103 assessment for eighth graders: NAEP (2011), 58.
103 [Individual demonstrates ability]: From “English Language Arts Standards » Reading: Literature » Grade 11-12,” Common Core State Standards Initiative, http://goo.gl/JPOsKL.
105 Two esteemed reading researchers: Beck & Juel (1995).
106 Louis Golstein and Cathy Browman: Browman & Goldstein (1990).
106 The translation from one code: Werker & Tees (1999). Phonology is also a branch of linguistics that addresses a much more extensive range of phenomena than the pronunciations and sounds of words.
106 acoustic blobs: Infants rapidly acquire strong language-specific sensitivity to differences between initial sounds in syllables such as “bin” and “din.” They nonetheless treat the syllables alike in word-learning experiments (Stager & Werker 1997). The sounds can be distinguished phonetically (based on smaller features) but aren’t yet treated as discrete phonemic segments.
107 “lexical restructuring”: Metsala & Walley (1998).
107 Phonemes are abstractions: Phonemes (and the letters that represent them) involve a further abstraction because they ignore systematic variation in how they are pronounced. Both “pin” and “spin” contain the phoneme /p/, written with the same letter, but the sound is aspirated in “pin” and unaspirated in “spin.” These variants of the phoneme (called allophones) are not represented in this writing system. In other languages, the allophones are distinct phonemes with their own letters.
107 speech consists of phonemes: Fowler (1991).
108 Studies in English: McCardle, Scarborough, & Catts (2001).
108 Letter songs may be common: Horace Mann (1844, 91, 92) also found letter names perplexing: “I am satisfied that our greatest error, in teaching children to read, lies in beginning with the alphabet;—in giving them what are called the ‘Names of the Letters,’ a ,b, c, &c.” Later: “Although in former reports and publications I have dwelt at length upon what seems to me the absurdity of teaching to read by beginning with the alphabet, yet I feel constrained to recur to the subject again,—being persuaded that no thorough reform will ever be effected in our schools until this practice is abolished.”
108 The names teach the child: Some educators now favor associating letters with sounds rather than names on the view that the sound “b” is more relevant to reading than the name “bee.” The initial sound in “bat” is not pronounced “bee” to be sure, but it is not the “buh” we conventionally produce as the letter’s pronunciation either. Consonants cannot be pronounced in isolation; they require a vowel, which is truncated when we say “b.” A recording of the sound “b” produced in isolation, spliced onto a recording of “at,” yields a weird syllable, not “bat.”
108 a beginning reader has to learn: The font on the left in Figure 6.1 was created by David Rumelhart, a brilliant and inspirational figure in modern cognitive science (Rumelhart & Siple 1974).
109 Recognizing letters: Lake, Salakhutdinov, & Tenenbaum (2015). See a model learn to recognize handwritten digits here: http://goo.gl/S2Y2DZ. Explanation here: “Explanation of the Digit Movies,” University of Toronto Computer Science, http://goo.gl/AbNxE2. For a fascinating discussion of how a deep learning network handled the letter categorization problem for fifty fonts, see Erik Bernhardsson, “Analyzing 50k Fonts Using Deep Neural Networks,” erikbern.com, January 20, 2016, https://goo.gl/e8S2w5.
109 role of names in forming categories: Lupyan, Rakison, & McClelland (2007).
110 narrow sense of knowing: The Peabody Picture Vocabulary Test is a widely used assessment of this type; Dunn & Dunn (2015).
110 Charles Perfetti: Perfetti (2007).
111 Much of our knowledge: MacDonald, Pearlmutter, & Seidenberg (1994).
111 The cup could hold coffee: Seidenberg (1997).
111 Five- to six-year-old English learners: Beck & McKeown (1991).
111 This question assumes great importance: Hoff (2003).
111 effective teaching procedures: Beck, McKeown, & Kucan (2013).
112 “You shall know a word”: Firth (1957).
112 of little use: The limited relevance of language statistics was the point of Chomsky’s famous sentence “Colorless green ideas sleep furiously.” The statistical properties of the word combinations were said to be so low as to be indistinguishable from “Furiously sleep ideas green colorless,” which is word salad. Chomsky was wrong, however. See this post for an overview of the research by computational linguist Fernando Perreira: “Colorless Green Probability Estimates,” Language Log, October 4, 2003. We also implemented a simple computational model that distinguished between the two by tracking expected semantic properties rather than specific words: Allen & Seidenberg (1999).
112 the effects of brain injury on reading: Landauer & Dumais (1997).
