What Reading Does for the Mind

ANNE E. CUNNINGHAM and KEITH E. STANOVICH

concept to reading, we see that very early in What Reading the reading process poor readers, who experience greater difficulty in breaking the spelling- Does for the Mind to-sound code, begin to be exposed to much less text than their more skilled peers (Allington, 1984; Biemiller, 1977–1978). Anne E. Cunningham is visiting associate professor in cognition Further exacerbating the problem is the fact and development in the graduate school of education at the that less-skilled readers often find themselves University of California, Berkeley. Her research examines the in materials that are too difficult for them cognitive and motivational processes that underlie reading ability and the cognitive consequences of reading skill and engagement. (Allington, 1977, 1983, 1984; Gambrell, Keith E. Stanovich is professor of applied psychology at the Wilson, & Gantt, 1981). The combination of University of Toronto/Ontario Institute for Studies in Education. deficient decoding skills, lack of practice, and His recent awards include the Sylvia Scribner Award from the difficult materials results in unrewarding early American Educational Research Association and the Oscar S. reading experiences that lead to less involve- Causey Award from the National Research Conference for his ment in reading-related activities. Lack of distinguished and substantial contributions to literacy research. exposure and practice on the part of the less- This research was supported by a Spencer Foundation Small skilled reader delays the development of auto- Grant to Anne E. Cunningham and grant No. 410-95-0315 from maticity and speed at the word recognition the Social Sciences and Humanities Research Council of Canada level. Slow, capacity-draining word recognition to Keith E. Stanovich. processes require cognitive resources that should be allocated to comprehension. Thus, Reading has cognitive consequences that extend reading for meaning is hindered; unrewarding beyond its immediate task of lifting meaning reading experiences multiply; and practice is from a particular passage. Furthermore, these avoided or merely tolerated without real cogni- consequences are reciprocal and exponential in tive involvement. nature. Accumulated over time—spiraling either upward or downward—they carry profound The disparity in the reading experiences of chil- implications for the development of a wide dren of varying skill may have many other con- range of cognitive capabilities. sequences for their future reading and cognitive development. As skill develops and word recog- Concern about the reciprocal influences of reading achievement has been elucidated through discussions of so-called “Matthew effects” in academic achievement (Stanovich, 1986; Walberg & Tsai, 1983). The term Journal of Direct Instruction, Vol. 1, No. 2, pp. 137–149. “Matthew effects” is taken from the Biblical Reprinted with permission from The American Federation passage that describes a rich-get-richer and of Teachers. American Educator, Vol. 22, No. 1–2, pp. 8–15. Address correspondence to Anne E. Cunningham and Keith poor-get-poorer phenomenon. Applying this E. Stanovich at [email protected].

Journal of Direct Instruction 137 nition becomes less resource demanding and Theoretical Reasons more automatic, more general language skills, such as vocabulary, background knowledge, to Expect Positive familiarity with complex syntactic structures, etc., become the limiting factor on reading abili- Cognitive Consequences ty (Chall, 1983; Sticht, 1979). But the sheer from Reading Volume volume of reading done by the better reader has the potential to provide an advantage even here In certain very important cognitive domains, if—as our research suggests—reading a lot there are strong theoretical reasons to expect a serves to develop these very skills and knowl- positive and unique effect of avid reading. Vocabulary development provides a case in edge bases (Cunningham & Stanovich, 1997; point. Most theorists are agreed that the bulk of Echols, West, Stanovich, & Zehr, 1996; vocabulary growth during a child’s lifetime Stanovich & Cunningham, 1992, 1993). From occurs indirectly through language exposure the standpoint of a reciprocal model of reading rather than through direct teaching (Miller & development, this means that many cognitive Gildea, 1987; Nagy & Anderson, 1984; Nagy, differences observed between readers of differ- Herman, & Anderson, 1985; Sternberg, 1985, ing skill may in fact be consequences of differen- 1987). Furthermore, many researchers are con- tial practice that itself resulted from early differ- vinced that reading volume, rather than oral lan- ences in the speed of initial reading acquisition. guage, is the prime contributor to individual dif- The increased reading experiences of children ferences in children’s vocabularies (Hayes, 1988; who master the spelling-to-sound code early Hayes & Ahrens, 1988; Nagy & Anderson, 1984; thus might have important positive feedback Nagy & Herman, 1987; Stanovich, 1986). effects that are denied the slowly progressing The theoretical reasons for believing that read- reader. In our research, we have begun to ing volume is a particularly effective way of explore these reciprocal effects by examining expanding a child’s vocabulary derive from the the role that reading volume plays in shaping differences in the statistical distributions of the mind and will share many of our findings in words that have been found between print and this article. oral language. Some of these differences are illustrated in Table 1, which displays the results We should say at the outset that the complexity of some of the research of Hayes and Ahrens of some of the work we will describe in this arti- (1988), who have analyzed the distributions of cle was necessitated in large part by the fact that words used in various contexts. it is difficult to tease apart the unique contribution that reading volume affords. One of the dif- The table illustrates the three different cate- ficulties is that levels of reading volume are cor- gories of language that were analyzed: written related with many other cognitive and behavioral language sampled from genres as difficult as sci- characteristics. Avid readers tend to be different entific articles and as simple as preschool books; from nonreaders on a wide variety of cognitive words spoken on television shows of various skills, behavioral habits, and background vari- types; and adult speech in two contexts varying ables (Guthrie, Schafer, & Hutchinson, 1991; in formality. The words used in the different Kaestle, 1991; Zill & Winglee, 1990). Attributing contexts were analyzed according to a standard any particular outcome to reading volume is thus frequency count of English (Carroll, Davies, & extremely difficult. Richman, 1971). This frequency count ranks

