This study is about the idea that reading is good for you. The notion seems uncontroversial; most people would agree that reading is a useful learning experience. Certainly, the teacherly arguments are familiar: Reading takes us beyond ourselves; we broaden our perspectives, learn new facts and come to a better understanding of the world and our place in it. Furthermore, so the argument goes, there is an important fringe benefit: reading increases our vocabulary knowledge. Texts introduce us to new words, and in many cases, we can deduce their meanings from the written context. Presumably, we remember some of these new meaning associations, especially if we continue to read and meet the new items in context again. It seems reasonable to assume that this beneficial by-product of reading is also available to learners reading in a second language. Indeed, it has been claimed that reading in an L2 is one of the main ways language learners acquire new vocabulary knowledge (Krashen, 1989).
Get your grade
or your money back
using our Essay Writing Service!
But in spite of its common-sense appeal, good experimental evidence for the view that L2 reading has vocabulary learning benefits has been hard to come by. Most of the available studies report only tiny gains in vocabulary knowledge as a result of L2 reading.
In our view, the slim evidence these studies provide is unsatisfactory on at least two counts. First, the slight gains seem at odds with the strong claims made for the power of reading to impart new vocabulary knowledge. This inconsistency raises the following question: Have the claims for the impact of reading in an L2 been exaggerated, or have experiments failed to capture learning effects adequately? A main goal of the thesis is to conduct a series of experiments that we believe will provide a more convincing answer to the question of how much vocabulary knowledge a learner can be expected to acquire in a reading event. With a clearer answer to this question than has been available so far, we can arrive at a better sense of the power of the process.
A second problem is that most of the available studies offer little in the way of explanation of the vocabulary learning that occurs as a by-product of reading. Much as we might expect, they indicate that learners do indeed pick up some knowledge of new vocabulary items from reading a text, but the studies have little to say about how or why the words are learned. A number of important questions remain unanswered: How much does a learner know about a word after meeting it in context? How long does the knowledge last, and what features of the text contribute to learning? So, in addition to answering questions about the products of reading, the thesis also seeks to answer these questions about the process of learning vocabulary through reading. A better understanding of how vocabulary learning proceeds can provide teachers and readers useful information about the types and amounts of reading that lead to optimum vocabulary learning results.
So although the basic question of whether reading is beneficial for vocabulary acquisition has already been answered in the affirmative, it is clear that some questions about learning through reading have not been answered satisfactorily and others have not been answered at all. It is our belief that the main key to gaining better insights is better experimental methodology. In the literature review in Chapter 3, we will argue that experimental design flaws and the use of insensitve measures are at least partly to blame for the thin evidence and weak explanatory power of previous investigations. Then, in our own experiments we will attempt overcome these methodological problems. So, to the extent that a thesis has a plot, our story is one of successive methodological improvements. Design innovations in one experiment bring new insights but also new problems which must be addressed in the next.
But before we turn to studies of L2 readers and the products and processes of vocabulary learning, we will take a closer look at our original common sense proposition, the notion that people learn new words through reading even when they are under no obligation to learn them. The next sections outline the logical basis for this assumption and trace the development of what has come to be known as the incidental vocabulary learning hypothesis (Nagy, Herman and Anderson, 1985).
Always on Time
Marked to Standard
1.2.1 The default position: explaining L1 vocabulary acquisition
The belief that meanings can be acquired from written contexts without the help of teachers has been documented as far back as St. Augustine (Nagy, et al., 1985). The logic underpinning this widely held view is as follows: Most people cannot recall being taught very many words of their first language either at home or at school, yet they know thousands. This knowledge must have come from somewhere, and this "somewhere" can only be the ambient language environment. In the case of very young children learning their first language, this environment consists of spoken language, but in our adult experience we are aware of knowing words that we have never heard anyone say, and that we can have encountered only in our reading.
Reading appears to play an important role in the L1 vocabulary development of school age children. The amount of vocabulary growth children achieve during their school years is dramatic. According to a conservative estimate based on figures by Goulden, Nation and Read (1990), speakers of English know roughly 20,000 word families by the time they reach university age, which suggests an average growth in vocabulary size of around 1000 per year. Nagy, Anderson and Herman (1987) assume a much higher figure of as many as 3000 words per year between the third and twelfth year of schooling. But the number that could possibly be taught in class in a given year is in the order of hundreds, not thousands of words. Beck, McKeown and Omanson (1987) reckon that the number of English words that could be taught effectively in a year of schooling amounts to 400 at most. Even with very approximate figures and disagreement among researchers about how to count word units, the overall picture is clear: People know many more words than they could possibly have been taught. Therefore, for want of a better explanation, the enormous increase in L1 vocabulary knowledge during the school years must be due to incidental acquisition through reading. This line of argument is known as the "default" explanation of vocabulary acquisition.
