Errata etc.

Typos, errors, things I could have explained better explained better.  Will be corrected in the paperback version.

1  The Hebrew examples on p. 49 are printed wrong. Unusual editing error: Somewhere between my ms. and the published version, one of the copy-editors (there were several)  overcorrected them, reversing the order, which I did not catch. The perils of working in languages you don’t read.  Spotted by @allenshull on Twitter; thanks Allen, alas.  Here are the correct examples and others from the K-T-B group, taken from Wikipedia’s K-T-B page–which is where I got the examples in the first place.

ktb

2  Katakana vs. hiragana:  Mortifying! My colleague Karalyn Patterson has caught the fact that the descriptions of the two types of kana on p. 49 are reversed.   It’s katakana that are used for loanwords, scientific terms, and other specialized vocabulary, and hiragana for native words and grammatical elements for which there are no kanji. ごめんなさい, which means “I’m sorry,” according to Google translate.  Katakana are written with sharper, more angular strokes, as the text correctly states.

I’m afraid I rushed to rewrite the text when a grayscale version of this figure and its caption had to be deleted at the last moment: japanese-scripts-copy

3  Genes and SES

Genetic and environmental factors interact in complex ways.  Genes influence reading achievement via their impact on the development of relevant neural structures and circuits.  I briefly discussed some behavioral genetic research in the chapter on how well America reads (p. 246).  With hindsight, I would have written the section differently.  Here is a clarification.

In behavioral genetic (“twin”) studies, the relative contributions of environmental and genetic influences on reading ability vary with socioeconomic status (SES).

For low SES children, environment is the major factor; heritability (the genetic component) is low. For higher SES children the pattern is the reversed.

Diane Ravitch and others argue that the schools do well for middle and upper income children. Low achievement is due to poverty and thus an economic and social justice issue, not an educational one. I reject this argument for many reasons spelled out in the book. Poverty is indeed a major determinant of educational achievement. However, short of eliminating poverty, this argument offers nothing for the millions of low SES children in school.

Citing the behavioral genetic results, I wrote:

“This evidence puts Ravitch and colleagues’ arguments about the health of US education in an even worse light.  …. The system gets a high grade for mid- to high-SES children—the group for whom education has less influence than parentage. The system is failing students for whom the environment, of which education is a major part, exerts a greater influence. Ravitch and friends would take credit for successes in the segment of the population where educational quality has less impact and relinquish responsibility for the children for whom educational quality matters most.”

This description leaves out an important detail about this method: the behavioral genetic analysis only accounts for the variability of a characteristic (such as IQ or reading ability) within a specified population.  Say that Massachusetts 8th graders score an average of X on a math assessment.  The individual scores vary around this mean. The studies attempt to explain why these scores vary.  They do not address why the mean was at a particular level, or why it differs from the mean for 8th graders in some other state.

For children from lower income backgrounds, the environment explains more of the variability in their scores than does heritability. It’s not that they don’t have genes, only that the genetic component doesn’t account for why scores vary within this group.

For higher SES children, the impact of SES is consistently positive and it accounts for less of the variability in scores within this group. The estimated heritability therefore goes up. So the text should have said, “The system gets a high grade for mid- to high-SES children—the group for whom education explains less about achievement than parentage”.

These results may seem puzzling because SES is clearly relevant to differences between the groups—why the mean is higher for the higher SES group, for example. However, the behavioral genetic analysis does not address the causes of between-group differences.  It only applies to the variability of the scores within groups.

Quirks of behavioral genetic studies aside, the main point is that environmental factors affect educational outcomes for all children.  For lower SES children, many aspects of the environment are negative, including ones that can’t be easily or quickly changed. For these children it is especially important to invest in education, one thing that can be changed and have a huge impact.