138 Summer 2001 the 86,741 different word forms in English while talking or watching television is illustrat- according to their frequency of occurrence in a ed in the second column of Table 1. The col- large corpus of written English. So, for example, umn lists how many rare words per 1000 are the word “the” is ranked number 1, the 10th contained in each of the categories. A rare word most frequent word is “it,” the word “know” is is defined as one with a rank lower than 10,000; ranked 100, the word “pass” is ranked 1,000, roughly a word that is outside the vocabulary of the word “vibrate” is 5,000th in frequency, the a fourth to sixth grader. For vocabulary growth word “shrimp” is 9,000th in frequency, and the to occur after the middle grades, children must word “amplifier” is 16,000th in frequency. The be exposed to words that are rare by this defini- first column, labeled Rank of Median Word, is tion. Again, it is print that provides many more simply the frequency rank of the average word such word-learning opportunities. Children’s (after a small correction) in each of the categories. So, for example, the average word in children’s books was ranked 627th most frequent in Table 1 the Carroll et al. word count; the average word in popular magazines was ranked 1,399th most Selected Statistics for Major Sources frequent; and the average word in the abstracts of scientific articles had, not surprisingly, a very of Spoken and Written Language low rank (4,389). (Sample Means)

What is immediately apparent is how lexically Rank Rare impoverished is most speech, as compared to of Words written language. With the exception of the Median per special situation of courtroom testimony, aver- Word l000 age frequency of the words in all the samples of I. Printed texts oral speech is quite low, hovering in the Abstracts of scientific articles 4389 128.0 400–600 range of ranks. Newspapers 1690 68.3 Popular magazines 1399 65.7 The relative rarity of the words in children’s Adult books 1058 52.7 books is, in fact, greater than that in all of the Comic books 867 53.5 adult conversation, except for the courtroom Children’s books 627 30.9 testimony. Indeed, the words used in children’s Preschool books 578 16.3 books are considerably rarer than those in the speech on prime-time adult television. The cat- II. Television texts egories of adult reading matter contain words Popular prime-time adult shows 490 22.7 that are two or three times rarer than those Popular prime-time children’s 543 20.2 heard on television. shows Cartoon shows 598 30.8 These relative differences in word rarity have Mr. Rogers and Sesame Street 413 2.0 direct implications for vocabulary development. If most vocabulary is acquired outside of formal III.Adult speech Expert witness testimony 1008 28.4 teaching, then the only opportunities to acquire College graduates to friends, 496 17.3 new words occur when an individual is exposed spouses to a word in written or oral language that is outside his/her current vocabulary. That this will Adapted from Hayes and Ahrens (1988). happen vastly more often while reading than

Journal of Direct Instruction 139 books have 50 percent more rare words in them of written and spoken words reveals this argu- than does adult prime-time television and the ment to be patently false. Table 2 presents a list conversation of college graduates. Popular maga- of words that do not occur at all in two large cor- zines have roughly three times as many opportu- pora of oral language (Berger, 1977; Brown, nities for new word learning as does prime time 1984), but that have appreciable frequencies in television and adult conversation. Assurances by a written frequency count (Francis & Kucera, some educators that “What they read and write 1982). The words participation, luxury, maneuver, may make people smarter, but so will any activi- provoke, reluctantly, relinquish, portray, equate, hor- ty that engages the mind, including interesting mone, exposure, display, invariably, dominance, literal, conversation” (Smith, 1989) are overstated, at legitimate, and infinite are not unnecessary append- least when applied to the domain of vocabulary ages, concocted to exclude those who are unfa- learning. The data in Table 1 indicate that con- miliar with them. They are words that are neces- versation is not a substitute for reading. sary to make critical distinctions in the physical and social world in which we live. Without such It is sometimes argued or implied that the type lexical tools, one will be severely disadvantaged of words present in print but not represented in in attaining one’s goals in an advanced society speech are unnecessary words—jargon, academ- such as ours. As Olson (1986) notes: ic doublespeak, elitist terms of social advantage, or words used to maintain the status of the It is easy to show that sensitivity to the users but that serve no real functional purpose. subtleties of language are crucial to some A consideration of the frequency distributions undertakings. A person who does not clearly see the difference between an expression of intention and a promise or Table 2 between a mistake and an accident, or between a falsehood and a lie, should Examples of words that do not appear avoid a legal career or, for that matter, a in two large corpora of oral language theological one. (Berger, 1977; Brown, 1984) but that The large differences in lexical richness between speech and print are a major source of have appreciable frequencies in written individual differences in vocabulary develop- texts (Carroll, Davies & Richman, ment. These differences are created by the large variability among children in exposure to 1971; Francis & Kucera, 1982): literacy. Table 3 presents the data from a study of the out-of-school time use by fifth graders display literal conducted by Anderson, Wilson, and Fielding dominance legitimate (1988). From diaries that the children filled out dominant luxury daily over several months’ time, the investiga- exposure maneuver tors estimated how many minutes per day that equate participation individuals were engaged in reading and other equation portray activities while not in school. The table indi- gravity provoke cates that the child at the 50th percentile in hormone relinquish amount of independent reading was reading infinite reluctantly approximately 4.6 minutes per day, or about a invariably half an hour per week, over six times as much as the child at the 20th percentile in amount of