Although some words must also be learned from encounters with spoken language, analyses of corpora indicate that spoken language is not particularly rich in low frequency vocabulary (e.g. Meara, Lightbown & Halter, 1995; Ure, 1971; West & Stanovich, 1991). Common words and their many senses may be learned from speech but written text, which contains far more low frequency items, must necessarily be the source for the learning of the many less common items that the average person knows. The importance of written input can be gauged from a study by West and Stanovich (1991). In this investigation, American university students were tested on their ability to recognize names of magazines and authors (an indicator of exposure to print) and names of TV celebrities (an indicator of exposure to oral input). Results indicated that exposure to oral input facilitated far less vocabulary learning than reading did. High levels of magazine and author recognition were associated with strong performance on a test of vocabulary size.
The default position for L1 vocabulary learning through reading has been neatly summarized in paper by Landauer and Dumais (1997) as follows:
A typical American seventh grader knows the meanings of 10-15 words today that she didn't know yesterday. She must have acquired most of them as a result of reading, because: a) the majority of English words are used only in print, b) she already knew well almost all of the words she would have encountered in speech, and c) she learned less than one word by direct instruction. (p. 211)
1.2.2 The default position and L2 vocabulary acquisition
Even though learners may have difficulty reading in an L2, it is widely assumed that L2 learners will experience the word learning benefits of reading much as L1 readers do. Current language teaching methodology appears to endorse the notion; for instance, a recent manual for training teachers of English asks prospective teachers to evaluate the following statement: "Reading widely is one of the best ways to learn another language" (Willis, 1996, p. 8). In the discussion that follows, it is clear that readers are expected to agree with the statement (though the author cautions against learning only from reading). The vocabulary learning benefits of L2 reading are assumed to be considerable:
This Essay is
a Student's Work
Many successful learners find that reading is an excellent way of extending vocabulary, learning new phrases and consolidating grammar. Like extensive listening, reading provides rich exposure to language in use. (p. 8)
The assumption that second language vocabulary will be learned incidentally is also evident in materials designed for language learners. Raptis (1997) observes that activities designed to develop the skill of inferring meanings of unfamiliar words from context feature prominently and repeatedly in textbooks for learners of English. The message to the learner - either stated or implicit - is that there are more words to learn than can be taught in a language course or looked up in a dictionary, and therefore learners themselves must become responsible for their vocabulary growth. Rather than attempting to teach all the words, the premise is that training students how to work out the meanings of unfamiliar words from context will help them to acquire vocabulary incidentally as they read in the new language.
The fact that some L2 learners succeed in becoming very proficient in their second language and know many more words than could be taught in a course of instructed language study provides support for the default explanation of L2 vocabulary acquisition A study by Milton and Meara (1995) found that the English vocabulary size of learners in a study exchange program increased dramatically during a stay in the United Kingdom. The mean increase was 1326 words in six months, or about 2500 words per year. Since the participants needed to have a good knowledge of English to qualify for the program, they probably already knew most of the common words used in spoken interaction when they arrived. Therefore, we can assume that most of the items they learned after arriving in the UK were less common words likely to be encountered in reading.
1.2.3 Empirical evidence for acquiring L1 vocabulary through reading
Although the default position offers a reasonable explanation for how we come to know many more words than we have heard in speech or learned at school, it is no more than a logical inference. In the mid 1980s, a number of researchers set out to see whether classroom reading tasks resulted in demonstrable vocabulary knowledge gains. This research (Jenkins, Stein & Wysocki, 1984; Nagy, Anderson & Herman, 1987; Nagy, Herman & Anderson, 1985) succeeded in providing convincing evidence of a link between actual reading events and the learning of new words. The investigations are discussed in detail in the literature review in Chapter 3. Here, for the purposes of our historical overview, we will outline them briefly and highlight the main issues they raise.