If I could add one analogy:  the students who graduate from highly selective colleges such as the Ivies do well in the world, as indexed by income and other measures of well-being.  The quality of the college experience may have had relatively little to do with this.  The students who are admitted to such programs already have many advantages–they are disproportionately higher SES, they have attended better schools and received better counseling, they have demonstrated accomplishments. They are already highly likely to succeed. Outcomes for students who have fewer of these advantages going in depend more on the quality of the college experience.

The story is similar for the younger students in the behavioral genetic studies. For children from advantageous, higher SES backgrounds, environment accounts for relatively little of the residual variability in their performance. Under these conditions, it’s possible to detect the impact of variation on the genetic side.  For children from lower SES backgrounds, environment (which subsumes a huge number of highly variable factors) matters a great deal.  Against these effects, the impact of genetic variation is small.

Behavioral genetic studies have yielded important findings over the years, but they measure genetic effects indirectly (via “heritability”).  The approach is being superseded by studies of the actual genetic mechanisms, a field that is advancing rapidly though these mechanisms are ferociously complex.

For more information:

Willcutt et al. (2010). Understanding the complex etiologies of developmental disorders: Behavioral and molecular genetic approaches.  https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2953861.

Another fancy typo: The acknowledgement to my mother (!) on p. 308 has a kind of typo that psycholinguists often teach in class. You’ll probably see it, though I didn’t, nor did the copyeditors.

me-meHere’s another example. It may take a minute to find it.

as-as-error

Both examples involve the failure to detect when the last word in one line is repeated at the start of the next one.  It’s always a small word like ME or AS.  These are a type of repetition blindness.

The newspaper one is a double error. It was meant to be as AG secretary, as in “agriculture secretary,” as in “Doyle picks co-op vet as ag secretary”.  Someone wrote AS instead of AG, and then the AS AS repetition was missed.  (It was noticed when we called in to the newspaper to ask about it.)

The one in the book was a mistake,  not included for instructional purposes!  I’m sure that other inadvertent Easter Eggs of this sort will be found eventually.

There are a few other typos in the book that resulted from my editing the ms. up until the very last moment.  And then the moment after that one. Copy-editors were a great help but glitches happen.  As it says in the acknowledgements, the errors that remain are my responsibility. Sorry, readers.

5 Tone matters. In the Chinese examples on p. 48, the pronunciation of the foal character is given as jū in the figure, which is correct, but as jù in the text, which is wrong. The phonetic in foal is pronounced jù, but as in many cases it is only a partial clue to the pronunciation of the word, which has a different tone.

6 Text statistics. About the Landauer and Dumais theory of word learning on pp. 112-113. The theory is as described, but their own method of analyzing language statistics was somewhat different.  There should have been an endnote. Their method (called LSA) found the patterns in samples of text, which could be anything from newspaper articles to freshman English essays.  This “big bag of words” approach did not rely on information about the order of words, just co-occurrences.

The lion/tiger/lynx example invokes a simpler method, looking at the words that precede and follow a word.  Even simple trigrams—three word sequences—provide strong clues about word meanings.  Jon Willits, a former student of mine now at UC Riverside (languagelearninglab.org), conducted some excellent work using trigrams.

The important point is Frith’s, that similar words appear in similar contexts.  Contexts range along a continuum from very narrow (the immediately adjacent words) to wider (the words that co-occur anywhere within a window of a sentence or two) to much wider (the big bag of words).

People are good at tracking co-occurrences over the narrower contexts; about longer texts, not so clear.

Finnish, anyone?  I mentioned case marking to illustrate the complexity of Finnish morphology (p. 136). Grammatical case isn’t explained very well. The Wikipedia page on Finnish noun cases (yes, there is one) is much better.

8 Easter Egg or just a little too obscure?  p. 252: Thirteen Ways of Looking at a Blackbird, a poem by Wallace Stevens.

9 Number of words in English?  p. 89, the text says: “The Oxford English Dictionary has entries for about 170,000 words (in the entire history of the language)”.  Actually, 170,000 is the number of words in current usage. There are entries for another 47,000 or so archaic words.  As explained on the Oxford dictionaries web site:  https://goo.gl/i8SgyD.