140 Summer 2001 reading time (less than a minute daily). Or, to even more potent source of the rich-get-richer take another example, the child at the 80th per- and poor-get-poorer achievement patterns. centile in amount of independent reading time Therefore, we have sought to examine the (14.2 minutes) was reading over twenty times unique contribution that independent or out-of- as much as the child at the 20th percentile. school reading makes toward reading ability, Anderson et al. (1988) estimated the children’s aspects of verbal intelligence, and general reading rates and used these, in conjunction knowledge about the world. As part of this with the amount of reading in minutes per day, research program, our research group has pio- to extrapolate a figure for the number of words neered the use of a measure of reading volume that the children at various percentiles were that has some unique advantages in investiga- reading. These figures, presented in the far tions of this kind (Cunningham and Stanovich, right of the table, illustrate the enormous differ- 1990; Stanovich and West, 1989). ences in word exposure that are generated by children’s differential proclivities toward read- In all, we developed two measures of adults’ ing. For example, the average child at the 90th reading volume and one for children’s reading percentile reads almost two million words per volume. Briefly, the children’s measure, named year outside of school, more than 200 times the Title Recognition Test (TRT), requires more words than the child at the 10th per- children to pick out the titles of popular chil- centile, who reads just 8,000 words outside of dren’s books from a list of titles that includes school during a year. To put it another way, the entire year’s out-of-school reading for the child at the 10th percentile amounts to just two days Table 3 reading for the child at the 90th percentile! These dramatic differences, combined with the Variation in Amount lexical richness of print, act to create large vocabulary differences among children. of Independent Reading

Independent Words Read Examining the Consequences Reading Per Year of Differential Degrees % Minutes Per Day 98 65.0 4,358,000 of Reading Volume 90 21.1 1,823,000 It is one thing to speculate on how these differ- 80 14.2 1,146,000 ences in reading volume may result in specific 70 9.6 622,000 cognitive consequences in domains like vocabu- 60 6.5 432,000 lary; it is another to demonstrate that these 50 4.6 282,000 effects are occurring. In our research, we have 40 3.2 200,000 sought empirical evidence for the specific 30 1.3 106,000 effects of reading volume, effects that do not 20 0.7 21,000 simply result from the higher cognitive abilities 10 0.1 8,000 and skills of the more avid reader. Although 20.00 there are considerable differences in amount of Adapted from Anderson, Wilson, and Fielding reading volume in school, it is likely that differ- (1988). ences in out-of-school reading volume are an

Journal of Direct Instruction 141 equal numbers of made-up titles. This task is hierarchical multiple regression to solve the inter- easy to administer to large numbers of children, pretive problem that avid readers excel in most it does not make large cognitive demands, and domains of verbal learning and that, therefore, its results are reliable—it is not possible for our measures of reading volume might be spuri- children to distort their responses toward what ously correlated to a host of abilities they perceive as socially desirable answers. (Cunningham & Stanovich, 1990, 1991; Because the number of wrong answers can be Stanovich & Cunningham, 1992, 1993; counted against correct ones, it is possible to Stanovich & West, 1989). We have found that remove the effects of guessing from the results even when performance is statistically equated (see Cunningham & Stanovich, 1990; 1991; and for reading comprehension and general ability, Stanovich and West, 1989 for a full description reading volume is still a very powerful predictor of these instruments and a discussion of the of vocabulary and knowledge differences. Thus, logic behind them). The adults’ measures, we believe that reading volume is not simply an named the Author Recognition and Magazine indirect indicator of ability; it is actually a Recognition Test, have the same task require- potentially separable, independent source of ments and are described fully in Stanovich and cognitive differences. West (1989).