The studies investigated the vocabulary gains of school age English speaking subjects reading short texts in their native language. In the 1985 study by Nagy et al., eighth-grade participants read one of two 1000-word texts each containing 15 unfamiliar words. After reading the text, they were tested on their knowledge of 30 words, 15 from the text they had read and 15 from the other text which they had not read. Comparisons of the reading and non-reading conditions indicated that participants picked up knowledge of new word meanings as a result of exposure to the experimental texts. Mean gains were very small, on the order of two or three words, but were found to be statistically significant.
Extrapolating from these findings, Nagy and his colleagues determined that the probability that a subject will be able to produce a full definition of a word that he or she has encountered once in a reading passage amounts to 10 percent. They calculated that the chances of being able to recognize a correct definition in a multiple-choice format are 15 percent. In other words, about every tenth encounter with an unfamiliar word in a reading text can be said to result in a learning event. Nagy and his colleagues go on to consider what this might mean in numbers of words acquired in a year. They estimate that the typical school age child reads about 1 million words per year. By applying their probabilities to this approximation and estimates of how often unknown words would occur, they arrive at growth figures of 3125 to 4875 words per year (Nagy et al., 1985, p. 250). These figures coincide rather neatly with prior estimates based on what would have to be achieved on a yearly basis in order to arrive at an adult-sized vocabulary. Thus, their vocabulary learning results appear to give substance to the claims of the default position.
On the basis of this study, Nagy et al. proposed the incidental acquisition hypothesis which posits that children learn the vast majority of the words they eventually come to know in their native language through exposure to them in meaningful contexts. The term incidental describes the word learning that occurs as a by-product while readers are actually devoting most of their attention to comprehending the information content of a text. Incidental acquisition is defined in the cognitive psychology literature (Anderson, 1990) as learning activities that individuals engage in when they are not intent on retaining the presented material (e.g. in preparation for a test).
To substantiate the claims of the incidental vocabulary learning hypothesis, Nagy and his colleagues conducted further studies of child vocabulary acquisition. In addition to considering a wider range of ages and reading abilities, a 1987 study improved on the earlier one by involving more participants, more reading passages, more target items, and a longer delay between reading and testing. Under these stiffer conditions, results still confirmed the finding of a small but significant amount of word growth (two or three words) that could be ascribed only to reading the texts. Growth occurred across ability and age groups.
Additional confirmation for the incidental vocabulary learning hypothesis through reading comes from a study by Jenkins, Stein and Wysocki (1984) that investigated school-age native speakers of English using a methodology similar to that of Nagy and his colleagues. They too found small but significant gains, on the order of one or two words. However, these results are less convincing because Jenkins et al. had their subjects read passages that were specially constructed to provide full support for target words so that meanings could be worked out easily. Therefore it can be argued that these findings would not apply to reading natural texts, which cannot be depended on to always provide full meaning support.
An earlier study by Saragi, Nation and Meister (1978) also provides evidence of subjects acquiring new vocabulary through reading, but like the study by Jenkins et al., it makes use of an atypical text. In this study, native speakers of English read a whole novel, Anthony Burgess's A Clockwork Orange, which contains a large number of Russian-based "nadsat" words devised by the author. Scores on a surprise posttest showed that participants were able to correctly identify the meanings of most of the nadsat items. The mean number of words acquired was 68.4, amounting to about three quarters of the 90 words tested. Some of these words occurred more than 100 times in the text and frequency of occurrence was found to be a factor in the sizable gains subjects made. Again, findings based on reading a special text may not generalize to acquisition that might occur reading a more "normal" one. But even though this study and the one by Jenkins et al. (1984) may have offered unusually favorable opportunities for incidental vocabulary acquisition, there is little doubt that subjects learned new words from reading the text. These studies, along with the meticulous work done by Nagy and his colleagues, offer convincing evidence that learners do acquire vocabulary knowledge as they read.
1.2.4 Empirical evidence for acquiring L2 vocabulary through reading
L2 investigations modeled on the L1 research discussed above began to appear in the late 1980s. The first published paper (Pitts, White and Krashen, 1989) was a replication of Saragi, Nation and Meister's 1978 study. In this experiment, learners of English read a passage from A Clockwork Orange for an hour and then took a multiple-choice test on 30 nadsat items. Comparisons to scores of testees who had not read the text indicated mean gains of about two items. An earlier unpublished study by Ferris (1988) used the same read-and-test methodology but a longer, book-length reading treatment. She found that the difference in mean gains between readers and non-readers amounted to about seven words. Other experiments with shorter texts (e.g. Day, Omura & Hiramatsu, 1991; Hulstijn, 1992) report gains of just one, two or three words. Dupuy and Krashen (1993) report a larger gain of almost seven words, but the text was supplemented with a video and input sources were not distinguished in this study. Generally, the studies confirm what we would expect: Learners can and do acquire new word knowledge incidentally through comprehension-focused reading in a second language.
