked by these forces. Depending on the net result of change forces and conservative forces, orthographies will wind up considerably short of optimality.
Of course, the optimality hypothesis may be taken to mean that, by now, all orthographies have reached the equilibrium state proposed by Seidenberg and are now optimal. But to avoid test by presumption (optimality is asserted to have resulted in the world's current distribution of writing), it remains to develop optimization algorithms that will reveal the equilibrium point for any realistically large vocabulary for a given language.
To return to the weaker claim of adaptability: Adaptations can occur. Some are gradual, as when Chinese characters moved from more iconic to more abstract over time. (Similar changes occurred in Middle Eastern writing systems.) Some are rather sudden, as when King Sejong invented an alphabet for Korean in 1446. It is interesting for the adaptations argument that this pure Hangul alphabet, with one-to-one mapping of graphs to phonemes, gave way to a certain degree of one-to-many mappings in order to allow words that shared root morphemes but not exact pronunciations to be represented by a single spelling (Perfetti Reference Perfetti2003).
In the absence of an enlightened monarch, attempts at sudden change (spelling reform) meet strong resistance. In 1906, Andrew Carnegie's Simplified Spelling Board made recommendations for English spelling change with mixed success. While thru has become a competitor to through, the recommendations to spell tapped as tapt and wish as wisht died the death of the morphologically opaque.
Finally, whether writing systems adapt to language optimally or only roughly, there is the question of what makes a theory of reading universal. Models that allow neuronal flexibly in coding graph order are not wrong just because languages vary in their tolerance for graph order. The neural resources for reading are modified through experience with specific languages and orthographies (Bolger et al. Reference Bolger, Perfetti and Schneider2005; Paulesu et al. Reference Paulesu, McCrory, Fazio, Menoncello, Brunswick, Cappa, Cotelli, Cossu, Corte, Lorusso, Pesenti, Gallagher, Perani, Price, Frith and Frith2000) and both brain circuits and reading behavior show “accommodation” (Perfetti et al. Reference Perfetti, Liu, Fiez, Nelson, Bolger and Tan2007) to the mapping principles and their orthographic implementation. The neural basis of alphabetic reading certainly can allow adaption to the specific demands of order-coding. Chinese complicates matters by placing weak demands on serial order of graph-to-speech mapping and strong demands on spatial relations. These variations do not rule out a universal mechanism that accommodates such differences.
It is intriguing to consider that these differences among languages and writing systems could fall out from a single high-level universal based on optimizing language-writing mappings. However, specific models that can actually account for these differences more directly, for example, by specifying reading's behavioral and brain adaptations to experience, have the major burden of explanation.