A score on the Title Recognition Test, of course, is not an absolute measure of children’s reading Reading Volume volume and previous literacy experiences, but it as a Contributor does provide us with an index of the relative differences in reading volume. This index enables to Growth in Verbal Skills us to ask what effects reading volume (rather than general reading comprehension and word In several studies, we have attempted to link decoding ability) has on intelligence, vocabulary, children’s reading volume to specific cognitive spelling, and children’s general knowledge. In outcomes after controlling for relevant general short, it enables us to ask the question, does abilities such as IQ. In a study of fourth-, fifth-, reading—in and of itself—shape the quality of and sixth-grade children, we examined whether our mind? reading volume accounts for differences in vocabulary development once controls for both The titles appearing on the TRT were selected general intelligence and specific verbal abilities from a sample of book titles generated in pilot were invoked (Cunningham & Stanovich, investigations by groups of children ranging in 1991). We employed multiple measures of age from second grade through high school. In vocabulary and controlled for the effects of age selecting the items that appear on any one ver- and intelligence. We also controlled for the sion of the TRT, an attempt was made to effect of another ability that may be more close- choose titles that were not prominent parts of ly linked to vocabulary acquisition mechanisms: classroom reading activities in these particular decoding ability. Decoding skill might mediate a schools. Because we wanted the TRT to probe relationship between reading volume and a vari- out-of-school rather than school-directed read- able like vocabulary size in numerous ways. ing, an attempt was made to choose titles that High levels of decoding skill, certainly a con- were not used in the school curriculum. tributor to greater reading volume, might provide relatively complete contexts for figuring In our technical reports on this work, we have out the meaning of words during reading. Thus, used a powerful statistical technique known as reading volume and vocabulary might be linked

142 Summer 2001 via their connection to decoding ability: Good reading volume. The results of our study again decoders read a lot and have the best context attest to the potency of reading volume. We available for inferring new words. This poten- found that reading volume made a significant tial linkage was accounted for by statistically contribution to multiple measures of vocabulary, controlling for decoding ability prior to inves- general knowledge, spelling, and verbal fluency tigating reading volume. But we found that even after reading comprehension ability and even after accounting for general intelligence nonverbal ability had been partialed out. and decoding ability, reading volume con- tributed significantly and independently to One way of demonstrating the conservative vocabulary knowledge in fourth-, fifth-, and nature of these analyses is illustrated in a longi- sixth-grade children. tudinal study that we have conducted (Cipielewski & Stanovich, 1992). We addressed These findings demonstrate that reading vol- the question of whether reading volume can ume, although clearly a consequence of devel- predict individual differences in growth in read- oped reading ability, is itself a significant con- ing comprehension from third grade to fifth tributor to the development of other aspects of grade. We found that reading volume predicted verbal intelligence. Such rich get-richer (and of variance in fifth-grade reading comprehension course their converse, poor-get poorer) effects ability after third-grade reading comprehension are becoming of increasing concern in the edu- scores had been removed. Thus, in removing cational community (Adams, 1990; Chall, 1989) the contribution of reading comprehension in and are playing an increasingly prominent role our adult studies, we are undoubtedly removing in theories of individual differences in reading some of the variance in variables such as vocab- ability and growth (Anderson, et al., 1988; ulary and general knowledge that is rightfully Chall, Jacobs, & Baldwin, 1990; Hayes, 1988; attributed to reading volume. Hayes & Ahrens, 1988; Juel, 1988, 1994; Stanovich 1986, 1989, 1993).

In a study we conducted involving college stu- Reading Volume and dents, we employed an even more stringent test Declarative Knowledge of whether reading volume is a unique predictor of verbal skill (Stanovich & Cunningham, 1992). In other studies, we have focused even more In this study we examined many of the same directly on content knowledge by addressing variables as in our study of fourth- to sixth-grade the issue of “Where Does Knowledge Come students. However, we decided to stack the From?”. Stanovich and Cunningham (1993) deck against reading volume by first removing examined general ability, reading volume, and any contribution of reading ability and general exposure to other media sources as determi- intelligence. By structuring the analyses in this nants of individual differences in content way, we did not mean to imply that reading vol- knowledge. This study contained a particularly ume is not a determinant of reading comprehen- stringent test of the role of reading volume and sion ability. Indeed, we argue that there are individual differences in knowledge acquisition grounds for believing that reading volume facili- among 268 college students. We administered tates growth in comprehension ability. However, five different measures of general knowledge to we wanted to construct the most conservative the students. Then we stacked the deck against analysis possible by deliberately allowing the reading volume once again by statistically enter- comprehension measure to steal some variance ing four measures of general ability before look- that is rightfully attributed to the measure of ing at the contribution of reading volume: high