113 According to the statistical learning theory: The “bootstrapping” concept was introduced by Lila Gleitman (1990) in a highly influential account of how children learn about the syntactic structures associated with words, but it applies to learning many things.
113 This account of word learning also explains: Carey (1978).
113 children are born knowing: The innate concepts theory comes from Fodor (1975). The statistical learning account doesn’t exclude the possibility that human biology biases infants’ learning toward certain types of generalizations.
113 Computational analyses of the performance: Zhu (2005). It should be possible to use statistical methods to determine which words should be selected for explicit instruction to achieve the greatest impact, given the current state of a child’s vocabulary. Words can be represented in a multidimensional semantic space, with the distance between them indicating degree of similarity. This space can be mapped for samples of words (e.g., the ones that a third grader should know). Call this the goal. Learning a word also affects other words with which it overlaps. Given an index of the current state of a third grader’s vocabulary, it could be determined which word would have the biggest impact on moving the state of their knowledge toward the goal. This would be a good thing because vocabulary instruction is time-consuming, and teaching time is limited.
114 The late Martin Gardner: Douglas Martin, “Martin Gardner, Puzzler and Polymath, Dies at 95,” New York Times, May 23, 2010.
115 Kanzi, a bonobo: Savage-Rumbaugh et al. (1986).
115 The scientists who conducted: Seidenberg & Petitto (1987).
115 Research has clearly established: Justice & Ezell (2002).
115 Some children do learn: Given the overwhelming evidence that children benefit from reading instruction and are hindered by poor or absent instruction, it astonishes me when individuals with no discernible expertise but a popular platform declare, “Children teach themselves to read.” Gray (2010), the author of an introductory psychology textbook.
115 Reading to children: This keen observation is due to Maryellen MacDonald, who shared it with me some years ago, as well as the duckling and hungry caterpillar examples. Montag & MacDonald (2015), Montag et al. (2015), and Cameron-Faulkner & Noble (2013) document that children’s books contain a higher proportion of complex sentence types (passives, relative clauses, conjoined clauses) and more diverse vocabulary than does parental speech to children. Montag & MacDonald (2015) found that children’s reading experience affected how often they used certain complex sentence structures when speaking.
116 At thirty words: On children’s speech and parental speech to children, see Snow (1977); Huttenlocher et al. (1991).
116 You talk better too: See Calvin and Hobbes: http://goo.gl/GRc8wB.
116 A low-income family: Neuman & Celano (2001).
116 Reach Out and Read program: http://www.reachoutandread.org.
117 Kindergarten is sometimes: Bassok & Rorem (2014) report data indicating increases in teachers’ emphasis on reading in kindergarten. Whether children benefit from this earlier emphasis hasn’t been established.
117 Within the pre-K to grade two: Snow et al. (1977); Cunningham & Zibulsky (2013). These findings run counter to the intuition that starting as early as possible is advantageous. From 2008 to 2012, credulous parents could purchase “Your Baby Can Read,” devised by kinesiologist Robert Titzer. This was “Mega-Reading” for children, sold on a late-night infomercial. Eventually the FTC caught up with it, decided it had no educational value, and enjoined Titzer and his company from selling products that claimed to teach babies to read. Titzer was fined $185 million, equal to the gross sales for “Your Baby Can Read” (!), with the stipulation that the fine would be suspended after $300,000 was paid (!!). For the FTC judgment, see “Your Baby Can LLC, et al.,” FTC, https://goo.gl/WliG4L. As in the Mega-Reading case, the plaintiff continues to ply his trade, selling products including “Your Baby Can Learn!,” “Your Child Can Read!,” and “Your Baby/Child Can Discover!” Caveat emptor.
117 Fred Morrison: Morrison, Alberts, & Griffith (1997).
Design of enrollment cutoff study.
118 Asked to pick out: Ehri (1995).
118 “a simple view of reading”: Hoover & Gough (1990); Gough, Hoover, & Peterson (1996). The main weakness in Gough’s theory is that it did not make sufficient room for the ways that the components influence each other. Vocabulary, for example, is jointly determined by spoken language and reading. Vocabulary can also be considered a component of both basic skills and comprehension.
120 such difficulties are commonly reported: Hulme et al. (2007).
121 At the neural level: Munakata & McClelland (2003).
121 It should also be appreciated: Escoffier & Di Giacomo’s “Take Away the A” (2014) is a phonemic awareness alphabet book. “snow without the s falls now,” it says, but “snow” without the “s” says “no”.