However, none of the L2 studies we are aware of follow Nagy et al. (1985) in reporting incidental vocabulary gains in terms of probabilities. That is, they do not draw on findings to arrive at conclusions about the chances of new words being picked up incidentally by L2 readers. But it is possible to use the numbers of words tested in the studies and the mean gains reported to calculate the probability of tested word being picked up incidentally. Probabilities can then be expressed as pick-up rates. We analyzed five experiments following this procedure; the results are shown in Table 1.1 with approximate pick-up rates appearing in the last column. The rates range from 1 word correctly identified per 5 tested to 1 per 17. Taken together, the incidental word learning gains reported in these studies suggest that adult L2 learners pick up about 1 new word in 8. Thus, the L2 pick-up rate appears to be broadly consistent with the 1-in-10 rate (Nagy et al., 1985) established for L1 learners.
Word learning results in studies of incidental acquisition of L2 reading
No. & type of item tested
Mean no. of words learned
Probability of pick up
Approximate pick-up rate
1 in 7
Pitts, White & Krashen (1989)
1 in 17
Day, Omura & Hiramatsu (1991)
1 in 17
Hulstijn (1992) exp. 1
1 in 13
Dupuy & Krashen (1993)
1 in 5
Clearly these figure must be treated with caution. Our meta-analysis of L2 reading experiments is rough at best; in taking studies together, the analysis ignores differences in test types, reading treatments, languages and participants. Also, as we will see in Chapter 3, the rates are derived from experiments with serious methodological weaknesses. But despite these problems, we can safely conclude that the probability of learning a new L2 word from a single reading encounter is low. This raises the troubling question of what L2 learners can realistically expect to gain from extensive reading.
2. Are substantial benefits possible?
In the studies of L2 incidental vocabulary acquisition discussed above, word learning results are based on fairly small amounts of reading. In some of the experiments, participants read only a page or two of text. Therefore it is not surprising that the numbers of words the L2 readers learned were small. Far more worrisome is the fact that the probabilities derived from these studies appear to be very low. With a pick-up rate of only 1 new word in 8, acquiring knowledge of a substantial number of words depends on doing a very great deal of L2 reading. Nagy et al (1985) determined that at the acquisition rate of 1 new word in 10, L1 learners could learn the large number of words most adult native speakers know if they read one million words each year of their elementary education. Nagy and his colleagues claim that the million per annum figure is a reasonable estimate of the number of words school children typically encounter in a year.
Although we may question whether child learners actually accomplish this amount of reading in a year, it is clear that what may be a challenging goal for L1 readers amounts to a truly daunting task for L2 readers. Language learners are highly unlikely to read a million words a year in their L2, especially in the beginning stages when reading large amounts of text is hard work (Meara, 1988). A quick count based on multiplying words per page by numbers of pages in three novels on my shelf suggests that a reader would need to read 10 full length novels in order to encounter a million words, hardly a realistic goal for any but the most advanced L2 learners. Many learners have months rather than years to devote to acquiring an L2. It is clear that most will be hard pressed to read in the volume required for substantial incidental vocabulary learning benefits to accrue.
But is the picture really as dismal as this? We believe that there are at least two reasons for hope. First, the discussion of probabilities has focused on the chances of retaining a word's meaning in a single reading encounter, but perhaps a second or third reading encounter enhances the probability of retention dramatically. This may sound like optimistic speculation, but it is also a testable claim that a systematic exploration of the effects of multiple reading encounters could verify. Secondly, it is also possible that methodologically flawed experiments have underestimated the power of reading encounters with new words. For instance, meeting an unfamiliar L2 word in context may contribute to a considerable amount of learning of a type that simply does not register as a correct response on a word-knowledge test. This possibility can be explored by developing more sensitive measures than have been used in previous investigations. In brief, carefully designed experiments should be able to provide a clearer picture of the vocabulary learning benefits of reading in a second language than has been available so far.