Journal of Direct Instruction 143 school grade-point average, performance on an States senators were, more likely to know how intelligence test, an SAT-type mathematics test, many teaspoons are equivalent to one table- and an adult reading comprehension test. This spoon, were more likely to know what a stroke set of tasks surely exhausts the variance attrib- was, and what a closed shop in a factory was, etc. utable to any general ability construct; and, as One would be hard pressed to deny that at least one would expect, we found that general ability some of this knowledge is relevant to living in accounted for a substantial proportion of vari- the United States in the late 20th century. ance in the composite measure of general knowledge. Next we entered a composite meas- In other questions asked of these same stu- ure of exposure to television, but it did not dents, we attempted to probe areas that we account for any additional variance. However, a thought might be characterized by misinforma- composite index of reading volume accounted tion. We then attempted to trace the “cognitive for a substantial 37.1 percent of the variance anatomy” of this misinformation. One such when entered after the four ability measures question concerned the sizes of the world’s and television exposure. major religions and was designed to assess awareness of the multicultural nature of the This pattern was replicated in each of the five modern world. The question was phrased as fol- measures of general knowledge we employed, lows: “The 1986 Encyclopedia Britannica esti- including a homemade instrument we called the mates that there are approximately nine hun- Practical Knowledge Test. This task was dred million people in the world (not just the designed to address the criticism that our other United States) who identify themselves as measures of general knowledge were too aca- Christians. How many people in the world (not demic—that they tapped knowledge that was just the United States) do you think identify too esoteric or elitist and that was not useful in themselves as ?” Space was then provided on daily life. We didn’t think this was true; many the form for the subjects to make estimates of items on these measures were mundane and the number of Moslems, Jews, Buddhists, concrete questions such as “In what part of the Hindus, etc. body does the infection called pneumonia occur?” Nevertheless, in the Practical We will focus here on the estimates of Moslem Knowledge Test, we made an effort to devise and Jewish people because of our a priori questions that were directly relevant to daily liv- hypothesis that availability effects caused by ing in a technological society in the late twenti- televised coverage of Israel in the U.S. had eth century; for example, What does the carbu- skewed the perception of this ratio. While our retor in an automobile do? If a substance is car- sample’s median estimate of the number of cinogenic, it means that it is? After the Federal Jewish people (20 million) was quite close to Reserve Board raises the prime lending rate, the the actual figure of 18 million according to the interest that you will be asked to pay on a car 1990 Universal Almanac, the number of estimated loan will generally increase/decrease/ stay the Moslems—a mean of 10 million—was startlingly same? What vitamin is highly concentrated in low (817 million is the estimate in the Universal citrus fruits? When a stock exchange is in a “bear Almanac). For each participant in our study, we market,” what is happening? and so forth. calculated the ratio of the Moslem to Jewish estimates to see how many students were aware The results indicated that the more avid readers of the fact that the number of Moslems is an in our study—regardless of their general abili- order of magnitude larger (the actual estimated ties—knew more about how a carburetor worked, ratio is approximately 33:1 according to the were more likely to know who their United World Almanac; 45:1 according to the Universal

144 Summer 2001 Almanac). The median ratio in our sample was America is most often negatively associated 0.5. That is, 69.3 percent of our sample thought with knowledge acquisition. that there were more Jewish people in the world than Moslems. In another study, Stanovich, West, & Harrison (1995) examined a much older population in This level of inaccuracy is startling given that order to investigate the extent to which age- approximately 40 percent of our sample of 268 related growth in knowledge can be accounted students were attending one of the most selec- for by differences in reading volume. Although tive public institutions of higher education in much research effort has been expended on the United States (the University of California, describing cumulative growth in crystallized Berkeley). We have explored the correlates of intelligence (e.g., acquired knowledge such as this particular misconception in a variety of vocabulary and general information), we know ways. We looked at the performance on this little about the experiences that relate to question as a function of students’ level of read- knowledge growth in older individuals. For ing volume and television watching. We example, educational experience (years in observed a clear effect of reading volume on the school) is a predictor of intellectual functioning scores on the question and a significant effect of in older individuals (e.g., Schwartzman, Gold, television viewing, but the effects were in oppo- Andres, Arbuckle, & Chaikelson, 1987). It is site directions! Reading volume was associated assumed that education (which is received early with higher scores on the question, but televi- in life) in part determines the extent and quali- sion exposure was associated with lower scores. ty of many intellectual activities later in life. Scores among the group high in reading volume And it is presumably this intellectual activity as and low in television exposure were highest, one ages that is so crucial to the preservation of and the lowest scores were achieved by those cognitive capacities. Thus, while considerable high in television exposure and low in reading development of cognitive skills and abilities can volume. Our analyses confirmed that these result from formal educational experiences, it is relationships were not due to differences in the lifetime use of these skills that is assumed general ability. to have the beneficial effect.

Similarly, we have analyzed a variety of other In this study, Stanovich, et al. (1995) examined misconceptions in a number of other different the performance of college students and senior domains—including knowledge of World War citizens on general knowledge, vocabulary, II, the world’s languages, and the components working memory, syllogistic reasoning, and sev- of the federal budget—and all of them repli- eral measures of reading volume. The older cate the pattern shown for this question. The individuals outperformed the college students cognitive anatomy of misinformation appears to on the measures of general knowledge and be one of too little exposure to print (or read- vocabulary, but did significantly less well than ing) and over-reliance on television for infor- the college subjects on the working memory mation about the world. Although television and syllogistic reasoning tasks. This dissocia- viewing can have positive associations with tion between fluid intelligence (all-purpose knowledge when the viewing is confined to general problem-solving capacity) and crystal- public television, news, and/or documentary lized intelligence (general knowledge and material (Hall, Chiarello, & Edmondson, 1996; vocabulary) is a standard finding in the litera- West & Stanovich, 1991; West et al., 1993), ture (Baltes, 1987; Horn & Hofer, 1992; familiarity with the prime time television Salthouse, 1988). However, a series of analyses material that defines mass viewing in North indicated that when measures of reading vol-

Journal of Direct Instruction 145 ume were used as control variables, the positive first grade predicted these cognitive outcomes relationships between age and vocabulary and in the eleventh grade. We interpreted the read- age and declarative knowledge were eliminated ing volume measures administered in the (in contrast, the negative relationships between eleventh grade as cumulative indicators of vari- age and fluid abilities were largely unchanged). ance in reading volume that had taken place Thus, the results of this study are consistent many years earlier. Thus, we viewed the meas- with the conjecture that—in the domain of ver- ures as in some sense retrospective indicators bal abilities—reading a lot can even help to tapping the cumulative experiences and habits compensate for the normally deleterious effects of the students some distance in time before of aging! (See also, Smith, 1996.) actual assessment. As a result, we were able to examine how far this retrospective feature could be stretched. How Do We Become We addressed the question of whether the speed Avid Readers? of initial reading acquisition in the first grade could predict later tendencies to engage in Moving back again to the other end of the age reading activities even after differences in gen- spectrum, we switch focus to the question: eral cognitive abilities were controlled, as some Given that lifelong reading habits are such models of Matthew effects in educational strong predictors of verbal cognitive growth, achievement would predict (Chall, Jacobs, & what is it that predicts these habits? We’ve been Baldwin, 1990; Juel, 1994; Stanovich, 1986). We looking at reading volume as a predictor of read- statistically removed the contribution of ing comprehension and cognitive ability, but eleventh-grade reading comprehension ability, what predicts reading volume or avid reading? in order to remove the direct association between reading volume and current reading It is generally agreed that comprehension ability ability. Then we examined the contribution of and reading volume are in a reciprocal relation- three standardized measures of first grade read- ship. In an attempt to tease apart this reciprocal ing ability (decoding, word recognition, and relationship, we explored the linkages between comprehension) and observed that all three children’s first-grade reading and cognitive abili- measures predicted eleventh-grade reading vol- ties and eleventh-grade outcomes in a unique ume even after eleventh-grade reading compre- ten-year longitudinal study (Cunningham and hension ability had been partialed out! In con- Stanovich, 1997). Most of our earlier studies trast, we observed that first grade intelligence involved assessing contemporaneous relations, measures do not uniquely predict eleventh-grade but in this study, we examined the performance reading volume in the same way. Thus, this of a sample of students who had been tested as study showed us that an early start in reading is first graders (see Stanovich, Cunningham, and important in predicting a lifetime of literacy Feeman, 1984). About one half of these stu- experience—and this is true regardless of the dents were available ten years later for testing level of reading comprehension ability that the as eleventh graders. At this time, we adminis- individual eventually attains. tered a set of reading comprehension, cognitive ability, vocabulary, and general knowledge tasks, This is a stunning finding because it means as well as several measures of reading volume. that students who get off to a fast start in Additionally, some standardized test scores from reading are more likely to read more over the the intervening period were available. We were years, and, furthermore, this very act of read- therefore able to examine what variables in the ing can help children compensate for modest

146 Summer 2001 levels of cognitive ability by building their comprehension skills will build vocabulary and vocabulary and general knowledge. In other cognitive structures through reading. words, ability is not the only variable that counts in the development of intellectual We can thus elicit two crucial messages from functioning. Those who read a lot will our research findings. First, it is difficult to enhance their verbal intelligence; that is, overstate the importance of getting children off reading will make them smarter. to an early successful start in reading. We must ensure that students’ decoding and word recognition abilities are progressing solidly. Those The Reciprocal Effects who read well are likely to read more, thus set- of Reading Volume ting an upward spiral into motion. We can begin to sketch a view of the reciprocal Second, we should provide all children, regard- influences of early reading acquisition and read- less of their achievement levels, with as many ing volume as determinants of later reading reading experiences as possible. Indeed, this comprehension and other cognitive abilities. becomes doubly imperative for precisely those Early success at reading acquisition is one of the children whose verbal abilities are most in need keys that unlocks a lifetime of reading habits. of bolstering, for it is the very act of reading The subsequent exercise of this habit serves to that can build those capacities. An encouraging further develop reading comprehension ability message for teachers of low-achieving students in an interlocking positive feedback logic (Juel, is implicit here. We often despair of changing Griffith, & Gough, 1986; Juel, 1988; Snow, our students’ abilities, but there is at least one Barnes, Chandler, Goodman, & Hemphill, 1991; partially malleable habit that will itself develop Stanovich, 1986, 1993). Although it is difficult abilities—reading! to tease apart, we have attempted to trace the increasing divergence in children’s reading abili- References ty, as well as other cognitive outcomes, by exam- Adams, M. J. (1990). Beginning to read: Thinking and learn- ining both sides of the important role of recipro- ing about print. Cambridge, MA: MIT Press. cal causation. Our longitudinal study has per- Allington, R. L. (1977). If they don’t read much, how mitted us to observe these effects, whereby they ever gonna get good? Journal of Reading, 21, children who get out of the gate quickly—who 57–61. crack the spelling-to-sound code early on— Allington, R. L. (1983). The reading instruction provided readers of differing reading abilities. The appear to enter into a positive feedback loop. Elementary School Journal, 83, 548–559. One of the benefits of these reciprocating Allington, R. L. (1984). Content coverage and contextu- effects may be a level of participation in literacy al reading in reading groups. Journal of Reading activities that leads to a lifetime habit of read- Behavior, 16, 85–96. ing and thus sets the stage for future opportuni- Anderson, R. C., Wilson, P. T., & Fielding, L. G. (1988). ties—opportunities not enjoyed by children who Growth in reading and how children spend their time enter into this feedback loop more slowly. outside of school. Reading Research Quarterly, 23, 285–303. Baltes, P. B. (1987).Theoretical propositions of life-span A positive dimension of our research is that all developmental psychology: On the dynamics of our studies have demonstrated that reading between growth and decline. Developmental Psychology, yields significant dividends for everyone—not 23, 611–626. just for the “smart kids” or the more able read- Berger, K. W. (1977). The most common 100,000 words used in ers. Even the child with limited reading and conversations. Kent, Ohio: Herald Publishing House.

Journal of Direct Instruction 147 Biemiller, A. (1977–1978). Relationships between oral Hayes, D. P. (1988). Speaking and writing: Distinct pat- reading rates for letters, words, and simple text in terns of word choice. Journal of Memory and Language, the development of reading achievement. Reading 27, 572–585. Research Quarterly, 13, 223–253. Hayes, D. P. & Ahrens, M. (1988). Vocabulary simplifica- Brown, G. D. (1984). A frequency count of 190,000 tion for children: A special case of ‘motherese.’ words in the London-Lund Corpus of English Journal of Child Language, 15, 395–410. Conversation. Behavior Research Methods, Instruments, & Hall, V. C., Chiarello, K., & Edmondson, B. (1996). Computers, 16, 502–532. Deciding where knowledge comes from depends on Carroll, J. B., Davies, P., & Richman, B. (1971). Word fre- where you look. Journal of Educational Psychology, 88, quency book. Boston: Houghton Mifflin. 305–313. Horn, J. L., & Hofer, S. (1992). Major abilities and Chall, J. S. (1983). Stages of reading development. New development in the adult period. In R. J. Sternberg York: McGraw-Hill. & C.A. Berg (Eds.), Intellectual development (pp. 44–99). Chall. J. S. (1989) Learning to read: The great debate Cambridge: Cambridge University Press. twenty years later. A response to “Debunking the Juel, C. (1988). Learning to read and write: A longitudi- great phonics myth.” Phi Delta Kappan, 71, 521–538. nal study of fifty-four children from first through Chall, J. S., Jacobs, V., & Baldwin, L. (1990). The reading fourth grade. Journal of Educational Psychology, 80, crisis: Why poor children fall behind. Cambridge, MA: 437–447. Harvard University Press. Juel, C. (1994). Learning to read and write in one elementary Cipielewski, J., & Stanovich, K. E. (1992). Predicting school. New York: Springer-Verlag. growth in reading ability from children’s exposure to Juel, C., Griffith, P. L., & Gough, P. B. (1986). print. Journal of Experimental Child Psychology, 54, Acquisition of literacy: A longitudinal study of chil- 74–89. dren in first and second grade. Journal of Educational Cunningham, A. E., & Stanovich, K. E. (1990). Assessing Psychology, 78, 243–255. print exposure and orthographic processing skill in Kaestle, C. F. (1991). Literacy in the United States. New children: A quick measure of reading experience. Haven, CT.: Yale Press. Journal of Educational Psychology, 82, 733–740. Miller, G. A., & Gildea, P. M. (1987). How children learn Cunningham, A. E., & Stanovich, K. E. (1991).Tracking words. Scientific American, 257(3),94–99. the unique effects of print exposure in children: Nagy, W. E., & Anderson, R. C. (1984). How many words Associations with vocabulary, general knowledge, and are there in printed school English? Reading Research spelling. Journal of Educational Psychology, 83, 264–274. Quarterly, 19, 304–330. Cunningham, A. E., & Stanovich, K. E. (1997). Early Nagy, W. E., Herman, P. A. (1987). Breadth and depth of reading acquisition and its relation to reading experi- vocabulary knowledge: Implications for acquisition ence and ability 10 years later. Developmental and instruction. In M. G. McKeown & M. E. Curtis Psychology, 33(6), 934–945. (Eds.), The nature of vocabulary acquisition (pp. 19–35). Hillsdale, NJ: Erlbaum. Echols, L. D., West, R. F. Stanovich, K. E., & Zehr, K. S. Nagy, W. E., Herman, P. A. & Anderson, R. C. (1985). (1996). Using children’s literacy activities to predict Learning words from context. Reading Research growth in verbal cognitive skills: A longitudinal inves- Quarterly, 20, 233–253. tigation. Journal of Educational Psychology, 88, 296–304. Olson, D. R. (1986). Intelligence and literacy: The rela- Francis, W. N., & Kucera, H. (1982). Frequency analysis of tionships between intelligence and the technologies English usage: Lexicon and grammar. Boston: Houghton of representation and communication. In R. J. Mifflin. Sternberg & R. K. Wagner (Eds.), Practical intelligence Gambrell, L. B., Wilson, R. M., & Gantt, W. N. (1981). (pp. 338–360). Cambridge: Cambridge University Classroom observations of task-attending behaviors of Press. good and poor readers. Journal of Educational Research, Salthouse, T. A. (1988). Resource-reduction interpreta- 74, 400–404. tions of cognitive aging. Developmental Review, 8, Guthrie, J.T., Schafer, W. D. & Hutchinson, S. R. (1991). 238–272. Relation document literacy and prose literacy to Schwartzman, A., Gold, D., Andres, D., Arbuckle, T., & occupational and societal characteristics of young Chaikelson, J. (1987). Stability of intelligence: A 40- black and white adults. Reading Research Quarterly, 26, year follow-up. Canadian Journal of Psychology, 41, 30–48. 244–256.

148 Summer 2001 Smith, F. (1989). Overselling literacy. Phi Delta Kappan, Stanovich, K. E., & West, R. F., & Harrison, M. (1995). 70(5), 354. Knowledge growth and maintenance across the life Smith, M. C. (1996). Differences in adults’ reading prac- span: The role of print exposure. Developmental tices and literacy proficiencies. Reading Research Psychology, 31, 811–826. Quarterly, 31, 196–219. Sternberg, R. J. (1985). Beyond IQ: A triarchic theory of Snow, C. E., Barnes, W. S., Chandler, J., Goodman, L., & human intelligence. Cambridge: Cambridge University Hemphill, L. (1991). Unfulfilled expectations: Home and Press. school influences on literacy. Cambridge, MA: Harvard Sternberg, R. J. (1987). Most vocabulary is learned from University Press. context. In M. G. McKeown & M. E. Curtis (Eds.), Stanovich, K. E. (1986). Matthew effects in reading: The nature of vocabulary acquisition (pp. 89–105). Some consequences of individual differences in the Hillsdale, NJ: Erlbaum Associates. acquisition of literacy. Reading Research Quarterly, 21, Sticht, T. (1979). Applications of the audread model to 36–407. reading evaluation and instruction. In L. B. Resnick Stanovich, K. E. (1989). Has the learning disabilities & P. Weaver (Eds.), Theory and practice of early reading field lost its intelligence? Journal of Learning (pp. 209–226). Hillsdale, NJ: Erlbaum Associates. Disabilities, 22, 487–492. Walberg, H. J., & Tsai, S. (1983). Matthew effects in Stanovich K. E. (1993). Does reading make you smarter? education. American Educational Research Journal, 20, Literacy and the development of verbal intelligence. 359–373. In H. Reese (Ed.), Advances of child development and West, R. F., & Stanovich, K. E. (1991).The incidental behavior, 24, 133–180. San Diego, CA: Academic acquisition of information from reading. Psychological Press. Science, 2, 32–330. Stanovich, K. E., & Cunningham, A. E. (1992). Studying West, R. F., Stanovich, K. E., & Mitchell, H. R. (1993). the consequences of literacy within a literate society: Reading in the real world and its correlates. Reading The cognitive correlates of print exposure. Memory & Research Quarterly, 28, 34–50. Cognition, 20, 51–68. Zill, N., & Winglee, W. (1990). Who reads literature? Stanovich, K. E., & Cunningham, A. E. (1993). Where Cabin John, MD: Seven Locks Press. does knowledge come from? Specific associations between print exposure and information acquisition. Journal of Educational Psychology, 85, 211–229. Stanovich, K. E., Cunningham, A. E., & Feeman, D. J. (1984). Intelligence, cognitive skills, and early reading progress. Reading Research Quarterly, 19, 278–303. Stanovich, K. E., & West, R. F. (1989). Exposure to print and orthographic processing. Reading Research Quarterly, 24, 402–433.

Journal of Direct Instruction 149